From 3e80fa3dbc943de9b784fedc202ba38cf238f46d Mon Sep 17 00:00:00 2001 From: David Parks Date: Mon, 2 Nov 2009 19:55:37 +0000 Subject: Sync up with render-pipeline-7 ignore-dead-branch --- indra/newview/llvovolume.cpp | 183 +++++++++++++++++++++++++++++++++++-------- 1 file changed, 150 insertions(+), 33 deletions(-) (limited to 'indra/newview/llvovolume.cpp') diff --git a/indra/newview/llvovolume.cpp b/indra/newview/llvovolume.cpp index 583246c23e..ef3244b199 100644 --- a/indra/newview/llvovolume.cpp +++ b/indra/newview/llvovolume.cpp @@ -36,6 +36,8 @@ #include "llvovolume.h" +#include + #include "llviewercontrol.h" #include "lldir.h" #include "llflexibleobject.h" @@ -61,6 +63,7 @@ #include "llsky.h" #include "llviewercamera.h" #include "llviewertexturelist.h" +#include "llviewerobjectlist.h" #include "llviewerregion.h" #include "llviewertextureanim.h" #include "llworld.h" @@ -90,6 +93,12 @@ LLPointer LLVOVolume::sObjectMediaNavigateClient = static LLFastTimer::DeclareTimer FTM_GEN_TRIANGLES("Generate Triangles"); static LLFastTimer::DeclareTimer FTM_GEN_VOLUME("Generate Volumes"); +static LLFastTimer::DeclareTimer FTM_BUILD_MESH("Mesh"); +static LLFastTimer::DeclareTimer FTM_MESH_VFS("VFS"); +static LLFastTimer::DeclareTimer FTM_MESH_STREAM("Stream"); +static LLFastTimer::DeclareTimer FTM_MESH_FACES("Faces"); +static LLFastTimer::DeclareTimer FTM_VOLUME_TEXTURES("Volume Textures"); + LLVOVolume::LLVOVolume(const LLUUID &id, const LLPCode pcode, LLViewerRegion *regionp) : LLViewerObject(id, pcode, regionp), @@ -513,6 +522,7 @@ void LLVOVolume::updateTextures(LLAgent &agent) void LLVOVolume::updateTextures() { + LLFastTimer ftm(FTM_VOLUME_TEXTURES); // Update the pixel area of all faces if (!gPipeline.hasRenderType(LLPipeline::RENDER_TYPE_SIMPLE)) @@ -599,36 +609,60 @@ void LLVOVolume::updateTextures() if (isSculpted()) { LLSculptParams *sculpt_params = (LLSculptParams *)getParameterEntry(LLNetworkData::PARAMS_SCULPT); - LLUUID id = sculpt_params->getSculptTexture(); - mSculptTexture = LLViewerTextureManager::getFetchedTexture(id, TRUE, FALSE, LLViewerTexture::LOD_TEXTURE); - if (mSculptTexture.notNull()) - { - S32 lod = llmin(mLOD, 3); - F32 lodf = ((F32)(lod + 1.0f)/4.f); - F32 tex_size = lodf * MAX_SCULPT_REZ; - mSculptTexture->addTextureStats(2.f * tex_size * tex_size); - mSculptTexture->setBoostLevel(llmax((S32)mSculptTexture->getBoostLevel(), - (S32)LLViewerTexture::BOOST_SCULPTED)); - mSculptTexture->setForSculpt() ; - } - - S32 texture_discard = mSculptTexture->getDiscardLevel(); //try to match the texture - S32 current_discard = mSculptLevel; + LLUUID id = sculpt_params->getSculptTexture(); + U8 sculpt_type = sculpt_params->getSculptType(); - if (texture_discard >= 0 && //texture has some data available - (texture_discard < current_discard || //texture has more data than last rebuild - current_discard < 0)) //no previous rebuild + if ((sculpt_type & LL_SCULPT_TYPE_MASK) == LL_SCULPT_TYPE_MESH) + // mesh is a mesh { - gPipeline.markRebuild(mDrawable, LLDrawable::REBUILD_VOLUME, FALSE); - mSculptChanged = TRUE; + if (mSculptLevel == -2) + { + // get the asset please + gPipeline.loadMesh(this, id); + /*gAssetStorage->getAssetData(id, LLAssetType::AT_MESH, (LLGetAssetCallback)NULL, NULL, TRUE); + + if (gAssetStorage->hasLocalAsset(id, LLAssetType::AT_MESH)) + { + gPipeline.markRebuild(mDrawable, LLDrawable::REBUILD_VOLUME, FALSE); + mSculptChanged = TRUE; + }*/ + } } - if (gPipeline.hasRenderDebugMask(LLPipeline::RENDER_DEBUG_SCULPTED)) + else + // mesh is a sculptie + { + mSculptTexture = LLViewerTextureManager::getFetchedTexture(id, TRUE, FALSE, LLViewerTexture::LOD_TEXTURE); + + if (mSculptTexture.notNull()) + { + S32 lod = llmin(mLOD, 3); + F32 lodf = ((F32)(lod + 1.0f)/4.f); + F32 tex_size = lodf * MAX_SCULPT_REZ; + mSculptTexture->addTextureStats(2.f * tex_size * tex_size); + mSculptTexture->setBoostLevel(llmax((S32)mSculptTexture->getBoostLevel(), + (S32)LLViewerTexture::BOOST_SCULPTED)); + mSculptTexture->setForSculpt() ; + } + + S32 texture_discard = mSculptTexture->getDiscardLevel(); //try to match the texture + S32 current_discard = mSculptLevel; + + if (texture_discard >= 0 && //texture has some data available + (texture_discard < current_discard || //texture has more data than last rebuild + current_discard < 0)) //no previous rebuild + { + gPipeline.markRebuild(mDrawable, LLDrawable::REBUILD_VOLUME, FALSE); + mSculptChanged = TRUE; + } + + if (gPipeline.hasRenderDebugMask(LLPipeline::RENDER_DEBUG_SCULPTED)) { setDebugText(llformat("T%d C%d V%d\n%dx%d", texture_discard, current_discard, getVolume()->getSculptLevel(), mSculptTexture->getHeight(), mSculptTexture->getWidth())); } + } } if (getLightTextureID().notNull()) @@ -771,8 +805,10 @@ LLDrawable *LLVOVolume::createDrawable(LLPipeline *pipeline) } -BOOL LLVOVolume::setVolume(const LLVolumeParams &volume_params, const S32 detail, bool unique_volume) +BOOL LLVOVolume::setVolume(const LLVolumeParams ¶ms, const S32 detail, bool unique_volume) { + LLVolumeParams volume_params = params; + // Check if we need to change implementations bool is_flexible = (volume_params.getPathParams().getCurveType() == LL_PCODE_PATH_FLEXIBLE); if (is_flexible) @@ -811,21 +847,39 @@ BOOL LLVOVolume::setVolume(const LLVolumeParams &volume_params, const S32 detail if (isSculpted()) { - mSculptTexture = LLViewerTextureManager::getFetchedTexture(volume_params.getSculptID(), TRUE, FALSE, LLViewerTexture::LOD_TEXTURE); - if (mSculptTexture.notNull()) + // if it's a mesh + if ((volume_params.getSculptType() & LL_SCULPT_TYPE_MASK) == LL_SCULPT_TYPE_MESH) { - //ignore sculpt GL usage since bao fixed this in a separate branch - if (!gGLActive) + if (getVolume()->getNumVolumeFaces() == 0) { - gGLActive = TRUE; - sculpt(); - gGLActive = FALSE; + //mesh is not loaded, request pipeline load this mesh + LLUUID asset_id = volume_params.getSculptID(); + gPipeline.loadMesh(this, asset_id); } else { - sculpt(); + mSculptLevel = 1; + } + } + else // otherwise is sculptie + { + mSculptTexture = LLViewerTextureManager::getFetchedTexture(volume_params.getSculptID(), TRUE, FALSE, LLViewerTexture::LOD_TEXTURE); + + if (mSculptTexture.notNull()) + { + //ignore sculpt GL usage since bao fixed this in a separate branch + if (!gGLActive) + { + gGLActive = TRUE; + sculpt(); + gGLActive = FALSE; + } + else + { + sculpt(); + } + mSculptLevel = getVolume()->getSculptLevel(); } - mSculptLevel = getVolume()->getSculptLevel(); } } else @@ -838,6 +892,14 @@ BOOL LLVOVolume::setVolume(const LLVolumeParams &volume_params, const S32 detail return FALSE; } + + +void LLVOVolume::notifyMeshLoaded() +{ + mSculptChanged = TRUE; + gPipeline.markRebuild(mDrawable, LLDrawable::REBUILD_GEOMETRY); +} + // sculpt replaces generate() for sculpted surfaces void LLVOVolume::sculpt() { @@ -1039,6 +1101,11 @@ void LLVOVolume::updateFaceFlags() for (S32 i = 0; i < getVolume()->getNumFaces(); i++) { LLFace *face = mDrawable->getFace(i); + if (!face) + { + return; + } + BOOL fullbright = getTE(i)->getFullbright(); face->clearState(LLFace::FULLBRIGHT | LLFace::HUD_RENDER | LLFace::LIGHT); @@ -1111,6 +1178,10 @@ BOOL LLVOVolume::genBBoxes(BOOL force_global) for (S32 i = 0; i < getVolume()->getNumFaces(); i++) { LLFace *face = mDrawable->getFace(i); + if (!face) + { + continue; + } res &= face->genVolumeBBoxes(*getVolume(), i, mRelativeXform, mRelativeXformInvTrans, (mVolumeImpl && mVolumeImpl->isVolumeGlobal()) || force_global); @@ -2218,6 +2289,23 @@ BOOL LLVOVolume::isSculpted() const return FALSE; } +BOOL LLVOVolume::isMesh() const +{ + if (isSculpted()) + { + LLSculptParams *sculpt_params = (LLSculptParams *)getParameterEntry(LLNetworkData::PARAMS_SCULPT); + U8 sculpt_type = sculpt_params->getSculptType(); + + if ((sculpt_type & LL_SCULPT_TYPE_MASK) == LL_SCULPT_TYPE_MESH) + // mesh is a mesh + { + return TRUE; + } + } + + return FALSE; +} + BOOL LLVOVolume::hasLightTexture() const { if (getParameterEntryInUse(LLNetworkData::PARAMS_LIGHT_IMAGE)) @@ -2423,6 +2511,30 @@ F32 LLVOVolume::getBinRadius() { F32 radius; + F32 scale = 1.f; + + if (isSculpted()) + { + LLSculptParams *sculpt_params = (LLSculptParams *)getParameterEntry(LLNetworkData::PARAMS_SCULPT); + LLUUID id = sculpt_params->getSculptTexture(); + U8 sculpt_type = sculpt_params->getSculptType(); + + if ((sculpt_type & LL_SCULPT_TYPE_MASK) == LL_SCULPT_TYPE_MESH) + // mesh is a mesh + { + LLVolume* volume = getVolume(); + U32 vert_count = 0; + + for (S32 i = 0; i < volume->getNumVolumeFaces(); ++i) + { + const LLVolumeFace& face = volume->getVolumeFace(i); + vert_count += face.mVertices.size(); + } + + scale = 1.f/llmax(vert_count/1024.f, 1.f); + } + } + const LLVector3* ext = mDrawable->getSpatialExtents(); BOOL shrink_wrap = mDrawable->isAnimating(); @@ -2481,7 +2593,7 @@ F32 LLVOVolume::getBinRadius() radius = 8.f; } - return llclamp(radius, 0.5f, 256.f); + return llclamp(radius*scale, 0.5f, 256.f); } const LLVector3 LLVOVolume::getPivotPositionAgent() const @@ -3071,6 +3183,11 @@ void LLVolumeGeometryManager::rebuildMesh(LLSpatialGroup* group) face->getGeometryVolume(*volume, face->getTEOffset(), vobj->getRelativeXform(), vobj->getRelativeXformInvTrans(), face->getGeomIndex()); } + + if (!face) + { + llerrs << "WTF?" << llendl; + } } drawablep->clearState(LLDrawable::REBUILD_ALL); @@ -3354,7 +3471,7 @@ void LLVolumeGeometryManager::genDrawInfo(LLSpatialGroup* group, U32 mask, std:: } else { - if (LLPipeline::sRenderDeferred && te->getBumpmap()) + if (LLPipeline::sRenderDeferred && LLPipeline::sRenderBump && te->getBumpmap()) { registerFace(group, facep, LLRenderPass::PASS_BUMP); } -- cgit v1.3 From 10069e0e13e3214ba9320fdce915440b2e12f938 Mon Sep 17 00:00:00 2001 From: Dave Parks Date: Thu, 5 Nov 2009 19:58:10 -0600 Subject: Fix for prims all being 0 lod. Fix for dangling prim references. --- indra/llmath/llvolume.cpp | 3 ++- indra/llprimitive/llprimitive.cpp | 8 +++++++- indra/newview/llvovolume.cpp | 2 +- indra/newview/pipeline.cpp | 2 ++ 4 files changed, 12 insertions(+), 3 deletions(-) (limited to 'indra/newview/llvovolume.cpp') diff --git a/indra/llmath/llvolume.cpp b/indra/llmath/llvolume.cpp index ddd1b4b3db..33a8d33ce1 100644 --- a/indra/llmath/llvolume.cpp +++ b/indra/llmath/llvolume.cpp @@ -1692,9 +1692,10 @@ LLVolume::LLVolume(const LLVolumeParams ¶ms, const F32 detail, const BOOL ge mLODScaleBias.setVec(1,1,1); + generate(); + if (mParams.getSculptID().isNull() && params.getSculptType() == LL_SCULPT_TYPE_NONE) { - generate(); createVolumeFaces(); } } diff --git a/indra/llprimitive/llprimitive.cpp b/indra/llprimitive/llprimitive.cpp index 52265e7ad5..340f60ed1a 100644 --- a/indra/llprimitive/llprimitive.cpp +++ b/indra/llprimitive/llprimitive.cpp @@ -744,7 +744,10 @@ BOOL LLPrimitive::setVolume(const LLVolumeParams &volume_params, const S32 detai return TRUE; } -#if 0 +#if 0 + // #if 0'd out by davep + // this is a lot of cruft to set texture entry values that just stay the same for LOD switch + // or immediately get overridden by an object update message, also crashes occasionally U32 old_face_mask = mVolumep->mFaceMask; S32 face_bit = 0; @@ -943,6 +946,9 @@ BOOL LLPrimitive::setVolume(const LLVolumeParams &volume_params, const S32 detai } } #else + // build the new object + sVolumeManager->unrefVolume(mVolumep); + mVolumep = volumep; setNumTEs(mVolumep->getNumFaces()); #endif return TRUE; diff --git a/indra/newview/llvovolume.cpp b/indra/newview/llvovolume.cpp index 78fd371087..7666009a62 100644 --- a/indra/newview/llvovolume.cpp +++ b/indra/newview/llvovolume.cpp @@ -919,7 +919,7 @@ BOOL LLVOVolume::setVolume(const LLVolumeParams ¶ms, const S32 detail, bool { //mesh is not loaded, request pipeline load this mesh LLUUID asset_id = volume_params.getSculptID(); - gPipeline.loadMesh(this, asset_id); + gPipeline.loadMesh(this, asset_id, detail); } else { diff --git a/indra/newview/pipeline.cpp b/indra/newview/pipeline.cpp index 4ac95fa939..3554ceb7f0 100644 --- a/indra/newview/pipeline.cpp +++ b/indra/newview/pipeline.cpp @@ -8949,6 +8949,7 @@ void LLPipeline::loadMesh(LLVOVolume* vobj, LLUUID mesh_id, S32 detail) group->derefLOD(lod); return; } + group->derefLOD(lod); } //no lower LOD is a available, is a higher lod available? @@ -8961,6 +8962,7 @@ void LLPipeline::loadMesh(LLVOVolume* vobj, LLUUID mesh_id, S32 detail) group->derefLOD(lod); return; } + group->derefLOD(lod); } } } -- cgit v1.3 From 62233f22469cdc66042fc7bbbbd367dbb7212fde Mon Sep 17 00:00:00 2001 From: Dave Parks Date: Tue, 24 Nov 2009 07:38:04 -0600 Subject: Fix for copying of tetrahedrons in place of mesh LODs. Fix for bad tetrahedron bounding box. Bad fix for simultaneous loading of multiple LODs. --- indra/llmath/llvolume.cpp | 31 +++++++++++++++++++++++++-- indra/llmath/llvolume.h | 5 ++++- indra/llmath/llvolumemgr.cpp | 13 ++++++++++++ indra/llmath/llvolumemgr.h | 1 + indra/newview/llvovolume.cpp | 31 ++++----------------------- indra/newview/llvovolume.h | 1 - indra/newview/pipeline.cpp | 50 +++++++++++++++++++++++++++++++++++--------- indra/newview/pipeline.h | 6 ++++-- 8 files changed, 95 insertions(+), 43 deletions(-) (limited to 'indra/newview/llvovolume.cpp') diff --git a/indra/llmath/llvolume.cpp b/indra/llmath/llvolume.cpp index 84da1b3c62..515b1061f9 100644 --- a/indra/llmath/llvolume.cpp +++ b/indra/llmath/llvolume.cpp @@ -1676,6 +1676,7 @@ LLVolume::LLVolume(const LLVolumeParams ¶ms, const F32 detail, const BOOL ge mFaceMask = 0x0; mDetail = detail; mSculptLevel = -2; + mIsTetrahedron = FALSE; mLODScaleBias.setVec(1,1,1); // set defaults @@ -1905,7 +1906,7 @@ BOOL LLVolume::createVolumeFacesFromFile(const std::string& file_name) BOOL LLVolume::createVolumeFacesFromStream(std::istream& is) { mSculptLevel = -1; // default is an error occured - + LLSD header; { if (!LLSDSerialize::deserialize(header, is, 1024*1024*1024)) @@ -2048,6 +2049,11 @@ BOOL LLVolume::createVolumeFacesFromStream(std::istream& is) { U32 face_count = mdl.size(); + if (face_count == 0) + { + llerrs << "WTF?" << llendl; + } + mVolumeFaces.resize(face_count); for (U32 i = 0; i < face_count; ++i) @@ -2063,8 +2069,9 @@ BOOL LLVolume::createVolumeFacesFromStream(std::istream& is) //copy out indices face.mIndices.resize(idx.size()/2); - if (idx.empty()) + if (idx.empty() || face.mIndices.size() < 3) { //why is there an empty index list? + llerrs <<"WTF?" << llendl; continue; } @@ -2150,6 +2157,11 @@ void tetrahedron_set_normal(LLVolumeFace::VertexData* cv) cv[2].mNormal = nrm; } +BOOL LLVolume::isTetrahedron() +{ + return mIsTetrahedron; +} + void LLVolume::makeTetrahedron() { mVolumeFaces.clear(); @@ -2165,6 +2177,9 @@ void LLVolume::makeTetrahedron() LLVector3(x,-x,-x) }; + face.mExtents[0].setVec(-x,-x,-x); + face.mExtents[1].setVec(x,x,x); + LLVolumeFace::VertexData cv[3]; //set texture coordinates @@ -2225,12 +2240,19 @@ void LLVolume::makeTetrahedron() mVolumeFaces.push_back(face); mSculptLevel = 0; + mIsTetrahedron = TRUE; } void LLVolume::copyVolumeFaces(LLVolume* volume) { + if (volume->isTetrahedron()) + { + llerrs << "WTF?" << llendl; + } + mVolumeFaces = volume->mVolumeFaces; mSculptLevel = 0; + mIsTetrahedron = FALSE; } S32 const LL_SCULPT_MESH_MAX_FACES = 8; @@ -2615,6 +2637,11 @@ void LLVolume::sculpt(U16 sculpt_width, U16 sculpt_height, S8 sculpt_components, LLMemType m1(LLMemType::MTYPE_VOLUME); U8 sculpt_type = mParams.getSculptType(); + if (sculpt_type & LL_SCULPT_TYPE_MASK == LL_SCULPT_TYPE_MESH) + { + llerrs << "WTF?" << llendl; + } + BOOL data_is_empty = FALSE; if (sculpt_width == 0 || sculpt_height == 0 || sculpt_components < 3 || sculpt_data == NULL) diff --git a/indra/llmath/llvolume.h b/indra/llmath/llvolume.h index bf2854ede9..8e57f2e280 100644 --- a/indra/llmath/llvolume.h +++ b/indra/llmath/llvolume.h @@ -904,7 +904,8 @@ public: BOOL isUnique() const { return mUnique; } S32 getSculptLevel() const { return mSculptLevel; } - + void setSculptLevel(S32 level) { mSculptLevel = level; } + S32 *getTriangleIndices(U32 &num_indices) const; // returns number of triangle indeces required for path/profile mesh @@ -970,11 +971,13 @@ public: virtual BOOL createVolumeFacesFromFile(const std::string& file_name); virtual BOOL createVolumeFacesFromStream(std::istream& is); virtual void makeTetrahedron(); + virtual BOOL isTetrahedron(); protected: BOOL mUnique; F32 mDetail; S32 mSculptLevel; + BOOL mIsTetrahedron; LLVolumeParams mParams; LLPath *mPathp; diff --git a/indra/llmath/llvolumemgr.cpp b/indra/llmath/llvolumemgr.cpp index 61c5a0adc9..419e0015ba 100644 --- a/indra/llmath/llvolumemgr.cpp +++ b/indra/llmath/llvolumemgr.cpp @@ -375,6 +375,19 @@ F32 LLVolumeLODGroup::getVolumeScaleFromDetail(const S32 detail) return mDetailScales[detail]; } +S32 LLVolumeLODGroup::getVolumeDetailFromScale(const F32 detail) +{ + for (S32 i = 1; i < 4; i++) + { + if (mDetailScales[i] > detail) + { + return i-1; + } + } + + return 3; +} + F32 LLVolumeLODGroup::dump() { F32 usage = 0.f; diff --git a/indra/llmath/llvolumemgr.h b/indra/llmath/llvolumemgr.h index a78ea76a1a..f5dc4cd748 100644 --- a/indra/llmath/llvolumemgr.h +++ b/indra/llmath/llvolumemgr.h @@ -59,6 +59,7 @@ public: static S32 getDetailFromTan(const F32 tan_angle); static void getDetailProximity(const F32 tan_angle, F32 &to_lower, F32& to_higher); static F32 getVolumeScaleFromDetail(const S32 detail); + static S32 getVolumeDetailFromScale(F32 scale); LLVolume* refLOD(const S32 detail); BOOL derefLOD(LLVolume *volumep); diff --git a/indra/newview/llvovolume.cpp b/indra/newview/llvovolume.cpp index 4c126d8fd9..64c2e9e8ec 100644 --- a/indra/newview/llvovolume.cpp +++ b/indra/newview/llvovolume.cpp @@ -163,7 +163,6 @@ LLVOVolume::LLVOVolume(const LLUUID &id, const LLPCode pcode, LLViewerRegion *re mRelativeXformInvTrans.setIdentity(); mLOD = MIN_LOD; - mMeshSculptLevel = -2; mTextureAnimp = NULL; mVObjRadius = LLVector3(1,1,0.5f).length(); mNumFaces = 0; @@ -682,25 +681,7 @@ void LLVOVolume::updateTextureVirtualSize() LLUUID id = sculpt_params->getSculptTexture(); U8 sculpt_type = sculpt_params->getSculptType(); - if ((sculpt_type & LL_SCULPT_TYPE_MASK) == LL_SCULPT_TYPE_MESH) - // mesh is a mesh - { - if (mMeshSculptLevel == -2) - { - // get the asset please - gPipeline.loadMesh(this, id); - /*gAssetStorage->getAssetData(id, LLAssetType::AT_MESH, (LLGetAssetCallback)NULL, NULL, TRUE); - - if (gAssetStorage->hasLocalAsset(id, LLAssetType::AT_MESH)) - { - gPipeline.markRebuild(mDrawable, LLDrawable::REBUILD_VOLUME, FALSE); - mSculptChanged = TRUE; - }*/ - } - } - - else - // mesh is a sculptie + if ((sculpt_type & LL_SCULPT_TYPE_MASK) != LL_SCULPT_TYPE_MESH) { mSculptTexture = LLViewerTextureManager::getFetchedTexture(id, TRUE, LLViewerTexture::BOOST_NONE, LLViewerTexture::LOD_TEXTURE); @@ -902,14 +883,10 @@ BOOL LLVOVolume::setVolume(const LLVolumeParams ¶ms, const S32 detail, bool if ((volume_params.getSculptType() & LL_SCULPT_TYPE_MASK) == LL_SCULPT_TYPE_MESH) { if (getVolume()->getNumVolumeFaces() == 0) - { - //mesh is not loaded, request pipeline load this mesh + { + //load request not yet issued, request pipeline load this mesh LLUUID asset_id = volume_params.getSculptID(); - gPipeline.loadMesh(this, asset_id, detail); - } - else - { - mMeshSculptLevel = 1; + gPipeline.loadMesh(this, asset_id, LLVolumeLODGroup::getVolumeDetailFromScale(getVolume()->getDetail())); } } else // otherwise is sculptie diff --git a/indra/newview/llvovolume.h b/indra/newview/llvovolume.h index 4b247f3778..f14a71130b 100644 --- a/indra/newview/llvovolume.h +++ b/indra/newview/llvovolume.h @@ -298,7 +298,6 @@ private: LLFrameTimer mTextureUpdateTimer; S32 mLOD; BOOL mLODChanged; - S32 mMeshSculptLevel; BOOL mSculptChanged; F32 mSpotLightPriority; LLMatrix4 mRelativeXform; diff --git a/indra/newview/pipeline.cpp b/indra/newview/pipeline.cpp index 3410ad6559..0ee619244b 100644 --- a/indra/newview/pipeline.cpp +++ b/indra/newview/pipeline.cpp @@ -8913,23 +8913,29 @@ LLCullResult::sg_list_t::iterator LLPipeline::endAlphaGroups() void LLPipeline::loadMesh(LLVOVolume* vobj, LLUUID mesh_id, S32 detail) { + + if (detail < 0 || detail > 4) + { + return; + } + { LLMutexLock lock(mMeshMutex); //add volume to list of loading meshes - mesh_load_map::iterator iter = mLoadingMeshes.find(mesh_id); - if (iter != mLoadingMeshes.end()) + mesh_load_map::iterator iter = mLoadingMeshes[detail].find(mesh_id); + if (iter != mLoadingMeshes[detail].end()) { //request pending for this mesh, append volume id to list iter->second.insert(vobj->getID()); return; } //first request for this mesh - mLoadingMeshes[mesh_id].insert(vobj->getID()); + mLoadingMeshes[detail][mesh_id].insert(vobj->getID()); } if (gAssetStorage->hasLocalAsset(mesh_id, LLAssetType::AT_MESH)) { //already have asset, load desired LOD in background - mPendingMeshes.push_back(new LLMeshThread(mesh_id, vobj->getVolume())); + mPendingMeshes.push_back(new LLMeshThread(mesh_id, vobj->getVolume(), detail)); } else { //fetch asset and load when done @@ -8952,7 +8958,7 @@ void LLPipeline::loadMesh(LLVOVolume* vobj, LLUUID mesh_id, S32 detail) for (S32 i = detail-1; i >= 0; --i) { LLVolume* lod = group->refLOD(i); - if (lod && lod->getNumVolumeFaces() > 0) + if (lod && !lod->isTetrahedron() && lod->getNumVolumeFaces() > 0) { volume->copyVolumeFaces(lod); group->derefLOD(lod); @@ -8966,7 +8972,7 @@ void LLPipeline::loadMesh(LLVOVolume* vobj, LLUUID mesh_id, S32 detail) for (S32 i = detail+1; i < 4; ++i) { LLVolume* lod = group->refLOD(i); - if (lod && lod->getNumVolumeFaces() > 0) + if (lod && !lod->isTetrahedron() && lod->getNumVolumeFaces() > 0) { volume->copyVolumeFaces(lod); group->derefLOD(lod); @@ -8976,6 +8982,10 @@ void LLPipeline::loadMesh(LLVOVolume* vobj, LLUUID mesh_id, S32 detail) group->derefLOD(lod); } } + else + { + llerrs << "WTF?" << llendl; + } //nothing found, so make a tetrahedron volume->makeTetrahedron(); @@ -8992,12 +9002,22 @@ void LLPipeline::getMeshAssetCallback(LLVFS *vfs, } -LLPipeline::LLMeshThread::LLMeshThread(LLUUID mesh_id, LLVolume* target) +LLPipeline::LLMeshThread::LLMeshThread(LLUUID mesh_id, LLVolume* target, S32 detail) : LLThread("mesh_loading_thread") { mMeshID = mesh_id; mVolume = NULL; mDetail = target->getDetail(); + + if (detail == -1) + { + mDetailIndex = LLVolumeLODGroup::getVolumeDetailFromScale(target->getDetail()); + } + else + { + mDetailIndex = detail; + } + mTargetVolume = target; } @@ -9073,6 +9093,10 @@ void LLPipeline::notifyLoadedMeshes() for (std::list::iterator iter = mLoadedMeshes.begin(); iter != mLoadedMeshes.end(); ++iter) { //for each mesh done loading + + + + LLMeshThread* mesh = *iter; if (!mesh->isStopped()) @@ -9081,10 +9105,12 @@ void LLPipeline::notifyLoadedMeshes() continue; } + S32 detail = mesh->mDetailIndex; + //get list of objects waiting to be notified this mesh is loaded - mesh_load_map::iterator obj_iter = mLoadingMeshes.find(mesh->mMeshID); + mesh_load_map::iterator obj_iter = mLoadingMeshes[detail].find(mesh->mMeshID); - if (mesh->mVolume && obj_iter != mLoadingMeshes.end()) + if (mesh->mVolume && obj_iter != mLoadingMeshes[detail].end()) { //make sure target volume is still valid BOOL valid = FALSE; @@ -9109,6 +9135,10 @@ void LLPipeline::notifyLoadedMeshes() { mesh->mTargetVolume->copyVolumeFaces(mesh->mVolume); } + else + { + llwarns << "Mesh loading returned empty volume." << llendl; + } for (std::set::iterator vobj_iter = obj_iter->second.begin(); vobj_iter != obj_iter->second.end(); ++vobj_iter) { @@ -9120,7 +9150,7 @@ void LLPipeline::notifyLoadedMeshes() } } - mLoadingMeshes.erase(mesh->mMeshID); + mLoadingMeshes[detail].erase(mesh->mMeshID); } delete mesh; diff --git a/indra/newview/pipeline.h b/indra/newview/pipeline.h index 23e98aa0d6..a31379a209 100644 --- a/indra/newview/pipeline.h +++ b/indra/newview/pipeline.h @@ -676,8 +676,9 @@ public: protected: std::vector mSelectedFaces; + typedef std::map > mesh_load_map; - mesh_load_map mLoadingMeshes; + mesh_load_map mLoadingMeshes[4]; LLMutex* mMeshMutex; @@ -688,7 +689,8 @@ protected: LLVolume* mTargetVolume; LLUUID mMeshID; F32 mDetail; - LLMeshThread(LLUUID mesh_id, LLVolume* target); + S32 mDetailIndex; + LLMeshThread(LLUUID mesh_id, LLVolume* target, S32 detail = -1); ~LLMeshThread(); void run(); }; -- cgit v1.3 From 6d66910c6e2fbb25bf8b5c7b90e795f350342104 Mon Sep 17 00:00:00 2001 From: Dave Parks Date: Wed, 25 Nov 2009 11:35:41 -0600 Subject: Fix for spam on invalid mesh asset. Fix for index buffer overflow spam and crash in llvertexbuffer. --- indra/llmath/llvolume.cpp | 5 ----- indra/newview/llvovolume.cpp | 5 +++-- indra/newview/pipeline.cpp | 18 ++++++------------ 3 files changed, 9 insertions(+), 19 deletions(-) (limited to 'indra/newview/llvovolume.cpp') diff --git a/indra/llmath/llvolume.cpp b/indra/llmath/llvolume.cpp index 515b1061f9..3e547aec6f 100644 --- a/indra/llmath/llvolume.cpp +++ b/indra/llmath/llvolume.cpp @@ -2245,11 +2245,6 @@ void LLVolume::makeTetrahedron() void LLVolume::copyVolumeFaces(LLVolume* volume) { - if (volume->isTetrahedron()) - { - llerrs << "WTF?" << llendl; - } - mVolumeFaces = volume->mVolumeFaces; mSculptLevel = 0; mIsTetrahedron = FALSE; diff --git a/indra/newview/llvovolume.cpp b/indra/newview/llvovolume.cpp index 64c2e9e8ec..3e9db86cfa 100644 --- a/indra/newview/llvovolume.cpp +++ b/indra/newview/llvovolume.cpp @@ -882,7 +882,7 @@ BOOL LLVOVolume::setVolume(const LLVolumeParams ¶ms, const S32 detail, bool // if it's a mesh if ((volume_params.getSculptType() & LL_SCULPT_TYPE_MASK) == LL_SCULPT_TYPE_MESH) { - if (getVolume()->getNumVolumeFaces() == 0) + if (getVolume()->getNumVolumeFaces() == 0 || getVolume()->isTetrahedron()) { //load request not yet issued, request pipeline load this mesh LLUUID asset_id = volume_params.getSculptID(); @@ -924,7 +924,8 @@ BOOL LLVOVolume::setVolume(const LLVolumeParams ¶ms, const S32 detail, bool void LLVOVolume::notifyMeshLoaded() { mSculptChanged = TRUE; - gPipeline.markRebuild(mDrawable, LLDrawable::REBUILD_GEOMETRY); + gPipeline.markRebuild(mDrawable, LLDrawable::REBUILD_GEOMETRY, TRUE); + dirtySpatialGroup(TRUE); } // sculpt replaces generate() for sculpted surfaces diff --git a/indra/newview/pipeline.cpp b/indra/newview/pipeline.cpp index 0ee619244b..af3a35615c 100644 --- a/indra/newview/pipeline.cpp +++ b/indra/newview/pipeline.cpp @@ -1807,8 +1807,6 @@ void LLPipeline::rebuildPriorityGroups() assertInitialized(); - notifyLoadedMeshes(); - // Iterate through all drawables on the priority build queue, for (LLSpatialGroup::sg_list_t::iterator iter = mGroupQ1.begin(); iter != mGroupQ1.end(); ++iter) @@ -1888,6 +1886,8 @@ void LLPipeline::updateGeom(F32 max_dtime) // for now, only LLVOVolume does this to throttle LOD changes LLVOVolume::preUpdateGeom(); + notifyLoadedMeshes(); + // Iterate through all drawables on the priority build queue, for (LLDrawable::drawable_list_t::iterator iter = mBuildQ1.begin(); iter != mBuildQ1.end();) @@ -8913,7 +8913,6 @@ LLCullResult::sg_list_t::iterator LLPipeline::endAlphaGroups() void LLPipeline::loadMesh(LLVOVolume* vobj, LLUUID mesh_id, S32 detail) { - if (detail < 0 || detail > 4) { return; @@ -9093,10 +9092,6 @@ void LLPipeline::notifyLoadedMeshes() for (std::list::iterator iter = mLoadedMeshes.begin(); iter != mLoadedMeshes.end(); ++iter) { //for each mesh done loading - - - - LLMeshThread* mesh = *iter; if (!mesh->isStopped()) @@ -9131,14 +9126,13 @@ void LLPipeline::notifyLoadedMeshes() if (valid) { - if (mesh->mVolume->getNumVolumeFaces() > 0) - { - mesh->mTargetVolume->copyVolumeFaces(mesh->mVolume); - } - else + if (mesh->mVolume->getNumVolumeFaces() <= 0) { llwarns << "Mesh loading returned empty volume." << llendl; + mesh->mVolume->makeTetrahedron(); } + + mesh->mTargetVolume->copyVolumeFaces(mesh->mVolume); for (std::set::iterator vobj_iter = obj_iter->second.begin(); vobj_iter != obj_iter->second.end(); ++vobj_iter) { -- cgit v1.3 From 695969c77066de5032bdc9caefecf9b32b076b2f Mon Sep 17 00:00:00 2001 From: Dave Parks Date: Fri, 11 Dec 2009 14:47:11 -0600 Subject: HTTP Mesh fetch FTW.. still busted --- indra/llcommon/llthread.cpp | 31 +++- indra/llcommon/llthread.h | 22 +++ indra/llmath/llvolume.cpp | 7 +- indra/llmath/llvolume.h | 2 + indra/newview/CMakeLists.txt | 2 + indra/newview/app_settings/settings.xml | 2 +- indra/newview/llappviewer.cpp | 8 + indra/newview/llviewerregion.cpp | 1 + indra/newview/llvovolume.cpp | 15 +- indra/newview/pipeline.cpp | 255 +------------------------------- indra/newview/pipeline.h | 34 +---- 11 files changed, 93 insertions(+), 286 deletions(-) (limited to 'indra/newview/llvovolume.cpp') diff --git a/indra/llcommon/llthread.cpp b/indra/llcommon/llthread.cpp index b1175836b7..df7ea214cc 100644 --- a/indra/llcommon/llthread.cpp +++ b/indra/llcommon/llthread.cpp @@ -62,6 +62,9 @@ // //---------------------------------------------------------------------------- +U32 ll_thread_local sThreadID = 0; +U32 LLThread::sIDIter = 0; + // // Handed to the APR thread creation function // @@ -72,6 +75,8 @@ void *APR_THREAD_FUNC LLThread::staticRun(apr_thread_t *apr_threadp, void *datap // Set thread state to running threadp->mStatus = RUNNING; + sThreadID = threadp->mID; + // Run the user supplied function threadp->run(); @@ -90,6 +95,8 @@ LLThread::LLThread(const std::string& name, apr_pool_t *poolp) : mAPRThreadp(NULL), mStatus(STOPPED) { + mID = ++sIDIter; + // Thread creation probably CAN be paranoid about APR being initialized, if necessary if (poolp) { @@ -273,7 +280,7 @@ void LLThread::wakeLocked() //============================================================================ LLMutex::LLMutex(apr_pool_t *poolp) : - mAPRMutexp(NULL) + mAPRMutexp(NULL), mCount(0), mLockingThread(NO_THREAD) { //if (poolp) //{ @@ -305,7 +312,14 @@ LLMutex::~LLMutex() void LLMutex::lock() { + if (mLockingThread == sThreadID) + { //redundant lock + mCount++; + return; + } + apr_thread_mutex_lock(mAPRMutexp); + #if MUTEX_DEBUG // Have to have the lock before we can access the debug info U32 id = LLThread::currentID(); @@ -313,10 +327,18 @@ void LLMutex::lock() llerrs << "Already locked in Thread: " << id << llendl; mIsLocked[id] = TRUE; #endif + + mLockingThread = sThreadID; } void LLMutex::unlock() { + if (mCount > 0) + { //not the root unlock + mCount--; + return; + } + #if MUTEX_DEBUG // Access the debug info while we have the lock U32 id = LLThread::currentID(); @@ -324,6 +346,8 @@ void LLMutex::unlock() llerrs << "Not locked in Thread: " << id << llendl; mIsLocked[id] = FALSE; #endif + + mLockingThread = NO_THREAD; apr_thread_mutex_unlock(mAPRMutexp); } @@ -341,6 +365,11 @@ bool LLMutex::isLocked() } } +U32 LLMutex::lockingThread() const +{ + return mLockingThread; +} + //============================================================================ LLCondition::LLCondition(apr_pool_t *poolp) : diff --git a/indra/llcommon/llthread.h b/indra/llcommon/llthread.h index d8aa90de2e..2d553b5258 100644 --- a/indra/llcommon/llthread.h +++ b/indra/llcommon/llthread.h @@ -40,8 +40,17 @@ class LLThread; class LLMutex; class LLCondition; +#if LL_WINDOWS +#define ll_thread_local __declspec(thread) +#else +#define ll_thread_local __thread +#endif + class LL_COMMON_API LLThread { +private: + static U32 sIDIter; + public: typedef enum e_thread_status { @@ -82,6 +91,8 @@ public: apr_pool_t *getAPRPool() { return mAPRPoolp; } LLVolatileAPRPool* getLocalAPRFilePool() { return mLocalAPRFilePoolp ; } + U32 getID() const { return mID; } + private: BOOL mPaused; @@ -96,6 +107,7 @@ protected: apr_pool_t *mAPRPoolp; BOOL mIsLocalPool; EThreadStatus mStatus; + U32 mID; //a local apr_pool for APRFile operations in this thread. If it exists, LLAPRFile::sAPRFilePoolp should not be used. //Note: this pool is used by APRFile ONLY, do NOT use it for any other purposes. @@ -133,17 +145,27 @@ protected: class LL_COMMON_API LLMutex { public: + typedef enum + { + NO_THREAD = 0xFFFFFFFF + } e_locking_thread; + LLMutex(apr_pool_t *apr_poolp); // NULL pool constructs a new pool for the mutex ~LLMutex(); void lock(); // blocks void unlock(); bool isLocked(); // non-blocking, but does do a lock/unlock so not free + U32 lockingThread() const; //get ID of locking thread protected: apr_thread_mutex_t *mAPRMutexp; + mutable U32 mCount; + mutable U32 mLockingThread; + apr_pool_t *mAPRPoolp; BOOL mIsLocalPool; + S32 mLockCount; #if MUTEX_DEBUG std::map mIsLocked; #endif diff --git a/indra/llmath/llvolume.cpp b/indra/llmath/llvolume.cpp index fb2de92e35..44ff173502 100644 --- a/indra/llmath/llvolume.cpp +++ b/indra/llmath/llvolume.cpp @@ -1952,8 +1952,12 @@ BOOL LLVolume::createVolumeFacesFromStream(std::istream& is) } is.seekg(header[nm[lod]]["offset"].asInteger(), std::ios_base::cur); - + return unpackVolumeFaces(is, header[nm[lod]]["size"].asInteger()); +} + +BOOL LLVolume::unpackVolumeFaces(std::istream& is, S32 size) +{ U8* result = NULL; U32 cur_size = 0; @@ -1964,7 +1968,6 @@ BOOL LLVolume::createVolumeFacesFromStream(std::istream& is) const U32 CHUNK = 65536; - S32 size = header[nm[lod]]["size"].asInteger(); U8 *in = new U8[size]; is.read((char*) in, size); diff --git a/indra/llmath/llvolume.h b/indra/llmath/llvolume.h index 59c60ccd92..9970b24a94 100644 --- a/indra/llmath/llvolume.h +++ b/indra/llmath/llvolume.h @@ -971,6 +971,8 @@ protected: public: virtual BOOL createVolumeFacesFromFile(const std::string& file_name); virtual BOOL createVolumeFacesFromStream(std::istream& is); + virtual BOOL unpackVolumeFaces(std::istream& is, S32 size); + virtual void makeTetrahedron(); virtual BOOL isTetrahedron(); diff --git a/indra/newview/CMakeLists.txt b/indra/newview/CMakeLists.txt index 23f8b4b9ca..52adc740e3 100644 --- a/indra/newview/CMakeLists.txt +++ b/indra/newview/CMakeLists.txt @@ -284,6 +284,7 @@ set(viewer_SOURCE_FILES llmediaremotectrl.cpp llmemoryview.cpp llmenucommands.cpp + llmeshrepository.cpp llmetricperformancetester.cpp llmimetypes.cpp llmorphview.cpp @@ -789,6 +790,7 @@ set(viewer_HEADER_FILES llmediaremotectrl.h llmemoryview.h llmenucommands.h + llmeshrepository.h llmetricperformancetester.h llmimetypes.h llmorphview.h diff --git a/indra/newview/app_settings/settings.xml b/indra/newview/app_settings/settings.xml index 4ed2927ce6..087472661f 100644 --- a/indra/newview/app_settings/settings.xml +++ b/indra/newview/app_settings/settings.xml @@ -9998,7 +9998,7 @@ Type Boolean Value - 1 + 0 UseStartScreen diff --git a/indra/newview/llappviewer.cpp b/indra/newview/llappviewer.cpp index ddc818172d..a56e86c979 100644 --- a/indra/newview/llappviewer.cpp +++ b/indra/newview/llappviewer.cpp @@ -48,6 +48,7 @@ #include "llwindow.h" #include "llviewerstats.h" #include "llmd5.h" +#include "llmeshrepository.h" #include "llpumpio.h" #include "llmimetypes.h" #include "llslurl.h" @@ -1266,6 +1267,9 @@ bool LLAppViewer::cleanup() llinfos << "Cleaning Up" << llendflush; + // shut down mesh streamer + gMeshRepo.shutdown(); + // Must clean up texture references before viewer window is destroyed. LLHUDManager::getInstance()->updateEffects(); LLHUDObject::updateAll(); @@ -1678,6 +1682,9 @@ bool LLAppViewer::initThreads() mFastTimerLogThread->start(); } + // Mesh streaming and caching + gMeshRepo.init(); + // *FIX: no error handling here! return true; } @@ -2394,6 +2401,7 @@ bool LLAppViewer::initWindow() gSavedSettings.saveToFile( gSavedSettings.getString("ClientSettingsFile"), TRUE ); gPipeline.init(); + stop_glerror(); gViewerWindow->initGLDefaults(); diff --git a/indra/newview/llviewerregion.cpp b/indra/newview/llviewerregion.cpp index 8317837520..f173149bf4 100644 --- a/indra/newview/llviewerregion.cpp +++ b/indra/newview/llviewerregion.cpp @@ -1432,6 +1432,7 @@ void LLViewerRegion::setSeedCapability(const std::string& url) capabilityNames.append("FetchLib"); capabilityNames.append("FetchLibDescendents"); capabilityNames.append("GetTexture"); + capabilityNames.append("GetMesh"); capabilityNames.append("GroupProposalBallot"); capabilityNames.append("HomeLocation"); capabilityNames.append("MapLayer"); diff --git a/indra/newview/llvovolume.cpp b/indra/newview/llvovolume.cpp index 7debfb9186..f66f0c2d72 100644 --- a/indra/newview/llvovolume.cpp +++ b/indra/newview/llvovolume.cpp @@ -70,6 +70,7 @@ #include "llsdutil.h" #include "llmediaentry.h" #include "llmediadataclient.h" +#include "llmeshrepository.h" #include "llagent.h" const S32 MIN_QUIET_FRAMES_COALESCE = 30; @@ -861,6 +862,16 @@ BOOL LLVOVolume::setVolume(const LLVolumeParams ¶ms, const S32 detail, bool { LLVolumeParams volume_params = params; + if (isSculpted()) + { + // if it's a mesh + if ((volume_params.getSculptType() & LL_SCULPT_TYPE_MASK) == LL_SCULPT_TYPE_MESH) + { //meshes might not have all LODs, get the force detail to best existing LOD + LLUUID mesh_id = params.getSculptID(); + mLOD = gMeshRepo.getActualMeshLOD(mesh_id, mLOD); + } + } + // Check if we need to change implementations bool is_flexible = (volume_params.getPathParams().getCurveType() == LL_PCODE_PATH_FLEXIBLE); if (is_flexible) @@ -888,6 +899,8 @@ BOOL LLVOVolume::setVolume(const LLVolumeParams ¶ms, const S32 detail, bool } } + + if ((LLPrimitive::setVolume(volume_params, mLOD, (mVolumeImpl && mVolumeImpl->isVolumeUnique()))) || mSculptChanged) { mFaceMappingChanged = TRUE; @@ -906,7 +919,7 @@ BOOL LLVOVolume::setVolume(const LLVolumeParams ¶ms, const S32 detail, bool { //load request not yet issued, request pipeline load this mesh LLUUID asset_id = volume_params.getSculptID(); - gPipeline.loadMesh(this, asset_id, LLVolumeLODGroup::getVolumeDetailFromScale(getVolume()->getDetail())); + gMeshRepo.loadMesh(this, asset_id, LLVolumeLODGroup::getVolumeDetailFromScale(getVolume()->getDetail())); } } else // otherwise is sculptie diff --git a/indra/newview/pipeline.cpp b/indra/newview/pipeline.cpp index 610804c5eb..b37645d2de 100644 --- a/indra/newview/pipeline.cpp +++ b/indra/newview/pipeline.cpp @@ -70,6 +70,7 @@ #include "llgldbg.h" #include "llhudmanager.h" #include "lllightconstants.h" +#include "llmeshrepository.h" #include "llresmgr.h" #include "llselectmgr.h" #include "llsky.h" @@ -103,6 +104,7 @@ #include "llspatialpartition.h" #include "llmutelist.h" #include "lltoolpie.h" +#include "llcurl.h" #ifdef _DEBUG @@ -342,8 +344,6 @@ LLPipeline::LLPipeline() : mGlowPool(NULL), mBumpPool(NULL), mWLSkyPool(NULL), - mMeshMutex(NULL), - mMeshThreadCount(0), mLightMask(0), mLightMovingMask(0), mLightingDetail(0), @@ -403,7 +403,6 @@ void LLPipeline::init() stop_glerror(); - mMeshMutex = new LLMutex(NULL); for (U32 i = 0; i < 2; ++i) { mSpotLightFade[i] = 1.f; @@ -478,9 +477,6 @@ void LLPipeline::cleanup() //delete mWLSkyPool; mWLSkyPool = NULL; - delete mMeshMutex; - mMeshMutex = NULL; - releaseGLBuffers(); mBloomImagep = NULL; @@ -1802,6 +1798,8 @@ void LLPipeline::rebuildPriorityGroups() assertInitialized(); + gMeshRepo.notifyLoadedMeshes(); + // Iterate through all drawables on the priority build queue, for (LLSpatialGroup::sg_list_t::iterator iter = mGroupQ1.begin(); iter != mGroupQ1.end(); ++iter) @@ -1881,8 +1879,6 @@ void LLPipeline::updateGeom(F32 max_dtime) // for now, only LLVOVolume does this to throttle LOD changes LLVOVolume::preUpdateGeom(); - notifyLoadedMeshes(); - // Iterate through all drawables on the priority build queue, for (LLDrawable::drawable_list_t::iterator iter = mBuildQ1.begin(); iter != mBuildQ1.end();) @@ -8912,246 +8908,3 @@ LLCullResult::sg_list_t::iterator LLPipeline::endAlphaGroups() } -void LLPipeline::loadMesh(LLVOVolume* vobj, LLUUID mesh_id, S32 detail) -{ - if (detail < 0 || detail > 4) - { - return; - } - - { - LLMutexLock lock(mMeshMutex); - //add volume to list of loading meshes - mesh_load_map::iterator iter = mLoadingMeshes[detail].find(mesh_id); - if (iter != mLoadingMeshes[detail].end()) - { //request pending for this mesh, append volume id to list - iter->second.insert(vobj->getID()); - return; - } - - //first request for this mesh - mLoadingMeshes[detail][mesh_id].insert(vobj->getID()); - } - - if (gAssetStorage->hasLocalAsset(mesh_id, LLAssetType::AT_MESH)) - { //already have asset, load desired LOD in background - mPendingMeshes.push_back(new LLMeshThread(mesh_id, vobj->getVolume(), detail)); - } - else - { //fetch asset and load when done - gAssetStorage->getAssetData(mesh_id, LLAssetType::AT_MESH, - getMeshAssetCallback, vobj->getVolume(), TRUE); - } - - //do a quick search to see if we can't display something while we wait for this mesh to load - LLVolume* volume = vobj->getVolume(); - - if (volume) - { - LLVolumeParams params = volume->getParams(); - - LLVolumeLODGroup* group = LLPrimitive::getVolumeManager()->getGroup(params); - - if (group) - { - //first see what the next lowest LOD available might be - for (S32 i = detail-1; i >= 0; --i) - { - LLVolume* lod = group->refLOD(i); - if (lod && !lod->isTetrahedron() && lod->getNumVolumeFaces() > 0) - { - volume->copyVolumeFaces(lod); - group->derefLOD(lod); - return; - } - - group->derefLOD(lod); - } - - //no lower LOD is a available, is a higher lod available? - for (S32 i = detail+1; i < 4; ++i) - { - LLVolume* lod = group->refLOD(i); - if (lod && !lod->isTetrahedron() && lod->getNumVolumeFaces() > 0) - { - volume->copyVolumeFaces(lod); - group->derefLOD(lod); - return; - } - - group->derefLOD(lod); - } - } - else - { - llerrs << "WTF?" << llendl; - } - - //nothing found, so make a tetrahedron - volume->makeTetrahedron(); - } -} - -//static -void LLPipeline::getMeshAssetCallback(LLVFS *vfs, - const LLUUID& asset_uuid, - LLAssetType::EType type, - void* user_data, S32 status, LLExtStat ext_status) -{ - gPipeline.mPendingMeshes.push_back(new LLMeshThread(asset_uuid, (LLVolume*) user_data)); -} - - -LLPipeline::LLMeshThread::LLMeshThread(LLUUID mesh_id, LLVolume* target, S32 detail) -: LLThread("mesh_loading_thread") -{ - mMeshID = mesh_id; - mVolume = NULL; - mDetail = target->getDetail(); - - if (detail == -1) - { - mDetailIndex = LLVolumeLODGroup::getVolumeDetailFromScale(target->getDetail()); - } - else - { - mDetailIndex = detail; - } - - mTargetVolume = target; -} - -LLPipeline::LLMeshThread::~LLMeshThread() -{ - -} - -void LLPipeline::LLMeshThread::run() -{ - if (!gAssetStorage || LLApp::instance()->isQuitting()) - { - return; - } - - char* buffer = NULL; - S32 size = 0; - - LLVFS* vfs = gAssetStorage->mVFS; - - { - LLVFile file(vfs, mMeshID, LLAssetType::AT_MESH, LLVFile::READ); - file.waitForLock(VFSLOCK_READ); - size = file.getSize(); - - if (size == 0) - { - gPipeline.meshLoaded(this); - return; - } - - buffer = new char[size]; - file.read((U8*)&buffer[0], size); - } - - { - std::string buffer_string(buffer, size); - std::istringstream buffer_stream(buffer_string); - - { - LLVolumeParams volume_params; - volume_params.setType(LL_PCODE_PROFILE_SQUARE, LL_PCODE_PATH_LINE); - volume_params.setSculptID(mMeshID, LL_SCULPT_TYPE_MESH); - mVolume = new LLVolume(volume_params, mDetail); - mVolume->createVolumeFacesFromStream(buffer_stream); - } - } - delete[] buffer; - - gPipeline.meshLoaded(this); -} - -void LLPipeline::meshLoaded(LLPipeline::LLMeshThread* mesh_thread) -{ - LLMutexLock lock(mMeshMutex); - mLoadedMeshes.push_back(mesh_thread); -} - -void LLPipeline::notifyLoadedMeshes() -{ //called from main thread - - U32 max_thread_count = llmax(gSavedSettings.getU32("MeshThreadCount"), (U32) 1); - while (mMeshThreadCount < max_thread_count && !mPendingMeshes.empty()) - { - LLMeshThread* mesh_thread = mPendingMeshes.front(); - mesh_thread->start(); - ++mMeshThreadCount; - mPendingMeshes.pop_front(); - } - - LLMutexLock lock(mMeshMutex); - std::list stopping_threads; - - for (std::list::iterator iter = mLoadedMeshes.begin(); iter != mLoadedMeshes.end(); ++iter) - { //for each mesh done loading - LLMeshThread* mesh = *iter; - - if (!mesh->isStopped()) - { //don't process a LLMeshThread until it's stopped - stopping_threads.push_back(mesh); - continue; - } - - S32 detail = mesh->mDetailIndex; - - //get list of objects waiting to be notified this mesh is loaded - mesh_load_map::iterator obj_iter = mLoadingMeshes[detail].find(mesh->mMeshID); - - if (mesh->mVolume && obj_iter != mLoadingMeshes[detail].end()) - { - //make sure target volume is still valid - BOOL valid = FALSE; - - for (std::set::iterator vobj_iter = obj_iter->second.begin(); vobj_iter != obj_iter->second.end(); ++vobj_iter) - { - LLVOVolume* vobj = (LLVOVolume*) gObjectList.findObject(*vobj_iter); - - if (vobj) - { - if (vobj->getVolume() == mesh->mTargetVolume) - { - valid = TRUE; - } - } - } - - - if (valid) - { - if (mesh->mVolume->getNumVolumeFaces() <= 0) - { - llwarns << "Mesh loading returned empty volume." << llendl; - mesh->mVolume->makeTetrahedron(); - } - - mesh->mTargetVolume->copyVolumeFaces(mesh->mVolume); - - for (std::set::iterator vobj_iter = obj_iter->second.begin(); vobj_iter != obj_iter->second.end(); ++vobj_iter) - { - LLVOVolume* vobj = (LLVOVolume*) gObjectList.findObject(*vobj_iter); - if (vobj) - { - vobj->notifyMeshLoaded(); - } - } - } - - mLoadingMeshes[detail].erase(mesh->mMeshID); - } - - delete mesh; - --mMeshThreadCount; - } - - mLoadedMeshes = stopping_threads; -} - diff --git a/indra/newview/pipeline.h b/indra/newview/pipeline.h index f41f6173a9..0b040acf51 100644 --- a/indra/newview/pipeline.h +++ b/indra/newview/pipeline.h @@ -58,6 +58,8 @@ class LLCubeMap; class LLCullResult; class LLVOAvatar; class LLGLSLShader; +class LLCurlRequest; +class LLMeshResponder; typedef enum e_avatar_skinning_method { @@ -278,10 +280,6 @@ public: LLCullResult::sg_list_t::iterator beginAlphaGroups(); LLCullResult::sg_list_t::iterator endAlphaGroups(); - - //mesh management functions - void loadMesh(LLVOVolume* volume, LLUUID mesh_id, S32 detail = 0); - void addTrianglesDrawn(S32 count); BOOL hasRenderType(const U32 type) const { return (type && (mRenderTypeMask & (1< > mesh_load_map; mesh_load_map mLoadingMeshes[4]; - LLMutex* mMeshMutex; - - class LLMeshThread : public LLThread - { - public: - LLPointer mVolume; - LLVolume* mTargetVolume; - LLUUID mMeshID; - F32 mDetail; - S32 mDetailIndex; - LLMeshThread(LLUUID mesh_id, LLVolume* target, S32 detail = -1); - ~LLMeshThread(); - void run(); - }; - - static void getMeshAssetCallback(LLVFS *vfs, - const LLUUID& asset_uuid, - LLAssetType::EType type, - void* user_data, S32 status, LLExtStat ext_status); - - std::list mLoadedMeshes; - std::list mPendingMeshes; - U32 mMeshThreadCount; - - void meshLoaded(LLMeshThread* mesh_thread); - void notifyLoadedMeshes(); + typedef std::list mesh_response_list; + mesh_response_list mMeshResponseList; LLPointer mFaceSelectImagep; LLPointer mBloomImagep; -- cgit v1.3 From 081fa98a47d2b592ada0fbb049ff959ac2cd6294 Mon Sep 17 00:00:00 2001 From: Dave Parks Date: Tue, 15 Dec 2009 17:43:05 -0600 Subject: HTTP Mesh transfer relatively blocking-free. --- indra/llmath/llvolume.cpp | 2 +- indra/newview/app_settings/settings.xml | 11 +++++++++++ indra/newview/llvovolume.cpp | 20 ++++++++++++++++---- 3 files changed, 28 insertions(+), 5 deletions(-) (limited to 'indra/newview/llvovolume.cpp') diff --git a/indra/llmath/llvolume.cpp b/indra/llmath/llvolume.cpp index 44ff173502..844918432d 100644 --- a/indra/llmath/llvolume.cpp +++ b/indra/llmath/llvolume.cpp @@ -2171,7 +2171,7 @@ void LLVolume::makeTetrahedron() LLVolumeFace face; - F32 x = 0.5f; + F32 x = 0.25f; LLVector3 p[] = { //unit tetrahedron corners LLVector3(x,x,x), diff --git a/indra/newview/app_settings/settings.xml b/indra/newview/app_settings/settings.xml index a4142b41b0..4f17775dfd 100644 --- a/indra/newview/app_settings/settings.xml +++ b/indra/newview/app_settings/settings.xml @@ -7463,6 +7463,17 @@ Value 8 + MeshMaxConcurrentRequests + + Comment + Number of threads to use for loading meshes. + Persist + 1 + Type + U32 + Value + 32 + SafeMode diff --git a/indra/newview/llvovolume.cpp b/indra/newview/llvovolume.cpp index f66f0c2d72..ea0aa54751 100644 --- a/indra/newview/llvovolume.cpp +++ b/indra/newview/llvovolume.cpp @@ -862,13 +862,21 @@ BOOL LLVOVolume::setVolume(const LLVolumeParams ¶ms, const S32 detail, bool { LLVolumeParams volume_params = params; + S32 lod = mLOD; + if (isSculpted()) { // if it's a mesh if ((volume_params.getSculptType() & LL_SCULPT_TYPE_MASK) == LL_SCULPT_TYPE_MESH) { //meshes might not have all LODs, get the force detail to best existing LOD LLUUID mesh_id = params.getSculptID(); - mLOD = gMeshRepo.getActualMeshLOD(mesh_id, mLOD); + + //profile and path params don't matter for meshes + volume_params = LLVolumeParams(); + volume_params.setType(LL_PCODE_PROFILE_SQUARE, LL_PCODE_PATH_LINE); + volume_params.setSculptID(mesh_id, LL_SCULPT_TYPE_MESH); + + lod = gMeshRepo.getActualMeshLOD(mesh_id, lod); } } @@ -901,13 +909,13 @@ BOOL LLVOVolume::setVolume(const LLVolumeParams ¶ms, const S32 detail, bool - if ((LLPrimitive::setVolume(volume_params, mLOD, (mVolumeImpl && mVolumeImpl->isVolumeUnique()))) || mSculptChanged) + if ((LLPrimitive::setVolume(volume_params, lod, (mVolumeImpl && mVolumeImpl->isVolumeUnique()))) || mSculptChanged) { mFaceMappingChanged = TRUE; if (mVolumeImpl) { - mVolumeImpl->onSetVolume(volume_params, detail); + mVolumeImpl->onSetVolume(volume_params, mLOD); } if (isSculpted()) @@ -919,7 +927,11 @@ BOOL LLVOVolume::setVolume(const LLVolumeParams ¶ms, const S32 detail, bool { //load request not yet issued, request pipeline load this mesh LLUUID asset_id = volume_params.getSculptID(); - gMeshRepo.loadMesh(this, asset_id, LLVolumeLODGroup::getVolumeDetailFromScale(getVolume()->getDetail())); + S32 available_lod = gMeshRepo.loadMesh(this, asset_id, lod); + if (available_lod != lod) + { + LLPrimitive::setVolume(volume_params, available_lod); + } } } else // otherwise is sculptie -- cgit v1.3 From 024c0ebe19588f8452bae7ea01756fd7b4b30540 Mon Sep 17 00:00:00 2001 From: "Karl Stiefvater (qarl)" Date: Fri, 29 Jan 2010 14:33:04 -0600 Subject: enable mirror and invert flags for meshes. --- indra/llmath/llvolume.cpp | 59 +++++++++++++++++++++++++++++++++++++++++--- indra/llmath/llvolume.h | 7 +++--- indra/newview/llvovolume.cpp | 7 +++--- 3 files changed, 62 insertions(+), 11 deletions(-) (limited to 'indra/newview/llvovolume.cpp') diff --git a/indra/llmath/llvolume.cpp b/indra/llmath/llvolume.cpp index de32070da1..0328c09c9a 100644 --- a/indra/llmath/llvolume.cpp +++ b/indra/llmath/llvolume.cpp @@ -1694,7 +1694,7 @@ LLVolume::LLVolume(const LLVolumeParams ¶ms, const F32 detail, const BOOL ge generate(); - if (mParams.getSculptID().isNull() && params.getSculptType() == LL_SCULPT_TYPE_NONE) + if (mParams.getSculptID().isNull() && mParams.getSculptType() == LL_SCULPT_TYPE_NONE) { createVolumeFaces(); } @@ -2142,6 +2142,59 @@ BOOL LLVolume::unpackVolumeFaces(std::istream& is, S32 size) (F32) t[1] / 65535.f * tc_range.mV[1] + min_tc.mV[1]); } + + // modifier flags? + BOOL do_mirror = (mParams.getSculptType() & LL_SCULPT_FLAG_MIRROR); + BOOL do_invert = (mParams.getSculptType() &LL_SCULPT_FLAG_INVERT); + + + // translate to actions: + BOOL do_reflect_x = FALSE; + BOOL do_reverse_triangles = FALSE; + BOOL do_invert_normals = FALSE; + + if (do_mirror) + { + do_reflect_x = TRUE; + do_reverse_triangles = !do_reverse_triangles; + } + + if (do_invert) + { + do_invert_normals = TRUE; + do_reverse_triangles = !do_reverse_triangles; + } + + // now do the work + + if (do_reflect_x) + { + for (S32 i = 0; i < face.mVertices.size(); i++) + { + face.mVertices[i].mPosition.mV[VX] *= -1.0f; + face.mVertices[i].mNormal.mV[VX] *= -1.0f; + } + } + + if (do_invert_normals) + { + for (S32 i = 0; i < face.mVertices.size(); i++) + { + face.mVertices[i].mNormal *= -1.0f; + } + } + + if (do_reverse_triangles) + { + for (U32 j = 0; j < face.mIndices.size(); j += 3) + { + // swap the 2nd and 3rd index + S32 swap = face.mIndices[j+1]; + face.mIndices[j+1] = face.mIndices[j+2]; + face.mIndices[j+2] = swap; + } + } + } } @@ -3838,7 +3891,7 @@ void LLVolume::generateSilhouetteVertices(std::vector &vertices, normals.clear(); segments.clear(); - if (mParams.getSculptType() == LL_SCULPT_TYPE_MESH) + if ((mParams.getSculptType() & LL_SCULPT_TYPE_MASK) == LL_SCULPT_TYPE_MESH) { return; } @@ -5500,7 +5553,7 @@ BOOL LLVolumeFace::createSide(LLVolume* volume, BOOL partial_build) { mIndices.resize(num_indices); - if (volume->getParams().getSculptType() != LL_SCULPT_TYPE_MESH) + if ((volume->getParams().getSculptType() & LL_SCULPT_TYPE_MASK) != LL_SCULPT_TYPE_MESH) { mEdge.resize(num_indices); } diff --git a/indra/llmath/llvolume.h b/indra/llmath/llvolume.h index e3ab648fe3..0bc64f4487 100644 --- a/indra/llmath/llvolume.h +++ b/indra/llmath/llvolume.h @@ -183,12 +183,11 @@ const U8 LL_SCULPT_TYPE_SPHERE = 1; const U8 LL_SCULPT_TYPE_TORUS = 2; const U8 LL_SCULPT_TYPE_PLANE = 3; const U8 LL_SCULPT_TYPE_CYLINDER = 4; - -const U8 LL_SCULPT_TYPE_MASK = LL_SCULPT_TYPE_SPHERE | LL_SCULPT_TYPE_TORUS | LL_SCULPT_TYPE_PLANE | LL_SCULPT_TYPE_CYLINDER; - -// need to change this (these) names const U8 LL_SCULPT_TYPE_MESH = 5; +const U8 LL_SCULPT_TYPE_MASK = LL_SCULPT_TYPE_SPHERE | LL_SCULPT_TYPE_TORUS | LL_SCULPT_TYPE_PLANE | + LL_SCULPT_TYPE_CYLINDER | LL_SCULPT_TYPE_MESH; + const U8 LL_SCULPT_FLAG_INVERT = 64; const U8 LL_SCULPT_FLAG_MIRROR = 128; diff --git a/indra/newview/llvovolume.cpp b/indra/newview/llvovolume.cpp index 4acf3cb6fb..3f1ddc0ee7 100644 --- a/indra/newview/llvovolume.cpp +++ b/indra/newview/llvovolume.cpp @@ -905,14 +905,13 @@ BOOL LLVOVolume::setVolume(const LLVolumeParams ¶ms, const S32 detail, bool // if it's a mesh if ((volume_params.getSculptType() & LL_SCULPT_TYPE_MASK) == LL_SCULPT_TYPE_MESH) { //meshes might not have all LODs, get the force detail to best existing LOD + LLUUID mesh_id = params.getSculptID(); //profile and path params don't matter for meshes - volume_params = LLVolumeParams(); volume_params.setType(LL_PCODE_PROFILE_SQUARE, LL_PCODE_PATH_LINE); - volume_params.setSculptID(mesh_id, LL_SCULPT_TYPE_MESH); - lod = gMeshRepo.getActualMeshLOD(mesh_id, lod); + lod = gMeshRepo.getActualMeshLOD(volume_params, lod); } } @@ -963,7 +962,7 @@ BOOL LLVOVolume::setVolume(const LLVolumeParams ¶ms, const S32 detail, bool { //load request not yet issued, request pipeline load this mesh LLUUID asset_id = volume_params.getSculptID(); - S32 available_lod = gMeshRepo.loadMesh(this, asset_id, lod); + S32 available_lod = gMeshRepo.loadMesh(this, volume_params, lod); if (available_lod != lod) { LLPrimitive::setVolume(volume_params, available_lod); -- cgit v1.3 From e8899c4f4c2ae95f108b9bbf543e401debc9c588 Mon Sep 17 00:00:00 2001 From: Tofu Linden Date: Mon, 1 Feb 2010 17:29:08 +0000 Subject: start to abstract-out the 'is a spotlight' question so one day we can perhaps make the UI hurt less. --- indra/llprimitive/llprimitive.h | 1 + indra/newview/llpanelvolume.cpp | 4 ++-- indra/newview/llvovolume.cpp | 7 +++++++ indra/newview/llvovolume.h | 1 + indra/newview/pipeline.cpp | 6 +++--- 5 files changed, 14 insertions(+), 5 deletions(-) (limited to 'indra/newview/llvovolume.cpp') diff --git a/indra/llprimitive/llprimitive.h b/indra/llprimitive/llprimitive.h index 4db7aa7261..c25f9cf4a8 100644 --- a/indra/llprimitive/llprimitive.h +++ b/indra/llprimitive/llprimitive.h @@ -286,6 +286,7 @@ public: void setLightTexture(const LLUUID& id) { mLightTexture = id; } LLUUID getLightTexture() const { return mLightTexture; } + bool isLightSpotlight() const { return !mLightTexture.isNull(); } void setParams(const LLVector3& params) { mParams = params; } LLVector3 getParams() const { return mParams; } diff --git a/indra/newview/llpanelvolume.cpp b/indra/newview/llpanelvolume.cpp index fbe68b4d92..8b01637239 100644 --- a/indra/newview/llpanelvolume.cpp +++ b/indra/newview/llpanelvolume.cpp @@ -574,7 +574,7 @@ void LLPanelVolume::onCommitLight( LLUICtrl* ctrl, void* userdata ) LLUUID id = LightTextureCtrl->getImageAssetID(); if (id.notNull()) { - if (volobjp->getLightTextureID().isNull()) + if (!volobjp->isLightSpotlight()) { //this commit is making this a spot light, set UI to default params volobjp->setLightTextureID(id); LLVector3 spot_params = volobjp->getSpotLightParams(); @@ -591,7 +591,7 @@ void LLPanelVolume::onCommitLight( LLUICtrl* ctrl, void* userdata ) volobjp->setSpotLightParams(spot_params); } } - else if (volobjp->getLightTextureID().notNull()) + else if (volobjp->isLightSpotlight()) { //no longer a spot light volobjp->setLightTextureID(id); //self->childDisable("Light FOV"); diff --git a/indra/newview/llvovolume.cpp b/indra/newview/llvovolume.cpp index 3f1ddc0ee7..305c0a19d6 100644 --- a/indra/newview/llvovolume.cpp +++ b/indra/newview/llvovolume.cpp @@ -2462,6 +2462,13 @@ void LLVOVolume::updateSpotLightPriority() } +bool LLVOVolume::isLightSpotlight() +{ + LLLightImageParams* params = (LLLightImageParams*) getParameterEntry(LLNetworkData::PARAMS_LIGHT_IMAGE); + return params->isLightSpotlight(); +} + + LLViewerTexture* LLVOVolume::getLightTexture() { LLUUID id = getLightTextureID(); diff --git a/indra/newview/llvovolume.h b/indra/newview/llvovolume.h index e414d9da4d..280ffea76d 100644 --- a/indra/newview/llvovolume.h +++ b/indra/newview/llvovolume.h @@ -218,6 +218,7 @@ public: LLColor3 getLightBaseColor() const; // not scaled by intensity LLColor3 getLightColor() const; // scaled by intensity LLUUID getLightTextureID() const; + bool isLightSpotlight() const; LLVector3 getSpotLightParams() const; void updateSpotLightPriority(); F32 getSpotLightPriority() const; diff --git a/indra/newview/pipeline.cpp b/indra/newview/pipeline.cpp index 2ad0bad11f..cb0baf77f1 100644 --- a/indra/newview/pipeline.cpp +++ b/indra/newview/pipeline.cpp @@ -4584,7 +4584,7 @@ void LLPipeline::setupHWLights(LLDrawPool* pool) glLightf (gllight, GL_CONSTANT_ATTENUATION, 0.0f); glLightf (gllight, GL_LINEAR_ATTENUATION, linatten); glLightf (gllight, GL_QUADRATIC_ATTENUATION, 0.0f); - if (light->getLightTexture()) // directional (spot-)light + if (light->isLightSpotlight()) // directional (spot-)light { LLVector3 spotparams = light->getSpotLightParams(); LLQuaternion quat = light->getRenderRotation(); @@ -6693,7 +6693,7 @@ void LLPipeline::renderDeferredLighting() { //draw box if camera is outside box if (render_local) { - if (volume->getLightTexture()) + if (volume->isLightSpotlight()) { drawablep->getVOVolume()->updateSpotLightPriority(); spot_lights.push_back(drawablep); @@ -6710,7 +6710,7 @@ void LLPipeline::renderDeferredLighting() } else if (render_fullscreen) { - if (volume->getLightTexture()) + if (volume->isLightSpotlight()) { drawablep->getVOVolume()->updateSpotLightPriority(); fullscreen_spot_lights.push_back(drawablep); -- cgit v1.3 From 3dc32880ae0f2f0f5499a7f271df666759fc220a Mon Sep 17 00:00:00 2001 From: Tofu Linden Date: Mon, 1 Feb 2010 17:50:55 +0000 Subject: const fix. --- indra/newview/llvovolume.cpp | 2 +- indra/newview/pipeline.cpp | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-) (limited to 'indra/newview/llvovolume.cpp') diff --git a/indra/newview/llvovolume.cpp b/indra/newview/llvovolume.cpp index 305c0a19d6..a6caa2362b 100644 --- a/indra/newview/llvovolume.cpp +++ b/indra/newview/llvovolume.cpp @@ -2462,7 +2462,7 @@ void LLVOVolume::updateSpotLightPriority() } -bool LLVOVolume::isLightSpotlight() +bool LLVOVolume::isLightSpotlight() const { LLLightImageParams* params = (LLLightImageParams*) getParameterEntry(LLNetworkData::PARAMS_LIGHT_IMAGE); return params->isLightSpotlight(); diff --git a/indra/newview/pipeline.cpp b/indra/newview/pipeline.cpp index cb0baf77f1..b25c7c9212 100644 --- a/indra/newview/pipeline.cpp +++ b/indra/newview/pipeline.cpp @@ -4573,7 +4573,7 @@ void LLPipeline::setupHWLights(LLDrawPool* pool) F32 light_radius = llmax(light->getLightRadius(), 0.001f); - F32 x = (3.f * (1.f + light->getLightFalloff())); + F32 x = (3.f * (1.f + light->getLightFalloff())); // why this magic? probably trying to match a historic behavior. float linatten = x / (light_radius); // % of brightness at radius mHWLightColors[cur_light] = light_color; -- cgit v1.3 From cc5920bed7790db49a062fb98ce54191f06ed98c Mon Sep 17 00:00:00 2001 From: Tofu Linden Date: Tue, 2 Feb 2010 22:45:30 +0000 Subject: Probably the real fix for isLightSpotlight() - PARAMS_LIGHT_IMAGE isn't guaranteed, duh. --- indra/newview/llvovolume.cpp | 6 +++++- 1 file changed, 5 insertions(+), 1 deletion(-) (limited to 'indra/newview/llvovolume.cpp') diff --git a/indra/newview/llvovolume.cpp b/indra/newview/llvovolume.cpp index a6caa2362b..be2744aede 100644 --- a/indra/newview/llvovolume.cpp +++ b/indra/newview/llvovolume.cpp @@ -2465,7 +2465,11 @@ void LLVOVolume::updateSpotLightPriority() bool LLVOVolume::isLightSpotlight() const { LLLightImageParams* params = (LLLightImageParams*) getParameterEntry(LLNetworkData::PARAMS_LIGHT_IMAGE); - return params->isLightSpotlight(); + if (params) + { + return params->isLightSpotlight(); + } + return false; } -- cgit v1.3 From 095a5e84408b47ef3c5610e111aefe51d77633ca Mon Sep 17 00:00:00 2001 From: Dave Parks Date: Sat, 6 Feb 2010 17:33:12 -0600 Subject: Draw prims using triangle strips instead of triangle lists. --- indra/llmath/llvolume.cpp | 201 +++++++++++++++++++++++++------- indra/llmath/llvolume.h | 4 +- indra/newview/app_settings/settings.xml | 11 ++ indra/newview/lldrawpool.cpp | 4 +- indra/newview/lldrawpoolalpha.cpp | 8 +- indra/newview/lldrawpoolbump.cpp | 8 +- indra/newview/lldrawpooltree.cpp | 4 +- indra/newview/llface.cpp | 25 +++- indra/newview/llspatialpartition.cpp | 3 +- indra/newview/llspatialpartition.h | 1 + indra/newview/llviewercontrol.cpp | 1 + indra/newview/llviewerjointmesh.cpp | 2 +- indra/newview/llvotree.cpp | 6 +- indra/newview/llvovolume.cpp | 16 ++- indra/newview/llvowlsky.cpp | 2 +- indra/newview/pipeline.cpp | 15 ++- indra/newview/pipeline.h | 4 +- 17 files changed, 242 insertions(+), 73 deletions(-) (limited to 'indra/newview/llvovolume.cpp') diff --git a/indra/llmath/llvolume.cpp b/indra/llmath/llvolume.cpp index df4c618ac1..cd7d7a12e3 100644 --- a/indra/llmath/llvolume.cpp +++ b/indra/llmath/llvolume.cpp @@ -4520,15 +4520,65 @@ BOOL LLVolumeFace::createUnCutCubeCap(LLVolume* volume, BOOL partial_build) if (!partial_build) { - int idxs[] = {0,1,(grid_size+1)+1,(grid_size+1)+1,(grid_size+1),0}; - for(int gx = 0;gx=0;i--)mIndices.push_back(vtop+(gy*(grid_size+1))+gx+idxs[i]); - }else{ - for(int i=0;i<6;i++)mIndices.push_back(vtop+(gy*(grid_size+1))+gx+idxs[i]); + mTriStrip.clear(); + S32 idxs[] = {0,1,(grid_size+1)+1,(grid_size+1)+1,(grid_size+1),0}; + for(S32 gx = 0;gx=0;i--) + { + mIndices.push_back(vtop+(gy*(grid_size+1))+gx+idxs[i]); + } + + if (gy == 0) + { + mTriStrip.push_back((gx+1)*(grid_size+1)); + mTriStrip.push_back((gx+1)*(grid_size+1)); + mTriStrip.push_back(gx*(grid_size+1)); + } + + mTriStrip.push_back(gy+1+(gx+1)*(grid_size+1)); + mTriStrip.push_back(gy+1+gx*(grid_size+1)); + + + if (gy == grid_size-1) + { + mTriStrip.push_back(gy+1+gx*(grid_size+1)); + } + } + else + { + for(S32 i=0;i<6;i++) + { + mIndices.push_back(vtop+(gy*(grid_size+1))+gx+idxs[i]); + } + + if (gy == 0) + { + mTriStrip.push_back(gx*(grid_size+1)); + mTriStrip.push_back(gx*(grid_size+1)); + mTriStrip.push_back((gx+1)*(grid_size+1)); + } + + mTriStrip.push_back(gy+1+gx*(grid_size+1)); + mTriStrip.push_back(gy+1+(gx+1)*(grid_size+1)); + + if (gy == grid_size-1) + { + mTriStrip.push_back(gy+1+(gx+1)*(grid_size+1)); + } } } + + } + + if (mTriStrip.size()%2 == 1) + { + mTriStrip.push_back(mTriStrip[mTriStrip.size()-1]); } } @@ -4770,6 +4820,8 @@ BOOL LLVolumeFace::createCap(LLVolume* volume, BOOL partial_build) pt2--; } } + + makeTriStrip(); } else { @@ -4874,67 +4926,108 @@ BOOL LLVolumeFace::createCap(LLVolume* volume, BOOL partial_build) pt2--; } } + + makeTriStrip(); } } else { // Not hollow, generate the triangle fan. + U16 v1 = 2; + U16 v2 = 1; + if (mTypeMask & TOP_MASK) { - if (mTypeMask & OPEN_MASK) - { - // SOLID OPEN TOP - // Generate indices - // This is a tri-fan, so we reuse the same first point for all triangles. - for (S32 i = 0; i < (num_vertices - 2); i++) - { - mIndices[3*i] = num_vertices - 1; - mIndices[3*i+1] = i; - mIndices[3*i+2] = i + 1; - } - } - else - { - // SOLID CLOSED TOP - for (S32 i = 0; i < (num_vertices - 2); i++) - { - //MSMSM fix these caps but only for the un-cut case - mIndices[3*i] = num_vertices - 1; - mIndices[3*i+1] = i; - mIndices[3*i+2] = i + 1; - } - } + v1 = 1; + v2 = 2; + } + + for (S32 i = 0; i < (num_vertices - 2); i++) + { + mIndices[3*i] = num_vertices - 1; + mIndices[3*i+v1] = i; + mIndices[3*i+v2] = i + 1; + } + + //make tri strip + if (mTypeMask & OPEN_MASK) + { + makeTriStrip(); } else { - if (mTypeMask & OPEN_MASK) + S32 j = num_vertices-2; + if (mTypeMask & TOP_MASK) { - // SOLID OPEN BOTTOM - // Generate indices - // This is a tri-fan, so we reuse the same first point for all triangles. - for (S32 i = 0; i < (num_vertices - 2); i++) + mTriStrip.push_back(0); + for (S32 i = 1; i <= j; ++i) { - mIndices[3*i] = num_vertices - 1; - mIndices[3*i+1] = i + 1; - mIndices[3*i+2] = i; + mTriStrip.push_back(i); + if (i != j) + { + mTriStrip.push_back(j); + } + --j; } } else { - // SOLID CLOSED BOTTOM - for (S32 i = 0; i < (num_vertices - 2); i++) + mTriStrip.push_back(j); + for (S32 i = 1; i <= j; ++i) { - //MSMSM fix these caps but only for the un-cut case - mIndices[3*i] = num_vertices - 1; - mIndices[3*i+1] = i + 1; - mIndices[3*i+2] = i; + if (i != j) + { + mTriStrip.push_back(j); + } + mTriStrip.push_back(i); + --j; } } + + mTriStrip.push_back(mTriStrip[mTriStrip.size()-1]); + + if (mTriStrip.size()%2 == 1) + { + mTriStrip.push_back(mTriStrip[mTriStrip.size()-1]); + } } } + return TRUE; } +void LLVolumeFace::makeTriStrip() +{ + for (U32 i = 0; i < mIndices.size(); i+=3) + { + U16 i0 = mIndices[i]; + U16 i1 = mIndices[i+1]; + U16 i2 = mIndices[i+2]; + + if ((i/3)%2 == 1) + { + mTriStrip.push_back(i0); + mTriStrip.push_back(i0); + mTriStrip.push_back(i1); + mTriStrip.push_back(i2); + mTriStrip.push_back(i2); + } + else + { + mTriStrip.push_back(i2); + mTriStrip.push_back(i2); + mTriStrip.push_back(i1); + mTriStrip.push_back(i0); + mTriStrip.push_back(i0); + } + } + + if (mTriStrip.size()%2 == 1) + { + mTriStrip.push_back(mTriStrip[mTriStrip.size()-1]); + } +} + void LLVolumeFace::createBinormals() { LLMemType m1(LLMemType::MTYPE_VOLUME); @@ -5135,9 +5228,14 @@ BOOL LLVolumeFace::createSide(LLVolume* volume, BOOL partial_build) if (!partial_build) { + mTriStrip.clear(); + // Now we generate the indices. for (t = 0; t < (mNumT-1); t++) { + //prepend terminating index to strip + mTriStrip.push_back(mNumS*t); + for (s = 0; s < (mNumS-1); s++) { mIndices[cur_index++] = s + mNumS*t; //bottom left @@ -5147,6 +5245,14 @@ BOOL LLVolumeFace::createSide(LLVolume* volume, BOOL partial_build) mIndices[cur_index++] = s+1 + mNumS*t; //bottom right mIndices[cur_index++] = s+1 + mNumS*(t+1); //top right + if (s == 0) + { + mTriStrip.push_back(s+mNumS*t); + mTriStrip.push_back(s+mNumS*(t+1)); + } + mTriStrip.push_back(s+1+mNumS*t); + mTriStrip.push_back(s+1+mNumS*(t+1)); + mEdge[cur_edge++] = (mNumS-1)*2*t+s*2+1; //bottom left/top right neighbor face if (t < mNumT-2) { //top right/top left neighbor face mEdge[cur_edge++] = (mNumS-1)*2*(t+1)+s*2+1; @@ -5187,6 +5293,13 @@ BOOL LLVolumeFace::createSide(LLVolume* volume, BOOL partial_build) } mEdge[cur_edge++] = (mNumS-1)*2*t+s*2; //top right/bottom left neighbor face } + //append terminating vertex to strip + mTriStrip.push_back(mNumS-1+mNumS*(t+1)); + } + + if (mTriStrip.size()%2 == 1) + { + mTriStrip.push_back(mTriStrip[mTriStrip.size()-1]); } } diff --git a/indra/llmath/llvolume.h b/indra/llmath/llvolume.h index 871b334452..d9f80f0e30 100644 --- a/indra/llmath/llvolume.h +++ b/indra/llmath/llvolume.h @@ -798,7 +798,8 @@ public: BOOL create(LLVolume* volume, BOOL partial_build = FALSE); void createBinormals(); - + void makeTriStrip(); + class VertexData { public: @@ -839,6 +840,7 @@ public: std::vector mVertices; std::vector mIndices; + std::vector mTriStrip; std::vector mEdge; private: diff --git a/indra/newview/app_settings/settings.xml b/indra/newview/app_settings/settings.xml index 8a447502b0..ce19dda9a5 100644 --- a/indra/newview/app_settings/settings.xml +++ b/indra/newview/app_settings/settings.xml @@ -7386,6 +7386,17 @@ Value 1 + RenderUseTriStrips + + Comment + Use triangle strips for rendering prims. + Persist + 1 + Type + Boolean + Value + 1 + RenderUseFarClip Comment diff --git a/indra/newview/lldrawpool.cpp b/indra/newview/lldrawpool.cpp index 95ddacb722..ef946ac49e 100644 --- a/indra/newview/lldrawpool.cpp +++ b/indra/newview/lldrawpool.cpp @@ -496,8 +496,8 @@ void LLRenderPass::pushBatch(LLDrawInfo& params, U32 mask, BOOL texture) params.mGroup->rebuildMesh(); } params.mVertexBuffer->setBuffer(mask); - params.mVertexBuffer->drawRange(LLRender::TRIANGLES, params.mStart, params.mEnd, params.mCount, params.mOffset); - gPipeline.addTrianglesDrawn(params.mCount/3); + params.mVertexBuffer->drawRange(params.mDrawMode, params.mStart, params.mEnd, params.mCount, params.mOffset); + gPipeline.addTrianglesDrawn(params.mCount, params.mDrawMode); } if (params.mTextureMatrix && texture && params.mTexture.notNull()) diff --git a/indra/newview/lldrawpoolalpha.cpp b/indra/newview/lldrawpoolalpha.cpp index 6d77361414..0e02bc4b01 100644 --- a/indra/newview/lldrawpoolalpha.cpp +++ b/indra/newview/lldrawpoolalpha.cpp @@ -250,8 +250,8 @@ void LLDrawPoolAlpha::renderAlphaHighlight(U32 mask) params.mGroup->rebuildMesh(); } params.mVertexBuffer->setBuffer(mask); - params.mVertexBuffer->drawRange(LLRender::TRIANGLES, params.mStart, params.mEnd, params.mCount, params.mOffset); - gPipeline.addTrianglesDrawn(params.mCount/3); + params.mVertexBuffer->drawRange(params.mDrawMode, params.mStart, params.mEnd, params.mCount, params.mOffset); + gPipeline.addTrianglesDrawn(params.mCount, params.mDrawMode); } } } @@ -380,8 +380,8 @@ void LLDrawPoolAlpha::renderAlpha(U32 mask) } params.mVertexBuffer->setBuffer(mask); - params.mVertexBuffer->drawRange(LLRender::TRIANGLES, params.mStart, params.mEnd, params.mCount, params.mOffset); - gPipeline.addTrianglesDrawn(params.mCount/3); + params.mVertexBuffer->drawRange(params.mDrawMode, params.mStart, params.mEnd, params.mCount, params.mOffset); + gPipeline.addTrianglesDrawn(params.mCount, params.mDrawMode); if (params.mTextureMatrix && params.mTexture.notNull()) { diff --git a/indra/newview/lldrawpoolbump.cpp b/indra/newview/lldrawpoolbump.cpp index 03a8b108e2..d09d4a412f 100644 --- a/indra/newview/lldrawpoolbump.cpp +++ b/indra/newview/lldrawpoolbump.cpp @@ -560,8 +560,8 @@ void LLDrawPoolBump::renderGroup(LLSpatialGroup* group, U32 type, U32 mask, BOOL params.mGroup->rebuildMesh(); } params.mVertexBuffer->setBuffer(mask); - params.mVertexBuffer->drawRange(LLRender::TRIANGLES, params.mStart, params.mEnd, params.mCount, params.mOffset); - gPipeline.addTrianglesDrawn(params.mCount/3); + params.mVertexBuffer->drawRange(params.mDrawMode, params.mStart, params.mEnd, params.mCount, params.mOffset); + gPipeline.addTrianglesDrawn(params.mCount, params.mDrawMode); } } @@ -1231,8 +1231,8 @@ void LLDrawPoolBump::pushBatch(LLDrawInfo& params, U32 mask, BOOL texture) params.mGroup->rebuildMesh(); } params.mVertexBuffer->setBuffer(mask); - params.mVertexBuffer->drawRange(LLRender::TRIANGLES, params.mStart, params.mEnd, params.mCount, params.mOffset); - gPipeline.addTrianglesDrawn(params.mCount/3); + params.mVertexBuffer->drawRange(params.mDrawMode, params.mStart, params.mEnd, params.mCount, params.mOffset); + gPipeline.addTrianglesDrawn(params.mCount, params.mDrawMode); if (params.mTextureMatrix) { if (mShiny) diff --git a/indra/newview/lldrawpooltree.cpp b/indra/newview/lldrawpooltree.cpp index 5521fb05a8..1a5d55d793 100644 --- a/indra/newview/lldrawpooltree.cpp +++ b/indra/newview/lldrawpooltree.cpp @@ -115,7 +115,7 @@ void LLDrawPoolTree::render(S32 pass) LLFace *face = *iter; face->mVertexBuffer->setBuffer(LLDrawPoolTree::VERTEX_DATA_MASK); face->mVertexBuffer->drawRange(LLRender::TRIANGLES, 0, face->mVertexBuffer->getRequestedVerts()-1, face->mVertexBuffer->getRequestedIndices(), 0); - gPipeline.addTrianglesDrawn(face->mVertexBuffer->getRequestedIndices()/3); + gPipeline.addTrianglesDrawn(face->mVertexBuffer->getRequestedIndices()); } } } @@ -237,7 +237,7 @@ void LLDrawPoolTree::renderForSelect() face->mVertexBuffer->setBuffer(LLDrawPoolTree::VERTEX_DATA_MASK); face->mVertexBuffer->drawRange(LLRender::TRIANGLES, 0, face->mVertexBuffer->getRequestedVerts()-1, face->mVertexBuffer->getRequestedIndices(), 0); - gPipeline.addTrianglesDrawn(face->mVertexBuffer->getRequestedIndices()/3); + gPipeline.addTrianglesDrawn(face->mVertexBuffer->getRequestedIndices()); } } } diff --git a/indra/newview/llface.cpp b/indra/newview/llface.cpp index 965ac1cad0..93840e077c 100644 --- a/indra/newview/llface.cpp +++ b/indra/newview/llface.cpp @@ -868,7 +868,7 @@ BOOL LLFace::getGeometryVolume(const LLVolume& volume, llpushcallstacks ; const LLVolumeFace &vf = volume.getVolumeFace(f); S32 num_vertices = (S32)vf.mVertices.size(); - S32 num_indices = (S32)vf.mIndices.size(); + S32 num_indices = LLPipeline::sUseTriStrips ? (S32)vf.mTriStrip.size() : (S32) vf.mIndices.size(); if (mVertexBuffer.notNull()) { @@ -1058,9 +1058,19 @@ BOOL LLFace::getGeometryVolume(const LLVolume& volume, if (full_rebuild) { mVertexBuffer->getIndexStrider(indicesp, mIndicesIndex); - for (U16 i = 0; i < num_indices; i++) + if (LLPipeline::sUseTriStrips) { - *indicesp++ = vf.mIndices[i] + index_offset; + for (U16 i = 0; i < num_indices; i++) + { + *indicesp++ = vf.mTriStrip[i] + index_offset; + } + } + else + { + for (U16 i = 0; i < num_indices; i++) + { + *indicesp++ = vf.mIndices[i] + index_offset; + } } } @@ -1572,8 +1582,13 @@ S32 LLFace::pushVertices(const U16* index_array) const { if (mIndicesCount) { - mVertexBuffer->drawRange(LLRender::TRIANGLES, mGeomIndex, mGeomIndex+mGeomCount-1, mIndicesCount, mIndicesIndex); - gPipeline.addTrianglesDrawn(mIndicesCount/3); + U32 render_type = LLRender::TRIANGLES; + if (mDrawInfo) + { + render_type = mDrawInfo->mDrawMode; + } + mVertexBuffer->drawRange(render_type, mGeomIndex, mGeomIndex+mGeomCount-1, mIndicesCount, mIndicesIndex); + gPipeline.addTrianglesDrawn(mIndicesCount, render_type); } return mIndicesCount; diff --git a/indra/newview/llspatialpartition.cpp b/indra/newview/llspatialpartition.cpp index 2a57d48f16..d6e9256fee 100644 --- a/indra/newview/llspatialpartition.cpp +++ b/indra/newview/llspatialpartition.cpp @@ -3380,7 +3380,8 @@ LLDrawInfo::LLDrawInfo(U16 start, U16 end, U32 count, U32 offset, mVSize(0.f), mGroup(NULL), mFace(NULL), - mDistance(0.f) + mDistance(0.f), + mDrawMode(LLRender::TRIANGLES) { mDebugColor = (rand() << 16) + rand(); if (mStart >= mVertexBuffer->getRequestedVerts() || diff --git a/indra/newview/llspatialpartition.h b/indra/newview/llspatialpartition.h index 64c2a9acbc..7896488379 100644 --- a/indra/newview/llspatialpartition.h +++ b/indra/newview/llspatialpartition.h @@ -93,6 +93,7 @@ public: LLFace* mFace; //associated face F32 mDistance; LLVector3 mExtents[2]; + U32 mDrawMode; struct CompareTexture { diff --git a/indra/newview/llviewercontrol.cpp b/indra/newview/llviewercontrol.cpp index 827d34138f..0b7f6900ed 100644 --- a/indra/newview/llviewercontrol.cpp +++ b/indra/newview/llviewercontrol.cpp @@ -510,6 +510,7 @@ void settings_setup_listeners() gSavedSettings.getControl("FirstPersonAvatarVisible")->getSignal()->connect(boost::bind(&handleRenderAvatarMouselookChanged, _2)); gSavedSettings.getControl("RenderFarClip")->getSignal()->connect(boost::bind(&handleRenderFarClipChanged, _2)); gSavedSettings.getControl("RenderTerrainDetail")->getSignal()->connect(boost::bind(&handleTerrainDetailChanged, _2)); + gSavedSettings.getControl("RenderUseTriStrips")->getSignal()->connect(boost::bind(&handleResetVertexBuffersChanged, _2)); gSavedSettings.getControl("RenderAnimateTrees")->getSignal()->connect(boost::bind(&handleResetVertexBuffersChanged, _2)); gSavedSettings.getControl("RenderAvatarVP")->getSignal()->connect(boost::bind(&handleSetShaderChanged, _2)); gSavedSettings.getControl("VertexShaderEnable")->getSignal()->connect(boost::bind(&handleSetShaderChanged, _2)); diff --git a/indra/newview/llviewerjointmesh.cpp b/indra/newview/llviewerjointmesh.cpp index 1a67fc0966..7225aa1523 100644 --- a/indra/newview/llviewerjointmesh.cpp +++ b/indra/newview/llviewerjointmesh.cpp @@ -626,7 +626,7 @@ U32 LLViewerJointMesh::drawShape( F32 pixelArea, BOOL first_pass, BOOL is_dummy) mFace->mVertexBuffer->drawRange(LLRender::TRIANGLES, start, end, count, offset); glPopMatrix(); } - gPipeline.addTrianglesDrawn(count/3); + gPipeline.addTrianglesDrawn(count); triangle_count += count; diff --git a/indra/newview/llvotree.cpp b/indra/newview/llvotree.cpp index 24f1c4bd24..55e2c58a52 100644 --- a/indra/newview/llvotree.cpp +++ b/indra/newview/llvotree.cpp @@ -1141,7 +1141,7 @@ U32 LLVOTree::drawBranchPipeline(LLMatrix4& matrix, U16* indicesp, S32 trunk_LOD glLoadMatrixf((F32*) scale_mat.mMatrix); glDrawElements(GL_TRIANGLES, sLODIndexCount[trunk_LOD], GL_UNSIGNED_SHORT, indicesp + sLODIndexOffset[trunk_LOD]); - gPipeline.addTrianglesDrawn(LEAF_INDICES/3); + gPipeline.addTrianglesDrawn(LEAF_INDICES); stop_glerror(); ret += sLODIndexCount[trunk_LOD]; } @@ -1191,7 +1191,7 @@ U32 LLVOTree::drawBranchPipeline(LLMatrix4& matrix, U16* indicesp, S32 trunk_LOD glLoadMatrixf((F32*) scale_mat.mMatrix); glDrawElements(GL_TRIANGLES, LEAF_INDICES, GL_UNSIGNED_SHORT, indicesp); - gPipeline.addTrianglesDrawn(LEAF_INDICES/3); + gPipeline.addTrianglesDrawn(LEAF_INDICES); stop_glerror(); ret += LEAF_INDICES; } @@ -1216,7 +1216,7 @@ U32 LLVOTree::drawBranchPipeline(LLMatrix4& matrix, U16* indicesp, S32 trunk_LOD glLoadMatrixf((F32*) scale_mat.mMatrix); glDrawElements(GL_TRIANGLES, LEAF_INDICES, GL_UNSIGNED_SHORT, indicesp); - gPipeline.addTrianglesDrawn(LEAF_INDICES/3); + gPipeline.addTrianglesDrawn(LEAF_INDICES); stop_glerror(); ret += LEAF_INDICES; diff --git a/indra/newview/llvovolume.cpp b/indra/newview/llvovolume.cpp index 25ed931fb5..b87ad7fc47 100644 --- a/indra/newview/llvovolume.cpp +++ b/indra/newview/llvovolume.cpp @@ -1514,7 +1514,14 @@ void LLVOVolume::updateFaceSize(S32 idx) else { const LLVolumeFace& vol_face = getVolume()->getVolumeFace(idx); - facep->setSize(vol_face.mVertices.size(), vol_face.mIndices.size()); + if (LLPipeline::sUseTriStrips) + { + facep->setSize(vol_face.mVertices.size(), vol_face.mTriStrip.size()); + } + else + { + facep->setSize(vol_face.mVertices.size(), vol_face.mIndices.size()); + } } } @@ -3264,6 +3271,11 @@ void LLVolumeGeometryManager::registerFace(LLSpatialGroup* group, LLFace* facep, draw_info->mExtents[0] = facep->mExtents[0]; draw_info->mExtents[1] = facep->mExtents[1]; validate_draw_info(*draw_info); + + if (LLPipeline::sUseTriStrips) + { + draw_info->mDrawMode = LLRender::TRIANGLE_STRIP; + } } } @@ -3348,7 +3360,7 @@ void LLVolumeGeometryManager::rebuildGeom(LLSpatialGroup* group) drawablep->updateFaceSize(i); LLFace* facep = drawablep->getFace(i); - if (cur_total > max_total) + if (cur_total > max_total || facep->getIndicesCount() <= 0 || facep->getGeomCount() <= 0) { facep->mVertexBuffer = NULL; facep->mLastVertexBuffer = NULL; diff --git a/indra/newview/llvowlsky.cpp b/indra/newview/llvowlsky.cpp index 0272a2ab34..1749ccef94 100644 --- a/indra/newview/llvowlsky.cpp +++ b/indra/newview/llvowlsky.cpp @@ -519,7 +519,7 @@ void LLVOWLSky::drawDome(void) LLRender::TRIANGLE_STRIP, 0, strips_segment->getRequestedVerts()-1, strips_segment->getRequestedIndices(), 0); - gPipeline.addTrianglesDrawn(strips_segment->getRequestedIndices() - 2); + gPipeline.addTrianglesDrawn(strips_segment->getRequestedIndices(), LLRender::TRIANGLE_STRIP); } #else diff --git a/indra/newview/pipeline.cpp b/indra/newview/pipeline.cpp index d5f87b73fe..91da9e39e0 100644 --- a/indra/newview/pipeline.cpp +++ b/indra/newview/pipeline.cpp @@ -270,6 +270,7 @@ BOOL LLPipeline::sDelayVBUpdate = TRUE; BOOL LLPipeline::sFastAlpha = TRUE; BOOL LLPipeline::sDisableShaders = FALSE; BOOL LLPipeline::sRenderBump = TRUE; +BOOL LLPipeline::sUseTriStrips = TRUE; BOOL LLPipeline::sUseFarClip = TRUE; BOOL LLPipeline::sShadowRender = FALSE; BOOL LLPipeline::sWaterReflections = FALSE; @@ -358,6 +359,7 @@ void LLPipeline::init() sDynamicLOD = gSavedSettings.getBOOL("RenderDynamicLOD"); sRenderBump = gSavedSettings.getBOOL("RenderObjectBump"); + sUseTriStrips = gSavedSettings.getBOOL("RenderUseTriStrips"); sRenderAttachedLights = gSavedSettings.getBOOL("RenderAttachedLights"); sRenderAttachedParticles = gSavedSettings.getBOOL("RenderAttachedParticles"); @@ -3495,9 +3497,19 @@ void LLPipeline::renderGeomShadow(LLCamera& camera) } -void LLPipeline::addTrianglesDrawn(S32 count) +void LLPipeline::addTrianglesDrawn(S32 index_count, U32 render_type) { assertInitialized(); + S32 count = 0; + if (render_type == LLRender::TRIANGLE_STRIP) + { + count = index_count-2; + } + else + { + count = index_count/3; + } + mTrianglesDrawn += count; mBatchCount++; mMaxBatchSize = llmax(mMaxBatchSize, count); @@ -5349,6 +5361,7 @@ void LLPipeline::resetVertexBuffers(LLDrawable* drawable) void LLPipeline::resetVertexBuffers() { sRenderBump = gSavedSettings.getBOOL("RenderObjectBump"); + sUseTriStrips = gSavedSettings.getBOOL("RenderUseTriStrips"); for (LLWorld::region_list_t::const_iterator iter = LLWorld::getInstance()->getRegionList().begin(); iter != LLWorld::getInstance()->getRegionList().end(); ++iter) diff --git a/indra/newview/pipeline.h b/indra/newview/pipeline.h index c5285943e8..1bf2ee941b 100644 --- a/indra/newview/pipeline.h +++ b/indra/newview/pipeline.h @@ -277,7 +277,7 @@ public: LLCullResult::sg_list_t::iterator beginAlphaGroups(); LLCullResult::sg_list_t::iterator endAlphaGroups(); - void addTrianglesDrawn(S32 count); + void addTrianglesDrawn(S32 index_count, U32 render_type = LLRender::TRIANGLES); BOOL hasRenderType(const U32 type) const { return (type && (mRenderTypeMask & (1< Date: Sat, 13 Feb 2010 12:48:42 +0000 Subject: Backed out changeset d89b00b44ab6 --- indra/newview/llagent.cpp | 3 + indra/newview/llappviewer.cpp | 8 +- indra/newview/lldrawable.cpp | 3 - indra/newview/llflexibleobject.cpp | 15 ++- indra/newview/llglsandbox.cpp | 24 +++-- indra/newview/llviewerobject.cpp | 3 + indra/newview/llviewerobjectlist.cpp | 171 ++++++++++++++++++----------------- indra/newview/llviewerobjectlist.h | 31 ++++--- indra/newview/llviewerwindow.cpp | 4 +- indra/newview/llvoclouds.cpp | 4 +- indra/newview/llvotextbubble.cpp | 3 + indra/newview/llvovolume.cpp | 17 +++- indra/newview/llworld.cpp | 3 + indra/newview/pipeline.cpp | 64 +++---------- indra/newview/pipeline.h | 5 +- 15 files changed, 185 insertions(+), 173 deletions(-) (limited to 'indra/newview/llvovolume.cpp') diff --git a/indra/newview/llagent.cpp b/indra/newview/llagent.cpp index 2354323a66..9f2186f7f7 100644 --- a/indra/newview/llagent.cpp +++ b/indra/newview/llagent.cpp @@ -3017,6 +3017,9 @@ void LLAgent::endAnimationUpdateUI() //----------------------------------------------------------------------------- void LLAgent::updateCamera() { + static LLFastTimer::DeclareTimer ftm("Camera"); + LLFastTimer t(ftm); + //Ventrella - changed camera_skyward to the new global "mCameraUpVector" mCameraUpVector = LLVector3::z_axis; //LLVector3 camera_skyward(0.f, 0.f, 1.f); diff --git a/indra/newview/llappviewer.cpp b/indra/newview/llappviewer.cpp index 2d694eefd3..9eb793783b 100644 --- a/indra/newview/llappviewer.cpp +++ b/indra/newview/llappviewer.cpp @@ -3599,13 +3599,15 @@ void LLAppViewer::idle() { // Handle pending gesture processing + static LLFastTimer::DeclareTimer ftm("Agent Position"); + LLFastTimer t(ftm); LLGestureManager::instance().update(); gAgent.updateAgentPosition(gFrameDTClamped, yaw, current_mouse.mX, current_mouse.mY); } { - LLFastTimer t(FTM_OBJECTLIST_UPDATE); // Actually "object update" + LLFastTimer t(FTM_OBJECTLIST_UPDATE); if (!(logoutRequestSent() && hasSavedFinalSnapshot())) { @@ -3639,6 +3641,8 @@ void LLAppViewer::idle() // { + static LLFastTimer::DeclareTimer ftm("HUD Effects"); + LLFastTimer t(ftm); LLSelectMgr::getInstance()->updateEffects(); LLHUDManager::getInstance()->cleanupEffects(); LLHUDManager::getInstance()->sendEffects(); @@ -3895,7 +3899,7 @@ void LLAppViewer::sendLogoutRequest() static F32 CheckMessagesMaxTime = CHECK_MESSAGES_DEFAULT_MAX_TIME; #endif -static LLFastTimer::DeclareTimer FTM_IDLE_NETWORK("Network"); +static LLFastTimer::DeclareTimer FTM_IDLE_NETWORK("Idle Network"); void LLAppViewer::idleNetwork() { diff --git a/indra/newview/lldrawable.cpp b/indra/newview/lldrawable.cpp index d60330024a..244fed791f 100644 --- a/indra/newview/lldrawable.cpp +++ b/indra/newview/lldrawable.cpp @@ -386,8 +386,6 @@ void LLDrawable::makeActive() mParent->makeActive(); } - gPipeline.setActive(this, TRUE); - //all child objects must also be active llassert_always(mVObjp); @@ -434,7 +432,6 @@ void LLDrawable::makeStatic(BOOL warning_enabled) if (isState(ACTIVE)) { clearState(ACTIVE); - gPipeline.setActive(this, FALSE); if (mParent.notNull() && mParent->isActive() && warning_enabled) { diff --git a/indra/newview/llflexibleobject.cpp b/indra/newview/llflexibleobject.cpp index aea2de8e92..561965d021 100644 --- a/indra/newview/llflexibleobject.cpp +++ b/indra/newview/llflexibleobject.cpp @@ -51,6 +51,9 @@ /*static*/ F32 LLVolumeImplFlexible::sUpdateFactor = 1.0f; +static LLFastTimer::DeclareTimer FTM_FLEXIBLE_REBUILD("Rebuild"); +static LLFastTimer::DeclareTimer FTM_DO_FLEXIBLE_UPDATE("Update"); + // LLFlexibleObjectData::pack/unpack now in llprimitive.cpp //----------------------------------------------- @@ -194,7 +197,6 @@ void LLVolumeImplFlexible::remapSections(LLFlexibleObjectSection *source, S32 so } } - //----------------------------------------------------------------------------- void LLVolumeImplFlexible::setAttributesOfAllSections(LLVector3* inScale) { @@ -363,6 +365,7 @@ inline S32 log2(S32 x) void LLVolumeImplFlexible::doFlexibleUpdate() { + LLFastTimer ftm(FTM_DO_FLEXIBLE_UPDATE); LLVolume* volume = mVO->getVolume(); LLPath *path = &volume->getPath(); if (mSimulateRes == 0) @@ -693,7 +696,10 @@ BOOL LLVolumeImplFlexible::doUpdateGeometry(LLDrawable *drawable) } volume->updateRelativeXform(); - doFlexibleUpdate(); + { + LLFastTimer t(FTM_DO_FLEXIBLE_UPDATE); + doFlexibleUpdate(); + } // Object may have been rotated, which means it needs a rebuild. See SL-47220 BOOL rotated = FALSE; @@ -710,7 +716,10 @@ BOOL LLVolumeImplFlexible::doUpdateGeometry(LLDrawable *drawable) volume->regenFaces(); volume->mDrawable->setState(LLDrawable::REBUILD_VOLUME); volume->dirtySpatialGroup(); - doFlexibleRebuild(); + { + LLFastTimer t(FTM_FLEXIBLE_REBUILD); + doFlexibleRebuild(); + } volume->genBBoxes(isVolumeGlobal()); } else if (!mUpdated || rotated) diff --git a/indra/newview/llglsandbox.cpp b/indra/newview/llglsandbox.cpp index 750a9d478f..8569e208eb 100644 --- a/indra/newview/llglsandbox.cpp +++ b/indra/newview/llglsandbox.cpp @@ -897,19 +897,21 @@ void LLViewerObjectList::renderObjectBeacons() S32 last_line_width = -1; // gGL.begin(LLRender::LINES); // Always happens in (line_width != last_line_width) - for (S32 i = 0; i < mDebugBeacons.count(); i++) + BOOL flush = FALSE; + for (std::vector::iterator iter = mDebugBeacons.begin(); iter != mDebugBeacons.end(); ++iter) { - const LLDebugBeacon &debug_beacon = mDebugBeacons[i]; + const LLDebugBeacon &debug_beacon = *iter; LLColor4 color = debug_beacon.mColor; color.mV[3] *= 0.25f; S32 line_width = debug_beacon.mLineWidth; if (line_width != last_line_width) { - if (i > 0) + if (flush) { gGL.end(); - gGL.flush(); } + flush = TRUE; + gGL.flush(); glLineWidth( (F32)line_width ); last_line_width = line_width; gGL.begin(LLRender::LINES); @@ -936,18 +938,20 @@ void LLViewerObjectList::renderObjectBeacons() S32 last_line_width = -1; // gGL.begin(LLRender::LINES); // Always happens in (line_width != last_line_width) - for (S32 i = 0; i < mDebugBeacons.count(); i++) + BOOL flush = FALSE; + for (std::vector::iterator iter = mDebugBeacons.begin(); iter != mDebugBeacons.end(); ++iter) { - const LLDebugBeacon &debug_beacon = mDebugBeacons[i]; + const LLDebugBeacon &debug_beacon = *iter; S32 line_width = debug_beacon.mLineWidth; if (line_width != last_line_width) { - if (i > 0) + if (flush) { gGL.end(); - gGL.flush(); } + flush = TRUE; + gGL.flush(); glLineWidth( (F32)line_width ); last_line_width = line_width; gGL.begin(LLRender::LINES); @@ -969,9 +973,9 @@ void LLViewerObjectList::renderObjectBeacons() gGL.flush(); glLineWidth(1.f); - for (S32 i = 0; i < mDebugBeacons.count(); i++) + for (std::vector::iterator iter = mDebugBeacons.begin(); iter != mDebugBeacons.end(); ++iter) { - LLDebugBeacon &debug_beacon = mDebugBeacons[i]; + LLDebugBeacon &debug_beacon = *iter; if (debug_beacon.mString == "") { continue; diff --git a/indra/newview/llviewerobject.cpp b/indra/newview/llviewerobject.cpp index 886f1d9ef5..4fdfc37d6c 100644 --- a/indra/newview/llviewerobject.cpp +++ b/indra/newview/llviewerobject.cpp @@ -2001,6 +2001,9 @@ BOOL LLViewerObject::isActive() const BOOL LLViewerObject::idleUpdate(LLAgent &agent, LLWorld &world, const F64 &time) { + static LLFastTimer::DeclareTimer ftm("Viewer Object"); + LLFastTimer t(ftm); + if (mDead) { // It's dead. Don't update it. diff --git a/indra/newview/llviewerobjectlist.cpp b/indra/newview/llviewerobjectlist.cpp index 96828ee1b6..6347090f71 100644 --- a/indra/newview/llviewerobjectlist.cpp +++ b/indra/newview/llviewerobjectlist.cpp @@ -93,7 +93,7 @@ extern LLPipeline gPipeline; // Statics for object lookup tables. U32 LLViewerObjectList::sSimulatorMachineIndex = 1; // Not zero deliberately, to speed up index check. -LLMap LLViewerObjectList::sIPAndPortToIndex; +std::map LLViewerObjectList::sIPAndPortToIndex; std::map LLViewerObjectList::sIndexAndLocalIDToUUID; LLViewerObjectList::LLViewerObjectList() @@ -571,10 +571,9 @@ void LLViewerObjectList::processCachedObjectUpdate(LLMessageSystem *mesgsys, void LLViewerObjectList::dirtyAllObjectInventory() { - S32 count = mObjects.count(); - for(S32 i = 0; i < count; ++i) + for (vobj_list_t::iterator iter = mObjects.begin(); iter != mObjects.end(); ++iter) { - mObjects[i]->dirtyInventory(); + (*iter)->dirtyInventory(); } } @@ -587,14 +586,14 @@ void LLViewerObjectList::updateApparentAngles(LLAgent &agent) S32 num_updates, max_value; if (NUM_BINS - 1 == mCurBin) { - num_updates = mObjects.count() - mCurLazyUpdateIndex; - max_value = mObjects.count(); + num_updates = (S32) mObjects.size() - mCurLazyUpdateIndex; + max_value = (S32) mObjects.size(); gTextureList.setUpdateStats(TRUE); } else { - num_updates = (mObjects.count() / NUM_BINS) + 1; - max_value = llmin(mObjects.count(), mCurLazyUpdateIndex + num_updates); + num_updates = ((S32) mObjects.size() / NUM_BINS) + 1; + max_value = llmin((S32) mObjects.size(), mCurLazyUpdateIndex + num_updates); } @@ -647,7 +646,7 @@ void LLViewerObjectList::updateApparentAngles(LLAgent &agent) } mCurLazyUpdateIndex = max_value; - if (mCurLazyUpdateIndex == mObjects.count()) + if (mCurLazyUpdateIndex == mObjects.size()) { mCurLazyUpdateIndex = 0; } @@ -694,20 +693,26 @@ void LLViewerObjectList::update(LLAgent &agent, LLWorld &world) // Make a copy of the list in case something in idleUpdate() messes with it std::vector idle_list; - idle_list.reserve( mActiveObjects.size() ); + + static LLFastTimer::DeclareTimer idle_copy("Idle Copy"); - for (std::set >::iterator active_iter = mActiveObjects.begin(); - active_iter != mActiveObjects.end(); active_iter++) { - objectp = *active_iter; - if (objectp) + LLFastTimer t(idle_copy); + idle_list.reserve( mActiveObjects.size() ); + + for (std::set >::iterator active_iter = mActiveObjects.begin(); + active_iter != mActiveObjects.end(); active_iter++) { - idle_list.push_back( objectp ); - } - else - { // There shouldn't be any NULL pointers in the list, but they have caused - // crashes before. This may be idleUpdate() messing with the list. - llwarns << "LLViewerObjectList::update has a NULL objectp" << llendl; + objectp = *active_iter; + if (objectp) + { + idle_list.push_back( objectp ); + } + else + { // There shouldn't be any NULL pointers in the list, but they have caused + // crashes before. This may be idleUpdate() messing with the list. + llwarns << "LLViewerObjectList::update has a NULL objectp" << llendl; + } } } @@ -807,7 +812,7 @@ void LLViewerObjectList::update(LLAgent &agent, LLWorld &world) } */ - LLViewerStats::getInstance()->mNumObjectsStat.addValue(mObjects.count()); + LLViewerStats::getInstance()->mNumObjectsStat.addValue((S32) mObjects.size()); LLViewerStats::getInstance()->mNumActiveObjectsStat.addValue(num_active_objects); LLViewerStats::getInstance()->mNumSizeCulledStat.addValue(mNumSizeCulled); LLViewerStats::getInstance()->mNumVisCulledStat.addValue(mNumVisCulled); @@ -815,9 +820,9 @@ void LLViewerObjectList::update(LLAgent &agent, LLWorld &world) void LLViewerObjectList::clearDebugText() { - for (S32 i = 0; i < mObjects.count(); i++) + for (vobj_list_t::iterator iter = mObjects.begin(); iter != mObjects.end(); ++iter) { - mObjects[i]->setDebugText(""); + (*iter)->setDebugText(""); } } @@ -856,7 +861,7 @@ void LLViewerObjectList::cleanupReferences(LLViewerObject *objectp) if (objectp->isOnMap()) { - mMapObjects.removeObj(objectp); + removeFromMap(objectp); } // Don't clean up mObject references, these will be cleaned up more efficiently later! @@ -913,10 +918,10 @@ void LLViewerObjectList::killObjects(LLViewerRegion *regionp) { LLViewerObject *objectp; - S32 i; - for (i = 0; i < mObjects.count(); i++) + + for (vobj_list_t::iterator iter = mObjects.begin(); iter != mObjects.end(); ++iter) { - objectp = mObjects[i]; + objectp = *iter; if (objectp->mRegionp == regionp) { @@ -933,10 +938,9 @@ void LLViewerObjectList::killAllObjects() // Used only on global destruction. LLViewerObject *objectp; - for (S32 i = 0; i < mObjects.count(); i++) + for (vobj_list_t::iterator iter = mObjects.begin(); iter != mObjects.end(); ++iter) { - objectp = mObjects[i]; - + objectp = *iter; killObject(objectp); llassert(objectp->isDead()); } @@ -945,7 +949,7 @@ void LLViewerObjectList::killAllObjects() if(!mObjects.empty()) { - llwarns << "LLViewerObjectList::killAllObjects still has entries in mObjects: " << mObjects.count() << llendl; + llwarns << "LLViewerObjectList::killAllObjects still has entries in mObjects: " << mObjects.size() << llendl; mObjects.clear(); } @@ -970,16 +974,15 @@ void LLViewerObjectList::cleanDeadObjects(BOOL use_timer) return; } - S32 i = 0; S32 num_removed = 0; LLViewerObject *objectp; - while (i < mObjects.count()) + for (vobj_list_t::iterator iter = mObjects.begin(); iter != mObjects.end(); ) { // Scan for all of the dead objects and remove any "global" references to them. - objectp = mObjects[i]; + objectp = *iter; if (objectp->isDead()) { - mObjects.remove(i); + iter = mObjects.erase(iter); num_removed++; if (num_removed == mNumDeadObjects) @@ -990,8 +993,7 @@ void LLViewerObjectList::cleanDeadObjects(BOOL use_timer) } else { - // iterate, this isn't a dead object. - i++; + ++iter; } } @@ -1041,12 +1043,11 @@ void LLViewerObjectList::shiftObjects(const LLVector3 &offset) } LLViewerObject *objectp; - S32 i; - for (i = 0; i < mObjects.count(); i++) + for (vobj_list_t::iterator iter = mObjects.begin(); iter != mObjects.end(); ++iter) { - objectp = getObject(i); + objectp = *iter; // There could be dead objects on the object list, so don't update stuff if the object is dead. - if (objectp) + if (!objectp->isDead()) { objectp->updatePositionCaches(); @@ -1076,9 +1077,9 @@ void LLViewerObjectList::renderObjectsForMap(LLNetMap &netmap) F32 max_radius = gSavedSettings.getF32("MiniMapPrimMaxRadius"); - for (S32 i = 0; i < mMapObjects.count(); i++) + for (vobj_list_t::iterator iter = mMapObjects.begin(); iter != mMapObjects.end(); ++iter) { - LLViewerObject* objectp = mMapObjects[i]; + LLViewerObject* objectp = *iter; if (!objectp->getRegion() || objectp->isOrphaned() || objectp->isAttachment()) { continue; @@ -1144,21 +1145,14 @@ void LLViewerObjectList::renderObjectBounds(const LLVector3 ¢er) { } -void LLViewerObjectList::renderObjectsForSelect(LLCamera &camera, const LLRect& screen_rect, BOOL pick_parcel_wall, BOOL render_transparent) -{ - generatePickList(camera); - renderPickList(screen_rect, pick_parcel_wall, render_transparent); -} - void LLViewerObjectList::generatePickList(LLCamera &camera) { LLViewerObject *objectp; S32 i; // Reset all of the GL names to zero. - for (i = 0; i < mObjects.count(); i++) + for (vobj_list_t::iterator iter = mObjects.begin(); iter != mObjects.end(); ++iter) { - objectp = mObjects[i]; - objectp->mGLName = 0; + (*iter)->mGLName = 0; } mSelectPickList.clear(); @@ -1321,17 +1315,19 @@ void LLViewerObjectList::addDebugBeacon(const LLVector3 &pos_agent, const LLColor4 &text_color, S32 line_width) { - LLDebugBeacon *beaconp = mDebugBeacons.reserve_block(1); - beaconp->mPositionAgent = pos_agent; - beaconp->mString = string; - beaconp->mColor = color; - beaconp->mTextColor = text_color; - beaconp->mLineWidth = line_width; + LLDebugBeacon beacon; + beacon.mPositionAgent = pos_agent; + beacon.mString = string; + beacon.mColor = color; + beacon.mTextColor = text_color; + beacon.mLineWidth = line_width; + + mDebugBeacons.push_back(beacon); } void LLViewerObjectList::resetObjectBeacons() { - mDebugBeacons.reset(); + mDebugBeacons.clear(); } LLViewerObject *LLViewerObjectList::createObjectViewer(const LLPCode pcode, LLViewerRegion *regionp) @@ -1349,7 +1345,7 @@ LLViewerObject *LLViewerObjectList::createObjectViewer(const LLPCode pcode, LLVi mUUIDObjectMap[fullid] = objectp; - mObjects.put(objectp); + mObjects.push_back(objectp); updateActive(objectp); @@ -1388,7 +1384,7 @@ LLViewerObject *LLViewerObjectList::createObject(const LLPCode pcode, LLViewerRe gMessageSystem->getSenderIP(), gMessageSystem->getSenderPort()); - mObjects.put(objectp); + mObjects.push_back(objectp); updateActive(objectp); @@ -1411,11 +1407,11 @@ LLViewerObject *LLViewerObjectList::replaceObject(const LLUUID &id, const LLPCod S32 LLViewerObjectList::findReferences(LLDrawable *drawablep) const { LLViewerObject *objectp; - S32 i; S32 num_refs = 0; - for (i = 0; i < mObjects.count(); i++) + + for (vobj_list_t::const_iterator iter = mObjects.begin(); iter != mObjects.end(); ++iter) { - objectp = mObjects[i]; + objectp = *iter; if (objectp->mDrawable.notNull()) { num_refs += objectp->mDrawable->findReferences(drawablep); @@ -1460,15 +1456,15 @@ void LLViewerObjectList::orphanize(LLViewerObject *childp, U32 parent_id, U32 ip // Unknown parent, add to orpaned child list U64 parent_info = getIndex(parent_id, ip, port); - if (-1 == mOrphanParents.find(parent_info)) + if (std::find(mOrphanParents.begin(), mOrphanParents.end(), parent_info) == mOrphanParents.end()) { - mOrphanParents.put(parent_info); + mOrphanParents.push_back(parent_info); } LLViewerObjectList::OrphanInfo oi(parent_info, childp->mID); - if (-1 == mOrphanChildren.find(oi)) + if (std::find(mOrphanChildren.begin(), mOrphanChildren.end(), oi) == mOrphanChildren.end()) { - mOrphanChildren.put(oi); + mOrphanChildren.push_back(oi); mNumOrphans++; } } @@ -1491,28 +1487,29 @@ void LLViewerObjectList::findOrphans(LLViewerObject* objectp, U32 ip, U32 port) // See if we are a parent of an orphan. // Note: This code is fairly inefficient but it should happen very rarely. // It can be sped up if this is somehow a performance issue... - if (0 == mOrphanParents.count()) + if (mOrphanParents.empty()) { // no known orphan parents return; } - if (-1 == mOrphanParents.find(getIndex(objectp->mLocalID, ip, port))) + if (std::find(mOrphanParents.begin(), mOrphanParents.end(), getIndex(objectp->mLocalID, ip, port)) == mOrphanParents.end()) { // did not find objectp in OrphanParent list return; } - S32 i; U64 parent_info = getIndex(objectp->mLocalID, ip, port); BOOL orphans_found = FALSE; // Iterate through the orphan list, and set parents of matching children. - for (i = 0; i < mOrphanChildren.count(); i++) - { - if (mOrphanChildren[i].mParentInfo != parent_info) + + for (std::vector::iterator iter = mOrphanChildren.begin(); iter != mOrphanChildren.end(); ) + { + if (iter->mParentInfo != parent_info) { + ++iter; continue; } - LLViewerObject *childp = findObject(mOrphanChildren[i].mChildInfo); + LLViewerObject *childp = findObject(iter->mChildInfo); if (childp) { if (childp == objectp) @@ -1546,29 +1543,35 @@ void LLViewerObjectList::findOrphans(LLViewerObject* objectp, U32 ip, U32 port) objectp->addChild(childp); orphans_found = TRUE; + ++iter; } else { llinfos << "Missing orphan child, removing from list" << llendl; - mOrphanChildren.remove(i); - i--; + + iter = mOrphanChildren.erase(iter); } } // Remove orphan parent and children from lists now that they've been found - mOrphanParents.remove(mOrphanParents.find(parent_info)); - - i = 0; - while (i < mOrphanChildren.count()) { - if (mOrphanChildren[i].mParentInfo == parent_info) + std::vector::iterator iter = std::find(mOrphanParents.begin(), mOrphanParents.end(), parent_info); + if (iter != mOrphanParents.end()) + { + mOrphanParents.erase(iter); + } + } + + for (std::vector::iterator iter = mOrphanChildren.begin(); iter != mOrphanChildren.end(); ) + { + if (iter->mParentInfo == parent_info) { - mOrphanChildren.remove(i); + iter = mOrphanChildren.erase(iter); mNumOrphans--; } else { - i++; + ++iter; } } diff --git a/indra/newview/llviewerobjectlist.h b/indra/newview/llviewerobjectlist.h index ace5c5038e..8d3d2c4b44 100644 --- a/indra/newview/llviewerobjectlist.h +++ b/indra/newview/llviewerobjectlist.h @@ -38,8 +38,6 @@ // common includes #include "llstat.h" -#include "lldarrayptr.h" -#include "llmap.h" // *TODO: switch to std::map #include "llstring.h" // project includes @@ -50,7 +48,7 @@ class LLNetMap; class LLDebugBeacon; const U32 CLOSE_BIN_SIZE = 10; -const U32 NUM_BINS = 16; +const U32 NUM_BINS = 128; // GL name = position in object list + GL_NAME_INDEX_OFFSET so that // we can have special numbers like zero. @@ -111,13 +109,12 @@ public: void updateAvatarVisibility(); // Selection related stuff - void renderObjectsForSelect(LLCamera &camera, const LLRect& screen_rect, BOOL pick_parcel_wall = FALSE, BOOL render_transparent = TRUE); void generatePickList(LLCamera &camera); void renderPickList(const LLRect& screen_rect, BOOL pick_parcel_wall, BOOL render_transparent); LLViewerObject *getSelectedObject(const U32 object_id); - inline S32 getNumObjects() { return mObjects.count(); } + inline S32 getNumObjects() { return (S32) mObjects.size(); } void addToMap(LLViewerObject *objectp); void removeFromMap(LLViewerObject *objectp); @@ -131,7 +128,7 @@ public: S32 findReferences(LLDrawable *drawablep) const; // Find references to drawable in all objects, and return value. - S32 getOrphanParentCount() const { return mOrphanParents.count(); } + S32 getOrphanParentCount() const { return (S32) mOrphanParents.size(); } S32 getOrphanCount() const { return mNumOrphans; } void orphanize(LLViewerObject *childp, U32 parent_id, U32 ip, U32 port); void findOrphans(LLViewerObject* objectp, U32 ip, U32 port); @@ -179,26 +176,28 @@ public: S32 mNumUnknownKills; S32 mNumDeadObjects; protected: - LLDynamicArray mOrphanParents; // LocalID/ip,port of orphaned objects - LLDynamicArray mOrphanChildren; // UUID's of orphaned objects + std::vector mOrphanParents; // LocalID/ip,port of orphaned objects + std::vector mOrphanChildren; // UUID's of orphaned objects S32 mNumOrphans; - LLDynamicArrayPtr, 256> mObjects; + typedef std::vector > vobj_list_t; + + vobj_list_t mObjects; std::set > mActiveObjects; - LLDynamicArrayPtr > mMapObjects; + vobj_list_t mMapObjects; typedef std::map > vo_map; vo_map mDeadObjects; // Need to keep multiple entries per UUID std::map > mUUIDObjectMap; - LLDynamicArray mDebugBeacons; + std::vector mDebugBeacons; S32 mCurLazyUpdateIndex; static U32 sSimulatorMachineIndex; - static LLMap sIPAndPortToIndex; + static std::map sIPAndPortToIndex; static std::map sIndexAndLocalIDToUUID; @@ -260,12 +259,16 @@ inline LLViewerObject *LLViewerObjectList::getObject(const S32 index) inline void LLViewerObjectList::addToMap(LLViewerObject *objectp) { - mMapObjects.put(objectp); + mMapObjects.push_back(objectp); } inline void LLViewerObjectList::removeFromMap(LLViewerObject *objectp) { - mMapObjects.removeObj(objectp); + std::vector >::iterator iter = std::find(mMapObjects.begin(), mMapObjects.end(), objectp); + if (iter != mMapObjects.end()) + { + mMapObjects.erase(iter); + } } diff --git a/indra/newview/llviewerwindow.cpp b/indra/newview/llviewerwindow.cpp index cd6b9e2c50..de4317b2de 100644 --- a/indra/newview/llviewerwindow.cpp +++ b/indra/newview/llviewerwindow.cpp @@ -2316,6 +2316,9 @@ void append_xui_tooltip(LLView* viewp, LLToolTip::Params& params) // event processing. void LLViewerWindow::updateUI() { + static LLFastTimer::DeclareTimer ftm("Update UI"); + LLFastTimer t(ftm); + static std::string last_handle_msg; // animate layout stacks so we have up to date rect for world view @@ -2895,7 +2898,6 @@ void LLViewerWindow::saveLastMouse(const LLCoordGL &point) // Must be called after displayObjects is called, which sets the mGLName parameter // NOTE: This function gets called 3 times: // render_ui_3d: FALSE, FALSE, TRUE -// renderObjectsForSelect: TRUE, pick_parcel_wall, FALSE // render_hud_elements: FALSE, FALSE, FALSE void LLViewerWindow::renderSelections( BOOL for_gl_pick, BOOL pick_parcel_walls, BOOL for_hud ) { diff --git a/indra/newview/llvoclouds.cpp b/indra/newview/llvoclouds.cpp index 177cb16c50..5153cef709 100644 --- a/indra/newview/llvoclouds.cpp +++ b/indra/newview/llvoclouds.cpp @@ -77,9 +77,11 @@ BOOL LLVOClouds::isActive() const return TRUE; } - BOOL LLVOClouds::idleUpdate(LLAgent &agent, LLWorld &world, const F64 &time) { + static LLFastTimer::DeclareTimer ftm("Idle Clouds"); + LLFastTimer t(ftm); + if (!(gPipeline.hasRenderType(LLPipeline::RENDER_TYPE_CLOUDS))) { return TRUE; diff --git a/indra/newview/llvotextbubble.cpp b/indra/newview/llvotextbubble.cpp index 75beab519e..428ef20006 100644 --- a/indra/newview/llvotextbubble.cpp +++ b/indra/newview/llvotextbubble.cpp @@ -84,6 +84,9 @@ BOOL LLVOTextBubble::isActive() const BOOL LLVOTextBubble::idleUpdate(LLAgent &agent, LLWorld &world, const F64 &time) { + static LLFastTimer::DeclareTimer ftm("Text Bubble"); + LLFastTimer t(ftm); + F32 dt = mUpdateTimer.getElapsedTimeF32(); // Die after a few seconds. if (dt > 1.5f) diff --git a/indra/newview/llvovolume.cpp b/indra/newview/llvovolume.cpp index bfe38c14ba..3cdf485d7d 100644 --- a/indra/newview/llvovolume.cpp +++ b/indra/newview/llvovolume.cpp @@ -597,6 +597,9 @@ BOOL LLVOVolume::idleUpdate(LLAgent &agent, LLWorld &world, const F64 &time) { LLViewerObject::idleUpdate(agent, world, time); + static LLFastTimer::DeclareTimer ftm("Volume"); + LLFastTimer t(ftm); + if (mDead || mDrawable.isNull()) { return TRUE; @@ -618,6 +621,18 @@ BOOL LLVOVolume::idleUpdate(LLAgent &agent, LLWorld &world, const F64 &time) mVolumeImpl->doIdleUpdate(agent, world, time); } + const S32 MAX_ACTIVE_OBJECT_QUIET_FRAMES = 40; + + if (mDrawable->isActive()) + { + if (mDrawable->isRoot() && + mDrawable->mQuietCount++ > MAX_ACTIVE_OBJECT_QUIET_FRAMES && + (!mDrawable->getParent() || !mDrawable->getParent()->isActive())) + { + mDrawable->makeStatic(); + } + } + return TRUE; } @@ -1035,7 +1050,7 @@ BOOL LLVOVolume::calcLOD() S32 cur_detail = 0; F32 radius = getVolume()->mLODScaleBias.scaledVec(getScale()).length(); - F32 distance = llmin(mDrawable->mDistanceWRTCamera, MAX_LOD_DISTANCE); + F32 distance = mDrawable->mDistanceWRTCamera; //llmin(mDrawable->mDistanceWRTCamera, MAX_LOD_DISTANCE); distance *= sDistanceFactor; F32 rampDist = LLVOVolume::sLODFactor * 2; diff --git a/indra/newview/llworld.cpp b/indra/newview/llworld.cpp index 118d7f8d08..d7e5b464a6 100644 --- a/indra/newview/llworld.cpp +++ b/indra/newview/llworld.cpp @@ -657,6 +657,9 @@ void LLWorld::updateParticles() void LLWorld::updateClouds(const F32 dt) { + static LLFastTimer::DeclareTimer ftm("World Clouds"); + LLFastTimer t(ftm); + if (gSavedSettings.getBOOL("FreezeTime") || !gSavedSettings.getBOOL("SkyUseClassicClouds")) { diff --git a/indra/newview/pipeline.cpp b/indra/newview/pipeline.cpp index 4f4fc83819..d5f87b73fe 100644 --- a/indra/newview/pipeline.cpp +++ b/indra/newview/pipeline.cpp @@ -116,7 +116,6 @@ const F32 BACKLIGHT_DAY_MAGNITUDE_AVATAR = 0.2f; const F32 BACKLIGHT_NIGHT_MAGNITUDE_AVATAR = 0.1f; const F32 BACKLIGHT_DAY_MAGNITUDE_OBJECT = 0.1f; const F32 BACKLIGHT_NIGHT_MAGNITUDE_OBJECT = 0.08f; -const S32 MAX_ACTIVE_OBJECT_QUIET_FRAMES = 40; const S32 MAX_OFFSCREEN_GEOMETRY_CHANGES_PER_FRAME = 10; const U32 REFLECTION_MAP_RES = 128; @@ -1411,38 +1410,26 @@ void LLPipeline::updateMove() assertInitialized(); - for (LLDrawable::drawable_set_t::iterator iter = mRetexturedList.begin(); - iter != mRetexturedList.end(); ++iter) { - LLDrawable* drawablep = *iter; - if (drawablep && !drawablep->isDead()) - { - drawablep->updateTexture(); - } - } - mRetexturedList.clear(); + static LLFastTimer::DeclareTimer ftm("Retexture"); + LLFastTimer t(ftm); - updateMovedList(mMovedList); - - for (LLDrawable::drawable_set_t::iterator iter = mActiveQ.begin(); - iter != mActiveQ.end(); ) - { - LLDrawable::drawable_set_t::iterator curiter = iter++; - LLDrawable* drawablep = *curiter; - if (drawablep && !drawablep->isDead()) + for (LLDrawable::drawable_set_t::iterator iter = mRetexturedList.begin(); + iter != mRetexturedList.end(); ++iter) { - if (drawablep->isRoot() && - drawablep->mQuietCount++ > MAX_ACTIVE_OBJECT_QUIET_FRAMES && - (!drawablep->getParent() || !drawablep->getParent()->isActive())) + LLDrawable* drawablep = *iter; + if (drawablep && !drawablep->isDead()) { - drawablep->makeStatic(); // removes drawable and its children from mActiveQ - iter = mActiveQ.upper_bound(drawablep); // next valid entry + drawablep->updateTexture(); } } - else - { - mActiveQ.erase(curiter); - } + mRetexturedList.clear(); + } + + { + static LLFastTimer::DeclareTimer ftm("Moved List"); + LLFastTimer t(ftm); + updateMovedList(mMovedList); } //balance octrees @@ -3058,12 +3045,6 @@ void LLPipeline::renderGeom(LLCamera& camera, BOOL forceVBOUpdate) } } - if (gPipeline.hasRenderDebugMask(LLPipeline::RENDER_DEBUG_PICKING)) - { - LLAppViewer::instance()->pingMainloopTimeout("Pipeline:RenderForSelect"); - gObjectList.renderObjectsForSelect(camera, gViewerWindow->getWindowRectScaled()); - } - else { LLFastTimer t(FTM_POOLS); @@ -4797,10 +4778,6 @@ void LLPipeline::findReferences(LLDrawable *drawablep) llinfos << "In mRetexturedList" << llendl; } - if (mActiveQ.find(drawablep) != mActiveQ.end()) - { - llinfos << "In mActiveQ" << llendl; - } if (std::find(mBuildQ1.begin(), mBuildQ1.end(), drawablep) != mBuildQ1.end()) { llinfos << "In mBuildQ1" << llendl; @@ -4957,19 +4934,6 @@ void LLPipeline::setLight(LLDrawable *drawablep, BOOL is_light) } } -void LLPipeline::setActive(LLDrawable *drawablep, BOOL active) -{ - assertInitialized(); - if (active) - { - mActiveQ.insert(drawablep); - } - else - { - mActiveQ.erase(drawablep); - } -} - //static void LLPipeline::toggleRenderType(U32 type) { diff --git a/indra/newview/pipeline.h b/indra/newview/pipeline.h index 67004a5f2d..c5285943e8 100644 --- a/indra/newview/pipeline.h +++ b/indra/newview/pipeline.h @@ -270,8 +270,7 @@ public: void shiftObjects(const LLVector3 &offset); void setLight(LLDrawable *drawablep, BOOL is_light); - void setActive(LLDrawable *drawablep, BOOL active); - + BOOL hasRenderBatches(const U32 type) const; LLCullResult::drawinfo_list_t::iterator beginRenderMap(U32 type); LLCullResult::drawinfo_list_t::iterator endRenderMap(U32 type); @@ -589,8 +588,6 @@ protected: LLViewerObject::vobj_list_t mCreateQ; - LLDrawable::drawable_set_t mActiveQ; - LLDrawable::drawable_set_t mRetexturedList; class HighlightItem -- cgit v1.3 From de88d6ced487fd55fa6f6bb860849979f031a363 Mon Sep 17 00:00:00 2001 From: Dave Parks Date: Wed, 17 Feb 2010 11:29:16 -0600 Subject: Switched program database on windows to edit and continue for non-release builds. Adding a debug watch to LLCurlRequest to avoid invalidating iterator on processing posts. Mesh bulk uploading rewrite work in progress. 404 icons for mesh assets. --- indra/cmake/00-Common.cmake | 6 +++--- indra/llmessage/llcurl.cpp | 9 +++++++++ indra/llmessage/llcurl.h | 3 ++- indra/newview/llvovolume.cpp | 15 ++++++++++++--- 4 files changed, 26 insertions(+), 7 deletions(-) (limited to 'indra/newview/llvovolume.cpp') diff --git a/indra/cmake/00-Common.cmake b/indra/cmake/00-Common.cmake index 49a1ec7f2c..2219311fc7 100644 --- a/indra/cmake/00-Common.cmake +++ b/indra/cmake/00-Common.cmake @@ -35,13 +35,13 @@ if (WINDOWS) # Don't build DLLs. set(BUILD_SHARED_LIBS OFF) - set(CMAKE_CXX_FLAGS_DEBUG "${CMAKE_CXX_FLAGS_DEBUG} /Od /Zi /MDd /MP" + set(CMAKE_CXX_FLAGS_DEBUG "${CMAKE_CXX_FLAGS_DEBUG} /Od /ZI /MDd /MP" CACHE STRING "C++ compiler debug options" FORCE) set(CMAKE_CXX_FLAGS_RELWITHDEBINFO - "${CMAKE_CXX_FLAGS_RELWITHDEBINFO} /Od /Zi /MD /MP" + "${CMAKE_CXX_FLAGS_RELWITHDEBINFO} /Od /ZI /MD /MP" CACHE STRING "C++ compiler release-with-debug options" FORCE) set(CMAKE_CXX_FLAGS_RELEASE - "${CMAKE_CXX_FLAGS_RELEASE} ${LL_CXX_FLAGS} /O2 /Zi /MD /MP" + "${CMAKE_CXX_FLAGS_RELEASE} ${LL_CXX_FLAGS} /O2 /MD /MP" CACHE STRING "C++ compiler release options" FORCE) set(CMAKE_CXX_STANDARD_LIBRARIES "") diff --git a/indra/llmessage/llcurl.cpp b/indra/llmessage/llcurl.cpp index b93b94cd25..5212ba1eca 100644 --- a/indra/llmessage/llcurl.cpp +++ b/indra/llmessage/llcurl.cpp @@ -711,6 +711,7 @@ LLCurlRequest::LLCurlRequest() : mActiveRequestCount(0) { mThreadID = LLThread::currentID(); + mProcessing = FALSE; } LLCurlRequest::~LLCurlRequest() @@ -745,6 +746,11 @@ LLCurl::Easy* LLCurlRequest::allocEasy() bool LLCurlRequest::addEasy(LLCurl::Easy* easy) { llassert_always(mActiveMulti); + + if (mProcessing) + { + llerrs << "Posting to a LLCurlRequest instance from within a responder is not allowed (causes DNS timeouts)." << llendl; + } bool res = mActiveMulti->addEasy(easy); return res; } @@ -835,6 +841,8 @@ S32 LLCurlRequest::process() { llassert_always(mThreadID == LLThread::currentID()); S32 res = 0; + + mProcessing = TRUE; for (curlmulti_set_t::iterator iter = mMultiSet.begin(); iter != mMultiSet.end(); ) { @@ -848,6 +856,7 @@ S32 LLCurlRequest::process() delete multi; } } + mProcessing = FALSE; return res; } diff --git a/indra/llmessage/llcurl.h b/indra/llmessage/llcurl.h index 4302c19113..5367643289 100644 --- a/indra/llmessage/llcurl.h +++ b/indra/llmessage/llcurl.h @@ -215,7 +215,7 @@ public: bool getByteRange(const std::string& url, const headers_t& headers, S32 offset, S32 length, LLCurl::ResponderPtr responder); bool post(const std::string& url, const headers_t& headers, const LLSD& data, LLCurl::ResponderPtr responder); bool post(const std::string& url, const headers_t& headers, const std::string& data, LLCurl::ResponderPtr responder); - + S32 process(); S32 getQueued(); @@ -229,6 +229,7 @@ private: curlmulti_set_t mMultiSet; LLCurl::Multi* mActiveMulti; S32 mActiveRequestCount; + BOOL mProcessing; U32 mThreadID; // debug }; diff --git a/indra/newview/llvovolume.cpp b/indra/newview/llvovolume.cpp index fb28b78daf..cb67890515 100644 --- a/indra/newview/llvovolume.cpp +++ b/indra/newview/llvovolume.cpp @@ -920,6 +920,8 @@ BOOL LLVOVolume::setVolume(const LLVolumeParams ¶ms, const S32 detail, bool S32 lod = mLOD; + BOOL is404 = FALSE; + if (isSculpted()) { // if it's a mesh @@ -932,6 +934,11 @@ BOOL LLVOVolume::setVolume(const LLVolumeParams ¶ms, const S32 detail, bool volume_params.setType(LL_PCODE_PROFILE_SQUARE, LL_PCODE_PATH_LINE); lod = gMeshRepo.getActualMeshLOD(volume_params, lod); + if (lod == -1) + { + is404 = TRUE; + lod = 0; + } } } @@ -962,7 +969,10 @@ BOOL LLVOVolume::setVolume(const LLVolumeParams ¶ms, const S32 detail, bool } } - + if (is404) + { + setIcon(LLViewerTextureManager::getFetchedTextureFromFile("icons/Inv_Mesh.png", TRUE, LLViewerTexture::BOOST_UI)); + } if ((LLPrimitive::setVolume(volume_params, lod, (mVolumeImpl && mVolumeImpl->isVolumeUnique()))) || mSculptChanged) { @@ -995,10 +1005,9 @@ BOOL LLVOVolume::setVolume(const LLVolumeParams ¶ms, const S32 detail, bool } else // otherwise is sculptie { - if (mSculptTexture.notNull()) { - sculpt(); + sculpt(); } } } -- cgit v1.3 From 56c0d25ec88df661a15bd1401efd4a9f67b86328 Mon Sep 17 00:00:00 2001 From: Dave Parks Date: Wed, 24 Feb 2010 23:36:55 -0600 Subject: Fix for attempting to load mesh as texture. Fix for failing to retry when mesh uploads fail. Fix for race condition deadlocking mesh fetching thread. --- indra/newview/llvovolume.cpp | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'indra/newview/llvovolume.cpp') diff --git a/indra/newview/llvovolume.cpp b/indra/newview/llvovolume.cpp index 562c733a3e..726673013b 100644 --- a/indra/newview/llvovolume.cpp +++ b/indra/newview/llvovolume.cpp @@ -1021,7 +1021,7 @@ void LLVOVolume::updateSculptTexture() { LLPointer old_sculpt = mSculptTexture; - if (isSculpted()) + if (isSculpted() && !isMesh()) { LLSculptParams *sculpt_params = (LLSculptParams *)getParameterEntry(LLNetworkData::PARAMS_SCULPT); LLUUID id = sculpt_params->getSculptTexture(); -- cgit v1.3 From ea6397fe4990b73e190391d61781c609fbd1f8c1 Mon Sep 17 00:00:00 2001 From: Dave Parks Date: Thu, 4 Mar 2010 15:30:15 -0600 Subject: Optimiziation pass. Added RenderUseStreamVBO to enable/disable usage of VBO's for streaming buffers. Faster traversal of LLCullResult members. Removal of llpushcallstacks from inner loops. Sprinkling in fast timers. --- indra/llrender/llglslshader.cpp | 6 ++ indra/llrender/llvertexbuffer.cpp | 6 ++ indra/llrender/llvertexbuffer.h | 2 + indra/newview/app_settings/settings.xml | 11 +++ indra/newview/lldrawable.cpp | 7 ++ indra/newview/lldrawpool.cpp | 2 +- indra/newview/lldrawpoolalpha.cpp | 132 ++++++++++++++++---------------- indra/newview/lldrawpoolbump.cpp | 4 +- indra/newview/llface.cpp | 4 +- indra/newview/llspatialpartition.cpp | 36 +++++++-- indra/newview/llspatialpartition.h | 7 ++ indra/newview/llviewercontrol.cpp | 1 + indra/newview/llvovolume.cpp | 6 +- indra/newview/pipeline.cpp | 21 ++++- 14 files changed, 160 insertions(+), 85 deletions(-) (limited to 'indra/newview/llvovolume.cpp') diff --git a/indra/llrender/llglslshader.cpp b/indra/llrender/llglslshader.cpp index ca92cb6580..2b3179116d 100644 --- a/indra/llrender/llglslshader.cpp +++ b/indra/llrender/llglslshader.cpp @@ -343,8 +343,11 @@ BOOL LLGLSLShader::link(BOOL suppress_errors) return LLShaderMgr::instance()->linkProgramObject(mProgramObject, suppress_errors); } +static LLFastTimer::DeclareTimer FTM_BIND_SHADER("Bind Shader"); + void LLGLSLShader::bind() { + LLFastTimer ftm(FTM_BIND_SHADER); if (gGLManager.mHasShaderObjects) { glUseProgramObjectARB(mProgramObject); @@ -357,8 +360,11 @@ void LLGLSLShader::bind() } } +static LLFastTimer::DeclareTimer FTM_UNBIND_SHADER("Unbind Shader"); + void LLGLSLShader::unbind() { + LLFastTimer ftm(FTM_UNBIND_SHADER); if (gGLManager.mHasShaderObjects) { stop_glerror(); diff --git a/indra/llrender/llvertexbuffer.cpp b/indra/llrender/llvertexbuffer.cpp index 415d2f603b..4064e688e8 100644 --- a/indra/llrender/llvertexbuffer.cpp +++ b/indra/llrender/llvertexbuffer.cpp @@ -61,6 +61,7 @@ BOOL LLVertexBuffer::sVBOActive = FALSE; BOOL LLVertexBuffer::sIBOActive = FALSE; U32 LLVertexBuffer::sAllocatedBytes = 0; BOOL LLVertexBuffer::sMapped = FALSE; +BOOL LLVertexBuffer::sUseStreamDraw = TRUE; std::vector LLVertexBuffer::sDeleteList; @@ -381,6 +382,11 @@ LLVertexBuffer::LLVertexBuffer(U32 typemask, S32 usage) : { mUsage = 0 ; } + + if (mUsage == GL_STREAM_DRAW_ARB && !sUseStreamDraw) + { + mUsage = 0; + } S32 stride = calcStride(typemask, mOffsets); diff --git a/indra/llrender/llvertexbuffer.h b/indra/llrender/llvertexbuffer.h index b785a22976..e2fecdffef 100644 --- a/indra/llrender/llvertexbuffer.h +++ b/indra/llrender/llvertexbuffer.h @@ -83,6 +83,8 @@ public: static LLVBOPool sDynamicVBOPool; static LLVBOPool sStreamIBOPool; static LLVBOPool sDynamicIBOPool; + + static BOOL sUseStreamDraw; static void initClass(bool use_vbo); static void cleanupClass(); diff --git a/indra/newview/app_settings/settings.xml b/indra/newview/app_settings/settings.xml index 0238fab5af..e234af699b 100644 --- a/indra/newview/app_settings/settings.xml +++ b/indra/newview/app_settings/settings.xml @@ -7531,6 +7531,17 @@ Value 1 + RenderUseStreamVBO + + Comment + Use VBO's for stream buffers + Persist + 1 + Type + Boolean + Value + 1 + RenderVolumeLODFactor Comment diff --git a/indra/newview/lldrawable.cpp b/indra/newview/lldrawable.cpp index 390e950d75..fd9cb18a42 100644 --- a/indra/newview/lldrawable.cpp +++ b/indra/newview/lldrawable.cpp @@ -683,8 +683,11 @@ BOOL LLDrawable::updateMoveDamped() return done_moving; } +static LLFastTimer::DeclareTimer FTM_UPDATE_DISTANCE("Update Distance"); + void LLDrawable::updateDistance(LLCamera& camera, bool force_update) { + LLFastTimer t(FTM_UPDATE_DISTANCE); if (LLViewerCamera::sCurCameraID != LLViewerCamera::CAMERA_WORLD) { llerrs << "WTF?" << llendl; @@ -1317,8 +1320,12 @@ void LLSpatialBridge::setVisible(LLCamera& camera_in, std::vector* } } +static LLFastTimer::DeclareTimer FTM_BRIDGE_DISTANCE_UPDATE("Bridge Distance"); + void LLSpatialBridge::updateDistance(LLCamera& camera_in, bool force_update) { + LLFastTimer t(FTM_BRIDGE_DISTANCE_UPDATE); + if (mDrawable == NULL) { markDead(); diff --git a/indra/newview/lldrawpool.cpp b/indra/newview/lldrawpool.cpp index ef946ac49e..ae30af3647 100644 --- a/indra/newview/lldrawpool.cpp +++ b/indra/newview/lldrawpool.cpp @@ -442,7 +442,6 @@ void LLRenderPass::renderTexture(U32 type, U32 mask) void LLRenderPass::pushBatches(U32 type, U32 mask, BOOL texture) { - llpushcallstacks ; for (LLCullResult::drawinfo_list_t::iterator i = gPipeline.beginRenderMap(type); i != gPipeline.endRenderMap(type); ++i) { LLDrawInfo* pparams = *i; @@ -475,6 +474,7 @@ void LLRenderPass::pushBatch(LLDrawInfo& params, U32 mask, BOOL texture) { if (params.mTexture.notNull()) { + params.mTexture->addTextureStats(params.mVSize); gGL.getTexUnit(0)->bind(params.mTexture, TRUE) ; if (params.mTextureMatrix) { diff --git a/indra/newview/lldrawpoolalpha.cpp b/indra/newview/lldrawpoolalpha.cpp index 875c9ac6a9..75973cfa54 100644 --- a/indra/newview/lldrawpoolalpha.cpp +++ b/indra/newview/lldrawpoolalpha.cpp @@ -294,96 +294,98 @@ void LLDrawPoolAlpha::renderAlpha(U32 mask) LLRenderPass::applyModelMatrix(params); - if (params.mFullbright) { - // Turn off lighting if it hasn't already been so. - if (light_enabled || !initialized_lighting) + if (params.mFullbright) + { + // Turn off lighting if it hasn't already been so. + if (light_enabled || !initialized_lighting) + { + initialized_lighting = TRUE; + if (use_shaders) + { + target_shader = fullbright_shader; + } + else + { + gPipeline.enableLightsFullbright(LLColor4(1,1,1,1)); + } + light_enabled = FALSE; + } + } + // Turn on lighting if it isn't already. + else if (!light_enabled || !initialized_lighting) { initialized_lighting = TRUE; if (use_shaders) { - target_shader = fullbright_shader; + target_shader = simple_shader; } else { - gPipeline.enableLightsFullbright(LLColor4(1,1,1,1)); + gPipeline.enableLightsDynamic(); } - light_enabled = FALSE; + light_enabled = TRUE; } - } - // Turn on lighting if it isn't already. - else if (!light_enabled || !initialized_lighting) - { - initialized_lighting = TRUE; - if (use_shaders) - { - target_shader = simple_shader; - } - else - { - gPipeline.enableLightsDynamic(); - } - light_enabled = TRUE; - } - // If we need shaders, and we're not ALREADY using the proper shader, then bind it - // (this way we won't rebind shaders unnecessarily). - if(use_shaders && (current_shader != target_shader)) - { - llassert(target_shader != NULL); - if (deferred_render && current_shader != NULL) - { - gPipeline.unbindDeferredShader(*current_shader); - diffuse_channel = 0; - } - current_shader = target_shader; - if (deferred_render) - { - gPipeline.bindDeferredShader(*current_shader); - diffuse_channel = current_shader->enableTexture(LLViewerShaderMgr::DIFFUSE_MAP); - } - else + // If we need shaders, and we're not ALREADY using the proper shader, then bind it + // (this way we won't rebind shaders unnecessarily). + if(use_shaders && (current_shader != target_shader)) { - current_shader->bind(); + llassert(target_shader != NULL); + if (deferred_render && current_shader != NULL) + { + gPipeline.unbindDeferredShader(*current_shader); + diffuse_channel = 0; + } + current_shader = target_shader; + if (deferred_render) + { + gPipeline.bindDeferredShader(*current_shader); + diffuse_channel = current_shader->enableTexture(LLViewerShaderMgr::DIFFUSE_MAP); + } + else + { + current_shader->bind(); + } } - } - else if (!use_shaders && current_shader != NULL) - { - if (deferred_render) + else if (!use_shaders && current_shader != NULL) { - gPipeline.unbindDeferredShader(*current_shader); - diffuse_channel = 0; + if (deferred_render) + { + gPipeline.unbindDeferredShader(*current_shader); + diffuse_channel = 0; + } + LLGLSLShader::bindNoShader(); + current_shader = NULL; } - LLGLSLShader::bindNoShader(); - current_shader = NULL; - } - - if (params.mGroup) - { - params.mGroup->rebuildMesh(); - } - - if (params.mTexture.notNull()) - { - gGL.getTexUnit(diffuse_channel)->bind(params.mTexture.get()); - if(params.mTexture.notNull()) + if (params.mGroup) { - params.mTexture->addTextureStats(params.mVSize); + params.mGroup->rebuildMesh(); } - if (params.mTextureMatrix) + + + if (params.mTexture.notNull()) { - gGL.getTexUnit(0)->activate(); - glMatrixMode(GL_TEXTURE); - glLoadMatrixf((GLfloat*) params.mTextureMatrix->mMatrix); - gPipeline.mTextureMatrixOps++; + gGL.getTexUnit(diffuse_channel)->bind(params.mTexture.get()); + if(params.mTexture.notNull()) + { + params.mTexture->addTextureStats(params.mVSize); + } + if (params.mTextureMatrix) + { + gGL.getTexUnit(0)->activate(); + glMatrixMode(GL_TEXTURE); + glLoadMatrixf((GLfloat*) params.mTextureMatrix->mMatrix); + gPipeline.mTextureMatrixOps++; + } } } params.mVertexBuffer->setBuffer(mask); params.mVertexBuffer->drawRange(params.mDrawMode, params.mStart, params.mEnd, params.mCount, params.mOffset); gPipeline.addTrianglesDrawn(params.mCount, params.mDrawMode); - + if (params.mTextureMatrix && params.mTexture.notNull()) { gGL.getTexUnit(0)->activate(); diff --git a/indra/newview/lldrawpoolbump.cpp b/indra/newview/lldrawpoolbump.cpp index a4a8dc80b5..8f3e775976 100644 --- a/indra/newview/lldrawpoolbump.cpp +++ b/indra/newview/lldrawpoolbump.cpp @@ -837,7 +837,6 @@ void LLBumpImageList::addTextureStats(U8 bump, const LLUUID& base_image_id, F32 void LLBumpImageList::updateImages() { - llpushcallstacks ; for (bump_image_map_t::iterator iter = mBrightnessEntries.begin(); iter != mBrightnessEntries.end(); ) { bump_image_map_t::iterator curiter = iter++; @@ -864,7 +863,7 @@ void LLBumpImageList::updateImages() } } } - llpushcallstacks ; + for (bump_image_map_t::iterator iter = mDarknessEntries.begin(); iter != mDarknessEntries.end(); ) { bump_image_map_t::iterator curiter = iter++; @@ -891,7 +890,6 @@ void LLBumpImageList::updateImages() } } } - llpushcallstacks ; } diff --git a/indra/newview/llface.cpp b/indra/newview/llface.cpp index 770811b587..53330e4d98 100644 --- a/indra/newview/llface.cpp +++ b/indra/newview/llface.cpp @@ -883,12 +883,14 @@ void LLFace::updateRebuildFlags() } } +static LLFastTimer::DeclareTimer FTM_FACE_GET_GEOM("Face Geom"); + BOOL LLFace::getGeometryVolume(const LLVolume& volume, const S32 &f, const LLMatrix4& mat_vert, const LLMatrix3& mat_normal, const U16 &index_offset) { - llpushcallstacks ; + LLFastTimer t(FTM_FACE_GET_GEOM); const LLVolumeFace &vf = volume.getVolumeFace(f); S32 num_vertices = (S32)vf.mVertices.size(); S32 num_indices = LLPipeline::sUseTriStrips ? (S32)vf.mTriStrip.size() : (S32) vf.mIndices.size(); diff --git a/indra/newview/llspatialpartition.cpp b/indra/newview/llspatialpartition.cpp index 19bf40b56a..3742f70df5 100644 --- a/indra/newview/llspatialpartition.cpp +++ b/indra/newview/llspatialpartition.cpp @@ -1513,6 +1513,7 @@ void LLSpatialGroup::checkOcclusion() { if (LLPipeline::sUseOcclusion > 1) { + LLFastTimer t(FTM_OCCLUSION_READBACK); LLSpatialGroup* parent = getParent(); if (parent && parent->isOcclusionState(LLSpatialGroup::OCCLUDED)) { //if the parent has been marked as occluded, the child is implicitly occluded @@ -1520,7 +1521,6 @@ void LLSpatialGroup::checkOcclusion() } else if (isOcclusionState(QUERY_PENDING)) { //otherwise, if a query is pending, read it back - LLFastTimer t(FTM_OCCLUSION_READBACK); GLuint res = 1; if (!isOcclusionState(DISCARD_QUERY) && mOcclusionQuery[LLViewerCamera::sCurCameraID]) { @@ -3422,11 +3422,23 @@ LLCullResult::LLCullResult() void LLCullResult::clear() { mVisibleGroupsSize = 0; + mVisibleGroupsEnd = mVisibleGroups.begin(); + mAlphaGroupsSize = 0; + mAlphaGroupsEnd = mAlphaGroups.begin(); + mOcclusionGroupsSize = 0; + mOcclusionGroupsEnd = mOcclusionGroups.begin(); + mDrawableGroupsSize = 0; + mDrawableGroupsEnd = mDrawableGroups.begin(); + mVisibleListSize = 0; + mVisibleListEnd = mVisibleList.begin(); + mVisibleBridgeSize = 0; + mVisibleBridgeEnd = mVisibleBridge.begin(); + for (U32 i = 0; i < LLRenderPass::NUM_RENDER_TYPES; i++) { @@ -3435,6 +3447,7 @@ void LLCullResult::clear() mRenderMap[i][j] = 0; } mRenderMapSize[i] = 0; + mRenderMapEnd[i] = mRenderMap[i].begin(); } } @@ -3445,7 +3458,7 @@ LLCullResult::sg_list_t::iterator LLCullResult::beginVisibleGroups() LLCullResult::sg_list_t::iterator LLCullResult::endVisibleGroups() { - return mVisibleGroups.begin() + mVisibleGroupsSize; + return mVisibleGroupsEnd; } LLCullResult::sg_list_t::iterator LLCullResult::beginAlphaGroups() @@ -3455,7 +3468,7 @@ LLCullResult::sg_list_t::iterator LLCullResult::beginAlphaGroups() LLCullResult::sg_list_t::iterator LLCullResult::endAlphaGroups() { - return mAlphaGroups.begin() + mAlphaGroupsSize; + return mAlphaGroupsEnd; } LLCullResult::sg_list_t::iterator LLCullResult::beginOcclusionGroups() @@ -3465,7 +3478,7 @@ LLCullResult::sg_list_t::iterator LLCullResult::beginOcclusionGroups() LLCullResult::sg_list_t::iterator LLCullResult::endOcclusionGroups() { - return mOcclusionGroups.begin() + mOcclusionGroupsSize; + return mOcclusionGroupsEnd; } LLCullResult::sg_list_t::iterator LLCullResult::beginDrawableGroups() @@ -3475,7 +3488,7 @@ LLCullResult::sg_list_t::iterator LLCullResult::beginDrawableGroups() LLCullResult::sg_list_t::iterator LLCullResult::endDrawableGroups() { - return mDrawableGroups.begin() + mDrawableGroupsSize; + return mDrawableGroupsEnd; } LLCullResult::drawable_list_t::iterator LLCullResult::beginVisibleList() @@ -3485,7 +3498,7 @@ LLCullResult::drawable_list_t::iterator LLCullResult::beginVisibleList() LLCullResult::drawable_list_t::iterator LLCullResult::endVisibleList() { - return mVisibleList.begin() + mVisibleListSize; + return mVisibleListEnd; } LLCullResult::bridge_list_t::iterator LLCullResult::beginVisibleBridge() @@ -3495,7 +3508,7 @@ LLCullResult::bridge_list_t::iterator LLCullResult::beginVisibleBridge() LLCullResult::bridge_list_t::iterator LLCullResult::endVisibleBridge() { - return mVisibleBridge.begin() + mVisibleBridgeSize; + return mVisibleBridgeEnd; } LLCullResult::drawinfo_list_t::iterator LLCullResult::beginRenderMap(U32 type) @@ -3505,7 +3518,7 @@ LLCullResult::drawinfo_list_t::iterator LLCullResult::beginRenderMap(U32 type) LLCullResult::drawinfo_list_t::iterator LLCullResult::endRenderMap(U32 type) { - return mRenderMap[type].begin() + mRenderMapSize[type]; + return mRenderMapEnd[type]; } void LLCullResult::pushVisibleGroup(LLSpatialGroup* group) @@ -3519,6 +3532,7 @@ void LLCullResult::pushVisibleGroup(LLSpatialGroup* group) mVisibleGroups.push_back(group); } ++mVisibleGroupsSize; + mVisibleGroupsEnd = mVisibleGroups.begin()+mVisibleGroupsSize; } void LLCullResult::pushAlphaGroup(LLSpatialGroup* group) @@ -3532,6 +3546,7 @@ void LLCullResult::pushAlphaGroup(LLSpatialGroup* group) mAlphaGroups.push_back(group); } ++mAlphaGroupsSize; + mAlphaGroupsEnd = mAlphaGroups.begin()+mAlphaGroupsSize; } void LLCullResult::pushOcclusionGroup(LLSpatialGroup* group) @@ -3545,6 +3560,7 @@ void LLCullResult::pushOcclusionGroup(LLSpatialGroup* group) mOcclusionGroups.push_back(group); } ++mOcclusionGroupsSize; + mOcclusionGroupsEnd = mOcclusionGroups.begin()+mOcclusionGroupsSize; } void LLCullResult::pushDrawableGroup(LLSpatialGroup* group) @@ -3558,6 +3574,7 @@ void LLCullResult::pushDrawableGroup(LLSpatialGroup* group) mDrawableGroups.push_back(group); } ++mDrawableGroupsSize; + mDrawableGroupsEnd = mDrawableGroups.begin()+mDrawableGroupsSize; } void LLCullResult::pushDrawable(LLDrawable* drawable) @@ -3571,6 +3588,7 @@ void LLCullResult::pushDrawable(LLDrawable* drawable) mVisibleList.push_back(drawable); } ++mVisibleListSize; + mVisibleListEnd = mVisibleList.begin()+mVisibleListSize; } void LLCullResult::pushBridge(LLSpatialBridge* bridge) @@ -3584,6 +3602,7 @@ void LLCullResult::pushBridge(LLSpatialBridge* bridge) mVisibleBridge.push_back(bridge); } ++mVisibleBridgeSize; + mVisibleBridgeEnd = mVisibleBridge.begin()+mVisibleBridgeSize; } void LLCullResult::pushDrawInfo(U32 type, LLDrawInfo* draw_info) @@ -3597,6 +3616,7 @@ void LLCullResult::pushDrawInfo(U32 type, LLDrawInfo* draw_info) mRenderMap[type].push_back(draw_info); } ++mRenderMapSize[type]; + mRenderMapEnd[type] = mRenderMap[type].begin() + mRenderMapSize[type]; } diff --git a/indra/newview/llspatialpartition.h b/indra/newview/llspatialpartition.h index c5719bebcc..ef6b915e5c 100644 --- a/indra/newview/llspatialpartition.h +++ b/indra/newview/llspatialpartition.h @@ -535,12 +535,19 @@ private: U32 mRenderMapSize[LLRenderPass::NUM_RENDER_TYPES]; sg_list_t mVisibleGroups; + sg_list_t::iterator mVisibleGroupsEnd; sg_list_t mAlphaGroups; + sg_list_t::iterator mAlphaGroupsEnd; sg_list_t mOcclusionGroups; + sg_list_t::iterator mOcclusionGroupsEnd; sg_list_t mDrawableGroups; + sg_list_t::iterator mDrawableGroupsEnd; drawable_list_t mVisibleList; + drawable_list_t::iterator mVisibleListEnd; bridge_list_t mVisibleBridge; + bridge_list_t::iterator mVisibleBridgeEnd; drawinfo_list_t mRenderMap[LLRenderPass::NUM_RENDER_TYPES]; + drawinfo_list_t::iterator mRenderMapEnd[LLRenderPass::NUM_RENDER_TYPES]; }; diff --git a/indra/newview/llviewercontrol.cpp b/indra/newview/llviewercontrol.cpp index 0b7f6900ed..1d99be9960 100644 --- a/indra/newview/llviewercontrol.cpp +++ b/indra/newview/llviewercontrol.cpp @@ -571,6 +571,7 @@ void settings_setup_listeners() gSavedSettings.getControl("MuteAmbient")->getSignal()->connect(boost::bind(&handleAudioVolumeChanged, _2)); gSavedSettings.getControl("MuteUI")->getSignal()->connect(boost::bind(&handleAudioVolumeChanged, _2)); gSavedSettings.getControl("RenderVBOEnable")->getSignal()->connect(boost::bind(&handleRenderUseVBOChanged, _2)); + gSavedSettings.getControl("RenderUseStreamVBO")->getSignal()->connect(boost::bind(&handleResetVertexBuffersChanged, _2)); gSavedSettings.getControl("WLSkyDetail")->getSignal()->connect(boost::bind(&handleWLSkyDetailChanged, _2)); gSavedSettings.getControl("RenderLightingDetail")->getSignal()->connect(boost::bind(&handleRenderLightingDetailChanged, _2)); gSavedSettings.getControl("NumpadControl")->getSignal()->connect(boost::bind(&handleNumpadControlChanged, _2)); diff --git a/indra/newview/llvovolume.cpp b/indra/newview/llvovolume.cpp index eb2e6e3d76..39db6196a3 100644 --- a/indra/newview/llvovolume.cpp +++ b/indra/newview/llvovolume.cpp @@ -3422,7 +3422,6 @@ static LLFastTimer::DeclareTimer FTM_REBUILD_VBO("VBO Rebuilt"); void LLVolumeGeometryManager::rebuildGeom(LLSpatialGroup* group) { - llpushcallstacks ; if (group->changeLOD()) { group->mLastUpdateDistance = group->mDistance; @@ -3651,9 +3650,10 @@ void LLVolumeGeometryManager::rebuildGeom(LLSpatialGroup* group) } static LLFastTimer::DeclareTimer FTM_VOLUME_GEOM("Volume Geometry"); +static LLFastTimer::DeclareTimer FTM_VOLUME_GEOM_PARTIAL("Terse Rebuild"); + void LLVolumeGeometryManager::rebuildMesh(LLSpatialGroup* group) { - llpushcallstacks ; llassert(group); if (group && group->isState(LLSpatialGroup::MESH_DIRTY) && !group->isState(LLSpatialGroup::GEOM_DIRTY)) { @@ -3664,6 +3664,7 @@ void LLVolumeGeometryManager::rebuildMesh(LLSpatialGroup* group) for (LLSpatialGroup::element_iter drawable_iter = group->getData().begin(); drawable_iter != group->getData().end(); ++drawable_iter) { + LLFastTimer t(FTM_VOLUME_GEOM_PARTIAL); LLDrawable* drawablep = *drawable_iter; if (drawablep->isDead() || drawablep->isState(LLDrawable::FORCE_INVISIBLE) ) @@ -3750,7 +3751,6 @@ void LLVolumeGeometryManager::rebuildMesh(LLSpatialGroup* group) void LLVolumeGeometryManager::genDrawInfo(LLSpatialGroup* group, U32 mask, std::vector& faces, BOOL distance_sort) { - llpushcallstacks ; //calculate maximum number of vertices to store in a single buffer U32 max_vertices = (gSavedSettings.getS32("RenderMaxVBOSize")*1024)/LLVertexBuffer::calcStride(group->mSpatialPartition->mVertexDataMask); max_vertices = llmin(max_vertices, (U32) 65535); diff --git a/indra/newview/pipeline.cpp b/indra/newview/pipeline.cpp index 43b2104135..eb3ad95586 100644 --- a/indra/newview/pipeline.cpp +++ b/indra/newview/pipeline.cpp @@ -362,6 +362,7 @@ void LLPipeline::init() sDynamicLOD = gSavedSettings.getBOOL("RenderDynamicLOD"); sRenderBump = gSavedSettings.getBOOL("RenderObjectBump"); sUseTriStrips = gSavedSettings.getBOOL("RenderUseTriStrips"); + LLVertexBuffer::sUseStreamDraw = gSavedSettings.getBOOL("RenderUseStreamVBO"); sRenderAttachedLights = gSavedSettings.getBOOL("RenderAttachedLights"); sRenderAttachedParticles = gSavedSettings.getBOOL("RenderAttachedParticles"); @@ -1739,8 +1740,12 @@ void LLPipeline::markOccluder(LLSpatialGroup* group) } } +static LLFastTimer::DeclareTimer FTM_DO_OCCLUSION("Do Occlusion"); + void LLPipeline::doOcclusion(LLCamera& camera) { + LLFastTimer t(FTM_DO_OCCLUSION); + LLVertexBuffer::unbind(); if (hasRenderDebugMask(LLPipeline::RENDER_DEBUG_OCCLUSION)) @@ -1816,7 +1821,6 @@ void LLPipeline::rebuildPriorityGroups() void LLPipeline::rebuildGroups() { - llpushcallstacks ; // Iterate through some drawables on the non-priority build queue S32 size = (S32) mGroupQ2.size(); S32 min_count = llclamp((S32) ((F32) (size * size)/4096*0.25f), 1, size); @@ -1961,9 +1965,13 @@ void LLPipeline::updateGeom(F32 max_dtime) updateMovedList(mMovedBridge); } +static LLFastTimer::DeclareTimer FTM_MARK_VISIBLE("Mark Visible"); + void LLPipeline::markVisible(LLDrawable *drawablep, LLCamera& camera) { LLMemType mt(LLMemType::MTYPE_PIPELINE_MARK_VISIBLE); + LLFastTimer t(FTM_MARK_VISIBLE); + if(!drawablep || drawablep->isDead()) { return; @@ -4704,8 +4712,12 @@ void LLPipeline::setupHWLights(LLDrawPool* pool) mLightMask = 0; } +static LLFastTimer::DeclareTimer FTM_ENABLE_LIGHTS("Enable Lights"); + void LLPipeline::enableLights(U32 mask) { + LLFastTimer ftm(FTM_ENABLE_LIGHTS); + assertInitialized(); if (mLightingDetail == 0) @@ -4813,16 +4825,16 @@ void LLPipeline::enableLightsFullbright(const LLColor4& color) enableLights(mask); glLightModelfv(GL_LIGHT_MODEL_AMBIENT,color.mV); - if (mLightingDetail >= 2) + /*if (mLightingDetail >= 2) { glColor4f(0.f, 0.f, 0.f, 1.f); // no local lighting by default - } + }*/ } void LLPipeline::disableLights() { enableLights(0); // no lighting (full bright) - glColor4f(1.f, 1.f, 1.f, 1.f); // lighting color = white by default + //glColor4f(1.f, 1.f, 1.f, 1.f); // lighting color = white by default } //============================================================================ @@ -5425,6 +5437,7 @@ void LLPipeline::resetVertexBuffers() { sRenderBump = gSavedSettings.getBOOL("RenderObjectBump"); sUseTriStrips = gSavedSettings.getBOOL("RenderUseTriStrips"); + LLVertexBuffer::sUseStreamDraw = gSavedSettings.getBOOL("RenderUseStreamVBO"); for (LLWorld::region_list_t::const_iterator iter = LLWorld::getInstance()->getRegionList().begin(); iter != LLWorld::getInstance()->getRegionList().end(); ++iter) -- cgit v1.3 From 1889d02569a58d08b6c452e609819a1961bf64e8 Mon Sep 17 00:00:00 2001 From: Dave Parks Date: Sat, 6 Mar 2010 00:42:39 -0600 Subject: Better bounding boxes for meshes. --- indra/newview/llvovolume.cpp | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'indra/newview/llvovolume.cpp') diff --git a/indra/newview/llvovolume.cpp b/indra/newview/llvovolume.cpp index 39db6196a3..1d3d943b06 100644 --- a/indra/newview/llvovolume.cpp +++ b/indra/newview/llvovolume.cpp @@ -1329,7 +1329,7 @@ BOOL LLVOVolume::genBBoxes(BOOL force_global) BOOL rebuild = mDrawable->isState(LLDrawable::REBUILD_VOLUME | LLDrawable::REBUILD_POSITION); - for (S32 i = 0; i < getVolume()->getNumFaces(); i++) + for (S32 i = 0; i < getVolume()->getNumVolumeFaces(); i++) { LLFace *face = mDrawable->getFace(i); if (!face) -- cgit v1.3 From 9beeb33f091083d10be3169930b9c205efc2821b Mon Sep 17 00:00:00 2001 From: Tofu Linden Date: Mon, 8 Mar 2010 12:28:33 +0000 Subject: fixes VWR-4214 Glow effect passes through alpha textures where alpha is actually "solid" and its many many dupes. In summary this does the following: * When laying down alpha, uses the new separated color-and-alpha blend func support to suppress existing glow in proportion to the opacity of the incoming fragments * Moves glowing-alpha support from the glow pool to the alpha pool, ensures that the glow pass for a vertex buffer happens after the alpha pass for the vertex buffer so that it doesn't suppress its *own* glow. Tested with all shader modes, FBO on/off, and deferred rendering. Tested with every permutation of alpha/non-alpha/glow/non-glow/fullbright/waterglow/particle interaction I could contrive so far. :) --- indra/newview/lldrawpoolalpha.cpp | 37 ++++++++++++++++++++++++++++++++++++- indra/newview/llvovolume.cpp | 16 +++++----------- 2 files changed, 41 insertions(+), 12 deletions(-) (limited to 'indra/newview/llvovolume.cpp') diff --git a/indra/newview/lldrawpoolalpha.cpp b/indra/newview/lldrawpoolalpha.cpp index 75973cfa54..9a778a063d 100644 --- a/indra/newview/lldrawpoolalpha.cpp +++ b/indra/newview/lldrawpoolalpha.cpp @@ -204,7 +204,12 @@ void LLDrawPoolAlpha::render(S32 pass) } LLGLDepthTest depth(GL_TRUE, LLDrawPoolWater::sSkipScreenCopy ? GL_TRUE : GL_FALSE); + + gGL.setColorMask(true, true); + gGL.blendFunc(LLRender::BF_SOURCE_ALPHA, LLRender::BF_ONE_MINUS_SOURCE_ALPHA, + LLRender::BF_ZERO, LLRender::BF_ONE_MINUS_SOURCE_ALPHA); renderAlpha(getVertexDataMask()); + gGL.setColorMask(true, false); if (deferred_render && current_shader != NULL) { @@ -283,9 +288,18 @@ void LLDrawPoolAlpha::renderAlpha(U32 mask) for (LLCullResult::sg_list_t::iterator i = gPipeline.beginAlphaGroups(); i != gPipeline.endAlphaGroups(); ++i) { LLSpatialGroup* group = *i; + llassert(group); + llassert(group->mSpatialPartition); + if (group->mSpatialPartition->mRenderByGroup && - !group->isDead()) + !group->isDead()) { + bool draw_glow_for_this_partition = mVertexShaderLevel > 0 && // no shaders = no glow. + // All particle systems seem to come off the wire with texture entries which claim that they glow. This is probably a bug in the data. Suppress. + group->mSpatialPartition->mPartitionType != LLViewerRegion::PARTITION_PARTICLE && + group->mSpatialPartition->mPartitionType != LLViewerRegion::PARTITION_CLOUD && + group->mSpatialPartition->mPartitionType != LLViewerRegion::PARTITION_HUD_PARTICLE; + LLSpatialGroup::drawmap_elem_t& draw_info = group->mDrawMap[LLRenderPass::PASS_ALPHA]; for (LLSpatialGroup::drawmap_elem_t::iterator k = draw_info.begin(); k != draw_info.end(); ++k) @@ -385,6 +399,27 @@ void LLDrawPoolAlpha::renderAlpha(U32 mask) params.mVertexBuffer->setBuffer(mask); params.mVertexBuffer->drawRange(params.mDrawMode, params.mStart, params.mEnd, params.mCount, params.mOffset); gPipeline.addTrianglesDrawn(params.mCount, params.mDrawMode); + + // If this alpha mesh has glow, then draw it a second time to add the destination-alpha (=glow). Interleaving these state-changing calls could be expensive, but glow must be drawn Z-sorted with alpha. + if (draw_glow_for_this_partition && + params.mGlowColor.mV[3] > 0) + { + // install glow-accumulating blend mode + gGL.blendFunc(LLRender::BF_ZERO, LLRender::BF_ONE, // don't touch color + LLRender::BF_ONE, LLRender::BF_ONE); // add to alpha (glow) + + // glow doesn't use vertex colors from the mesh data + params.mVertexBuffer->setBuffer(mask & ~LLVertexBuffer::MAP_COLOR); + glColor4ubv(params.mGlowColor.mV); + + // do the actual drawing, again + params.mVertexBuffer->drawRange(params.mDrawMode, params.mStart, params.mEnd, params.mCount, params.mOffset); + gPipeline.addTrianglesDrawn(params.mCount, params.mDrawMode); + + // restore our alpha blend mode + gGL.blendFunc(LLRender::BF_SOURCE_ALPHA, LLRender::BF_ONE_MINUS_SOURCE_ALPHA, + LLRender::BF_ZERO, LLRender::BF_ONE_MINUS_SOURCE_ALPHA); + } if (params.mTextureMatrix && params.mTexture.notNull()) { diff --git a/indra/newview/llvovolume.cpp b/indra/newview/llvovolume.cpp index 1d3d943b06..91aa994eab 100644 --- a/indra/newview/llvovolume.cpp +++ b/indra/newview/llvovolume.cpp @@ -3317,11 +3317,10 @@ void LLVolumeGeometryManager::registerFace(LLSpatialGroup* group, LLFace* facep, S32 idx = draw_vec.size()-1; - BOOL fullbright = (type == LLRenderPass::PASS_FULLBRIGHT) || - (type == LLRenderPass::PASS_INVISIBLE) || - (type == LLRenderPass::PASS_ALPHA ? facep->isState(LLFace::FULLBRIGHT) : FALSE); - + (type == LLRenderPass::PASS_INVISIBLE) || + (type == LLRenderPass::PASS_ALPHA ? facep->isState(LLFace::FULLBRIGHT) : FALSE); + if (!fullbright && type != LLRenderPass::PASS_GLOW && !facep->mVertexBuffer->hasDataType(LLVertexBuffer::TYPE_NORMAL)) { llwarns << "Non fullbright face has no normals!" << llendl; @@ -3350,12 +3349,7 @@ void LLVolumeGeometryManager::registerFace(LLSpatialGroup* group, LLFace* facep, LLViewerTexture* tex = facep->getTexture(); - U8 glow = 0; - - if (type == LLRenderPass::PASS_GLOW) - { - glow = (U8) (facep->getTextureEntry()->getGlow() * 255); - } + U8 glow = (U8) (facep->getTextureEntry()->getGlow() * 255); if (facep->mVertexBuffer.isNull()) { @@ -4001,7 +3995,7 @@ void LLVolumeGeometryManager::genDrawInfo(LLSpatialGroup* group, U32 mask, std:: } } - if (LLPipeline::sRenderGlow && te->getGlow() > 0.f) + if (!is_alpha && LLPipeline::sRenderGlow && te->getGlow() > 0.f) { registerFace(group, facep, LLRenderPass::PASS_GLOW); } -- cgit v1.3 From 4c022455685598bfd3ab90c4611e053fad72345d Mon Sep 17 00:00:00 2001 From: Tofu Linden Date: Tue, 9 Mar 2010 13:28:23 +0000 Subject: Disable alpha-masking for glowing surfaces. It seems beyond my ability to get right. Refactor some replicated and/or strangely-expressed logic. --- indra/newview/llspatialpartition.h | 5 +++-- indra/newview/llvovolume.cpp | 24 ++++++++++++++---------- 2 files changed, 17 insertions(+), 12 deletions(-) (limited to 'indra/newview/llvovolume.cpp') diff --git a/indra/newview/llspatialpartition.h b/indra/newview/llspatialpartition.h index 88fab01ea4..d74216de2d 100644 --- a/indra/newview/llspatialpartition.h +++ b/indra/newview/llspatialpartition.h @@ -614,14 +614,15 @@ public: //class for wrangling geometry out of volumes (implemented in LLVOVolume.cpp) class LLVolumeGeometryManager: public LLGeometryManager { -public: + public: virtual ~LLVolumeGeometryManager() { } virtual void rebuildGeom(LLSpatialGroup* group); virtual void rebuildMesh(LLSpatialGroup* group); virtual void getGeometry(LLSpatialGroup* group); void genDrawInfo(LLSpatialGroup* group, U32 mask, std::vector& faces, BOOL distance_sort = FALSE); void registerFace(LLSpatialGroup* group, LLFace* facep, U32 type); - + private: + bool canRenderAsMask(LLFace* facep); // logic helper }; //spatial partition that uses volume geometry manager (implemented in LLVOVolume.cpp) diff --git a/indra/newview/llvovolume.cpp b/indra/newview/llvovolume.cpp index 91aa994eab..90a06a966d 100644 --- a/indra/newview/llvovolume.cpp +++ b/indra/newview/llvovolume.cpp @@ -3319,7 +3319,7 @@ void LLVolumeGeometryManager::registerFace(LLSpatialGroup* group, LLFace* facep, BOOL fullbright = (type == LLRenderPass::PASS_FULLBRIGHT) || (type == LLRenderPass::PASS_INVISIBLE) || - (type == LLRenderPass::PASS_ALPHA ? facep->isState(LLFace::FULLBRIGHT) : FALSE); + (type == LLRenderPass::PASS_ALPHA && facep->isState(LLFace::FULLBRIGHT)); if (!fullbright && type != LLRenderPass::PASS_GLOW && !facep->mVertexBuffer->hasDataType(LLVertexBuffer::TYPE_NORMAL)) { @@ -3414,6 +3414,16 @@ void LLVolumeGeometryManager::getGeometry(LLSpatialGroup* group) static LLFastTimer::DeclareTimer FTM_REBUILD_VOLUME_VB("Volume"); static LLFastTimer::DeclareTimer FTM_REBUILD_VBO("VBO Rebuilt"); +bool LLVolumeGeometryManager::canRenderAsMask(LLFace* facep) +{ + const LLTextureEntry* te = facep->getTextureEntry(); + return (LLPipeline::sFastAlpha && + (te->getColor().mV[3] == 1.0f) && + (!te->getFullbright()) && // hack: alpha masking renders fullbright faces invisible, need to figure out why - for now, avoid + (te->getGlow() == 0.f) && // glowing masks are hard to implement - don't mask + facep->getTexture()->getIsAlphaMask()); +} + void LLVolumeGeometryManager::rebuildGeom(LLSpatialGroup* group) { if (group->changeLOD()) @@ -3539,10 +3549,7 @@ void LLVolumeGeometryManager::rebuildGeom(LLSpatialGroup* group) if (type == LLDrawPool::POOL_ALPHA) { - if (LLPipeline::sFastAlpha && - (te->getColor().mV[VW] == 1.0f) && - (!te->getFullbright()) && // hack: alpha masking renders fullbright faces invisible, need to figure out why - for now, avoid - facep->getTexture()->getIsAlphaMask()) + if (canRenderAsMask(facep)) { //can be treated as alpha mask simple_faces.push_back(facep); } @@ -3891,15 +3898,12 @@ void LLVolumeGeometryManager::genDrawInfo(LLSpatialGroup* group, U32 mask, std:: const LLTextureEntry* te = facep->getTextureEntry(); - BOOL is_alpha = facep->getPoolType() == LLDrawPool::POOL_ALPHA ? TRUE : FALSE; + BOOL is_alpha = (facep->getPoolType() == LLDrawPool::POOL_ALPHA) ? TRUE : FALSE; if (is_alpha) { // can we safely treat this as an alpha mask? - if (LLPipeline::sFastAlpha && - (te->getColor().mV[VW] == 1.0f) && - (!te->getFullbright()) && // hack: alpha masking renders fullbright faces invisible, need to figure out why - for now, avoid - facep->getTexture()->getIsAlphaMask()) + if (canRenderAsMask(facep)) { if (te->getFullbright()) { -- cgit v1.3 From 8da956d60926cb9c62c15e6766ea542049690777 Mon Sep 17 00:00:00 2001 From: Tofu Linden Date: Tue, 9 Mar 2010 13:45:53 +0000 Subject: We can enable alpha masking for fullbright faces as long as we're not in deferred mode. Deferred masked fullbrights are still broken (we fall back to blending instead of masking as a workaround), but they always have been. :/ --- indra/newview/llvovolume.cpp | 12 ++++++++---- 1 file changed, 8 insertions(+), 4 deletions(-) (limited to 'indra/newview/llvovolume.cpp') diff --git a/indra/newview/llvovolume.cpp b/indra/newview/llvovolume.cpp index 90a06a966d..1725879b7f 100644 --- a/indra/newview/llvovolume.cpp +++ b/indra/newview/llvovolume.cpp @@ -3417,11 +3417,15 @@ static LLFastTimer::DeclareTimer FTM_REBUILD_VBO("VBO Rebuilt"); bool LLVolumeGeometryManager::canRenderAsMask(LLFace* facep) { const LLTextureEntry* te = facep->getTextureEntry(); - return (LLPipeline::sFastAlpha && - (te->getColor().mV[3] == 1.0f) && - (!te->getFullbright()) && // hack: alpha masking renders fullbright faces invisible, need to figure out why - for now, avoid + return ( + LLPipeline::sFastAlpha && // do we want masks at all? + + (te->getColor().mV[3] == 1.0f) && // can't treat as mask if we have face alpha + !(LLPipeline::sRenderDeferred && te->getFullbright()) && // hack: alpha masking renders fullbright faces invisible in deferred rendering mode, need to figure out why - for now, avoid (te->getGlow() == 0.f) && // glowing masks are hard to implement - don't mask - facep->getTexture()->getIsAlphaMask()); + + facep->getTexture()->getIsAlphaMask() // texture actually qualifies for masking (lazily calculated but expensive) + ); } void LLVolumeGeometryManager::rebuildGeom(LLSpatialGroup* group) -- cgit v1.3 From 47ffcdb93d6e2ac1f9d497e43e0213c98d129254 Mon Sep 17 00:00:00 2001 From: Dave Parks Date: Tue, 6 Apr 2010 16:24:08 -0500 Subject: Rigged attachments (almost works). --- indra/llmath/llvolume.cpp | 22 +++++--- indra/llrender/llglslshader.cpp | 14 ++++- indra/llrender/llglslshader.h | 3 +- indra/llrender/llshadermgr.cpp | 8 +++ indra/newview/lldrawpoolavatar.cpp | 26 ++++++++- indra/newview/lldrawpoolavatar.h | 2 + indra/newview/llface.cpp | 18 +++++- indra/newview/llface.h | 4 +- indra/newview/llspatialpartition.h | 7 +++ indra/newview/llviewershadermgr.cpp | 25 +++++++++ indra/newview/llviewershadermgr.h | 3 + indra/newview/llvoavatar.cpp | 109 ++++++++++++++++++++++++++++++++++++ indra/newview/llvoavatar.h | 1 + indra/newview/llvovolume.cpp | 20 +++++++ 14 files changed, 247 insertions(+), 15 deletions(-) (limited to 'indra/newview/llvovolume.cpp') diff --git a/indra/llmath/llvolume.cpp b/indra/llmath/llvolume.cpp index c563af592f..fdd48b9e9e 100644 --- a/indra/llmath/llvolume.cpp +++ b/indra/llmath/llvolume.cpp @@ -2013,22 +2013,23 @@ bool LLVolume::unpackVolumeFaces(std::istream& is, S32 size) if (mdl[i].has("Weights")) { face.mWeights.resize(num_verts); + LLSD::Binary weights = mdl[i]["Weights"]; - LLSD::Binary::iterator iter = weights.begin(); + U32 idx = 0; U32 cur_vertex = 0; - while (iter != weights.end()) + while (idx < weights.size() && cur_vertex < num_verts) { - const S32 END_INFLUENCES = 0xFF; - U8 joint = *(iter++); + const U8 END_INFLUENCES = 0xFF; + U8 joint = weights[idx++]; U32 cur_influence = 0; while (joint != END_INFLUENCES) { - U16 influence = *(iter++); + U16 influence = weights[idx++]; influence = influence << 8; - influence |= *(iter++); + influence |= weights[idx++]; F32 w = llmin((F32) influence / 65535.f, 0.99999f); face.mWeights[cur_vertex].mV[cur_influence++] = (F32) joint + w; @@ -2039,13 +2040,18 @@ bool LLVolume::unpackVolumeFaces(std::istream& is, S32 size) } else { - joint = *(iter++); + joint = weights[idx++]; } } cur_vertex++; - iter++; } + + if (cur_vertex != num_verts || idx != weights.size()) + { + llwarns << "Vertex weight count does not match vertex count!" << llendl; + } + } LLVector3 min_pos; diff --git a/indra/llrender/llglslshader.cpp b/indra/llrender/llglslshader.cpp index ca92cb6580..949057df04 100644 --- a/indra/llrender/llglslshader.cpp +++ b/indra/llrender/llglslshader.cpp @@ -61,7 +61,7 @@ BOOL shouldChange(const LLVector4& v1, const LLVector4& v2) LLShaderFeatures::LLShaderFeatures() : calculatesLighting(false), isShiny(false), isFullbright(false), hasWaterFog(false), -hasTransport(false), hasSkinning(false), hasAtmospherics(false), isSpecular(false), +hasTransport(false), hasSkinning(false), hasObjectSkinning(false), hasAtmospherics(false), isSpecular(false), hasGamma(false), hasLighting(false), calculatesAtmospherics(false) { } @@ -717,6 +717,18 @@ GLint LLGLSLShader::getUniformLocation(const string& uniform) return -1; } +GLint LLGLSLShader::getAttribLocation(U32 attrib) +{ + if (attrib < mAttribute.size()) + { + return mAttribute[attrib]; + } + else + { + return -1; + } +} + void LLGLSLShader::uniform1i(const string& uniform, GLint v) { GLint location = getUniformLocation(uniform); diff --git a/indra/llrender/llglslshader.h b/indra/llrender/llglslshader.h index 166d4af04c..dc493ba162 100644 --- a/indra/llrender/llglslshader.h +++ b/indra/llrender/llglslshader.h @@ -48,6 +48,7 @@ public: bool hasWaterFog; // implies no gamma bool hasTransport; // implies no lighting (it's possible to have neither though) bool hasSkinning; + bool hasObjectSkinning; bool hasAtmospherics; bool hasGamma; @@ -109,7 +110,7 @@ public: void vertexAttrib4fv(U32 index, GLfloat* v); GLint getUniformLocation(const std::string& uniform); - + GLint getAttribLocation(U32 attrib); GLint mapUniformTextureChannel(GLint location, GLenum type); diff --git a/indra/llrender/llshadermgr.cpp b/indra/llrender/llshadermgr.cpp index 1286e91e49..23b76351eb 100644 --- a/indra/llrender/llshadermgr.cpp +++ b/indra/llrender/llshadermgr.cpp @@ -152,6 +152,14 @@ BOOL LLShaderMgr::attachShaderFeatures(LLGLSLShader * shader) return FALSE; } } + + if (features->hasObjectSkinning) + { + if (!shader->attachObject("avatar/objectSkinV.glsl")) + { + return FALSE; + } + } /////////////////////////////////////// // Attach Fragment Shader Features Next diff --git a/indra/newview/lldrawpoolavatar.cpp b/indra/newview/lldrawpoolavatar.cpp index 012e41383f..9311a5f60e 100644 --- a/indra/newview/lldrawpoolavatar.cpp +++ b/indra/newview/lldrawpoolavatar.cpp @@ -320,7 +320,7 @@ void LLDrawPoolAvatar::renderShadow(S32 pass) S32 LLDrawPoolAvatar::getNumPasses() { - return LLPipeline::sImpostorRender ? 1 : 3; + return LLPipeline::sImpostorRender ? 1 : 4; } void LLDrawPoolAvatar::render(S32 pass) @@ -357,6 +357,8 @@ void LLDrawPoolAvatar::beginRenderPass(S32 pass) break; case 2: beginSkinned(); + case 3: + beginRigged(); break; } } @@ -381,6 +383,10 @@ void LLDrawPoolAvatar::endRenderPass(S32 pass) break; case 2: endSkinned(); + break; + case 3: + endRigged(); + break; } } @@ -566,6 +572,18 @@ void LLDrawPoolAvatar::endSkinned() gGL.getTexUnit(0)->activate(); } +void LLDrawPoolAvatar::beginRigged() +{ + gSkinnedObjectSimpleProgram.bind(); + LLVertexBuffer::sWeight4Loc = gSkinnedObjectSimpleProgram.getAttribLocation(LLViewerShaderMgr::OBJECT_WEIGHT); +} + +void LLDrawPoolAvatar::endRigged() +{ + gSkinnedObjectSimpleProgram.unbind(); + LLVertexBuffer::sWeight4Loc = -1; +} + void LLDrawPoolAvatar::beginDeferredSkinned() { sShaderLevel = mVertexShaderLevel; @@ -711,6 +729,12 @@ void LLDrawPoolAvatar::renderAvatars(LLVOAvatar* single_avatar, S32 pass) avatarp->renderRigid(); return; } + + if (pass == 3) + { + avatarp->renderSkinnedAttachments(); + return; + } if (sShaderLevel > 0) { diff --git a/indra/newview/lldrawpoolavatar.h b/indra/newview/lldrawpoolavatar.h index b947943619..c43aa9b1e3 100644 --- a/indra/newview/lldrawpoolavatar.h +++ b/indra/newview/lldrawpoolavatar.h @@ -90,10 +90,12 @@ public: void beginRigid(); void beginFootShadow(); void beginSkinned(); + void beginRigged(); void endRigid(); void endFootShadow(); void endSkinned(); + void endRigged(); void beginDeferredImpostor(); void beginDeferredRigid(); diff --git a/indra/newview/llface.cpp b/indra/newview/llface.cpp index 53330e4d98..bc3e04db18 100644 --- a/indra/newview/llface.cpp +++ b/indra/newview/llface.cpp @@ -888,7 +888,8 @@ static LLFastTimer::DeclareTimer FTM_FACE_GET_GEOM("Face Geom"); BOOL LLFace::getGeometryVolume(const LLVolume& volume, const S32 &f, const LLMatrix4& mat_vert, const LLMatrix3& mat_normal, - const U16 &index_offset) + const U16 &index_offset, + bool force_rebuild) { LLFastTimer t(FTM_FACE_GET_GEOM); const LLVolumeFace &vf = volume.getVolumeFace(f); @@ -925,8 +926,9 @@ BOOL LLFace::getGeometryVolume(const LLVolume& volume, LLStrider colors; LLStrider binormals; LLStrider indicesp; + LLStrider weights; - BOOL full_rebuild = mDrawablep->isState(LLDrawable::REBUILD_VOLUME); + BOOL full_rebuild = force_rebuild || mDrawablep->isState(LLDrawable::REBUILD_VOLUME); BOOL global_volume = mDrawablep->getVOVolume()->isVolumeGlobal(); LLVector3 scale; @@ -944,6 +946,7 @@ BOOL LLFace::getGeometryVolume(const LLVolume& volume, BOOL rebuild_tcoord = full_rebuild || mDrawablep->isState(LLDrawable::REBUILD_TCOORD); BOOL rebuild_normal = rebuild_pos && mVertexBuffer->hasDataType(LLVertexBuffer::TYPE_NORMAL); BOOL rebuild_binormal = rebuild_pos && mVertexBuffer->hasDataType(LLVertexBuffer::TYPE_BINORMAL); + bool rebuild_weights = rebuild_pos && mVertexBuffer->hasDataType(LLVertexBuffer::TYPE_WEIGHT4); const LLTextureEntry *tep = mVObjp->getTE(f); U8 bump_code = tep ? tep->getBumpmap() : 0; @@ -960,7 +963,11 @@ BOOL LLFace::getGeometryVolume(const LLVolume& volume, { mVertexBuffer->getBinormalStrider(binormals, mGeomIndex); } - + if (rebuild_weights) + { + mVertexBuffer->getWeight4Strider(weights, mGeomIndex); + } + F32 tcoord_xoffset = 0.f ; F32 tcoord_yoffset = 0.f ; F32 tcoord_xscale = 1.f ; @@ -1338,6 +1345,11 @@ BOOL LLFace::getGeometryVolume(const LLVolume& volume, *binormals++ = binormal; } + if (rebuild_weights) + { + *weights++ = vf.mWeights[i]; + } + if (rebuild_color) { *colors++ = color; diff --git a/indra/newview/llface.h b/indra/newview/llface.h index 67dd97e6f7..06ec043c76 100644 --- a/indra/newview/llface.h +++ b/indra/newview/llface.h @@ -73,6 +73,7 @@ public: HUD_RENDER = 0x0008, USE_FACE_COLOR = 0x0010, TEXTURE_ANIM = 0x0020, + RIGGED = 0x0040, }; static void initClass(); @@ -145,7 +146,8 @@ public: BOOL getGeometryVolume(const LLVolume& volume, const S32 &f, const LLMatrix4& mat_vert, const LLMatrix3& mat_normal, - const U16 &index_offset); + const U16 &index_offset, + bool force_rebuild = false); // For avatar U16 getGeometryAvatar( diff --git a/indra/newview/llspatialpartition.h b/indra/newview/llspatialpartition.h index d74216de2d..b5e5967374 100644 --- a/indra/newview/llspatialpartition.h +++ b/indra/newview/llspatialpartition.h @@ -615,6 +615,13 @@ public: class LLVolumeGeometryManager: public LLGeometryManager { public: + typedef enum + { + NONE = 0, + BATCH_SORT, + DISTANCE_SORT + } eSortType; + virtual ~LLVolumeGeometryManager() { } virtual void rebuildGeom(LLSpatialGroup* group); virtual void rebuildMesh(LLSpatialGroup* group); diff --git a/indra/newview/llviewershadermgr.cpp b/indra/newview/llviewershadermgr.cpp index a0d0b9d490..8a68dd6ea7 100644 --- a/indra/newview/llviewershadermgr.cpp +++ b/indra/newview/llviewershadermgr.cpp @@ -77,6 +77,9 @@ LLGLSLShader gObjectFullbrightShinyProgram; LLGLSLShader gObjectShinyProgram; LLGLSLShader gObjectShinyWaterProgram; +//object hardware skinning shaders +LLGLSLShader gSkinnedObjectSimpleProgram; + //environment shaders LLGLSLShader gTerrainProgram; LLGLSLShader gTerrainWaterProgram; @@ -148,6 +151,7 @@ LLViewerShaderMgr::LLViewerShaderMgr() : mShaderList.push_back(&gObjectSimpleProgram); mShaderList.push_back(&gObjectFullbrightProgram); mShaderList.push_back(&gObjectFullbrightShinyProgram); + mShaderList.push_back(&gSkinnedObjectSimpleProgram); mShaderList.push_back(&gTerrainProgram); mShaderList.push_back(&gTerrainWaterProgram); mShaderList.push_back(&gObjectSimpleWaterProgram); @@ -195,6 +199,7 @@ void LLViewerShaderMgr::initAttribsAndUniforms(void) mReservedAttribs.push_back("materialColor"); mReservedAttribs.push_back("specularColor"); mReservedAttribs.push_back("binormal"); + mReservedAttribs.push_back("object_weight"); mAvatarAttribs.reserve(5); mAvatarAttribs.push_back("weight"); @@ -548,6 +553,9 @@ void LLViewerShaderMgr::unloadShaders() gObjectShinyProgram.unload(); gObjectFullbrightShinyProgram.unload(); gObjectShinyWaterProgram.unload(); + + gSkinnedObjectSimpleProgram.unload(); + gWaterProgram.unload(); gUnderWaterProgram.unload(); gTerrainProgram.unload(); @@ -625,6 +633,7 @@ BOOL LLViewerShaderMgr::loadBasicShaders() shaders.push_back( make_pair( "lighting/lightSpecularV.glsl", mVertexShaderLevel[SHADER_LIGHTING] ) ); shaders.push_back( make_pair( "windlight/atmosphericsV.glsl", mVertexShaderLevel[SHADER_WINDLIGHT] ) ); shaders.push_back( make_pair( "avatar/avatarSkinV.glsl", 1 ) ); + shaders.push_back( make_pair( "avatar/objectSkinV.glsl", 1 ) ); // We no longer have to bind the shaders to global glhandles, they are automatically added to a map now. for (U32 i = 0; i < shaders.size(); i++) @@ -1214,6 +1223,7 @@ BOOL LLViewerShaderMgr::loadShadersObject() gObjectSimpleWaterProgram.unload(); gObjectFullbrightProgram.unload(); gObjectFullbrightWaterProgram.unload(); + gSkinnedObjectSimpleProgram.unload(); return FALSE; } @@ -1323,6 +1333,21 @@ BOOL LLViewerShaderMgr::loadShadersObject() success = gObjectFullbrightShinyProgram.createShader(NULL, &mShinyUniforms); } + if (success) + { + gSkinnedObjectSimpleProgram.mName = "Skinned Simple Shader"; + gSkinnedObjectSimpleProgram.mFeatures.calculatesLighting = true; + gSkinnedObjectSimpleProgram.mFeatures.calculatesAtmospherics = true; + gSkinnedObjectSimpleProgram.mFeatures.hasGamma = true; + gSkinnedObjectSimpleProgram.mFeatures.hasAtmospherics = true; + gSkinnedObjectSimpleProgram.mFeatures.hasLighting = true; + gSkinnedObjectSimpleProgram.mFeatures.hasObjectSkinning = true; + gSkinnedObjectSimpleProgram.mShaderFiles.clear(); + gSkinnedObjectSimpleProgram.mShaderFiles.push_back(make_pair("objects/simpleSkinnedV.glsl", GL_VERTEX_SHADER_ARB)); + gSkinnedObjectSimpleProgram.mShaderFiles.push_back(make_pair("objects/simpleF.glsl", GL_FRAGMENT_SHADER_ARB)); + gSkinnedObjectSimpleProgram.mShaderLevel = mVertexShaderLevel[SHADER_OBJECT]; + success = gSkinnedObjectSimpleProgram.createShader(NULL, NULL); + } if( !success ) { diff --git a/indra/newview/llviewershadermgr.h b/indra/newview/llviewershadermgr.h index ac2b4624e0..83a650cdbc 100644 --- a/indra/newview/llviewershadermgr.h +++ b/indra/newview/llviewershadermgr.h @@ -82,6 +82,7 @@ public: MATERIAL_COLOR = 0, SPECULAR_COLOR, BINORMAL, + OBJECT_WEIGHT, END_RESERVED_ATTRIBS } eGLSLReservedAttribs; @@ -313,6 +314,8 @@ extern LLGLSLShader gObjectFullbrightShinyProgram; extern LLGLSLShader gObjectShinyProgram; extern LLGLSLShader gObjectShinyWaterProgram; +extern LLGLSLShader gSkinnedObjectSimpleProgram; + //environment shaders extern LLGLSLShader gTerrainProgram; extern LLGLSLShader gTerrainWaterProgram; diff --git a/indra/newview/llvoavatar.cpp b/indra/newview/llvoavatar.cpp index c400e8510e..535440692f 100644 --- a/indra/newview/llvoavatar.cpp +++ b/indra/newview/llvoavatar.cpp @@ -64,6 +64,7 @@ #include "llkeyframefallmotion.h" #include "llkeyframestandmotion.h" #include "llkeyframewalkmotion.h" +#include "llmeshrepository.h" #include "llmutelist.h" #include "llmoveview.h" #include "llquantize.h" @@ -79,6 +80,7 @@ #include "llviewermenu.h" #include "llviewerobjectlist.h" #include "llviewerparcelmgr.h" +#include "llviewershadermgr.h" #include "llviewerstats.h" #include "llvoavatarself.h" #include "llvovolume.h" @@ -3644,6 +3646,113 @@ bool LLVOAvatar::shouldAlphaMask() } +U32 LLVOAvatar::renderSkinnedAttachments() +{ + U32 num_indices = 0; + + const U32 data_mask = LLVertexBuffer::MAP_VERTEX | + LLVertexBuffer::MAP_NORMAL | + LLVertexBuffer::MAP_TEXCOORD0 | + LLVertexBuffer::MAP_WEIGHT4; + + for (attachment_map_t::const_iterator iter = mAttachmentPoints.begin(); + iter != mAttachmentPoints.end(); + ++iter) + { + LLViewerJointAttachment* attachment = iter->second; + for (LLViewerJointAttachment::attachedobjs_vec_t::iterator attachment_iter = attachment->mAttachedObjects.begin(); + attachment_iter != attachment->mAttachedObjects.end(); + ++attachment_iter) + { + const LLViewerObject* attached_object = (*attachment_iter); + if (attached_object && !attached_object->isHUDAttachment()) + { + const LLDrawable* drawable = attached_object->mDrawable; + if (drawable) + { + for (S32 i = 0; i < drawable->getNumFaces(); ++i) + { + LLFace* face = drawable->getFace(i); + if (face->isState(LLFace::RIGGED)) + { + LLVolume* volume = attached_object->getVolume(); + const LLVolumeFace& vol_face = volume->getVolumeFace(i); + + const LLMeshSkinInfo* skin = NULL; + LLVertexBuffer* buff = face->mVertexBuffer; + + if (!buff || + !buff->hasDataType(LLVertexBuffer::TYPE_WEIGHT4) || + buff->getRequestedVerts() != vol_face.mVertices.size()) + { + face->mVertexBuffer = NULL; + face->mLastVertexBuffer = NULL; + buff = NULL; + + LLUUID mesh_id = volume->getParams().getSculptID(); + if (mesh_id.notNull()) + { + skin = gMeshRepo.getSkinInfo(mesh_id); + if (skin) + { + face->mVertexBuffer = new LLVertexBuffer(data_mask, 0); + face->mVertexBuffer->allocateBuffer(vol_face.mVertices.size(), vol_face.mIndices.size(), true); + + face->setGeomIndex(0); + face->setIndicesIndex(0); + + U16 offset = 0; + + LLMatrix4 mat_vert = skin->mBindShapeMatrix; + LLMatrix3 mat_normal; + + face->getGeometryVolume(*volume, i, mat_vert, mat_normal, offset, true); + buff = face->mVertexBuffer; + } + } + } + + if (buff) + { + if (skin) + { + LLMatrix4 mat[64]; + + for (U32 i = 0; i < skin->mJointNames.size(); ++i) + { + LLJoint* joint = getJoint(skin->mJointNames[i]); + if (joint) + { + mat[i] = skin->mInvBindMatrix[i]; + mat[i] *= joint->getWorldMatrix(); + } + } + + gSkinnedObjectSimpleProgram.uniformMatrix4fv("matrixPalette", + skin->mJointNames.size(), + FALSE, + (GLfloat*) mat[0].mMatrix); + + buff->setBuffer(data_mask); + + U16 start = face->getGeomStart(); + U16 end = start + face->getGeomCount(); + S32 offset = face->getIndicesStart(); + U32 count = face->getIndicesCount(); + + buff->drawRange(LLRender::TRIANGLES, start, end, count, offset); + } + } + } + } + } + } + } + } + + return num_indices; +} + //----------------------------------------------------------------------------- // renderSkinned() //----------------------------------------------------------------------------- diff --git a/indra/newview/llvoavatar.h b/indra/newview/llvoavatar.h index d5485413f4..b0535a4a26 100644 --- a/indra/newview/llvoavatar.h +++ b/indra/newview/llvoavatar.h @@ -339,6 +339,7 @@ public: U32 renderImpostor(LLColor4U color = LLColor4U(255,255,255,255), S32 diffuse_channel = 0); U32 renderRigid(); U32 renderSkinned(EAvatarRenderPass pass); + U32 renderSkinnedAttachments(); U32 renderTransparent(BOOL first_pass); void renderCollisionVolumes(); static void deleteCachedImages(bool clearAll=true); diff --git a/indra/newview/llvovolume.cpp b/indra/newview/llvovolume.cpp index bc83e11fd2..56fb42bb89 100644 --- a/indra/newview/llvovolume.cpp +++ b/indra/newview/llvovolume.cpp @@ -3496,6 +3496,10 @@ void LLVolumeGeometryManager::rebuildGeom(LLSpatialGroup* group) drawablep->clearState(LLDrawable::HAS_ALPHA); + bool rigged = vobj->isAttachment() && + vobj->isMesh() && + gMeshRepo.getSkinInfo(vobj->getVolume()->getParams().getSculptID()); + //for each face for (S32 i = 0; i < drawablep->getNumFaces(); i++) { @@ -3503,6 +3507,22 @@ void LLVolumeGeometryManager::rebuildGeom(LLSpatialGroup* group) drawablep->updateFaceSize(i); LLFace* facep = drawablep->getFace(i); + if (rigged) + { + if (!facep->isState(LLFace::RIGGED)) + { + facep->mVertexBuffer = NULL; + facep->mLastVertexBuffer = NULL; + facep->setState(LLFace::RIGGED); + } + + continue; + } + else + { + facep->clearState(LLFace::RIGGED); + } + if (cur_total > max_total || facep->getIndicesCount() <= 0 || facep->getGeomCount() <= 0) { facep->mVertexBuffer = NULL; -- cgit v1.3 From 64112134be2a4b46c7ade1483e3d968ea1e3e81f Mon Sep 17 00:00:00 2001 From: Tofu Linden Date: Mon, 26 Apr 2010 14:42:06 +0100 Subject: EXT-7060 default auto alpha masking to ON for deferred rendering (only deferred rendering) --- indra/newview/app_settings/settings.xml | 15 +++++++++++++-- indra/newview/lldrawpoolalpha.cpp | 2 +- indra/newview/llface.cpp | 20 ++++++++++++++++++++ indra/newview/llface.h | 2 +- indra/newview/llspatialpartition.h | 2 -- indra/newview/llviewercontrol.cpp | 3 ++- indra/newview/llviewerdisplay.cpp | 3 ++- indra/newview/llvovolume.cpp | 21 +++------------------ indra/newview/pipeline.cpp | 3 ++- indra/newview/pipeline.h | 3 ++- indra/newview/skins/default/xui/en/menu_viewer.xml | 18 ++++++++++++++---- 11 files changed, 60 insertions(+), 32 deletions(-) (limited to 'indra/newview/llvovolume.cpp') diff --git a/indra/newview/app_settings/settings.xml b/indra/newview/app_settings/settings.xml index 8abb0f7106..a820d82f6f 100644 --- a/indra/newview/app_settings/settings.xml +++ b/indra/newview/app_settings/settings.xml @@ -6996,10 +6996,10 @@ Value 256.0 - RenderFastAlpha + RenderAutoMaskAlphaNonDeferred Comment - Use alpha masks where appropriate. + Use alpha masks where appropriate, in the non-deferred (non-'Lighting and Shadows') graphics mode Persist 1 Type @@ -7007,6 +7007,17 @@ Value 0 + RenderAutoMaskAlphaDeferred + + Comment + Use alpha masks where appropriate, in the deferred ('Lighting and Shadows') graphics mode + Persist + 1 + Type + Boolean + Value + 1 + RenderFastUI Comment diff --git a/indra/newview/lldrawpoolalpha.cpp b/indra/newview/lldrawpoolalpha.cpp index 7fd7cd3910..def463cb41 100644 --- a/indra/newview/lldrawpoolalpha.cpp +++ b/indra/newview/lldrawpoolalpha.cpp @@ -182,7 +182,7 @@ void LLDrawPoolAlpha::render(S32 pass) gGL.setColorMask(true, true); - if (LLPipeline::sFastAlpha && !deferred_render) + if (LLPipeline::sAutoMaskAlphaNonDeferred && !deferred_render) { mColorSFactor = LLRender::BF_ONE; // } mColorDFactor = LLRender::BF_ZERO; // } these are like disabling blend on the color channels, but we're still blending on the alpha channel so that we can suppress glow diff --git a/indra/newview/llface.cpp b/indra/newview/llface.cpp index bc3e04db18..7816418dc2 100644 --- a/indra/newview/llface.cpp +++ b/indra/newview/llface.cpp @@ -883,6 +883,26 @@ void LLFace::updateRebuildFlags() } } + +bool LLFace::canRenderAsMask() +{ + const LLTextureEntry* te = getTextureEntry(); + return ( + ( + (LLPipeline::sRenderDeferred && LLPipeline::sAutoMaskAlphaDeferred) || + + (!LLPipeline::sRenderDeferred && LLPipeline::sAutoMaskAlphaNonDeferred) + ) // do we want masks at all? + && + (te->getColor().mV[3] == 1.0f) && // can't treat as mask if we have face alpha + !(LLPipeline::sRenderDeferred && te->getFullbright()) && // hack: alpha masking renders fullbright faces invisible in deferred rendering mode, need to figure out why - for now, avoid + (te->getGlow() == 0.f) && // glowing masks are hard to implement - don't mask + + getTexture()->getIsAlphaMask() // texture actually qualifies for masking (lazily recalculated but expensive) + ); +} + + static LLFastTimer::DeclareTimer FTM_FACE_GET_GEOM("Face Geom"); BOOL LLFace::getGeometryVolume(const LLVolume& volume, diff --git a/indra/newview/llface.h b/indra/newview/llface.h index 06ec043c76..bbf8de04bc 100644 --- a/indra/newview/llface.h +++ b/indra/newview/llface.h @@ -139,10 +139,10 @@ public: void unsetFaceColor(); // switch back to material color const LLColor4& getFaceColor() const { return mFaceColor; } const LLColor4& getRenderColor() const; - //for volumes void updateRebuildFlags(); + bool canRenderAsMask(); // logic helper BOOL getGeometryVolume(const LLVolume& volume, const S32 &f, const LLMatrix4& mat_vert, const LLMatrix3& mat_normal, diff --git a/indra/newview/llspatialpartition.h b/indra/newview/llspatialpartition.h index b5e5967374..67c33d5b0f 100644 --- a/indra/newview/llspatialpartition.h +++ b/indra/newview/llspatialpartition.h @@ -628,8 +628,6 @@ class LLVolumeGeometryManager: public LLGeometryManager virtual void getGeometry(LLSpatialGroup* group); void genDrawInfo(LLSpatialGroup* group, U32 mask, std::vector& faces, BOOL distance_sort = FALSE); void registerFace(LLSpatialGroup* group, LLFace* facep, U32 type); - private: - bool canRenderAsMask(LLFace* facep); // logic helper }; //spatial partition that uses volume geometry manager (implemented in LLVOVolume.cpp) diff --git a/indra/newview/llviewercontrol.cpp b/indra/newview/llviewercontrol.cpp index 3fa6b0c8c2..33fb3d0f0f 100644 --- a/indra/newview/llviewercontrol.cpp +++ b/indra/newview/llviewercontrol.cpp @@ -532,7 +532,8 @@ void settings_setup_listeners() gSavedSettings.getControl("RenderMaxPartCount")->getSignal()->connect(boost::bind(&handleMaxPartCountChanged, _2)); gSavedSettings.getControl("RenderDynamicLOD")->getSignal()->connect(boost::bind(&handleRenderDynamicLODChanged, _2)); gSavedSettings.getControl("RenderDebugTextureBind")->getSignal()->connect(boost::bind(&handleResetVertexBuffersChanged, _2)); - gSavedSettings.getControl("RenderFastAlpha")->getSignal()->connect(boost::bind(&handleResetVertexBuffersChanged, _2)); + gSavedSettings.getControl("RenderAutoMaskAlphaDeferred")->getSignal()->connect(boost::bind(&handleResetVertexBuffersChanged, _2)); + gSavedSettings.getControl("RenderAutoMaskAlphaNonDeferred")->getSignal()->connect(boost::bind(&handleResetVertexBuffersChanged, _2)); gSavedSettings.getControl("RenderObjectBump")->getSignal()->connect(boost::bind(&handleResetVertexBuffersChanged, _2)); gSavedSettings.getControl("RenderMaxVBOSize")->getSignal()->connect(boost::bind(&handleResetVertexBuffersChanged, _2)); gSavedSettings.getControl("RenderUseFBO")->getSignal()->connect(boost::bind(&handleRenderUseFBOChanged, _2)); diff --git a/indra/newview/llviewerdisplay.cpp b/indra/newview/llviewerdisplay.cpp index 5be5dc0444..18558425ac 100644 --- a/indra/newview/llviewerdisplay.cpp +++ b/indra/newview/llviewerdisplay.cpp @@ -596,7 +596,8 @@ void display(BOOL rebuild, F32 zoom_factor, int subfield, BOOL for_snapshot) LLPipeline::sUseOcclusion = 3; } - LLPipeline::sFastAlpha = gSavedSettings.getBOOL("RenderFastAlpha"); + LLPipeline::sAutoMaskAlphaDeferred = gSavedSettings.getBOOL("RenderAutoMaskAlphaDeferred"); + LLPipeline::sAutoMaskAlphaNonDeferred = gSavedSettings.getBOOL("RenderAutoMaskAlphaNonDeferred"); LLPipeline::sUseFarClip = gSavedSettings.getBOOL("RenderUseFarClip"); LLVOAvatar::sMaxVisible = gSavedSettings.getS32("RenderAvatarMaxVisible"); LLPipeline::sDelayVBUpdate = gSavedSettings.getBOOL("RenderDelayVBUpdate"); diff --git a/indra/newview/llvovolume.cpp b/indra/newview/llvovolume.cpp index 56fb42bb89..96f69b3676 100644 --- a/indra/newview/llvovolume.cpp +++ b/indra/newview/llvovolume.cpp @@ -3042,9 +3042,7 @@ F32 LLVOVolume::getBinRadius() { LLFace* face = mDrawable->getFace(i); if (face->getPoolType() == LLDrawPool::POOL_ALPHA && - (!LLPipeline::sFastAlpha || - face->getFaceColor().mV[3] != 1.f || - !face->getTexture()->getIsAlphaMask())) + !face->canRenderAsMask()) { alpha_wrap = TRUE; break; @@ -3417,19 +3415,6 @@ void LLVolumeGeometryManager::getGeometry(LLSpatialGroup* group) static LLFastTimer::DeclareTimer FTM_REBUILD_VOLUME_VB("Volume"); static LLFastTimer::DeclareTimer FTM_REBUILD_VBO("VBO Rebuilt"); -bool LLVolumeGeometryManager::canRenderAsMask(LLFace* facep) -{ - const LLTextureEntry* te = facep->getTextureEntry(); - return ( - LLPipeline::sFastAlpha && // do we want masks at all? - - (te->getColor().mV[3] == 1.0f) && // can't treat as mask if we have face alpha - !(LLPipeline::sRenderDeferred && te->getFullbright()) && // hack: alpha masking renders fullbright faces invisible in deferred rendering mode, need to figure out why - for now, avoid - (te->getGlow() == 0.f) && // glowing masks are hard to implement - don't mask - - facep->getTexture()->getIsAlphaMask() // texture actually qualifies for masking (lazily calculated but expensive) - ); -} void LLVolumeGeometryManager::rebuildGeom(LLSpatialGroup* group) { @@ -3576,7 +3561,7 @@ void LLVolumeGeometryManager::rebuildGeom(LLSpatialGroup* group) if (type == LLDrawPool::POOL_ALPHA) { - if (canRenderAsMask(facep)) + if (facep->canRenderAsMask()) { //can be treated as alpha mask simple_faces.push_back(facep); } @@ -3930,7 +3915,7 @@ void LLVolumeGeometryManager::genDrawInfo(LLSpatialGroup* group, U32 mask, std:: if (is_alpha) { // can we safely treat this as an alpha mask? - if (canRenderAsMask(facep)) + if (facep->canRenderAsMask()) { if (te->getFullbright()) { diff --git a/indra/newview/pipeline.cpp b/indra/newview/pipeline.cpp index 5a173d8460..3df9a119bf 100644 --- a/indra/newview/pipeline.cpp +++ b/indra/newview/pipeline.cpp @@ -271,7 +271,8 @@ BOOL LLPipeline::sRenderHighlight = TRUE; BOOL LLPipeline::sForceOldBakedUpload = FALSE; S32 LLPipeline::sUseOcclusion = 0; BOOL LLPipeline::sDelayVBUpdate = TRUE; -BOOL LLPipeline::sFastAlpha = TRUE; +BOOL LLPipeline::sAutoMaskAlphaDeferred = TRUE; +BOOL LLPipeline::sAutoMaskAlphaNonDeferred = FALSE; BOOL LLPipeline::sDisableShaders = FALSE; BOOL LLPipeline::sRenderBump = TRUE; BOOL LLPipeline::sUseTriStrips = TRUE; diff --git a/indra/newview/pipeline.h b/indra/newview/pipeline.h index 8bdc635d05..be878ae667 100644 --- a/indra/newview/pipeline.h +++ b/indra/newview/pipeline.h @@ -450,7 +450,8 @@ public: static BOOL sForceOldBakedUpload; // If true will not use capabilities to upload baked textures. static S32 sUseOcclusion; // 0 = no occlusion, 1 = read only, 2 = read/write static BOOL sDelayVBUpdate; - static BOOL sFastAlpha; + static BOOL sAutoMaskAlphaDeferred; + static BOOL sAutoMaskAlphaNonDeferred; static BOOL sDisableShaders; // if TRUE, rendering will be done without shaders static BOOL sRenderBump; static BOOL sUseTriStrips; diff --git a/indra/newview/skins/default/xui/en/menu_viewer.xml b/indra/newview/skins/default/xui/en/menu_viewer.xml index bb57464741..6a5f9ed8f8 100644 --- a/indra/newview/skins/default/xui/en/menu_viewer.xml +++ b/indra/newview/skins/default/xui/en/menu_viewer.xml @@ -2148,14 +2148,24 @@ parameter="RenderDebugPipeline" /> + label="Automatic Alpha Masks (deferred)" + name="Automatic Alpha Masks (deferred)"> + parameter="RenderAutoMaskAlphaDeferred" /> + parameter="RenderAutoMaskAlphaDeferred" /> + + + + Date: Sat, 1 May 2010 00:45:44 -0500 Subject: Rigged attachment integration WIP. --- indra/llrender/llglslshader.cpp | 2 +- .../shaders/class1/deferred/diffuseV.glsl | 2 +- indra/newview/lldrawpool.cpp | 5 - indra/newview/lldrawpool.h | 2 - indra/newview/lldrawpoolavatar.cpp | 244 ++++++++++++++++++++- indra/newview/lldrawpoolavatar.h | 42 +++- indra/newview/lldrawpoolbump.cpp | 54 +++-- indra/newview/lldrawpoolbump.h | 4 + indra/newview/llface.cpp | 5 + indra/newview/llface.h | 1 + indra/newview/llviewerobject.cpp | 19 ++ indra/newview/llviewerobject.h | 1 + indra/newview/llviewershadermgr.cpp | 27 ++- indra/newview/llviewershadermgr.h | 1 + indra/newview/llvoavatar.cpp | 99 +-------- indra/newview/llvovolume.cpp | 66 +++++- 16 files changed, 440 insertions(+), 134 deletions(-) (limited to 'indra/newview/llvovolume.cpp') diff --git a/indra/llrender/llglslshader.cpp b/indra/llrender/llglslshader.cpp index 949057df04..9256e3959c 100644 --- a/indra/llrender/llglslshader.cpp +++ b/indra/llrender/llglslshader.cpp @@ -124,7 +124,7 @@ BOOL LLGLSLShader::createShader(vector * attributes, { GLhandleARB shaderhandle = LLShaderMgr::instance()->loadShaderFile((*fileIter).first, mShaderLevel, (*fileIter).second); LL_DEBUGS("ShaderLoading") << "SHADER FILE: " << (*fileIter).first << " mShaderLevel=" << mShaderLevel << LL_ENDL; - if (mShaderLevel > 0) + if (shaderhandle > 0) { attachObject(shaderhandle); } diff --git a/indra/newview/app_settings/shaders/class1/deferred/diffuseV.glsl b/indra/newview/app_settings/shaders/class1/deferred/diffuseV.glsl index 44468cdfa2..b458842657 100644 --- a/indra/newview/app_settings/shaders/class1/deferred/diffuseV.glsl +++ b/indra/newview/app_settings/shaders/class1/deferred/diffuseV.glsl @@ -13,7 +13,7 @@ void main() gl_Position = gl_ModelViewProjectionMatrix * gl_Vertex; gl_TexCoord[0] = gl_TextureMatrix[0] * gl_MultiTexCoord0; - vary_normal = normalize(gl_NormalMatrix * gl_Normal); + vary_nomral = normalize(gl_NormalMatrix * gl_Normal); gl_FrontColor = gl_Color; } diff --git a/indra/newview/lldrawpool.cpp b/indra/newview/lldrawpool.cpp index ae30af3647..a99b80d618 100644 --- a/indra/newview/lldrawpool.cpp +++ b/indra/newview/lldrawpool.cpp @@ -248,11 +248,6 @@ void LLFacePool::dirtyTextures(const std::set& textures { } -BOOL LLFacePool::moveFace(LLFace *face, LLDrawPool *poolp, BOOL copy_data) -{ - return TRUE; -} - // static S32 LLFacePool::drawLoop(face_array_t& face_list) { diff --git a/indra/newview/lldrawpool.h b/indra/newview/lldrawpool.h index 67870c10e9..e46d503db3 100644 --- a/indra/newview/lldrawpool.h +++ b/indra/newview/lldrawpool.h @@ -187,8 +187,6 @@ public: virtual void resetDrawOrders(); void resetAll(); - BOOL moveFace(LLFace *face, LLDrawPool *poolp, BOOL copy_data = FALSE); - void destroy(); void buildEdges(); diff --git a/indra/newview/lldrawpoolavatar.cpp b/indra/newview/lldrawpoolavatar.cpp index 62226383a4..02c7e3bb6f 100644 --- a/indra/newview/lldrawpoolavatar.cpp +++ b/indra/newview/lldrawpoolavatar.cpp @@ -39,13 +39,17 @@ #include "m3math.h" #include "lldrawable.h" +#include "lldrawpoolbump.h" #include "llface.h" +#include "llmeshrepository.h" #include "llsky.h" #include "llviewercamera.h" #include "llviewerregion.h" #include "noise.h" #include "pipeline.h" #include "llviewershadermgr.h" +#include "llvovolume.h" +#include "llvolume.h" #include "llappviewer.h" #include "llrendersphere.h" #include "llviewerpartsim.h" @@ -94,6 +98,8 @@ static BOOL sRenderingSkinned = FALSE; S32 normal_channel = -1; S32 specular_channel = -1; S32 diffuse_channel = -1; +S32 cube_channel = -1; + static LLFastTimer::DeclareTimer FTM_SHADOW_AVATAR("Avatar Shadow"); @@ -358,7 +364,7 @@ S32 LLDrawPoolAvatar::getNumPasses() } else if (getVertexShaderLevel() > 0) { - return 4; + return 5; } else { @@ -402,7 +408,10 @@ void LLDrawPoolAvatar::beginRenderPass(S32 pass) beginSkinned(); break; case 3: - beginRigged(); + beginRiggedSimple(); + break; + case 4: + beginRiggedShinySimple(); break; } } @@ -429,7 +438,10 @@ void LLDrawPoolAvatar::endRenderPass(S32 pass) endSkinned(); break; case 3: - endRigged(); + endRiggedSimple(); + break; + case 4: + endRiggedShinySimple(); break; } } @@ -616,14 +628,15 @@ void LLDrawPoolAvatar::endSkinned() gGL.getTexUnit(0)->activate(); } -void LLDrawPoolAvatar::beginRigged() +void LLDrawPoolAvatar::beginRiggedSimple() { sVertexProgram = &gSkinnedObjectSimpleProgram; + diffuse_channel = 0; gSkinnedObjectSimpleProgram.bind(); LLVertexBuffer::sWeight4Loc = gSkinnedObjectSimpleProgram.getAttribLocation(LLViewerShaderMgr::OBJECT_WEIGHT); } -void LLDrawPoolAvatar::endRigged() +void LLDrawPoolAvatar::endRiggedSimple() { sVertexProgram = NULL; LLVertexBuffer::unbind(); @@ -631,6 +644,23 @@ void LLDrawPoolAvatar::endRigged() LLVertexBuffer::sWeight4Loc = -1; } +void LLDrawPoolAvatar::beginRiggedShinySimple() +{ + sVertexProgram = &gSkinnedObjectShinySimpleProgram; + sVertexProgram->bind(); + LLDrawPoolBump::bindCubeMap(sVertexProgram, 2, diffuse_channel, cube_channel, false); + LLVertexBuffer::sWeight4Loc = sVertexProgram->getAttribLocation(LLViewerShaderMgr::OBJECT_WEIGHT); +} + +void LLDrawPoolAvatar::endRiggedShinySimple() +{ + LLVertexBuffer::unbind(); + LLDrawPoolBump::unbindCubeMap(sVertexProgram, 2, diffuse_channel, cube_channel, false); + sVertexProgram->unbind(); + sVertexProgram = NULL; + LLVertexBuffer::sWeight4Loc = -1; +} + void LLDrawPoolAvatar::beginDeferredRigged() { sVertexProgram = &gDeferredSkinnedDiffuseProgram; @@ -790,9 +820,16 @@ void LLDrawPoolAvatar::renderAvatars(LLVOAvatar* single_avatar, S32 pass) if (pass == 3) { - avatarp->renderSkinnedAttachments(); + renderRiggedSimple(avatarp); + return; + } + + if (pass == 4) + { + renderRiggedShinySimple(avatarp); return; } + if (sShaderLevel > 0) { @@ -830,13 +867,146 @@ void LLDrawPoolAvatar::renderAvatars(LLVOAvatar* single_avatar, S32 pass) } } +void LLDrawPoolAvatar::updateRiggedFaceVertexBuffer(LLFace* face, const LLMeshSkinInfo* skin, LLVolume* volume, const LLVolumeFace& vol_face, U32 data_mask) +{ + LLVertexBuffer* buff = face->mVertexBuffer; + + if (!buff || + !buff->hasDataType(LLVertexBuffer::TYPE_WEIGHT4) || + buff->getRequestedVerts() != vol_face.mVertices.size()) + { + face->setGeomIndex(0); + face->setIndicesIndex(0); + face->setSize(vol_face.mVertices.size(), vol_face.mIndices.size()); + + face->mVertexBuffer = new LLVertexBuffer(data_mask, 0); + face->mVertexBuffer->allocateBuffer(vol_face.mVertices.size(), vol_face.mIndices.size(), true); + + U16 offset = 0; + + LLMatrix4 mat_vert = skin->mBindShapeMatrix; + glh::matrix4f m((F32*) mat_vert.mMatrix); + m = m.inverse().transpose(); + + F32 mat3[] = + { m.m[0], m.m[1], m.m[2], + m.m[4], m.m[5], m.m[6], + m.m[8], m.m[9], m.m[10] }; + + LLMatrix3 mat_normal(mat3); + + face->getGeometryVolume(*volume, face->getTEOffset(), mat_vert, mat_normal, offset, true); + buff = face->mVertexBuffer; + } +} + +void LLDrawPoolAvatar::renderRigged(LLVOAvatar* avatar, U32 type, const U32 data_mask) +{ + for (U32 i = 0; i < mRiggedFace[type].size(); ++i) + { + LLFace* face = mRiggedFace[type][i]; + LLDrawable* drawable = face->getDrawable(); + if (!drawable) + { + continue; + } + + LLVOVolume* vobj = drawable->getVOVolume(); + + if (!vobj) + { + continue; + } + + LLVolume* volume = vobj->getVolume(); + S32 te = face->getTEOffset(); + + if (!volume || volume->getNumVolumeFaces() <= te) + { + continue; + } + + LLUUID mesh_id = volume->getParams().getSculptID(); + if (mesh_id.isNull()) + { + continue; + } + + const LLMeshSkinInfo* skin = gMeshRepo.getSkinInfo(mesh_id); + if (!skin) + { + continue; + } + + const LLVolumeFace& vol_face = volume->getVolumeFace(te); + updateRiggedFaceVertexBuffer(face, skin, volume, vol_face, data_mask); + + LLVertexBuffer* buff = face->mVertexBuffer; + + if (buff) + { + LLMatrix4 mat[64]; + + for (U32 i = 0; i < skin->mJointNames.size(); ++i) + { + LLJoint* joint = avatar->getJoint(skin->mJointNames[i]); + if (joint) + { + mat[i] = skin->mInvBindMatrix[i]; + mat[i] *= joint->getWorldMatrix(); + } + } + + LLDrawPoolAvatar::sVertexProgram->uniformMatrix4fv("matrixPalette", + skin->mJointNames.size(), + FALSE, + (GLfloat*) mat[0].mMatrix); + LLDrawPoolAvatar::sVertexProgram->uniformMatrix4fv("matrixPalette[0]", + skin->mJointNames.size(), + FALSE, + (GLfloat*) mat[0].mMatrix); + + buff->setBuffer(data_mask); + + U16 start = face->getGeomStart(); + U16 end = start + face->getGeomCount()-1; + S32 offset = face->getIndicesStart(); + U32 count = face->getIndicesCount(); + + gGL.getTexUnit(0)->bind(face->getTexture()); + buff->drawRange(LLRender::TRIANGLES, start, end, count, offset); + } + } +} + +void LLDrawPoolAvatar::renderRiggedSimple(LLVOAvatar* avatar) +{ + const U32 data_mask = LLVertexBuffer::MAP_VERTEX | + LLVertexBuffer::MAP_NORMAL | + LLVertexBuffer::MAP_TEXCOORD0 | + LLVertexBuffer::MAP_COLOR | + LLVertexBuffer::MAP_WEIGHT4; + + renderRigged(avatar, RIGGED_SIMPLE, data_mask); +} + + +void LLDrawPoolAvatar::renderRiggedShinySimple(LLVOAvatar* avatar) +{ + const U32 data_mask = LLVertexBuffer::MAP_VERTEX | + LLVertexBuffer::MAP_NORMAL | + LLVertexBuffer::MAP_TEXCOORD0 | + LLVertexBuffer::MAP_COLOR | + LLVertexBuffer::MAP_WEIGHT4; + + renderRigged(avatar, RIGGED_SHINY_SIMPLE, data_mask); +} + //----------------------------------------------------------------------------- // renderForSelect() //----------------------------------------------------------------------------- void LLDrawPoolAvatar::renderForSelect() { - - if (mDrawFace.empty()) { return; @@ -930,6 +1100,64 @@ LLColor3 LLDrawPoolAvatar::getDebugColor() const return LLColor3(0.f, 1.f, 0.f); } +void LLDrawPoolAvatar::addRiggedFace(LLFace* facep, U32 type) +{ + if (facep->getReferenceIndex() != -1) + { + llerrs << "Tried to add a rigged face that's referenced elsewhere." << llendl; + } + + if (type >= NUM_RIGGED_PASSES) + { + llerrs << "Invalid rigged face type." << llendl; + } + + facep->setReferenceIndex(mRiggedFace[type].size()); + facep->mDrawPoolp = this; + mRiggedFace[type].push_back(facep); +} + +void LLDrawPoolAvatar::removeRiggedFace(LLFace* facep, U32 type) +{ + S32 index = facep->getReferenceIndex(); + if (index == -1) + { + llerrs << "Tried to remove rigged face with invalid index." << llendl; + } + + if (type > RIGGED_UNKNOWN) + { + llerrs << "Invalid rigged face type." << llendl; + } + + facep->setReferenceIndex(-1); + facep->mDrawPoolp = NULL; + + if (type == RIGGED_UNKNOWN) + { + for (U32 i = 0; i < NUM_RIGGED_PASSES; ++i) + { + if (mRiggedFace[i].size() > index && mRiggedFace[i][index] == facep) + { + type = i; + break; + } + } + } + + if (type >= NUM_RIGGED_PASSES) + { + llerrs << "Could not find face for removal from current drawpool." << llendl; + } + + mRiggedFace[type].erase(mRiggedFace[type].begin()+index); + + for (S32 i = index; i < mRiggedFace[type].size(); ++i) + { //bump indexes of currently held faces down after removal + mRiggedFace[type][i]->setReferenceIndex(i); + } +} + LLVertexBufferAvatar::LLVertexBufferAvatar() : LLVertexBuffer(sDataMask, GL_STREAM_DRAW_ARB) //avatars are always stream draw due to morph targets diff --git a/indra/newview/lldrawpoolavatar.h b/indra/newview/lldrawpoolavatar.h index b42cc54622..0ebb035f2a 100644 --- a/indra/newview/lldrawpoolavatar.h +++ b/indra/newview/lldrawpoolavatar.h @@ -37,6 +37,11 @@ class LLVOAvatar; class LLGLSLShader; +class LLFace; +class LLMeshSkinInfo; +class LLVolume; +class LLVolumeFace; + class LLDrawPoolAvatar : public LLFacePool { @@ -91,12 +96,14 @@ public: void beginRigid(); void beginImpostor(); void beginSkinned(); - void beginRigged(); - + void beginRiggedSimple(); + void beginRiggedShinySimple(); + void endRigid(); void endImpostor(); void endSkinned(); - void endRigged(); + void endRiggedSimple(); + void endRiggedShinySimple(); void beginDeferredImpostor(); void beginDeferredRigid(); @@ -108,11 +115,40 @@ public: void endDeferredSkinned(); void endDeferredRigged(); + void updateRiggedFaceVertexBuffer(LLFace* facep, + const LLMeshSkinInfo* skin, + LLVolume* volume, + const LLVolumeFace& vol_face, + U32 data_mask); + + void renderRigged(LLVOAvatar* avatar, U32 type, const U32 data_mask); + void renderRiggedSimple(LLVOAvatar* avatar); + void renderRiggedShinySimple(LLVOAvatar* avatar); + /*virtual*/ LLViewerTexture *getDebugTexture(); /*virtual*/ LLColor3 getDebugColor() const; // For AGP debug display void renderAvatars(LLVOAvatar *single_avatar, S32 pass = -1); // renders only one avatar if single_avatar is not null. + typedef enum + { + RIGGED_SIMPLE = 0, + RIGGED_SHINY_SIMPLE, + RIGGED_SHINY_FULLBRIGHT, + RIGGED_SHINY_BUMP, + RIGGED_BUMP, + RIGGED_FULLBRIGHT, + RIGGED_ALPHA, + NUM_RIGGED_PASSES, + RIGGED_UNKNOWN, + } eRiggedPass; + + + void addRiggedFace(LLFace* facep, U32 type); + void removeRiggedFace(LLFace* facep, U32 type = RIGGED_UNKNOWN); + + std::vector mRiggedFace[NUM_RIGGED_PASSES]; + static BOOL sSkipOpaque; static BOOL sSkipTransparent; static LLGLSLShader* sVertexProgram; diff --git a/indra/newview/lldrawpoolbump.cpp b/indra/newview/lldrawpoolbump.cpp index 8f3e775976..2f449fa42f 100644 --- a/indra/newview/lldrawpoolbump.cpp +++ b/indra/newview/lldrawpoolbump.cpp @@ -323,30 +323,43 @@ void LLDrawPoolBump::beginShiny(bool invisible) sVertexMask = VERTEX_MASK_SHINY | LLVertexBuffer::MAP_TEXCOORD0; } - if (LLPipeline::sUnderWaterRender) + if (getVertexShaderLevel() > 0) { - shader = &gObjectShinyWaterProgram; + if (LLPipeline::sUnderWaterRender) + { + shader = &gObjectShinyWaterProgram; + } + else + { + shader = &gObjectShinyProgram; + } + shader->bind(); } else { - shader = &gObjectShinyProgram; + shader = NULL; } + bindCubeMap(shader, mVertexShaderLevel, diffuse_channel, cube_channel, invisible); +} + +//static +void LLDrawPoolBump::bindCubeMap(LLGLSLShader* shader, S32 shader_level, S32& diffuse_channel, S32& cube_channel, bool invisible) +{ LLCubeMap* cube_map = gSky.mVOSkyp ? gSky.mVOSkyp->getCubeMap() : NULL; if( cube_map ) { - if (!invisible && LLViewerShaderMgr::instance()->getVertexShaderLevel(LLViewerShaderMgr::SHADER_OBJECT) > 0 ) + if (!invisible && shader ) { LLMatrix4 mat; mat.initRows(LLVector4(gGLModelView+0), LLVector4(gGLModelView+4), LLVector4(gGLModelView+8), LLVector4(gGLModelView+12)); - shader->bind(); LLVector3 vec = LLVector3(gShinyOrigin) * mat; LLVector4 vec4(vec, gShinyOrigin.mV[3]); shader->uniform4fv(LLViewerShaderMgr::SHINY_ORIGIN, 1, vec4.mV); - if (mVertexShaderLevel > 1) + if (shader_level > 1) { cube_map->setMatrix(1); // Make sure that texture coord generation happens for tex unit 1, as that's the one we use for @@ -408,22 +421,16 @@ void LLDrawPoolBump::renderShiny(bool invisible) } } -void LLDrawPoolBump::endShiny(bool invisible) +//static +void LLDrawPoolBump::unbindCubeMap(LLGLSLShader* shader, S32 shader_level, S32& diffuse_channel, S32& cube_channel, bool invisible) { - LLFastTimer t(FTM_RENDER_SHINY); - if ((!invisible && !gPipeline.hasRenderBatches(LLRenderPass::PASS_SHINY))|| - (invisible && !gPipeline.hasRenderBatches(LLRenderPass::PASS_INVISI_SHINY))) - { - return; - } - LLCubeMap* cube_map = gSky.mVOSkyp ? gSky.mVOSkyp->getCubeMap() : NULL; if( cube_map ) { cube_map->disable(); cube_map->restoreMatrix(); - if (!invisible && mVertexShaderLevel > 1) + if (!invisible && shader_level > 1) { shader->disableTexture(LLViewerShaderMgr::ENVIRONMENT_MAP, LLTexUnit::TT_CUBE_MAP); @@ -434,7 +441,6 @@ void LLDrawPoolBump::endShiny(bool invisible) shader->disableTexture(LLViewerShaderMgr::DIFFUSE_MAP); } } - shader->unbind(); } } gGL.getTexUnit(diffuse_channel)->disable(); @@ -442,6 +448,22 @@ void LLDrawPoolBump::endShiny(bool invisible) gGL.getTexUnit(0)->unbind(LLTexUnit::TT_TEXTURE); gGL.getTexUnit(0)->setTextureBlendType(LLTexUnit::TB_MULT); +} + +void LLDrawPoolBump::endShiny(bool invisible) +{ + LLFastTimer t(FTM_RENDER_SHINY); + if ((!invisible && !gPipeline.hasRenderBatches(LLRenderPass::PASS_SHINY))|| + (invisible && !gPipeline.hasRenderBatches(LLRenderPass::PASS_INVISI_SHINY))) + { + return; + } + + unbindCubeMap(shader, mVertexShaderLevel, diffuse_channel, cube_channel, invisible); + if (shader) + { + shader->unbind(); + } diffuse_channel = -1; cube_channel = 0; diff --git a/indra/newview/lldrawpoolbump.h b/indra/newview/lldrawpoolbump.h index 2019f1df26..89bbefe778 100644 --- a/indra/newview/lldrawpoolbump.h +++ b/indra/newview/lldrawpoolbump.h @@ -41,6 +41,7 @@ class LLImageRaw; class LLSpatialGroup; class LLDrawInfo; +class LLGLSLShader; class LLViewerFetchedTexture; class LLDrawPoolBump : public LLRenderPass @@ -79,6 +80,9 @@ public: void renderBump(); void endBump(); + static void bindCubeMap(LLGLSLShader* shader, S32 shader_level, S32& diffuse_channel, S32& cube_channel, bool invisible); + static void unbindCubeMap(LLGLSLShader* shader, S32 shader_level, S32& diffuse_channel, S32& cube_channel, bool invisible); + virtual S32 getNumDeferredPasses(); /*virtual*/ void beginDeferredPass(S32 pass); /*virtual*/ void endDeferredPass(S32 pass); diff --git a/indra/newview/llface.cpp b/indra/newview/llface.cpp index 7866e49bae..0e0b8447ca 100644 --- a/indra/newview/llface.cpp +++ b/indra/newview/llface.cpp @@ -205,7 +205,12 @@ void LLFace::destroy() if (mDrawPoolp) { LLFastTimer t(FTM_DESTROY_DRAWPOOL); + + if (this->isState(LLFace::RIGGED) && mDrawPoolp->getType() == LLDrawPool::POOL_AVATAR) mDrawPoolp->removeFace(this); + + + mDrawPoolp = NULL; } diff --git a/indra/newview/llface.h b/indra/newview/llface.h index bbf8de04bc..2b8fdf2e58 100644 --- a/indra/newview/llface.h +++ b/indra/newview/llface.h @@ -234,6 +234,7 @@ public: private: friend class LLGeometryManager; friend class LLVolumeGeometryManager; + friend class LLDrawPoolAvatar; U32 mState; LLFacePool* mDrawPoolp; diff --git a/indra/newview/llviewerobject.cpp b/indra/newview/llviewerobject.cpp index 6bd3ceb8a8..3aecd0175d 100644 --- a/indra/newview/llviewerobject.cpp +++ b/indra/newview/llviewerobject.cpp @@ -5219,6 +5219,25 @@ void LLViewerObject::resetChildrenPosition(const LLVector3& offset, BOOL simplif return ; } +//virtual +LLVOAvatar* LLViewerObject::getAvatar() const +{ + if (isAttachment()) + { + LLViewerObject* vobj = (LLViewerObject*) getParent(); + + while (vobj && !vobj->asAvatar()) + { + vobj = (LLViewerObject*) vobj->getParent(); + } + + return (LLVOAvatar*) vobj; + } + + return NULL; +} + + class ObjectPhysicsProperties : public LLHTTPNode { public: diff --git a/indra/newview/llviewerobject.h b/indra/newview/llviewerobject.h index 594d7a0827..0fd0cbfa60 100644 --- a/indra/newview/llviewerobject.h +++ b/indra/newview/llviewerobject.h @@ -181,6 +181,7 @@ public: void setOnActiveList(BOOL on_active) { mOnActiveList = on_active; } virtual BOOL isAttachment() const { return FALSE; } + virtual LLVOAvatar* getAvatar() const; //get the avatar this object is attached to, or NULL if object is not an attachment virtual BOOL isHUDAttachment() const { return FALSE; } virtual void updateRadius() {}; virtual F32 getVObjRadius() const; // default implemenation is mDrawable->getRadius() diff --git a/indra/newview/llviewershadermgr.cpp b/indra/newview/llviewershadermgr.cpp index d978e856a6..fe68d6eaa4 100644 --- a/indra/newview/llviewershadermgr.cpp +++ b/indra/newview/llviewershadermgr.cpp @@ -79,6 +79,7 @@ LLGLSLShader gObjectShinyWaterProgram; //object hardware skinning shaders LLGLSLShader gSkinnedObjectSimpleProgram; +LLGLSLShader gSkinnedObjectShinySimpleProgram; //environment shaders LLGLSLShader gTerrainProgram; @@ -154,6 +155,7 @@ LLViewerShaderMgr::LLViewerShaderMgr() : mShaderList.push_back(&gObjectFullbrightProgram); mShaderList.push_back(&gObjectFullbrightShinyProgram); mShaderList.push_back(&gSkinnedObjectSimpleProgram); + mShaderList.push_back(&gSkinnedObjectShinySimpleProgram); mShaderList.push_back(&gTerrainProgram); mShaderList.push_back(&gTerrainWaterProgram); mShaderList.push_back(&gObjectSimpleWaterProgram); @@ -505,6 +507,9 @@ void LLViewerShaderMgr::setShaders() if (!loadShadersDeferred()) { gSavedSettings.setBOOL("RenderDeferred", FALSE); + reentrance = false; + setShaders(); + return; } #endif } @@ -557,6 +562,8 @@ void LLViewerShaderMgr::unloadShaders() gObjectShinyWaterProgram.unload(); gSkinnedObjectSimpleProgram.unload(); + gSkinnedObjectShinySimpleProgram.unload(); + gWaterProgram.unload(); gUnderWaterProgram.unload(); @@ -917,7 +924,7 @@ BOOL LLViewerShaderMgr::loadShadersDeferred() gDeferredGIProgram.unload(); gDeferredGIFinalProgram.unload(); gDeferredWaterProgram.unload(); - return FALSE; + return TRUE; } mVertexShaderLevel[SHADER_AVATAR] = 1; @@ -1251,6 +1258,8 @@ BOOL LLViewerShaderMgr::loadShadersObject() gObjectFullbrightProgram.unload(); gObjectFullbrightWaterProgram.unload(); gSkinnedObjectSimpleProgram.unload(); + gSkinnedObjectShinySimpleProgram.unload(); + return FALSE; } @@ -1376,6 +1385,22 @@ BOOL LLViewerShaderMgr::loadShadersObject() success = gSkinnedObjectSimpleProgram.createShader(NULL, NULL); } + if (success) + { + gSkinnedObjectShinySimpleProgram.mName = "Skinned Shiny Simple Shader"; + gSkinnedObjectShinySimpleProgram.mFeatures.calculatesLighting = true; + gSkinnedObjectShinySimpleProgram.mFeatures.calculatesAtmospherics = true; + gSkinnedObjectShinySimpleProgram.mFeatures.hasGamma = true; + gSkinnedObjectShinySimpleProgram.mFeatures.hasAtmospherics = true; + gSkinnedObjectShinySimpleProgram.mFeatures.hasObjectSkinning = true; + gSkinnedObjectShinySimpleProgram.mFeatures.isShiny = true; + gSkinnedObjectShinySimpleProgram.mShaderFiles.clear(); + gSkinnedObjectShinySimpleProgram.mShaderFiles.push_back(make_pair("objects/shinySimpleSkinnedV.glsl", GL_VERTEX_SHADER_ARB)); + gSkinnedObjectShinySimpleProgram.mShaderFiles.push_back(make_pair("objects/shinyF.glsl", GL_FRAGMENT_SHADER_ARB)); + gSkinnedObjectShinySimpleProgram.mShaderLevel = mVertexShaderLevel[SHADER_OBJECT]; + success = gSkinnedObjectShinySimpleProgram.createShader(NULL, &mShinyUniforms); + } + if( !success ) { mVertexShaderLevel[SHADER_OBJECT] = 0; diff --git a/indra/newview/llviewershadermgr.h b/indra/newview/llviewershadermgr.h index b279a59777..beac5462e2 100644 --- a/indra/newview/llviewershadermgr.h +++ b/indra/newview/llviewershadermgr.h @@ -315,6 +315,7 @@ extern LLGLSLShader gObjectShinyProgram; extern LLGLSLShader gObjectShinyWaterProgram; extern LLGLSLShader gSkinnedObjectSimpleProgram; +extern LLGLSLShader gSkinnedObjectShinySimpleProgram; //environment shaders extern LLGLSLShader gTerrainProgram; diff --git a/indra/newview/llvoavatar.cpp b/indra/newview/llvoavatar.cpp index 7cdbebf4d1..c51a7d9cbb 100644 --- a/indra/newview/llvoavatar.cpp +++ b/indra/newview/llvoavatar.cpp @@ -3642,7 +3642,7 @@ bool LLVOAvatar::shouldAlphaMask() U32 LLVOAvatar::renderSkinnedAttachments() { - U32 num_indices = 0; + /*U32 num_indices = 0; const U32 data_mask = LLVertexBuffer::MAP_VERTEX | LLVertexBuffer::MAP_NORMAL | @@ -3670,107 +3670,14 @@ U32 LLVOAvatar::renderSkinnedAttachments() LLFace* face = drawable->getFace(i); if (face->isState(LLFace::RIGGED)) { - LLVolume* volume = attached_object->getVolume(); - if (!volume || volume->getNumVolumeFaces() <= i) - { - continue; - } - - const LLVolumeFace& vol_face = volume->getVolumeFace(i); - - const LLMeshSkinInfo* skin = NULL; - LLVertexBuffer* buff = face->mVertexBuffer; - LLUUID mesh_id = volume->getParams().getSculptID();; - - if (!buff || - !buff->hasDataType(LLVertexBuffer::TYPE_WEIGHT4) || - buff->getRequestedVerts() != vol_face.mVertices.size()) - { - face->mVertexBuffer = NULL; - face->mLastVertexBuffer = NULL; - buff = NULL; - - if (mesh_id.notNull()) - { - skin = gMeshRepo.getSkinInfo(mesh_id); - if (skin) - { - face->mVertexBuffer = new LLVertexBuffer(data_mask, 0); - face->mVertexBuffer->allocateBuffer(vol_face.mVertices.size(), vol_face.mIndices.size(), true); - - face->setGeomIndex(0); - face->setIndicesIndex(0); - face->setSize(vol_face.mVertices.size(), vol_face.mIndices.size()); - - U16 offset = 0; - - LLMatrix4 mat_vert = skin->mBindShapeMatrix; - glh::matrix4f m((F32*) mat_vert.mMatrix); - m = m.inverse().transpose(); - - F32 mat3[] = - { m.m[0], m.m[1], m.m[2], - m.m[4], m.m[5], m.m[6], - m.m[8], m.m[9], m.m[10] }; - - LLMatrix3 mat_normal(mat3); - - face->getGeometryVolume(*volume, i, mat_vert, mat_normal, offset, true); - buff = face->mVertexBuffer; - } - } - } - if (buff && mesh_id.notNull()) - { - if (!skin) - { - skin = gMeshRepo.getSkinInfo(mesh_id); - } - - if (skin) - { - LLMatrix4 mat[64]; - - for (U32 i = 0; i < skin->mJointNames.size(); ++i) - { - LLJoint* joint = getJoint(skin->mJointNames[i]); - if (joint) - { - mat[i] = skin->mInvBindMatrix[i]; - mat[i] *= joint->getWorldMatrix(); - } - } - - LLDrawPoolAvatar::sVertexProgram->uniformMatrix4fv("matrixPalette", - skin->mJointNames.size(), - FALSE, - (GLfloat*) mat[0].mMatrix); - LLDrawPoolAvatar::sVertexProgram->uniformMatrix4fv("matrixPalette[0]", - skin->mJointNames.size(), - FALSE, - (GLfloat*) mat[0].mMatrix); - - buff->setBuffer(data_mask); - - U16 start = face->getGeomStart(); - U16 end = start + face->getGeomCount()-1; - S32 offset = face->getIndicesStart(); - U32 count = face->getIndicesCount(); - - gGL.getTexUnit(0)->bind(face->getTexture()); - buff->drawRange(LLRender::TRIANGLES, start, end, count, offset); - - } - } - } - } } } } } - return num_indices; + return num_indices;*/ + return 0; } //----------------------------------------------------------------------------- diff --git a/indra/newview/llvovolume.cpp b/indra/newview/llvovolume.cpp index 96f69b3676..e49b33bd80 100644 --- a/indra/newview/llvovolume.cpp +++ b/indra/newview/llvovolume.cpp @@ -52,6 +52,7 @@ #include "object_flags.h" #include "llagentconstants.h" #include "lldrawable.h" +#include "lldrawpoolavatar.h" #include "lldrawpoolbump.h" #include "llface.h" #include "llspatialpartition.h" @@ -73,6 +74,8 @@ #include "llmeshrepository.h" #include "llagent.h" #include "llviewermediafocus.h" +#include "llvoavatar.h" + const S32 MIN_QUIET_FRAMES_COALESCE = 30; const F32 FORCE_SIMPLE_RENDER_AREA = 512.f; @@ -3415,6 +3418,33 @@ void LLVolumeGeometryManager::getGeometry(LLSpatialGroup* group) static LLFastTimer::DeclareTimer FTM_REBUILD_VOLUME_VB("Volume"); static LLFastTimer::DeclareTimer FTM_REBUILD_VBO("VBO Rebuilt"); +LLDrawPoolAvatar* get_avatar_drawpool(LLViewerObject* vobj) +{ + LLVOAvatar* avatar = vobj->getAvatar(); + + if (avatar) + { + LLDrawable* drawable = avatar->mDrawable; + if (drawable && drawable->getNumFaces() > 0) + { + LLFace* face = drawable->getFace(0); + if (face) + { + LLDrawPool* drawpool = face->getPool(); + if (drawpool) + { + if (drawpool->getType() == LLDrawPool::POOL_AVATAR) + { + return (LLDrawPoolAvatar*) drawpool; + } + } + } + } + } + + return NULL; +} + void LLVolumeGeometryManager::rebuildGeom(LLSpatialGroup* group) { @@ -3499,13 +3529,47 @@ void LLVolumeGeometryManager::rebuildGeom(LLSpatialGroup* group) facep->mVertexBuffer = NULL; facep->mLastVertexBuffer = NULL; facep->setState(LLFace::RIGGED); + + //get drawpool of avatar with rigged face + LLDrawPoolAvatar* pool = get_avatar_drawpool(vobj); + + if (pool) + { + const LLTextureEntry* te = facep->getTextureEntry(); + + //remove face from old pool if it exists + LLDrawPool* old_pool = facep->getPool(); + if (old_pool && old_pool->getType() == LLDrawPool::POOL_AVATAR) + { + ((LLDrawPoolAvatar*) old_pool)->removeRiggedFace(facep); + } + + //add face to new pool + if (te->getShiny()) + { + pool->addRiggedFace(facep, LLDrawPoolAvatar::RIGGED_SHINY_SIMPLE); + } + else + { + pool->addRiggedFace(facep, LLDrawPoolAvatar::RIGGED_SIMPLE); + } + } + } continue; } else { - facep->clearState(LLFace::RIGGED); + if (facep->isState(LLFace::RIGGED)) + { //face is not rigged but used to be, remove from rigged face pool + LLDrawPoolAvatar* pool = (LLDrawPoolAvatar*) facep->getPool(); + if (pool) + { + pool->removeRiggedFace(facep); + } + facep->clearState(LLFace::RIGGED); + } } if (cur_total > max_total || facep->getIndicesCount() <= 0 || facep->getGeomCount() <= 0) -- cgit v1.3 From 2f95a549a365ca2bedf7824014a687b3af88e20f Mon Sep 17 00:00:00 2001 From: Dave Parks Date: Sat, 1 May 2010 01:55:21 -0500 Subject: Fullbrigt skinned and fix for silly crash from not removing face references. --- .../shaders/class1/objects/shinyV.glsl | 2 +- indra/newview/lldrawpoolavatar.cpp | 40 +++++++++++++++++++++- indra/newview/lldrawpoolavatar.h | 5 ++- indra/newview/llface.cpp | 13 ++++--- indra/newview/llviewershadermgr.cpp | 19 ++++++++++ indra/newview/llviewershadermgr.h | 1 + indra/newview/llvovolume.cpp | 9 ++++- 7 files changed, 81 insertions(+), 8 deletions(-) (limited to 'indra/newview/llvovolume.cpp') diff --git a/indra/newview/app_settings/shaders/class1/objects/shinyV.glsl b/indra/newview/app_settings/shaders/class1/objects/shinyV.glsl index c2e1ddf734..101458c438 100644 --- a/indra/newview/app_settings/shaders/class1/objects/shinyV.glsl +++ b/indra/newview/app_settings/shaders/class1/objects/shinyV.glsl @@ -12,7 +12,7 @@ uniform vec4 origin; void main() { //transform vertex - gl_Position = ftransform(); //gl_ModelViewProjectionMatrix * gl_Vertex; + gl_Position = ftransform(); vec4 pos = (gl_ModelViewMatrix * gl_Vertex); vec3 norm = normalize(gl_NormalMatrix * gl_Normal); diff --git a/indra/newview/lldrawpoolavatar.cpp b/indra/newview/lldrawpoolavatar.cpp index 02c7e3bb6f..9463be6059 100644 --- a/indra/newview/lldrawpoolavatar.cpp +++ b/indra/newview/lldrawpoolavatar.cpp @@ -364,7 +364,7 @@ S32 LLDrawPoolAvatar::getNumPasses() } else if (getVertexShaderLevel() > 0) { - return 5; + return 6; } else { @@ -411,6 +411,9 @@ void LLDrawPoolAvatar::beginRenderPass(S32 pass) beginRiggedSimple(); break; case 4: + beginRiggedFullbright(); + break; + case 5: beginRiggedShinySimple(); break; } @@ -441,6 +444,9 @@ void LLDrawPoolAvatar::endRenderPass(S32 pass) endRiggedSimple(); break; case 4: + endRiggedFullbright(); + break; + case 5: endRiggedShinySimple(); break; } @@ -644,6 +650,22 @@ void LLDrawPoolAvatar::endRiggedSimple() LLVertexBuffer::sWeight4Loc = -1; } +void LLDrawPoolAvatar::beginRiggedFullbright() +{ + sVertexProgram = &gSkinnedObjectFullbrightProgram; + diffuse_channel = 0; + gSkinnedObjectFullbrightProgram.bind(); + LLVertexBuffer::sWeight4Loc = gSkinnedObjectFullbrightProgram.getAttribLocation(LLViewerShaderMgr::OBJECT_WEIGHT); +} + +void LLDrawPoolAvatar::endRiggedFullbright() +{ + sVertexProgram = NULL; + LLVertexBuffer::unbind(); + gSkinnedObjectFullbrightProgram.unbind(); + LLVertexBuffer::sWeight4Loc = -1; +} + void LLDrawPoolAvatar::beginRiggedShinySimple() { sVertexProgram = &gSkinnedObjectShinySimpleProgram; @@ -825,6 +847,12 @@ void LLDrawPoolAvatar::renderAvatars(LLVOAvatar* single_avatar, S32 pass) } if (pass == 4) + { + renderRiggedFullbright(avatarp); + return; + } + + if (pass == 5) { renderRiggedShinySimple(avatarp); return; @@ -990,6 +1018,16 @@ void LLDrawPoolAvatar::renderRiggedSimple(LLVOAvatar* avatar) renderRigged(avatar, RIGGED_SIMPLE, data_mask); } +void LLDrawPoolAvatar::renderRiggedFullbright(LLVOAvatar* avatar) +{ + const U32 data_mask = LLVertexBuffer::MAP_VERTEX | + LLVertexBuffer::MAP_TEXCOORD0 | + LLVertexBuffer::MAP_COLOR | + LLVertexBuffer::MAP_WEIGHT4; + + renderRigged(avatar, RIGGED_FULLBRIGHT, data_mask); +} + void LLDrawPoolAvatar::renderRiggedShinySimple(LLVOAvatar* avatar) { diff --git a/indra/newview/lldrawpoolavatar.h b/indra/newview/lldrawpoolavatar.h index 0ebb035f2a..8443069376 100644 --- a/indra/newview/lldrawpoolavatar.h +++ b/indra/newview/lldrawpoolavatar.h @@ -97,12 +97,14 @@ public: void beginImpostor(); void beginSkinned(); void beginRiggedSimple(); + void beginRiggedFullbright(); void beginRiggedShinySimple(); void endRigid(); void endImpostor(); void endSkinned(); void endRiggedSimple(); + void endRiggedFullbright(); void endRiggedShinySimple(); void beginDeferredImpostor(); @@ -123,6 +125,7 @@ public: void renderRigged(LLVOAvatar* avatar, U32 type, const U32 data_mask); void renderRiggedSimple(LLVOAvatar* avatar); + void renderRiggedFullbright(LLVOAvatar* avatar); void renderRiggedShinySimple(LLVOAvatar* avatar); /*virtual*/ LLViewerTexture *getDebugTexture(); @@ -133,11 +136,11 @@ public: typedef enum { RIGGED_SIMPLE = 0, + RIGGED_FULLBRIGHT, RIGGED_SHINY_SIMPLE, RIGGED_SHINY_FULLBRIGHT, RIGGED_SHINY_BUMP, RIGGED_BUMP, - RIGGED_FULLBRIGHT, RIGGED_ALPHA, NUM_RIGGED_PASSES, RIGGED_UNKNOWN, diff --git a/indra/newview/llface.cpp b/indra/newview/llface.cpp index 0e0b8447ca..9df692e787 100644 --- a/indra/newview/llface.cpp +++ b/indra/newview/llface.cpp @@ -41,6 +41,7 @@ #include "m3math.h" #include "v3color.h" +#include "lldrawpoolavatar.h" #include "lldrawpoolbump.h" #include "llgl.h" #include "llrender.h" @@ -207,10 +208,14 @@ void LLFace::destroy() LLFastTimer t(FTM_DESTROY_DRAWPOOL); if (this->isState(LLFace::RIGGED) && mDrawPoolp->getType() == LLDrawPool::POOL_AVATAR) - mDrawPoolp->removeFace(this); - - - + { + ((LLDrawPoolAvatar*) mDrawPoolp)->removeRiggedFace(this); + } + else + { + mDrawPoolp->removeFace(this); + } + mDrawPoolp = NULL; } diff --git a/indra/newview/llviewershadermgr.cpp b/indra/newview/llviewershadermgr.cpp index fe68d6eaa4..3cc5b4357a 100644 --- a/indra/newview/llviewershadermgr.cpp +++ b/indra/newview/llviewershadermgr.cpp @@ -79,6 +79,7 @@ LLGLSLShader gObjectShinyWaterProgram; //object hardware skinning shaders LLGLSLShader gSkinnedObjectSimpleProgram; +LLGLSLShader gSkinnedObjectFullbrightProgram; LLGLSLShader gSkinnedObjectShinySimpleProgram; //environment shaders @@ -155,6 +156,7 @@ LLViewerShaderMgr::LLViewerShaderMgr() : mShaderList.push_back(&gObjectFullbrightProgram); mShaderList.push_back(&gObjectFullbrightShinyProgram); mShaderList.push_back(&gSkinnedObjectSimpleProgram); + mShaderList.push_back(&gSkinnedObjectFullbrightProgram); mShaderList.push_back(&gSkinnedObjectShinySimpleProgram); mShaderList.push_back(&gTerrainProgram); mShaderList.push_back(&gTerrainWaterProgram); @@ -562,6 +564,7 @@ void LLViewerShaderMgr::unloadShaders() gObjectShinyWaterProgram.unload(); gSkinnedObjectSimpleProgram.unload(); + gSkinnedObjectFullbrightProgram.unload(); gSkinnedObjectShinySimpleProgram.unload(); @@ -1258,6 +1261,7 @@ BOOL LLViewerShaderMgr::loadShadersObject() gObjectFullbrightProgram.unload(); gObjectFullbrightWaterProgram.unload(); gSkinnedObjectSimpleProgram.unload(); + gSkinnedObjectFullbrightProgram.unload(); gSkinnedObjectShinySimpleProgram.unload(); return FALSE; @@ -1385,6 +1389,21 @@ BOOL LLViewerShaderMgr::loadShadersObject() success = gSkinnedObjectSimpleProgram.createShader(NULL, NULL); } + if (success) + { + gSkinnedObjectFullbrightProgram.mName = "Skinned Fullbright Shader"; + gSkinnedObjectFullbrightProgram.mFeatures.calculatesAtmospherics = true; + gSkinnedObjectFullbrightProgram.mFeatures.hasGamma = true; + gSkinnedObjectFullbrightProgram.mFeatures.hasTransport = true; + gSkinnedObjectFullbrightProgram.mFeatures.isFullbright = true; + gSkinnedObjectFullbrightProgram.mFeatures.hasObjectSkinning = true; + gSkinnedObjectFullbrightProgram.mShaderFiles.clear(); + gSkinnedObjectFullbrightProgram.mShaderFiles.push_back(make_pair("objects/fullbrightSkinnedV.glsl", GL_VERTEX_SHADER_ARB)); + gSkinnedObjectFullbrightProgram.mShaderFiles.push_back(make_pair("objects/fullbrightF.glsl", GL_FRAGMENT_SHADER_ARB)); + gSkinnedObjectFullbrightProgram.mShaderLevel = mVertexShaderLevel[SHADER_OBJECT]; + success = gSkinnedObjectFullbrightProgram.createShader(NULL, NULL); + } + if (success) { gSkinnedObjectShinySimpleProgram.mName = "Skinned Shiny Simple Shader"; diff --git a/indra/newview/llviewershadermgr.h b/indra/newview/llviewershadermgr.h index beac5462e2..b4fc336bf3 100644 --- a/indra/newview/llviewershadermgr.h +++ b/indra/newview/llviewershadermgr.h @@ -315,6 +315,7 @@ extern LLGLSLShader gObjectShinyProgram; extern LLGLSLShader gObjectShinyWaterProgram; extern LLGLSLShader gSkinnedObjectSimpleProgram; +extern LLGLSLShader gSkinnedObjectFullbrightProgram; extern LLGLSLShader gSkinnedObjectShinySimpleProgram; //environment shaders diff --git a/indra/newview/llvovolume.cpp b/indra/newview/llvovolume.cpp index e49b33bd80..3449c05be8 100644 --- a/indra/newview/llvovolume.cpp +++ b/indra/newview/llvovolume.cpp @@ -3551,7 +3551,14 @@ void LLVolumeGeometryManager::rebuildGeom(LLSpatialGroup* group) } else { - pool->addRiggedFace(facep, LLDrawPoolAvatar::RIGGED_SIMPLE); + if (te->getFullbright()) + { + pool->addRiggedFace(facep, LLDrawPoolAvatar::RIGGED_FULLBRIGHT); + } + else + { + pool->addRiggedFace(facep, LLDrawPoolAvatar::RIGGED_SIMPLE); + } } } -- cgit v1.3 From dc15fc5ba0fcf9907d834b523e6622d800d7ed4e Mon Sep 17 00:00:00 2001 From: Dave Parks Date: Sat, 1 May 2010 02:56:23 -0500 Subject: Fullbright shiny skinned. --- indra/newview/lldrawpoolavatar.cpp | 45 +++++++++++++++++++++++++++++++++++-- indra/newview/lldrawpoolavatar.h | 11 +++++---- indra/newview/llviewershadermgr.cpp | 20 +++++++++++++++++ indra/newview/llviewershadermgr.h | 1 + indra/newview/llvovolume.cpp | 9 +++++++- 5 files changed, 79 insertions(+), 7 deletions(-) (limited to 'indra/newview/llvovolume.cpp') diff --git a/indra/newview/lldrawpoolavatar.cpp b/indra/newview/lldrawpoolavatar.cpp index 9463be6059..8227e8ede2 100644 --- a/indra/newview/lldrawpoolavatar.cpp +++ b/indra/newview/lldrawpoolavatar.cpp @@ -364,7 +364,7 @@ S32 LLDrawPoolAvatar::getNumPasses() } else if (getVertexShaderLevel() > 0) { - return 6; + return 7; } else { @@ -416,6 +416,9 @@ void LLDrawPoolAvatar::beginRenderPass(S32 pass) case 5: beginRiggedShinySimple(); break; + case 6: + beginRiggedFullbrightShiny(); + break; } } @@ -449,6 +452,9 @@ void LLDrawPoolAvatar::endRenderPass(S32 pass) case 5: endRiggedShinySimple(); break; + case 6: + endRiggedFullbrightShiny(); + break; } } @@ -683,6 +689,24 @@ void LLDrawPoolAvatar::endRiggedShinySimple() LLVertexBuffer::sWeight4Loc = -1; } +void LLDrawPoolAvatar::beginRiggedFullbrightShiny() +{ + sVertexProgram = &gSkinnedObjectFullbrightShinyProgram; + sVertexProgram->bind(); + LLDrawPoolBump::bindCubeMap(sVertexProgram, 2, diffuse_channel, cube_channel, false); + LLVertexBuffer::sWeight4Loc = sVertexProgram->getAttribLocation(LLViewerShaderMgr::OBJECT_WEIGHT); +} + +void LLDrawPoolAvatar::endRiggedFullbrightShiny() +{ + LLVertexBuffer::unbind(); + LLDrawPoolBump::unbindCubeMap(sVertexProgram, 2, diffuse_channel, cube_channel, false); + sVertexProgram->unbind(); + sVertexProgram = NULL; + LLVertexBuffer::sWeight4Loc = -1; +} + + void LLDrawPoolAvatar::beginDeferredRigged() { sVertexProgram = &gDeferredSkinnedDiffuseProgram; @@ -858,6 +882,12 @@ void LLDrawPoolAvatar::renderAvatars(LLVOAvatar* single_avatar, S32 pass) return; } + if (pass == 6) + { + renderRiggedFullbrightShiny(avatarp); + return; + } + if (sShaderLevel > 0) { @@ -1037,7 +1067,18 @@ void LLDrawPoolAvatar::renderRiggedShinySimple(LLVOAvatar* avatar) LLVertexBuffer::MAP_COLOR | LLVertexBuffer::MAP_WEIGHT4; - renderRigged(avatar, RIGGED_SHINY_SIMPLE, data_mask); + renderRigged(avatar, RIGGED_SHINY, data_mask); +} + +void LLDrawPoolAvatar::renderRiggedFullbrightShiny(LLVOAvatar* avatar) +{ + const U32 data_mask = LLVertexBuffer::MAP_VERTEX | + LLVertexBuffer::MAP_NORMAL | + LLVertexBuffer::MAP_TEXCOORD0 | + LLVertexBuffer::MAP_COLOR | + LLVertexBuffer::MAP_WEIGHT4; + + renderRigged(avatar, RIGGED_FULLBRIGHT_SHINY, data_mask); } //----------------------------------------------------------------------------- diff --git a/indra/newview/lldrawpoolavatar.h b/indra/newview/lldrawpoolavatar.h index 8443069376..d758bb07f0 100644 --- a/indra/newview/lldrawpoolavatar.h +++ b/indra/newview/lldrawpoolavatar.h @@ -98,6 +98,7 @@ public: void beginSkinned(); void beginRiggedSimple(); void beginRiggedFullbright(); + void beginRiggedFullbrightShiny(); void beginRiggedShinySimple(); void endRigid(); @@ -105,6 +106,7 @@ public: void endSkinned(); void endRiggedSimple(); void endRiggedFullbright(); + void endRiggedFullbrightShiny(); void endRiggedShinySimple(); void beginDeferredImpostor(); @@ -127,6 +129,7 @@ public: void renderRiggedSimple(LLVOAvatar* avatar); void renderRiggedFullbright(LLVOAvatar* avatar); void renderRiggedShinySimple(LLVOAvatar* avatar); + void renderRiggedFullbrightShiny(LLVOAvatar* avatar); /*virtual*/ LLViewerTexture *getDebugTexture(); /*virtual*/ LLColor3 getDebugColor() const; // For AGP debug display @@ -137,11 +140,11 @@ public: { RIGGED_SIMPLE = 0, RIGGED_FULLBRIGHT, - RIGGED_SHINY_SIMPLE, - RIGGED_SHINY_FULLBRIGHT, - RIGGED_SHINY_BUMP, - RIGGED_BUMP, + RIGGED_SHINY, + RIGGED_FULLBRIGHT_SHINY, + RIGGED_GLOW, RIGGED_ALPHA, + RIGGED_FULLBRIGHT_ALPHA, NUM_RIGGED_PASSES, RIGGED_UNKNOWN, } eRiggedPass; diff --git a/indra/newview/llviewershadermgr.cpp b/indra/newview/llviewershadermgr.cpp index 3cc5b4357a..eafc52748e 100644 --- a/indra/newview/llviewershadermgr.cpp +++ b/indra/newview/llviewershadermgr.cpp @@ -80,6 +80,7 @@ LLGLSLShader gObjectShinyWaterProgram; //object hardware skinning shaders LLGLSLShader gSkinnedObjectSimpleProgram; LLGLSLShader gSkinnedObjectFullbrightProgram; +LLGLSLShader gSkinnedObjectFullbrightShinyProgram; LLGLSLShader gSkinnedObjectShinySimpleProgram; //environment shaders @@ -157,6 +158,7 @@ LLViewerShaderMgr::LLViewerShaderMgr() : mShaderList.push_back(&gObjectFullbrightShinyProgram); mShaderList.push_back(&gSkinnedObjectSimpleProgram); mShaderList.push_back(&gSkinnedObjectFullbrightProgram); + mShaderList.push_back(&gSkinnedObjectFullbrightShinyProgram); mShaderList.push_back(&gSkinnedObjectShinySimpleProgram); mShaderList.push_back(&gTerrainProgram); mShaderList.push_back(&gTerrainWaterProgram); @@ -565,6 +567,7 @@ void LLViewerShaderMgr::unloadShaders() gSkinnedObjectSimpleProgram.unload(); gSkinnedObjectFullbrightProgram.unload(); + gSkinnedObjectFullbrightShinyProgram.unload(); gSkinnedObjectShinySimpleProgram.unload(); @@ -1262,6 +1265,7 @@ BOOL LLViewerShaderMgr::loadShadersObject() gObjectFullbrightWaterProgram.unload(); gSkinnedObjectSimpleProgram.unload(); gSkinnedObjectFullbrightProgram.unload(); + gSkinnedObjectFullbrightShinyProgram.unload(); gSkinnedObjectShinySimpleProgram.unload(); return FALSE; @@ -1404,6 +1408,22 @@ BOOL LLViewerShaderMgr::loadShadersObject() success = gSkinnedObjectFullbrightProgram.createShader(NULL, NULL); } + if (success) + { + gSkinnedObjectFullbrightShinyProgram.mName = "Skinned Fullbright Shiny Shader"; + gSkinnedObjectFullbrightShinyProgram.mFeatures.calculatesAtmospherics = true; + gSkinnedObjectFullbrightShinyProgram.mFeatures.hasGamma = true; + gSkinnedObjectFullbrightShinyProgram.mFeatures.hasTransport = true; + gSkinnedObjectFullbrightShinyProgram.mFeatures.isShiny = true; + gSkinnedObjectFullbrightShinyProgram.mFeatures.isFullbright = true; + gSkinnedObjectFullbrightShinyProgram.mFeatures.hasObjectSkinning = true; + gSkinnedObjectFullbrightShinyProgram.mShaderFiles.clear(); + gSkinnedObjectFullbrightShinyProgram.mShaderFiles.push_back(make_pair("objects/fullbrightShinySkinnedV.glsl", GL_VERTEX_SHADER_ARB)); + gSkinnedObjectFullbrightShinyProgram.mShaderFiles.push_back(make_pair("objects/fullbrightShinyF.glsl", GL_FRAGMENT_SHADER_ARB)); + gSkinnedObjectFullbrightShinyProgram.mShaderLevel = mVertexShaderLevel[SHADER_OBJECT]; + success = gSkinnedObjectFullbrightShinyProgram.createShader(NULL, &mShinyUniforms); + } + if (success) { gSkinnedObjectShinySimpleProgram.mName = "Skinned Shiny Simple Shader"; diff --git a/indra/newview/llviewershadermgr.h b/indra/newview/llviewershadermgr.h index b4fc336bf3..bb28cd7ec2 100644 --- a/indra/newview/llviewershadermgr.h +++ b/indra/newview/llviewershadermgr.h @@ -316,6 +316,7 @@ extern LLGLSLShader gObjectShinyWaterProgram; extern LLGLSLShader gSkinnedObjectSimpleProgram; extern LLGLSLShader gSkinnedObjectFullbrightProgram; +extern LLGLSLShader gSkinnedObjectFullbrightShinyProgram; extern LLGLSLShader gSkinnedObjectShinySimpleProgram; //environment shaders diff --git a/indra/newview/llvovolume.cpp b/indra/newview/llvovolume.cpp index 3449c05be8..43252c553c 100644 --- a/indra/newview/llvovolume.cpp +++ b/indra/newview/llvovolume.cpp @@ -3547,7 +3547,14 @@ void LLVolumeGeometryManager::rebuildGeom(LLSpatialGroup* group) //add face to new pool if (te->getShiny()) { - pool->addRiggedFace(facep, LLDrawPoolAvatar::RIGGED_SHINY_SIMPLE); + if (te->getFullbright()) + { + pool->addRiggedFace(facep, LLDrawPoolAvatar::RIGGED_FULLBRIGHT_SHINY); + } + else + { + pool->addRiggedFace(facep, LLDrawPoolAvatar::RIGGED_SHINY); + } } else { -- cgit v1.3 From 58defe76b0eed6f773b961e8ba5cd3aeb856467a Mon Sep 17 00:00:00 2001 From: Dave Parks Date: Sun, 2 May 2010 00:28:30 -0500 Subject: Rigged glow. --- indra/newview/lldrawpoolavatar.cpp | 214 +++++++++++++++++++++++++++++++------ indra/newview/lldrawpoolavatar.h | 13 ++- indra/newview/llface.h | 2 + indra/newview/llvovolume.cpp | 21 +++- 4 files changed, 212 insertions(+), 38 deletions(-) (limited to 'indra/newview/llvovolume.cpp') diff --git a/indra/newview/lldrawpoolavatar.cpp b/indra/newview/lldrawpoolavatar.cpp index 8227e8ede2..866aea42c4 100644 --- a/indra/newview/lldrawpoolavatar.cpp +++ b/indra/newview/lldrawpoolavatar.cpp @@ -364,7 +364,7 @@ S32 LLDrawPoolAvatar::getNumPasses() } else if (getVertexShaderLevel() > 0) { - return 7; + return 10; } else { @@ -419,6 +419,15 @@ void LLDrawPoolAvatar::beginRenderPass(S32 pass) case 6: beginRiggedFullbrightShiny(); break; + case 7: + beginRiggedAlpha(); + break; + case 8: + beginRiggedFullbrightAlpha(); + break; + case 9: + beginRiggedGlow(); + break; } } @@ -455,6 +464,15 @@ void LLDrawPoolAvatar::endRenderPass(S32 pass) case 6: endRiggedFullbrightShiny(); break; + case 7: + endRiggedAlpha(); + break; + case 8: + endRiggedFullbrightAlpha(); + break; + case 9: + endRiggedGlow(); + break; } } @@ -656,6 +674,55 @@ void LLDrawPoolAvatar::endRiggedSimple() LLVertexBuffer::sWeight4Loc = -1; } +void LLDrawPoolAvatar::beginRiggedAlpha() +{ + sVertexProgram = &gSkinnedObjectSimpleProgram; + diffuse_channel = 0; + sVertexProgram->bind(); + LLVertexBuffer::sWeight4Loc = sVertexProgram->getAttribLocation(LLViewerShaderMgr::OBJECT_WEIGHT); +} + +void LLDrawPoolAvatar::endRiggedAlpha() +{ + sVertexProgram->unbind(); + sVertexProgram = NULL; + LLVertexBuffer::unbind(); + LLVertexBuffer::sWeight4Loc = -1; +} + + +void LLDrawPoolAvatar::beginRiggedFullbrightAlpha() +{ + sVertexProgram = &gSkinnedObjectFullbrightProgram; + diffuse_channel = 0; + sVertexProgram->bind(); + LLVertexBuffer::sWeight4Loc = sVertexProgram->getAttribLocation(LLViewerShaderMgr::OBJECT_WEIGHT); +} + +void LLDrawPoolAvatar::endRiggedFullbrightAlpha() +{ + sVertexProgram->unbind(); + sVertexProgram = NULL; + LLVertexBuffer::unbind(); + LLVertexBuffer::sWeight4Loc = -1; +} + +void LLDrawPoolAvatar::beginRiggedGlow() +{ + sVertexProgram = &gSkinnedObjectFullbrightProgram; + diffuse_channel = 0; + sVertexProgram->bind(); + LLVertexBuffer::sWeight4Loc = sVertexProgram->getAttribLocation(LLViewerShaderMgr::OBJECT_WEIGHT); +} + +void LLDrawPoolAvatar::endRiggedGlow() +{ + sVertexProgram->unbind(); + sVertexProgram = NULL; + LLVertexBuffer::unbind(); + LLVertexBuffer::sWeight4Loc = -1; +} + void LLDrawPoolAvatar::beginRiggedFullbright() { sVertexProgram = &gSkinnedObjectFullbrightProgram; @@ -888,6 +955,48 @@ void LLDrawPoolAvatar::renderAvatars(LLVOAvatar* single_avatar, S32 pass) return; } + if (pass >= 7 && pass <= 9) + { + LLGLEnable blend(GL_BLEND); + + gGL.setColorMask(true, true); + gGL.blendFunc(LLRender::BF_SOURCE_ALPHA, + LLRender::BF_ONE_MINUS_SOURCE_ALPHA, + LLRender::BF_ZERO, + LLRender::BF_ONE_MINUS_SOURCE_ALPHA); + + + if (pass == 7) + { + renderRiggedAlpha(avatarp); + return; + } + + if (pass == 8) + { + renderRiggedFullbrightAlpha(avatarp); + return; + } + } + + if (pass == 9) + { + LLGLEnable blend(GL_BLEND); + LLGLDisable test(GL_ALPHA_TEST); + gGL.flush(); + + LLGLEnable polyOffset(GL_POLYGON_OFFSET_FILL); + glPolygonOffset(-1.0f, -1.0f); + gGL.setSceneBlendType(LLRender::BT_ADD); + + LLGLDepthTest depth(GL_TRUE, GL_FALSE); + gGL.setColorMask(false, true); + + renderRiggedGlow(avatarp); + gGL.setColorMask(true, false); + gGL.setSceneBlendType(LLRender::BT_ALPHA); + } + if (sShaderLevel > 0) { @@ -958,7 +1067,7 @@ void LLDrawPoolAvatar::updateRiggedFaceVertexBuffer(LLFace* face, const LLMeshSk } } -void LLDrawPoolAvatar::renderRigged(LLVOAvatar* avatar, U32 type, const U32 data_mask) +void LLDrawPoolAvatar::renderRigged(LLVOAvatar* avatar, U32 type, const U32 data_mask, bool glow) { for (U32 i = 0; i < mRiggedFace[type].size(); ++i) { @@ -1031,6 +1140,11 @@ void LLDrawPoolAvatar::renderRigged(LLVOAvatar* avatar, U32 type, const U32 data S32 offset = face->getIndicesStart(); U32 count = face->getIndicesCount(); + if (glow) + { + glColor4f(0,0,0,face->getTextureEntry()->getGlow()); + } + gGL.getTexUnit(0)->bind(face->getTexture()); buff->drawRange(LLRender::TRIANGLES, start, end, count, offset); } @@ -1081,6 +1195,39 @@ void LLDrawPoolAvatar::renderRiggedFullbrightShiny(LLVOAvatar* avatar) renderRigged(avatar, RIGGED_FULLBRIGHT_SHINY, data_mask); } +void LLDrawPoolAvatar::renderRiggedAlpha(LLVOAvatar* avatar) +{ + const U32 data_mask = LLVertexBuffer::MAP_VERTEX | + LLVertexBuffer::MAP_NORMAL | + LLVertexBuffer::MAP_TEXCOORD0 | + LLVertexBuffer::MAP_COLOR | + LLVertexBuffer::MAP_WEIGHT4; + + renderRigged(avatar, RIGGED_ALPHA, data_mask); +} + +void LLDrawPoolAvatar::renderRiggedFullbrightAlpha(LLVOAvatar* avatar) +{ + const U32 data_mask = LLVertexBuffer::MAP_VERTEX | + LLVertexBuffer::MAP_TEXCOORD0 | + LLVertexBuffer::MAP_COLOR | + LLVertexBuffer::MAP_WEIGHT4; + + renderRigged(avatar, RIGGED_FULLBRIGHT_ALPHA, data_mask); +} + +void LLDrawPoolAvatar::renderRiggedGlow(LLVOAvatar* avatar) +{ + const U32 data_mask = LLVertexBuffer::MAP_VERTEX | + LLVertexBuffer::MAP_TEXCOORD0 | + LLVertexBuffer::MAP_WEIGHT4; + + renderRigged(avatar, RIGGED_GLOW, data_mask, true); +} + + + + //----------------------------------------------------------------------------- // renderForSelect() //----------------------------------------------------------------------------- @@ -1181,60 +1328,57 @@ LLColor3 LLDrawPoolAvatar::getDebugColor() const void LLDrawPoolAvatar::addRiggedFace(LLFace* facep, U32 type) { - if (facep->getReferenceIndex() != -1) + if (facep->mRiggedIndex.empty()) { - llerrs << "Tried to add a rigged face that's referenced elsewhere." << llendl; - } + facep->mRiggedIndex.resize(LLDrawPoolAvatar::NUM_RIGGED_PASSES); + for (U32 i = 0; i < facep->mRiggedIndex.size(); ++i) + { + facep->mRiggedIndex[i] = -1; + } + } if (type >= NUM_RIGGED_PASSES) { llerrs << "Invalid rigged face type." << llendl; } - facep->setReferenceIndex(mRiggedFace[type].size()); + if (facep->mRiggedIndex[type] != -1) + { + llerrs << "Tried to add a rigged face that's referenced elsewhere." << llendl; + } + + + facep->mRiggedIndex[type] = mRiggedFace[type].size(); facep->mDrawPoolp = this; mRiggedFace[type].push_back(facep); } -void LLDrawPoolAvatar::removeRiggedFace(LLFace* facep, U32 type) +void LLDrawPoolAvatar::removeRiggedFace(LLFace* facep) { - S32 index = facep->getReferenceIndex(); - if (index == -1) - { - llerrs << "Tried to remove rigged face with invalid index." << llendl; - } - - if (type > RIGGED_UNKNOWN) - { - llerrs << "Invalid rigged face type." << llendl; - } - - facep->setReferenceIndex(-1); + facep->mDrawPoolp = NULL; - if (type == RIGGED_UNKNOWN) + for (U32 i = 0; i < NUM_RIGGED_PASSES; ++i) { - for (U32 i = 0; i < NUM_RIGGED_PASSES; ++i) + S32 index = facep->mRiggedIndex[i]; + + if (index > -1) { if (mRiggedFace[i].size() > index && mRiggedFace[i][index] == facep) { - type = i; - break; + facep->mRiggedIndex[i] = -1; + mRiggedFace[i].erase(mRiggedFace[i].begin()+index); + for (U32 j = index; j < mRiggedFace[i].size(); ++j) + { //bump indexes down for faces referenced after erased face + mRiggedFace[i][j]->mRiggedIndex[i] = j; + } + } + else + { + llerrs << "Face reference data corrupt for rigged type " << i << llendl; } } } - - if (type >= NUM_RIGGED_PASSES) - { - llerrs << "Could not find face for removal from current drawpool." << llendl; - } - - mRiggedFace[type].erase(mRiggedFace[type].begin()+index); - - for (S32 i = index; i < mRiggedFace[type].size(); ++i) - { //bump indexes of currently held faces down after removal - mRiggedFace[type][i]->setReferenceIndex(i); - } } LLVertexBufferAvatar::LLVertexBufferAvatar() diff --git a/indra/newview/lldrawpoolavatar.h b/indra/newview/lldrawpoolavatar.h index d758bb07f0..59f9cf7ddb 100644 --- a/indra/newview/lldrawpoolavatar.h +++ b/indra/newview/lldrawpoolavatar.h @@ -100,6 +100,9 @@ public: void beginRiggedFullbright(); void beginRiggedFullbrightShiny(); void beginRiggedShinySimple(); + void beginRiggedAlpha(); + void beginRiggedFullbrightAlpha(); + void beginRiggedGlow(); void endRigid(); void endImpostor(); @@ -108,6 +111,9 @@ public: void endRiggedFullbright(); void endRiggedFullbrightShiny(); void endRiggedShinySimple(); + void endRiggedAlpha(); + void endRiggedFullbrightAlpha(); + void endRiggedGlow(); void beginDeferredImpostor(); void beginDeferredRigid(); @@ -125,11 +131,14 @@ public: const LLVolumeFace& vol_face, U32 data_mask); - void renderRigged(LLVOAvatar* avatar, U32 type, const U32 data_mask); + void renderRigged(LLVOAvatar* avatar, U32 type, const U32 data_mask, bool glow = false); void renderRiggedSimple(LLVOAvatar* avatar); + void renderRiggedAlpha(LLVOAvatar* avatar); + void renderRiggedFullbrightAlpha(LLVOAvatar* avatar); void renderRiggedFullbright(LLVOAvatar* avatar); void renderRiggedShinySimple(LLVOAvatar* avatar); void renderRiggedFullbrightShiny(LLVOAvatar* avatar); + void renderRiggedGlow(LLVOAvatar* avatar); /*virtual*/ LLViewerTexture *getDebugTexture(); /*virtual*/ LLColor3 getDebugColor() const; // For AGP debug display @@ -151,7 +160,7 @@ public: void addRiggedFace(LLFace* facep, U32 type); - void removeRiggedFace(LLFace* facep, U32 type = RIGGED_UNKNOWN); + void removeRiggedFace(LLFace* facep); std::vector mRiggedFace[NUM_RIGGED_PASSES]; diff --git a/indra/newview/llface.h b/indra/newview/llface.h index 2b8fdf2e58..f9e9c3e078 100644 --- a/indra/newview/llface.h +++ b/indra/newview/llface.h @@ -260,6 +260,8 @@ private: S32 mTEOffset; S32 mReferenceIndex; + std::vector mRiggedIndex; + F32 mVSize; F32 mPixelArea; diff --git a/indra/newview/llvovolume.cpp b/indra/newview/llvovolume.cpp index 43252c553c..e263f8d937 100644 --- a/indra/newview/llvovolume.cpp +++ b/indra/newview/llvovolume.cpp @@ -3545,7 +3545,21 @@ void LLVolumeGeometryManager::rebuildGeom(LLSpatialGroup* group) } //add face to new pool - if (te->getShiny()) + LLViewerTexture* tex = facep->getTexture(); + U32 type = gPipeline.getPoolTypeFromTE(te, tex); + + if (type == LLDrawPool::POOL_ALPHA) + { + if (te->getFullbright()) + { + pool->addRiggedFace(facep, LLDrawPoolAvatar::RIGGED_FULLBRIGHT_ALPHA); + } + else + { + pool->addRiggedFace(facep, LLDrawPoolAvatar::RIGGED_ALPHA); + } + } + else if (te->getShiny()) { if (te->getFullbright()) { @@ -3567,6 +3581,11 @@ void LLVolumeGeometryManager::rebuildGeom(LLSpatialGroup* group) pool->addRiggedFace(facep, LLDrawPoolAvatar::RIGGED_SIMPLE); } } + + if (te->getGlow()) + { + pool->addRiggedFace(facep, LLDrawPoolAvatar::RIGGED_GLOW); + } } } -- cgit v1.3 From eb283701afc7ecbe3009a9fb75be1dcb222a383b Mon Sep 17 00:00:00 2001 From: Dave Parks Date: Tue, 4 May 2010 00:45:28 -0500 Subject: Deferred pipeline integration of rigged attachments and cleanup. --- indra/llrender/llvertexbuffer.h | 2 +- .../shaders/class1/deferred/alphaF.glsl | 5 +- .../shaders/class1/deferred/alphaV.glsl | 2 +- .../shaders/class1/deferred/avatarAlphaV.glsl | 4 +- .../shaders/class1/deferred/diffuseV.glsl | 2 +- .../shaders/class2/deferred/alphaF.glsl | 2 +- indra/newview/lldrawpoolavatar.cpp | 302 +++++++++++++++------ indra/newview/lldrawpoolavatar.h | 51 +++- indra/newview/lldrawpoolbump.cpp | 27 +- indra/newview/lldrawpoolbump.h | 7 +- indra/newview/llviewershadermgr.cpp | 36 ++- indra/newview/llviewershadermgr.h | 2 + indra/newview/llvovolume.cpp | 15 + 13 files changed, 359 insertions(+), 98 deletions(-) (limited to 'indra/newview/llvovolume.cpp') diff --git a/indra/llrender/llvertexbuffer.h b/indra/llrender/llvertexbuffer.h index 225237215c..d1700aa54a 100644 --- a/indra/llrender/llvertexbuffer.h +++ b/indra/llrender/llvertexbuffer.h @@ -190,7 +190,7 @@ public: U8* getIndicesPointer() const { return useVBOs() ? NULL : mMappedIndexData; } U8* getVerticesPointer() const { return useVBOs() ? NULL : mMappedData; } S32 getStride() const { return mStride; } - S32 getTypeMask() const { return mTypeMask; } + U32 getTypeMask() const { return mTypeMask; } BOOL hasDataType(S32 type) const { return ((1 << type) & getTypeMask()) ? TRUE : FALSE; } S32 getSize() const { return mNumVerts*mStride; } S32 getIndicesSize() const { return mNumIndices * sizeof(U16); } diff --git a/indra/newview/app_settings/shaders/class1/deferred/alphaF.glsl b/indra/newview/app_settings/shaders/class1/deferred/alphaF.glsl index fea2e16090..6f027de6a2 100644 --- a/indra/newview/app_settings/shaders/class1/deferred/alphaF.glsl +++ b/indra/newview/app_settings/shaders/class1/deferred/alphaF.glsl @@ -22,7 +22,6 @@ varying vec3 vary_ambient; varying vec3 vary_directional; varying vec3 vary_fragcoord; varying vec3 vary_position; -varying vec3 vary_light; uniform mat4 inv_proj; @@ -55,8 +54,8 @@ void main() color.rgb = scaleSoftClip(color.rgb); - //gl_FragColor = gl_Color; - gl_FragColor = color; + gl_FragColor = gl_Color; + //gl_FragColor = color; //gl_FragColor = vec4(1,0,1,1); //gl_FragColor = vec4(1,0,1,1)*shadow; diff --git a/indra/newview/app_settings/shaders/class1/deferred/alphaV.glsl b/indra/newview/app_settings/shaders/class1/deferred/alphaV.glsl index 04e556c11a..43200f1b07 100644 --- a/indra/newview/app_settings/shaders/class1/deferred/alphaV.glsl +++ b/indra/newview/app_settings/shaders/class1/deferred/alphaV.glsl @@ -36,7 +36,7 @@ void main() vec4 pos = (gl_ModelViewMatrix * gl_Vertex); vec3 norm = normalize(gl_NormalMatrix * gl_Normal); - vary_position = pos.xyz + norm.xyz * (-pos.z/64.0*shadow_offset+shadow_bias); + vary_position = pos.xyz; calcAtmospherics(pos.xyz); diff --git a/indra/newview/app_settings/shaders/class1/deferred/avatarAlphaV.glsl b/indra/newview/app_settings/shaders/class1/deferred/avatarAlphaV.glsl index 650fbcc3f5..da1dafda36 100644 --- a/indra/newview/app_settings/shaders/class1/deferred/avatarAlphaV.glsl +++ b/indra/newview/app_settings/shaders/class1/deferred/avatarAlphaV.glsl @@ -17,7 +17,7 @@ vec3 atmosAffectDirectionalLight(float lightIntensity); vec3 scaleDownLight(vec3 light); vec3 scaleUpLight(vec3 light); -varying vec4 vary_position; +varying vec3 vary_position; varying vec3 vary_ambient; varying vec3 vary_directional; varying vec3 vary_normal; @@ -41,7 +41,7 @@ void main() norm = normalize(norm); gl_Position = gl_ProjectionMatrix * pos; - vary_position = pos; + vary_position = pos.xyz; vary_normal = norm; calcAtmospherics(pos.xyz); diff --git a/indra/newview/app_settings/shaders/class1/deferred/diffuseV.glsl b/indra/newview/app_settings/shaders/class1/deferred/diffuseV.glsl index b458842657..44468cdfa2 100644 --- a/indra/newview/app_settings/shaders/class1/deferred/diffuseV.glsl +++ b/indra/newview/app_settings/shaders/class1/deferred/diffuseV.glsl @@ -13,7 +13,7 @@ void main() gl_Position = gl_ModelViewProjectionMatrix * gl_Vertex; gl_TexCoord[0] = gl_TextureMatrix[0] * gl_MultiTexCoord0; - vary_nomral = normalize(gl_NormalMatrix * gl_Normal); + vary_normal = normalize(gl_NormalMatrix * gl_Normal); gl_FrontColor = gl_Color; } diff --git a/indra/newview/app_settings/shaders/class2/deferred/alphaF.glsl b/indra/newview/app_settings/shaders/class2/deferred/alphaF.glsl index 665fe16b43..fa6b4e2afb 100644 --- a/indra/newview/app_settings/shaders/class2/deferred/alphaF.glsl +++ b/indra/newview/app_settings/shaders/class2/deferred/alphaF.glsl @@ -115,7 +115,7 @@ void main() //gl_FragColor = gl_Color; gl_FragColor = color; - //gl_FragColor = vec4(1,0,1,1)*shadow; + //gl_FragColor = vec4(1,shadow,1,1); } diff --git a/indra/newview/lldrawpoolavatar.cpp b/indra/newview/lldrawpoolavatar.cpp index 866aea42c4..ac599caa5b 100644 --- a/indra/newview/lldrawpoolavatar.cpp +++ b/indra/newview/lldrawpoolavatar.cpp @@ -63,6 +63,8 @@ LLGLSLShader* LLDrawPoolAvatar::sVertexProgram = NULL; BOOL LLDrawPoolAvatar::sSkipOpaque = FALSE; BOOL LLDrawPoolAvatar::sSkipTransparent = FALSE; +static bool is_deferred_render = false; + extern BOOL gUseGLPick; F32 CLOTHING_GRAVITY_EFFECT = 0.7f; @@ -100,6 +102,18 @@ S32 specular_channel = -1; S32 diffuse_channel = -1; S32 cube_channel = -1; +static const U32 rigged_data_mask[] = { + LLDrawPoolAvatar::RIGGED_SIMPLE_MASK, + LLDrawPoolAvatar::RIGGED_FULLBRIGHT_MASK, + LLDrawPoolAvatar::RIGGED_SHINY_MASK, + LLDrawPoolAvatar::RIGGED_FULLBRIGHT_SHINY_MASK, + LLDrawPoolAvatar::RIGGED_GLOW_MASK, + LLDrawPoolAvatar::RIGGED_ALPHA_MASK, + LLDrawPoolAvatar::RIGGED_FULLBRIGHT_ALPHA_MASK, + LLDrawPoolAvatar::RIGGED_DEFERRED_BUMP_MASK, + LLDrawPoolAvatar::RIGGED_DEFERRED_SIMPLE_MASK, +}; + static LLFastTimer::DeclareTimer FTM_SHADOW_AVATAR("Avatar Shadow"); @@ -155,21 +169,17 @@ LLMatrix4& LLDrawPoolAvatar::getModelView() //----------------------------------------------------------------------------- -S32 LLDrawPoolAvatar::getNumDeferredPasses() -{ - return getNumPasses(); -} void LLDrawPoolAvatar::beginDeferredPass(S32 pass) { LLFastTimer t(FTM_RENDER_CHARACTERS); sSkipTransparent = TRUE; - + is_deferred_render = true; + if (LLPipeline::sImpostorRender) - { - beginDeferredSkinned(); - return; + { //impostor pass does not have rigid or impostor rendering + pass += 2; } switch (pass) @@ -184,7 +194,10 @@ void LLDrawPoolAvatar::beginDeferredPass(S32 pass) beginDeferredSkinned(); break; case 3: - beginDeferredRigged(); + beginDeferredRiggedSimple(); + break; + case 4: + beginDeferredRiggedBump(); break; } } @@ -194,11 +207,11 @@ void LLDrawPoolAvatar::endDeferredPass(S32 pass) LLFastTimer t(FTM_RENDER_CHARACTERS); sSkipTransparent = FALSE; + is_deferred_render = false; if (LLPipeline::sImpostorRender) { - endDeferredSkinned(); - return; + pass += 2; } switch (pass) @@ -213,7 +226,11 @@ void LLDrawPoolAvatar::endDeferredPass(S32 pass) endDeferredSkinned(); break; case 3: - endDeferredRigged(); + endDeferredRiggedSimple(); + break; + case 4: + endDeferredRiggedBump(); + break; } } @@ -224,10 +241,35 @@ void LLDrawPoolAvatar::renderDeferred(S32 pass) S32 LLDrawPoolAvatar::getNumPostDeferredPasses() { - return 1; + return 6; } void LLDrawPoolAvatar::beginPostDeferredPass(S32 pass) +{ + switch (pass) + { + case 0: + beginPostDeferredAlpha(); + break; + case 1: + beginRiggedFullbright(); + break; + case 2: + beginRiggedFullbrightShiny(); + break; + case 3: + beginDeferredRiggedAlpha(); + break; + case 4: + beginRiggedFullbrightAlpha(); + break; + case 5: + beginRiggedGlow(); + break; + } +} + +void LLDrawPoolAvatar::beginPostDeferredAlpha() { sSkipOpaque = TRUE; sShaderLevel = mVertexShaderLevel; @@ -240,7 +282,49 @@ void LLDrawPoolAvatar::beginPostDeferredPass(S32 pass) enable_vertex_weighting(sVertexProgram->mAttribute[LLViewerShaderMgr::AVATAR_WEIGHT]); } +void LLDrawPoolAvatar::beginDeferredRiggedAlpha() +{ + sVertexProgram = &gDeferredSkinnedAlphaProgram; + gPipeline.bindDeferredShader(*sVertexProgram); + diffuse_channel = sVertexProgram->enableTexture(LLViewerShaderMgr::DIFFUSE_MAP); + LLVertexBuffer::sWeight4Loc = sVertexProgram->getAttribLocation(LLViewerShaderMgr::OBJECT_WEIGHT); + gPipeline.enableLightsDynamic(); +} + +void LLDrawPoolAvatar::endDeferredRiggedAlpha() +{ + LLVertexBuffer::unbind(); + gPipeline.unbindDeferredShader(*sVertexProgram); + LLVertexBuffer::sWeight4Loc = -1; + sVertexProgram = NULL; +} + void LLDrawPoolAvatar::endPostDeferredPass(S32 pass) +{ + switch (pass) + { + case 0: + endPostDeferredAlpha(); + break; + case 1: + endRiggedFullbright(); + break; + case 2: + endRiggedFullbrightShiny(); + break; + case 3: + endDeferredRiggedAlpha(); + break; + case 4: + endRiggedFullbrightAlpha(); + break; + case 5: + endRiggedGlow(); + break; + } +} + +void LLDrawPoolAvatar::endPostDeferredAlpha() { // if we're in software-blending, remember to set the fence _after_ we draw so we wait till this rendering is done sRenderingSkinned = FALSE; @@ -254,7 +338,17 @@ void LLDrawPoolAvatar::endPostDeferredPass(S32 pass) void LLDrawPoolAvatar::renderPostDeferred(S32 pass) { - render(2); //pass 2 = skinned + const S32 actual_pass[] = + { //map post deferred pass numbers to what render() expects + 2, //skinned + 4, // rigged fullbright + 6, //rigged fullbright shiny + 7, //rigged alpha + 8, //rigged fullbright alpha + 9, //rigged glow + }; + + render(actual_pass[pass]); } @@ -288,6 +382,7 @@ void LLDrawPoolAvatar::beginShadowPass(S32 pass) else { sVertexProgram = &gDeferredAttachmentShadowProgram; + diffuse_channel = sVertexProgram->enableTexture(LLViewerShaderMgr::DIFFUSE_MAP); sVertexProgram->bind(); LLVertexBuffer::sWeight4Loc = sVertexProgram->getAttribLocation(LLViewerShaderMgr::OBJECT_WEIGHT); } @@ -352,7 +447,12 @@ void LLDrawPoolAvatar::renderShadow(S32 pass) } else { - avatarp->renderSkinnedAttachments(); + renderRigged(avatarp, RIGGED_SIMPLE); + renderRigged(avatarp, RIGGED_ALPHA); + renderRigged(avatarp, RIGGED_FULLBRIGHT); + renderRigged(avatarp, RIGGED_FULLBRIGHT_SHINY); + renderRigged(avatarp, RIGGED_SHINY); + renderRigged(avatarp, RIGGED_FULLBRIGHT_ALPHA); } } @@ -360,7 +460,7 @@ S32 LLDrawPoolAvatar::getNumPasses() { if (LLPipeline::sImpostorRender) { - return 1; + return 8; } else if (getVertexShaderLevel() > 0) { @@ -372,6 +472,19 @@ S32 LLDrawPoolAvatar::getNumPasses() } } +S32 LLDrawPoolAvatar::getNumDeferredPasses() +{ + if (LLPipeline::sImpostorRender) + { + return 3; + } + else + { + return 5; + } +} + + void LLDrawPoolAvatar::render(S32 pass) { LLFastTimer t(FTM_RENDER_CHARACTERS); @@ -391,9 +504,8 @@ void LLDrawPoolAvatar::beginRenderPass(S32 pass) LLVertexBuffer::unbind(); if (LLPipeline::sImpostorRender) - { - beginSkinned(); - return; + { //impostor render does not have impostors or rigid rendering + pass += 2; } switch (pass) @@ -437,8 +549,7 @@ void LLDrawPoolAvatar::endRenderPass(S32 pass) if (LLPipeline::sImpostorRender) { - endSkinned(); - return; + pass += 2; } switch (pass) @@ -536,8 +647,8 @@ void LLDrawPoolAvatar::beginDeferredImpostor() sVertexProgram = &gDeferredImpostorProgram; - normal_channel = sVertexProgram->enableTexture(LLViewerShaderMgr::DEFERRED_NORMAL); specular_channel = sVertexProgram->enableTexture(LLViewerShaderMgr::SPECULAR_MAP); + normal_channel = sVertexProgram->enableTexture(LLViewerShaderMgr::DEFERRED_NORMAL); diffuse_channel = sVertexProgram->enableTexture(LLViewerShaderMgr::DIFFUSE_MAP); sVertexProgram->bind(); @@ -774,18 +885,40 @@ void LLDrawPoolAvatar::endRiggedFullbrightShiny() } -void LLDrawPoolAvatar::beginDeferredRigged() +void LLDrawPoolAvatar::beginDeferredRiggedSimple() { sVertexProgram = &gDeferredSkinnedDiffuseProgram; + diffuse_channel = 0; + sVertexProgram->bind(); + LLVertexBuffer::sWeight4Loc = sVertexProgram->getAttribLocation(LLViewerShaderMgr::OBJECT_WEIGHT); +} + +void LLDrawPoolAvatar::endDeferredRiggedSimple() +{ + LLVertexBuffer::unbind(); + sVertexProgram->unbind(); + LLVertexBuffer::sWeight4Loc = -1; + sVertexProgram = NULL; +} + +void LLDrawPoolAvatar::beginDeferredRiggedBump() +{ + sVertexProgram = &gDeferredSkinnedBumpProgram; sVertexProgram->bind(); + normal_channel = sVertexProgram->enableTexture(LLViewerShaderMgr::BUMP_MAP); + diffuse_channel = sVertexProgram->enableTexture(LLViewerShaderMgr::DIFFUSE_MAP); LLVertexBuffer::sWeight4Loc = sVertexProgram->getAttribLocation(LLViewerShaderMgr::OBJECT_WEIGHT); } -void LLDrawPoolAvatar::endDeferredRigged() +void LLDrawPoolAvatar::endDeferredRiggedBump() { LLVertexBuffer::unbind(); + sVertexProgram->disableTexture(LLViewerShaderMgr::BUMP_MAP); + sVertexProgram->disableTexture(LLViewerShaderMgr::DIFFUSE_MAP); sVertexProgram->unbind(); LLVertexBuffer::sWeight4Loc = -1; + normal_channel = -1; + diffuse_channel = -1; sVertexProgram = NULL; } @@ -933,13 +1066,28 @@ void LLDrawPoolAvatar::renderAvatars(LLVOAvatar* single_avatar, S32 pass) if (pass == 3) { - renderRiggedSimple(avatarp); + if (is_deferred_render) + { + renderDeferredRiggedSimple(avatarp); + } + else + { + renderRiggedSimple(avatarp); + } return; } if (pass == 4) { - renderRiggedFullbright(avatarp); + if (is_deferred_render) + { + renderDeferredRiggedBump(avatarp); + } + else + { + renderRiggedFullbright(avatarp); + } + return; } @@ -995,6 +1143,7 @@ void LLDrawPoolAvatar::renderAvatars(LLVOAvatar* single_avatar, S32 pass) renderRiggedGlow(avatarp); gGL.setColorMask(true, false); gGL.setSceneBlendType(LLRender::BT_ALPHA); + return; } @@ -1034,12 +1183,21 @@ void LLDrawPoolAvatar::renderAvatars(LLVOAvatar* single_avatar, S32 pass) } } -void LLDrawPoolAvatar::updateRiggedFaceVertexBuffer(LLFace* face, const LLMeshSkinInfo* skin, LLVolume* volume, const LLVolumeFace& vol_face, U32 data_mask) +void LLDrawPoolAvatar::updateRiggedFaceVertexBuffer(LLFace* face, const LLMeshSkinInfo* skin, LLVolume* volume, const LLVolumeFace& vol_face) { + U32 data_mask = 0; + for (U32 i = 0; i < face->mRiggedIndex.size(); ++i) + { + if (face->mRiggedIndex[i] > -1) + { + data_mask |= rigged_data_mask[i]; + } + } + LLVertexBuffer* buff = face->mVertexBuffer; if (!buff || - !buff->hasDataType(LLVertexBuffer::TYPE_WEIGHT4) || + buff->getTypeMask() != data_mask || buff->getRequestedVerts() != vol_face.mVertices.size()) { face->setGeomIndex(0); @@ -1067,7 +1225,7 @@ void LLDrawPoolAvatar::updateRiggedFaceVertexBuffer(LLFace* face, const LLMeshSk } } -void LLDrawPoolAvatar::renderRigged(LLVOAvatar* avatar, U32 type, const U32 data_mask, bool glow) +void LLDrawPoolAvatar::renderRigged(LLVOAvatar* avatar, U32 type, bool glow) { for (U32 i = 0; i < mRiggedFace[type].size(); ++i) { @@ -1106,8 +1264,10 @@ void LLDrawPoolAvatar::renderRigged(LLVOAvatar* avatar, U32 type, const U32 data } const LLVolumeFace& vol_face = volume->getVolumeFace(te); - updateRiggedFaceVertexBuffer(face, skin, volume, vol_face, data_mask); + updateRiggedFaceVertexBuffer(face, skin, volume, vol_face); + U32 data_mask = rigged_data_mask[type]; + LLVertexBuffer* buff = face->mVertexBuffer; if (buff) @@ -1145,84 +1305,72 @@ void LLDrawPoolAvatar::renderRigged(LLVOAvatar* avatar, U32 type, const U32 data glColor4f(0,0,0,face->getTextureEntry()->getGlow()); } - gGL.getTexUnit(0)->bind(face->getTexture()); - buff->drawRange(LLRender::TRIANGLES, start, end, count, offset); + gGL.getTexUnit(diffuse_channel)->bind(face->getTexture()); + if (normal_channel > -1) + { + LLDrawPoolBump::bindBumpMap(face, normal_channel); + } + + if (face->mTextureMatrix) + { + glMatrixMode(GL_TEXTURE); + glLoadMatrixf((F32*) face->mTextureMatrix->mMatrix); + buff->drawRange(LLRender::TRIANGLES, start, end, count, offset); + glLoadIdentity(); + glMatrixMode(GL_MODELVIEW); + } + else + { + buff->drawRange(LLRender::TRIANGLES, start, end, count, offset); + } } } } -void LLDrawPoolAvatar::renderRiggedSimple(LLVOAvatar* avatar) +void LLDrawPoolAvatar::renderDeferredRiggedSimple(LLVOAvatar* avatar) { - const U32 data_mask = LLVertexBuffer::MAP_VERTEX | - LLVertexBuffer::MAP_NORMAL | - LLVertexBuffer::MAP_TEXCOORD0 | - LLVertexBuffer::MAP_COLOR | - LLVertexBuffer::MAP_WEIGHT4; + renderRigged(avatar, RIGGED_DEFERRED_SIMPLE); +} - renderRigged(avatar, RIGGED_SIMPLE, data_mask); +void LLDrawPoolAvatar::renderDeferredRiggedBump(LLVOAvatar* avatar) +{ + renderRigged(avatar, RIGGED_DEFERRED_BUMP); } -void LLDrawPoolAvatar::renderRiggedFullbright(LLVOAvatar* avatar) +void LLDrawPoolAvatar::renderRiggedSimple(LLVOAvatar* avatar) { - const U32 data_mask = LLVertexBuffer::MAP_VERTEX | - LLVertexBuffer::MAP_TEXCOORD0 | - LLVertexBuffer::MAP_COLOR | - LLVertexBuffer::MAP_WEIGHT4; + renderRigged(avatar, RIGGED_SIMPLE); +} - renderRigged(avatar, RIGGED_FULLBRIGHT, data_mask); +void LLDrawPoolAvatar::renderRiggedFullbright(LLVOAvatar* avatar) +{ + renderRigged(avatar, RIGGED_FULLBRIGHT); } void LLDrawPoolAvatar::renderRiggedShinySimple(LLVOAvatar* avatar) { - const U32 data_mask = LLVertexBuffer::MAP_VERTEX | - LLVertexBuffer::MAP_NORMAL | - LLVertexBuffer::MAP_TEXCOORD0 | - LLVertexBuffer::MAP_COLOR | - LLVertexBuffer::MAP_WEIGHT4; - - renderRigged(avatar, RIGGED_SHINY, data_mask); + renderRigged(avatar, RIGGED_SHINY); } void LLDrawPoolAvatar::renderRiggedFullbrightShiny(LLVOAvatar* avatar) { - const U32 data_mask = LLVertexBuffer::MAP_VERTEX | - LLVertexBuffer::MAP_NORMAL | - LLVertexBuffer::MAP_TEXCOORD0 | - LLVertexBuffer::MAP_COLOR | - LLVertexBuffer::MAP_WEIGHT4; - - renderRigged(avatar, RIGGED_FULLBRIGHT_SHINY, data_mask); + renderRigged(avatar, RIGGED_FULLBRIGHT_SHINY); } void LLDrawPoolAvatar::renderRiggedAlpha(LLVOAvatar* avatar) { - const U32 data_mask = LLVertexBuffer::MAP_VERTEX | - LLVertexBuffer::MAP_NORMAL | - LLVertexBuffer::MAP_TEXCOORD0 | - LLVertexBuffer::MAP_COLOR | - LLVertexBuffer::MAP_WEIGHT4; - - renderRigged(avatar, RIGGED_ALPHA, data_mask); + renderRigged(avatar, RIGGED_ALPHA); } void LLDrawPoolAvatar::renderRiggedFullbrightAlpha(LLVOAvatar* avatar) { - const U32 data_mask = LLVertexBuffer::MAP_VERTEX | - LLVertexBuffer::MAP_TEXCOORD0 | - LLVertexBuffer::MAP_COLOR | - LLVertexBuffer::MAP_WEIGHT4; - - renderRigged(avatar, RIGGED_FULLBRIGHT_ALPHA, data_mask); + renderRigged(avatar, RIGGED_FULLBRIGHT_ALPHA); } void LLDrawPoolAvatar::renderRiggedGlow(LLVOAvatar* avatar) { - const U32 data_mask = LLVertexBuffer::MAP_VERTEX | - LLVertexBuffer::MAP_TEXCOORD0 | - LLVertexBuffer::MAP_WEIGHT4; - - renderRigged(avatar, RIGGED_GLOW, data_mask, true); + renderRigged(avatar, RIGGED_GLOW, true); } diff --git a/indra/newview/lldrawpoolavatar.h b/indra/newview/lldrawpoolavatar.h index 59f9cf7ddb..bab6f01480 100644 --- a/indra/newview/lldrawpoolavatar.h +++ b/indra/newview/lldrawpoolavatar.h @@ -103,6 +103,8 @@ public: void beginRiggedAlpha(); void beginRiggedFullbrightAlpha(); void beginRiggedGlow(); + void beginPostDeferredAlpha(); + void beginDeferredRiggedAlpha(); void endRigid(); void endImpostor(); @@ -114,24 +116,27 @@ public: void endRiggedAlpha(); void endRiggedFullbrightAlpha(); void endRiggedGlow(); + void endPostDeferredAlpha(); + void endDeferredRiggedAlpha(); void beginDeferredImpostor(); void beginDeferredRigid(); void beginDeferredSkinned(); - void beginDeferredRigged(); + void beginDeferredRiggedSimple(); + void beginDeferredRiggedBump(); void endDeferredImpostor(); void endDeferredRigid(); void endDeferredSkinned(); - void endDeferredRigged(); + void endDeferredRiggedSimple(); + void endDeferredRiggedBump(); void updateRiggedFaceVertexBuffer(LLFace* facep, const LLMeshSkinInfo* skin, LLVolume* volume, - const LLVolumeFace& vol_face, - U32 data_mask); + const LLVolumeFace& vol_face); - void renderRigged(LLVOAvatar* avatar, U32 type, const U32 data_mask, bool glow = false); + void renderRigged(LLVOAvatar* avatar, U32 type, bool glow = false); void renderRiggedSimple(LLVOAvatar* avatar); void renderRiggedAlpha(LLVOAvatar* avatar); void renderRiggedFullbrightAlpha(LLVOAvatar* avatar); @@ -139,6 +144,8 @@ public: void renderRiggedShinySimple(LLVOAvatar* avatar); void renderRiggedFullbrightShiny(LLVOAvatar* avatar); void renderRiggedGlow(LLVOAvatar* avatar); + void renderDeferredRiggedSimple(LLVOAvatar* avatar); + void renderDeferredRiggedBump(LLVOAvatar* avatar); /*virtual*/ LLViewerTexture *getDebugTexture(); /*virtual*/ LLColor3 getDebugColor() const; // For AGP debug display @@ -154,11 +161,43 @@ public: RIGGED_GLOW, RIGGED_ALPHA, RIGGED_FULLBRIGHT_ALPHA, + RIGGED_DEFERRED_BUMP, + RIGGED_DEFERRED_SIMPLE, NUM_RIGGED_PASSES, RIGGED_UNKNOWN, } eRiggedPass; - + typedef enum + { + RIGGED_SIMPLE_MASK = LLVertexBuffer::MAP_VERTEX | + LLVertexBuffer::MAP_NORMAL | + LLVertexBuffer::MAP_TEXCOORD0 | + LLVertexBuffer::MAP_COLOR | + LLVertexBuffer::MAP_WEIGHT4, + RIGGED_FULLBRIGHT_MASK = LLVertexBuffer::MAP_VERTEX | + LLVertexBuffer::MAP_TEXCOORD0 | + LLVertexBuffer::MAP_COLOR | + LLVertexBuffer::MAP_WEIGHT4, + RIGGED_SHINY_MASK = RIGGED_SIMPLE_MASK, + RIGGED_FULLBRIGHT_SHINY_MASK = RIGGED_SIMPLE_MASK, + RIGGED_GLOW_MASK = LLVertexBuffer::MAP_VERTEX | + LLVertexBuffer::MAP_TEXCOORD0 | + LLVertexBuffer::MAP_WEIGHT4, + RIGGED_ALPHA_MASK = RIGGED_SIMPLE_MASK, + RIGGED_FULLBRIGHT_ALPHA_MASK = RIGGED_FULLBRIGHT_MASK, + RIGGED_DEFERRED_BUMP_MASK = LLVertexBuffer::MAP_VERTEX | + LLVertexBuffer::MAP_NORMAL | + LLVertexBuffer::MAP_TEXCOORD0 | + LLVertexBuffer::MAP_BINORMAL | + LLVertexBuffer::MAP_COLOR | + LLVertexBuffer::MAP_WEIGHT4, + RIGGED_DEFERRED_SIMPLE_MASK = LLVertexBuffer::MAP_VERTEX | + LLVertexBuffer::MAP_NORMAL | + LLVertexBuffer::MAP_TEXCOORD0 | + LLVertexBuffer::MAP_COLOR | + LLVertexBuffer::MAP_WEIGHT4, + } eRiggedDataMask; + void addRiggedFace(LLFace* facep, U32 type); void removeRiggedFace(LLFace* facep); diff --git a/indra/newview/lldrawpoolbump.cpp b/indra/newview/lldrawpoolbump.cpp index 2f449fa42f..906615ade8 100644 --- a/indra/newview/lldrawpoolbump.cpp +++ b/indra/newview/lldrawpoolbump.cpp @@ -591,18 +591,37 @@ void LLDrawPoolBump::renderGroup(LLSpatialGroup* group, U32 type, U32 mask, BOOL // static BOOL LLDrawPoolBump::bindBumpMap(LLDrawInfo& params, S32 channel) { - LLViewerTexture* bump = NULL; - U8 bump_code = params.mBump; + return bindBumpMap(bump_code, params.mTexture, params.mVSize, channel); +} + +//static +BOOL LLDrawPoolBump::bindBumpMap(LLFace* face, S32 channel) +{ + const LLTextureEntry* te = face->getTextureEntry(); + if (te) + { + U8 bump_code = te->getBumpmap(); + return bindBumpMap(bump_code, face->getTexture(), face->getVirtualSize(), channel); + } + + return FALSE; +} + +//static +BOOL LLDrawPoolBump::bindBumpMap(U8 bump_code, LLViewerTexture* texture, F32 vsize, S32 channel) +{ //Note: texture atlas does not support bump texture now. - LLViewerFetchedTexture* tex = LLViewerTextureManager::staticCastToFetchedTexture(params.mTexture) ; + LLViewerFetchedTexture* tex = LLViewerTextureManager::staticCastToFetchedTexture(texture) ; if(!tex) { //if the texture is not a fetched texture return FALSE; } + LLViewerTexture* bump = NULL; + switch( bump_code ) { case BE_NO_BUMP: @@ -616,7 +635,7 @@ BOOL LLDrawPoolBump::bindBumpMap(LLDrawInfo& params, S32 channel) if( bump_code < LLStandardBumpmap::sStandardBumpmapCount ) { bump = gStandardBumpmapList[bump_code].mImage; - gBumpImageList.addTextureStats(bump_code, tex->getID(), params.mVSize); + gBumpImageList.addTextureStats(bump_code, tex->getID(), vsize); } break; } diff --git a/indra/newview/lldrawpoolbump.h b/indra/newview/lldrawpoolbump.h index 89bbefe778..127c9efe85 100644 --- a/indra/newview/lldrawpoolbump.h +++ b/indra/newview/lldrawpoolbump.h @@ -93,7 +93,12 @@ public: /*virtual*/ void endPostDeferredPass(S32 pass); /*virtual*/ void renderPostDeferred(S32 pass); - BOOL bindBumpMap(LLDrawInfo& params, S32 channel = -2); + static BOOL bindBumpMap(LLDrawInfo& params, S32 channel = -2); + static BOOL bindBumpMap(LLFace* face, S32 channel = -2); + +private: + static BOOL bindBumpMap(U8 bump_code, LLViewerTexture* tex, F32 vsize, S32 channel); + }; enum EBumpEffect diff --git a/indra/newview/llviewershadermgr.cpp b/indra/newview/llviewershadermgr.cpp index eafc52748e..e64fdfc0b4 100644 --- a/indra/newview/llviewershadermgr.cpp +++ b/indra/newview/llviewershadermgr.cpp @@ -114,6 +114,8 @@ LLGLSLShader gDeferredEdgeProgram; LLGLSLShader gDeferredWaterProgram; LLGLSLShader gDeferredDiffuseProgram; LLGLSLShader gDeferredSkinnedDiffuseProgram; +LLGLSLShader gDeferredSkinnedBumpProgram; +LLGLSLShader gDeferredSkinnedAlphaProgram; LLGLSLShader gDeferredBumpProgram; LLGLSLShader gDeferredTerrainProgram; LLGLSLShader gDeferredTreeProgram; @@ -173,6 +175,7 @@ LLViewerShaderMgr::LLViewerShaderMgr() : mShaderList.push_back(&gDeferredLightProgram); mShaderList.push_back(&gDeferredMultiLightProgram); mShaderList.push_back(&gDeferredAlphaProgram); + mShaderList.push_back(&gDeferredSkinnedAlphaProgram); mShaderList.push_back(&gDeferredFullbrightProgram); mShaderList.push_back(&gDeferredPostGIProgram); mShaderList.push_back(&gDeferredEdgeProgram); @@ -591,6 +594,8 @@ void LLViewerShaderMgr::unloadShaders() gDeferredDiffuseProgram.unload(); gDeferredSkinnedDiffuseProgram.unload(); + gDeferredSkinnedBumpProgram.unload(); + gDeferredSkinnedAlphaProgram.unload(); mVertexShaderLevel[SHADER_LIGHTING] = 0; mVertexShaderLevel[SHADER_OBJECT] = 0; @@ -906,6 +911,8 @@ BOOL LLViewerShaderMgr::loadShadersDeferred() gDeferredTreeProgram.unload(); gDeferredDiffuseProgram.unload(); gDeferredSkinnedDiffuseProgram.unload(); + gDeferredSkinnedBumpProgram.unload(); + gDeferredSkinnedAlphaProgram.unload(); gDeferredBumpProgram.unload(); gDeferredImpostorProgram.unload(); gDeferredTerrainProgram.unload(); @@ -958,6 +965,33 @@ BOOL LLViewerShaderMgr::loadShadersDeferred() success = gDeferredSkinnedDiffuseProgram.createShader(NULL, NULL); } + if (success) + { + gDeferredSkinnedBumpProgram.mName = "Deferred Skinned Bump Shader"; + gDeferredSkinnedBumpProgram.mFeatures.hasObjectSkinning = true; + gDeferredSkinnedBumpProgram.mShaderFiles.clear(); + gDeferredSkinnedBumpProgram.mShaderFiles.push_back(make_pair("deferred/bumpSkinnedV.glsl", GL_VERTEX_SHADER_ARB)); + gDeferredSkinnedBumpProgram.mShaderFiles.push_back(make_pair("deferred/bumpF.glsl", GL_FRAGMENT_SHADER_ARB)); + gDeferredSkinnedBumpProgram.mShaderLevel = mVertexShaderLevel[SHADER_DEFERRED]; + success = gDeferredSkinnedBumpProgram.createShader(NULL, NULL); + } + + if (success) + { + gDeferredSkinnedAlphaProgram.mName = "Deferred Skinned Alpha Shader"; + gDeferredSkinnedAlphaProgram.mFeatures.hasObjectSkinning = true; + gDeferredSkinnedAlphaProgram.mFeatures.calculatesLighting = true; + gDeferredSkinnedAlphaProgram.mFeatures.calculatesAtmospherics = true; + gDeferredSkinnedAlphaProgram.mFeatures.hasGamma = true; + gDeferredSkinnedAlphaProgram.mFeatures.hasAtmospherics = true; + gDeferredSkinnedAlphaProgram.mFeatures.hasLighting = true; + gDeferredSkinnedAlphaProgram.mShaderFiles.clear(); + gDeferredSkinnedAlphaProgram.mShaderFiles.push_back(make_pair("deferred/alphaSkinnedV.glsl", GL_VERTEX_SHADER_ARB)); + gDeferredSkinnedAlphaProgram.mShaderFiles.push_back(make_pair("deferred/alphaF.glsl", GL_FRAGMENT_SHADER_ARB)); + gDeferredSkinnedAlphaProgram.mShaderLevel = mVertexShaderLevel[SHADER_DEFERRED]; + success = gDeferredSkinnedAlphaProgram.createShader(NULL, NULL); + } + if (success) { gDeferredBumpProgram.mName = "Deferred Bump Shader"; @@ -1176,7 +1210,7 @@ BOOL LLViewerShaderMgr::loadShadersDeferred() gDeferredAvatarAlphaProgram.mFeatures.hasLighting = true; gDeferredAvatarAlphaProgram.mShaderFiles.clear(); gDeferredAvatarAlphaProgram.mShaderFiles.push_back(make_pair("deferred/avatarAlphaV.glsl", GL_VERTEX_SHADER_ARB)); - gDeferredAvatarAlphaProgram.mShaderFiles.push_back(make_pair("deferred/avatarAlphaF.glsl", GL_FRAGMENT_SHADER_ARB)); + gDeferredAvatarAlphaProgram.mShaderFiles.push_back(make_pair("deferred/alphaF.glsl", GL_FRAGMENT_SHADER_ARB)); gDeferredAvatarAlphaProgram.mShaderLevel = mVertexShaderLevel[SHADER_DEFERRED]; success = gDeferredAvatarAlphaProgram.createShader(&mAvatarAttribs, &mAvatarUniforms); } diff --git a/indra/newview/llviewershadermgr.h b/indra/newview/llviewershadermgr.h index bb28cd7ec2..da16a38427 100644 --- a/indra/newview/llviewershadermgr.h +++ b/indra/newview/llviewershadermgr.h @@ -350,6 +350,8 @@ extern LLGLSLShader gDeferredEdgeProgram; extern LLGLSLShader gDeferredWaterProgram; extern LLGLSLShader gDeferredDiffuseProgram; extern LLGLSLShader gDeferredSkinnedDiffuseProgram; +extern LLGLSLShader gDeferredSkinnedBumpProgram; +extern LLGLSLShader gDeferredSkinnedAlphaProgram; extern LLGLSLShader gDeferredBumpProgram; extern LLGLSLShader gDeferredTerrainProgram; extern LLGLSLShader gDeferredTreeProgram; diff --git a/indra/newview/llvovolume.cpp b/indra/newview/llvovolume.cpp index e263f8d937..220634897e 100644 --- a/indra/newview/llvovolume.cpp +++ b/indra/newview/llvovolume.cpp @@ -3586,6 +3586,21 @@ void LLVolumeGeometryManager::rebuildGeom(LLSpatialGroup* group) { pool->addRiggedFace(facep, LLDrawPoolAvatar::RIGGED_GLOW); } + + if (LLPipeline::sRenderDeferred) + { + if (type != LLDrawPool::POOL_ALPHA && !te->getFullbright()) + { + if (te->getBumpmap()) + { + pool->addRiggedFace(facep, LLDrawPoolAvatar::RIGGED_DEFERRED_BUMP); + } + else + { + pool->addRiggedFace(facep, LLDrawPoolAvatar::RIGGED_DEFERRED_SIMPLE); + } + } + } } } -- cgit v1.3 From ddad6dd5ea88705a0b7db603e1785715593624de Mon Sep 17 00:00:00 2001 From: Tofu Linden Date: Tue, 4 May 2010 14:55:02 +0100 Subject: EXT-7161 Fullbright has no effect on shiny & bump mapped prim while Lighting and Shadows are enabled. (transplanted from 6f25e5a4f0e0d7bbe1ed3e14e7da4d371e290c67) --- indra/newview/llvovolume.cpp | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) (limited to 'indra/newview/llvovolume.cpp') diff --git a/indra/newview/llvovolume.cpp b/indra/newview/llvovolume.cpp index 220634897e..f71c3ce418 100644 --- a/indra/newview/llvovolume.cpp +++ b/indra/newview/llvovolume.cpp @@ -4060,13 +4060,13 @@ void LLVolumeGeometryManager::genDrawInfo(LLSpatialGroup* group, U32 mask, std:: } else if (LLPipeline::sRenderDeferred) { - if (te->getBumpmap()) + if (te->getFullbright()) { - registerFace(group, facep, LLRenderPass::PASS_BUMP); + registerFace(group, facep, LLRenderPass::PASS_FULLBRIGHT_SHINY); } - else if (te->getFullbright()) + else if (te->getBumpmap()) { - registerFace(group, facep, LLRenderPass::PASS_FULLBRIGHT_SHINY); + registerFace(group, facep, LLRenderPass::PASS_BUMP); } else { -- cgit v1.3 From 744f2e4951fc6fc1a1f08e97fc446e909c1fe39e Mon Sep 17 00:00:00 2001 From: Dave Parks Date: Tue, 4 May 2010 12:14:47 -0500 Subject: Cleanup from review. --- indra/newview/app_settings/shaders/class1/deferred/alphaF.glsl | 3 +-- indra/newview/lldrawpoolavatar.cpp | 2 +- indra/newview/llvovolume.cpp | 2 +- 3 files changed, 3 insertions(+), 4 deletions(-) (limited to 'indra/newview/llvovolume.cpp') diff --git a/indra/newview/app_settings/shaders/class1/deferred/alphaF.glsl b/indra/newview/app_settings/shaders/class1/deferred/alphaF.glsl index 6f027de6a2..7d18ae5036 100644 --- a/indra/newview/app_settings/shaders/class1/deferred/alphaF.glsl +++ b/indra/newview/app_settings/shaders/class1/deferred/alphaF.glsl @@ -54,8 +54,7 @@ void main() color.rgb = scaleSoftClip(color.rgb); - gl_FragColor = gl_Color; - //gl_FragColor = color; + gl_FragColor = color; //gl_FragColor = vec4(1,0,1,1); //gl_FragColor = vec4(1,0,1,1)*shadow; diff --git a/indra/newview/lldrawpoolavatar.cpp b/indra/newview/lldrawpoolavatar.cpp index ac599caa5b..a49cf8781e 100644 --- a/indra/newview/lldrawpoolavatar.cpp +++ b/indra/newview/lldrawpoolavatar.cpp @@ -1103,7 +1103,7 @@ void LLDrawPoolAvatar::renderAvatars(LLVOAvatar* single_avatar, S32 pass) return; } - if (pass >= 7 && pass <= 9) + if (pass >= 7 && pass < 9) { LLGLEnable blend(GL_BLEND); diff --git a/indra/newview/llvovolume.cpp b/indra/newview/llvovolume.cpp index 220634897e..1ed8da85a8 100644 --- a/indra/newview/llvovolume.cpp +++ b/indra/newview/llvovolume.cpp @@ -3418,7 +3418,7 @@ void LLVolumeGeometryManager::getGeometry(LLSpatialGroup* group) static LLFastTimer::DeclareTimer FTM_REBUILD_VOLUME_VB("Volume"); static LLFastTimer::DeclareTimer FTM_REBUILD_VBO("VBO Rebuilt"); -LLDrawPoolAvatar* get_avatar_drawpool(LLViewerObject* vobj) +static LLDrawPoolAvatar* get_avatar_drawpool(LLViewerObject* vobj) { LLVOAvatar* avatar = vobj->getAvatar(); -- cgit v1.3 From 9950c22a5568b55ff896734185e0e1181a93fcf0 Mon Sep 17 00:00:00 2001 From: Dave Parks Date: Wed, 12 May 2010 03:59:01 -0500 Subject: Fix for fullbright bump not working in deferred render. Fix for shadow artifacts around split frusta. Tweak for shadow aliasing with projectors. Fix for crash on exit in mesh thread. --- indra/newview/app_settings/settings.xml | 6 +- .../shaders/class2/deferred/sunLightF.glsl | 3 +- indra/newview/lldrawpool.h | 1 + indra/newview/lldrawpoolbump.cpp | 44 ++- indra/newview/lldrawpoolbump.h | 8 +- indra/newview/lldrawpoolsimple.cpp | 3 +- indra/newview/llviewercontrol.cpp | 3 + indra/newview/llviewerdisplay.cpp | 8 +- indra/newview/llvovolume.cpp | 39 +- indra/newview/pipeline.cpp | 394 ++++++++++++++------- indra/newview/pipeline.h | 30 +- 11 files changed, 364 insertions(+), 175 deletions(-) (limited to 'indra/newview/llvovolume.cpp') diff --git a/indra/newview/app_settings/settings.xml b/indra/newview/app_settings/settings.xml index 623f0f3b01..4c72b03b2e 100644 --- a/indra/newview/app_settings/settings.xml +++ b/indra/newview/app_settings/settings.xml @@ -6604,7 +6604,7 @@ Type F32 Value - -0.0005 + 0.0 RenderSpotShadowOffset @@ -6615,7 +6615,7 @@ Type F32 Value - -0.01 + 0.04 RenderShadowResolutionScale @@ -6751,7 +6751,7 @@ Type F32 Value - 1 + 8 RenderDeferred diff --git a/indra/newview/app_settings/shaders/class2/deferred/sunLightF.glsl b/indra/newview/app_settings/shaders/class2/deferred/sunLightF.glsl index 413835515a..794c4d2761 100644 --- a/indra/newview/app_settings/shaders/class2/deferred/sunLightF.glsl +++ b/indra/newview/app_settings/shaders/class2/deferred/sunLightF.glsl @@ -77,7 +77,6 @@ float pcfShadow(sampler2DRectShadow shadowMap, vec4 stc, float scl) float pcfShadow(sampler2DShadow shadowMap, vec4 stc, float scl) { - stc.z += spot_shadow_offset; stc.xyz /= stc.w; stc.z += spot_shadow_bias*scl; @@ -182,7 +181,7 @@ void main() gl_FragColor[0] = shadow; gl_FragColor[1] = 1.0; - spos = vec4(shadow_pos, 1.0); + spos = vec4(shadow_pos+norm*spot_shadow_offset, 1.0); //spotlight shadow 1 vec4 lpos = shadow_matrix[4]*spos; diff --git a/indra/newview/lldrawpool.h b/indra/newview/lldrawpool.h index e46d503db3..ec870f59c6 100644 --- a/indra/newview/lldrawpool.h +++ b/indra/newview/lldrawpool.h @@ -133,6 +133,7 @@ public: PASS_FULLBRIGHT_SHINY, PASS_SHINY, PASS_BUMP, + PASS_POST_BUMP, PASS_GLOW, PASS_ALPHA, PASS_ALPHA_MASK, diff --git a/indra/newview/lldrawpoolbump.cpp b/indra/newview/lldrawpoolbump.cpp index 906615ade8..967ce3fc5c 100644 --- a/indra/newview/lldrawpoolbump.cpp +++ b/indra/newview/lldrawpoolbump.cpp @@ -659,9 +659,9 @@ BOOL LLDrawPoolBump::bindBumpMap(U8 bump_code, LLViewerTexture* texture, F32 vsi } //static -void LLDrawPoolBump::beginBump() +void LLDrawPoolBump::beginBump(U32 pass) { - if (!gPipeline.hasRenderBatches(LLRenderPass::PASS_BUMP)) + if (!gPipeline.hasRenderBatches(pass)) { return; } @@ -704,9 +704,9 @@ void LLDrawPoolBump::beginBump() } //static -void LLDrawPoolBump::renderBump() +void LLDrawPoolBump::renderBump(U32 pass) { - if (!gPipeline.hasRenderBatches(LLRenderPass::PASS_BUMP)) + if (!gPipeline.hasRenderBatches(pass)) { return; } @@ -719,13 +719,13 @@ void LLDrawPoolBump::renderBump() /// Get rid of z-fighting with non-bump pass. LLGLEnable polyOffset(GL_POLYGON_OFFSET_FILL); glPolygonOffset(-1.0f, -1.0f); - renderBump(LLRenderPass::PASS_BUMP, sVertexMask); + renderBump(pass, sVertexMask); } //static -void LLDrawPoolBump::endBump() +void LLDrawPoolBump::endBump(U32 pass) { - if (!gPipeline.hasRenderBatches(LLRenderPass::PASS_BUMP)) + if (!gPipeline.hasRenderBatches(pass)) { return; } @@ -808,17 +808,41 @@ void LLDrawPoolBump::renderDeferred(S32 pass) void LLDrawPoolBump::beginPostDeferredPass(S32 pass) { - beginFullbrightShiny(); + switch (pass) + { + case 0: + beginFullbrightShiny(); + break; + case 1: + beginBump(LLRenderPass::PASS_POST_BUMP); + break; + } } void LLDrawPoolBump::endPostDeferredPass(S32 pass) { - endFullbrightShiny(); + switch (pass) + { + case 0: + endFullbrightShiny(); + break; + case 1: + endBump(LLRenderPass::PASS_POST_BUMP); + break; + } } void LLDrawPoolBump::renderPostDeferred(S32 pass) { - renderFullbrightShiny(); + switch (pass) + { + case 0: + renderFullbrightShiny(); + break; + case 1: + renderBump(LLRenderPass::PASS_POST_BUMP); + break; + } } //////////////////////////////////////////////////////////////// diff --git a/indra/newview/lldrawpoolbump.h b/indra/newview/lldrawpoolbump.h index 127c9efe85..a75653e68e 100644 --- a/indra/newview/lldrawpoolbump.h +++ b/indra/newview/lldrawpoolbump.h @@ -76,9 +76,9 @@ public: void renderFullbrightShiny(); void endFullbrightShiny(); - void beginBump(); - void renderBump(); - void endBump(); + void beginBump(U32 pass = LLRenderPass::PASS_BUMP); + void renderBump(U32 pass = LLRenderPass::PASS_BUMP); + void endBump(U32 pass = LLRenderPass::PASS_BUMP); static void bindCubeMap(LLGLSLShader* shader, S32 shader_level, S32& diffuse_channel, S32& cube_channel, bool invisible); static void unbindCubeMap(LLGLSLShader* shader, S32 shader_level, S32& diffuse_channel, S32& cube_channel, bool invisible); @@ -88,7 +88,7 @@ public: /*virtual*/ void endDeferredPass(S32 pass); /*virtual*/ void renderDeferred(S32 pass); - virtual S32 getNumPostDeferredPasses() { return 1; } + virtual S32 getNumPostDeferredPasses() { return 2; } /*virtual*/ void beginPostDeferredPass(S32 pass); /*virtual*/ void endPostDeferredPass(S32 pass); /*virtual*/ void renderPostDeferred(S32 pass); diff --git a/indra/newview/lldrawpoolsimple.cpp b/indra/newview/lldrawpoolsimple.cpp index c8e6f95ab6..91191287cd 100644 --- a/indra/newview/lldrawpoolsimple.cpp +++ b/indra/newview/lldrawpoolsimple.cpp @@ -151,7 +151,8 @@ void LLDrawPoolSimple::render(S32 pass) renderTexture(LLRenderPass::PASS_SIMPLE, getVertexDataMask()); if (LLPipeline::sRenderDeferred) - { + { //if deferred rendering is enabled, bump faces aren't reigstered as simple + //render bump faces here as simple so bump faces will appear under water renderTexture(LLRenderPass::PASS_BUMP, getVertexDataMask()); } } diff --git a/indra/newview/llviewercontrol.cpp b/indra/newview/llviewercontrol.cpp index a8004f26ed..701df86de4 100644 --- a/indra/newview/llviewercontrol.cpp +++ b/indra/newview/llviewercontrol.cpp @@ -505,6 +505,9 @@ void settings_setup_listeners() gSavedSettings.getControl("RenderAvatarVP")->getSignal()->connect(boost::bind(&handleSetShaderChanged, _2)); gSavedSettings.getControl("VertexShaderEnable")->getSignal()->connect(boost::bind(&handleSetShaderChanged, _2)); gSavedSettings.getControl("RenderUIBuffer")->getSignal()->connect(boost::bind(&handleReleaseGLBufferChanged, _2)); + gSavedSettings.getControl("RenderSpecularResX")->getSignal()->connect(boost::bind(&handleReleaseGLBufferChanged, _2)); + gSavedSettings.getControl("RenderSpecularResY")->getSignal()->connect(boost::bind(&handleReleaseGLBufferChanged, _2)); + gSavedSettings.getControl("RenderSpecularExponent")->getSignal()->connect(boost::bind(&handleReleaseGLBufferChanged, _2)); gSavedSettings.getControl("RenderFSAASamples")->getSignal()->connect(boost::bind(&handleReleaseGLBufferChanged, _2)); gSavedSettings.getControl("RenderShadowResolutionScale")->getSignal()->connect(boost::bind(&handleReleaseGLBufferChanged, _2)); gSavedSettings.getControl("RenderGlow")->getSignal()->connect(boost::bind(&handleReleaseGLBufferChanged, _2)); diff --git a/indra/newview/llviewerdisplay.cpp b/indra/newview/llviewerdisplay.cpp index 360cee9f73..3cd7af6433 100644 --- a/indra/newview/llviewerdisplay.cpp +++ b/indra/newview/llviewerdisplay.cpp @@ -946,9 +946,10 @@ void render_hud_attachments() bool render_particles = gPipeline.hasRenderType(LLPipeline::RENDER_TYPE_PARTICLES) && gSavedSettings.getBOOL("RenderHUDParticles"); //only render hud objects - U32 mask = gPipeline.getRenderTypeMask(); + gPipeline.pushRenderTypeMask(); + // turn off everything - gPipeline.setRenderTypeMask(0); + gPipeline.andRenderTypeMask(LLPipeline::END_RENDER_TYPES); // turn on HUD gPipeline.toggleRenderType(LLPipeline::RENDER_TYPE_HUD); // turn on HUD particles @@ -1002,7 +1003,8 @@ void render_hud_attachments() render_hud_elements(); //restore type mask - gPipeline.setRenderTypeMask(mask); + gPipeline.popRenderTypeMask(); + if (has_ui) { gPipeline.toggleRenderDebugFeature((void*) LLPipeline::RENDER_DEBUG_FEATURE_UI); diff --git a/indra/newview/llvovolume.cpp b/indra/newview/llvovolume.cpp index aaa6e8007f..b120f5abf3 100644 --- a/indra/newview/llvovolume.cpp +++ b/indra/newview/llvovolume.cpp @@ -3349,7 +3349,8 @@ void LLVolumeGeometryManager::registerFace(LLSpatialGroup* group, LLFace* facep, model_mat = &(drawable->getRegion()->mRenderMatrix); } - U8 bump = (type == LLRenderPass::PASS_BUMP ? facep->getTextureEntry()->getBumpmap() : 0); + + U8 bump = (type == LLRenderPass::PASS_BUMP || type == LLRenderPass::PASS_POST_BUMP) ? facep->getTextureEntry()->getBumpmap() : 0; LLViewerTexture* tex = facep->getTexture(); @@ -4052,66 +4053,76 @@ void LLVolumeGeometryManager::genDrawInfo(LLSpatialGroup* group, U32 mask, std:: && group->mSpatialPartition->mPartitionType != LLViewerRegion::PARTITION_HUD && LLPipeline::sRenderBump && te->getShiny()) - { + { //shiny if (tex->getPrimaryFormat() == GL_ALPHA) - { + { //invisiprim+shiny registerFace(group, facep, LLRenderPass::PASS_INVISI_SHINY); registerFace(group, facep, LLRenderPass::PASS_INVISIBLE); } else if (LLPipeline::sRenderDeferred) - { + { //deferred rendering if (te->getFullbright()) - { + { //register in post deferred fullbright shiny pass registerFace(group, facep, LLRenderPass::PASS_FULLBRIGHT_SHINY); + if (te->getBumpmap()) + { //register in post deferred bump pass + registerFace(group, facep, LLRenderPass::PASS_POST_BUMP); + } } else if (te->getBumpmap()) - { + { //register in deferred bump pass registerFace(group, facep, LLRenderPass::PASS_BUMP); } else - { + { //register in deferred simple pass (deferred simple includes shiny) llassert(mask & LLVertexBuffer::MAP_NORMAL); registerFace(group, facep, LLRenderPass::PASS_SIMPLE); } } else if (fullbright) - { + { //not deferred, register in standard fullbright shiny pass registerFace(group, facep, LLRenderPass::PASS_FULLBRIGHT_SHINY); } else - { + { //not deferred or fullbright, register in standard shiny pass registerFace(group, facep, LLRenderPass::PASS_SHINY); } } else - { + { //not alpha and not shiny if (!is_alpha && tex->getPrimaryFormat() == GL_ALPHA) - { + { //invisiprim registerFace(group, facep, LLRenderPass::PASS_INVISIBLE); } else if (fullbright) - { + { //fullbright registerFace(group, facep, LLRenderPass::PASS_FULLBRIGHT); + if (LLPipeline::sRenderDeferred && LLPipeline::sRenderBump && te->getBumpmap()) + { //if this is the deferred render and a bump map is present, register in post deferred bump + registerFace(group, facep, LLRenderPass::PASS_POST_BUMP); + } } else { if (LLPipeline::sRenderDeferred && LLPipeline::sRenderBump && te->getBumpmap()) - { + { //non-shiny or fullbright deferred bump registerFace(group, facep, LLRenderPass::PASS_BUMP); } else - { + { //all around simple llassert(mask & LLVertexBuffer::MAP_NORMAL); registerFace(group, facep, LLRenderPass::PASS_SIMPLE); } } + //not sure why this is here -- shiny HUD attachments maybe? -- davep 5/11/2010 if (!is_alpha && te->getShiny() && LLPipeline::sRenderBump) { registerFace(group, facep, LLRenderPass::PASS_SHINY); } } + //not sure why this is here, and looks like it might cause bump mapped objects to get rendered redundantly -- davep 5/11/2010 if (!is_alpha && !LLPipeline::sRenderDeferred) { llassert((mask & LLVertexBuffer::MAP_NORMAL) || fullbright); diff --git a/indra/newview/pipeline.cpp b/indra/newview/pipeline.cpp index da01ae44db..a066b7fc0d 100644 --- a/indra/newview/pipeline.cpp +++ b/indra/newview/pipeline.cpp @@ -331,7 +331,6 @@ LLPipeline::LLPipeline() : mInitialized(FALSE), mVertexShadersEnabled(FALSE), mVertexShadersLoaded(0), - mRenderTypeMask(0), mRenderDebugFeatureMask(0), mRenderDebugMask(0), mOldRenderDebugMask(0), @@ -385,7 +384,11 @@ void LLPipeline::init() LLViewerStats::getInstance()->mTrianglesDrawnStat.reset(); resetFrameStats(); - mRenderTypeMask = 0xffffffff; // All render types start on + for (U32 i = 0; i < NUM_RENDER_TYPES; ++i) + { + mRenderTypeEnabled[i] = TRUE; //all rendering types start enabled + } + mRenderDebugFeatureMask = 0xffffffff; // All debugging features on mRenderDebugMask = 0; // All debug starts off @@ -2217,14 +2220,13 @@ static LLFastTimer::DeclareTimer FTM_RESET_DRAWORDER("Reset Draw Order"); void LLPipeline::stateSort(LLCamera& camera, LLCullResult &result) { - const U32 face_mask = (1 << LLPipeline::RENDER_TYPE_AVATAR) | - (1 << LLPipeline::RENDER_TYPE_GROUND) | - (1 << LLPipeline::RENDER_TYPE_TERRAIN) | - (1 << LLPipeline::RENDER_TYPE_TREE) | - (1 << LLPipeline::RENDER_TYPE_SKY) | - (1 << LLPipeline::RENDER_TYPE_WATER); - - if (mRenderTypeMask & face_mask) + if (hasAnyRenderType(LLPipeline::RENDER_TYPE_AVATAR, + LLPipeline::RENDER_TYPE_GROUND, + LLPipeline::RENDER_TYPE_TERRAIN, + LLPipeline::RENDER_TYPE_TREE, + LLPipeline::RENDER_TYPE_SKY, + LLPipeline::RENDER_TYPE_WATER, + LLPipeline::END_RENDER_TYPES)) { //clear faces from face pools LLFastTimer t(FTM_RESET_DRAWORDER); @@ -5072,8 +5074,7 @@ void LLPipeline::setLight(LLDrawable *drawablep, BOOL is_light) //static void LLPipeline::toggleRenderType(U32 type) { - U32 bit = (1< 0) { //mask out selected geometry based on reflection detail if (detail < 4) { - mRenderTypeMask &= ~(1 << LLPipeline::RENDER_TYPE_PARTICLES); + clearRenderTypeMask(LLPipeline::RENDER_TYPE_PARTICLES, END_RENDER_TYPES); if (detail < 3) { - mRenderTypeMask &= ~(1 << LLPipeline::RENDER_TYPE_AVATAR); + clearRenderTypeMask(LLPipeline::RENDER_TYPE_AVATAR, END_RENDER_TYPES); if (detail < 2) { - mRenderTypeMask &= ~(1 << LLPipeline::RENDER_TYPE_VOLUME); + clearRenderTypeMask(LLPipeline::RENDER_TYPE_VOLUME, END_RENDER_TYPES); } } } @@ -7418,18 +7426,17 @@ void LLPipeline::generateWaterReflection(LLCamera& camera_in) stateSort(camera, ref_result); } - ref_mask = mRenderTypeMask; - mRenderTypeMask = mask; - } - if (LLDrawPoolWater::sNeedsDistortionUpdate) - { - mRenderTypeMask = ref_mask; - if (gSavedSettings.getS32("RenderReflectionDetail") > 0) + if (LLDrawPoolWater::sNeedsDistortionUpdate) { - gPipeline.grabReferences(ref_result); - LLGLUserClipPlane clip_plane(plane, mat, projection); - renderGeom(camera); + if (gSavedSettings.getS32("RenderReflectionDetail") > 0) + { + gPipeline.grabReferences(ref_result); + LLGLUserClipPlane clip_plane(plane, mat, projection); + renderGeom(camera); + } } + + gPipeline.popRenderTypeMask(); } glCullFace(GL_BACK); glPopMatrix(); @@ -7443,18 +7450,20 @@ void LLPipeline::generateWaterReflection(LLCamera& camera_in) if (last_update) { camera.setFar(camera_in.getFar()); - mRenderTypeMask = type_mask & (~(1<cameraUnderWater() ? FALSE : TRUE; if (LLPipeline::sUnderWaterRender) { - mRenderTypeMask &= ~((1<setup3DViewport(); - mRenderTypeMask = type_mask; + gPipeline.popRenderTypeMask(); LLDrawPoolWater::sNeedsReflectionUpdate = FALSE; LLDrawPoolWater::sNeedsDistortionUpdate = FALSE; LLViewerCamera::getInstance()->setUserClipPlane(LLPlane(-pnorm, -pd)); @@ -7951,21 +7960,22 @@ void LLPipeline::generateGI(LLCamera& camera, LLVector3& lightDir, std::vectorisMuted(avatar->getID()); + pushRenderTypeMask(); + if (muted) { - mask = 1 << LLPipeline::RENDER_TYPE_AVATAR; + andRenderTypeMask(LLPipeline::RENDER_TYPE_AVATAR, END_RENDER_TYPES); } else { - mask = (1<endAlphaGroups(); } +BOOL LLPipeline::hasRenderType(const U32 type) const +{ + return mRenderTypeEnabled[type]; +} + +void LLPipeline::setRenderTypeMask(U32 type, ...) +{ + va_list args; + + va_start(args, type); + while (type < END_RENDER_TYPES) + { + mRenderTypeEnabled[type] = TRUE; + type = va_arg(args, U32); + } + va_end(args); + + if (type > END_RENDER_TYPES) + { + llerrs << "Invalid render type." << llendl; + } +} + +BOOL LLPipeline::hasAnyRenderType(U32 type, ...) const +{ + va_list args; + + va_start(args, type); + while (type < END_RENDER_TYPES) + { + if (mRenderTypeEnabled[type]) + { + return TRUE; + } + type = va_arg(args, U32); + } + va_end(args); + + if (type > END_RENDER_TYPES) + { + llerrs << "Invalid render type." << llendl; + } + + return FALSE; +} + +void LLPipeline::pushRenderTypeMask() +{ + std::string cur_mask; + cur_mask.assign((const char*) mRenderTypeEnabled, sizeof(mRenderTypeEnabled)); + mRenderTypeEnableStack.push(cur_mask); +} + +void LLPipeline::popRenderTypeMask() +{ + if (mRenderTypeEnableStack.empty()) + { + llerrs << "Depleted render type stack." << llendl; + } + + memcpy(mRenderTypeEnabled, mRenderTypeEnableStack.top().data(), sizeof(mRenderTypeEnabled)); + mRenderTypeEnableStack.pop(); +} + +void LLPipeline::andRenderTypeMask(U32 type, ...) +{ + va_list args; + + BOOL tmp[NUM_RENDER_TYPES]; + for (U32 i = 0; i < NUM_RENDER_TYPES; ++i) + { + tmp[i] = FALSE; + } + + va_start(args, type); + while (type < END_RENDER_TYPES) + { + if (mRenderTypeEnabled[type]) + { + tmp[type] = TRUE; + } + + type = va_arg(args, U32); + } + va_end(args); + + if (type > END_RENDER_TYPES) + { + llerrs << "Invalid render type." << llendl; + } + + for (U32 i = 0; i < LLPipeline::NUM_RENDER_TYPES; ++i) + { + mRenderTypeEnabled[i] = tmp[i]; + } + +} + +void LLPipeline::clearRenderTypeMask(U32 type, ...) +{ + va_list args; + + va_start(args, type); + while (type < END_RENDER_TYPES) + { + mRenderTypeEnabled[type] = FALSE; + + type = va_arg(args, U32); + } + va_end(args); + + if (type > END_RENDER_TYPES) + { + llerrs << "Invalid render type." << llendl; + } +} + diff --git a/indra/newview/pipeline.h b/indra/newview/pipeline.h index be878ae667..2c14b94847 100644 --- a/indra/newview/pipeline.h +++ b/indra/newview/pipeline.h @@ -46,6 +46,8 @@ #include "lldrawable.h" #include "llrendertarget.h" +#include + class LLViewerTexture; class LLEdge; class LLFace; @@ -279,12 +281,25 @@ public: LLCullResult::sg_list_t::iterator beginAlphaGroups(); LLCullResult::sg_list_t::iterator endAlphaGroups(); + void addTrianglesDrawn(S32 index_count, U32 render_type = LLRender::TRIANGLES); - BOOL hasRenderType(const U32 type) const { return (type && (mRenderTypeMask & (1< mRenderTypeEnableStack; + U32 mRenderDebugFeatureMask; U32 mRenderDebugMask; -- cgit v1.3 From 696249e731bcc78aa30624b478b5ea0581456c63 Mon Sep 17 00:00:00 2001 From: Xiaohong Bao Date: Thu, 13 May 2010 14:43:24 -0600 Subject: EXT-7327: FIXED: viewer crash in LLVOVolume::syncMediaData on Meshopotamia 1 (transplanted from 8be9afe40b24ddde1ad5f93e3a6554519acc7dc7) --- indra/newview/llvovolume.cpp | 5 +++++ 1 file changed, 5 insertions(+) (limited to 'indra/newview/llvovolume.cpp') diff --git a/indra/newview/llvovolume.cpp b/indra/newview/llvovolume.cpp index b120f5abf3..396d716ae2 100644 --- a/indra/newview/llvovolume.cpp +++ b/indra/newview/llvovolume.cpp @@ -1939,6 +1939,11 @@ void LLVOVolume::syncMediaData(S32 texture_index, const LLSD &media_data, bool m } LLTextureEntry *te = getTE(texture_index); + if(!te) + { + return ; + } + LL_DEBUGS("MediaOnAPrim") << "BEFORE: texture_index = " << texture_index << " hasMedia = " << te->hasMedia() << " : " << ((NULL == te->getMediaData()) ? "NULL MEDIA DATA" : ll_pretty_print_sd(te->getMediaData()->asLLSD())) << llendl; -- cgit v1.3 From 387b7adbc4edea3f5c649b48b0714bf73d7af9da Mon Sep 17 00:00:00 2001 From: Dave Parks Date: Tue, 18 May 2010 16:59:01 -0500 Subject: Take a stab at baked sunlight to help low end hardware. --- indra/newview/app_settings/settings.xml | 39 ++++++++++++++++++++++++++++++++- indra/newview/llface.cpp | 23 ++++++++++++++++++- indra/newview/llviewercontrol.cpp | 2 ++ indra/newview/llvovolume.cpp | 10 ++++++--- indra/newview/pipeline.cpp | 4 ++++ indra/newview/pipeline.h | 2 ++ 6 files changed, 75 insertions(+), 5 deletions(-) (limited to 'indra/newview/llvovolume.cpp') diff --git a/indra/newview/app_settings/settings.xml b/indra/newview/app_settings/settings.xml index 4c72b03b2e..9b7cc04120 100644 --- a/indra/newview/app_settings/settings.xml +++ b/indra/newview/app_settings/settings.xml @@ -6313,6 +6313,7 @@ Value 0 + RenderAnimateRes Comment @@ -6324,7 +6325,31 @@ Value 0 - + + RenderBakeSunlight + + Comment + Bake sunlight into vertex buffers for static objects. + Persist + 1 + Type + Boolean + Value + 0 + + + RenderNoAlpha + + Comment + Disable rendering of alpha objects (render all alpha objects as alpha masks). + Persist + 1 + Type + Boolean + Value + 0 + + RenderAnimateTrees Comment @@ -6481,6 +6506,18 @@ 16.0 + RenderMinimumLODTriangleCount + + Comment + Triangle count threshold at which automatic LOD generation stops + Persist + 1 + Type + U32 + Value + 16 + + RenderEdgeDepthCutoff Comment diff --git a/indra/newview/llface.cpp b/indra/newview/llface.cpp index 9df692e787..365f2fbe1a 100644 --- a/indra/newview/llface.cpp +++ b/indra/newview/llface.cpp @@ -896,6 +896,11 @@ void LLFace::updateRebuildFlags() bool LLFace::canRenderAsMask() { + if (LLPipeline::sNoAlpha) + { + return true; + } + const LLTextureEntry* te = getTextureEntry(); return ( ( @@ -1205,6 +1210,10 @@ BOOL LLFace::getGeometryVolume(const LLVolume& volume, mVObjp->getVolume()->genBinormals(f); } + //if it's not fullbright and has no normals, bake sunlight based on face normal + bool bake_sunlight = !getTextureEntry()->getFullbright() && + !mVertexBuffer->hasDataType(LLVertexBuffer::TYPE_NORMAL); + for (S32 i = 0; i < num_vertices; i++) { if (rebuild_tcoord) @@ -1382,7 +1391,19 @@ BOOL LLFace::getGeometryVolume(const LLVolume& volume, if (rebuild_color) { - *colors++ = color; + if (bake_sunlight) + { + LLVector3 normal = vf.mVertices[i].mNormal * mat_normal; + normal.normVec(); + + F32 da = normal * gPipeline.mSunDir; + + *colors++ = LLColor4U(color.mV[0]*da, color.mV[1]*da, color.mV[2]*da, color.mV[3]); + } + else + { + *colors++ = color; + } } } diff --git a/indra/newview/llviewercontrol.cpp b/indra/newview/llviewercontrol.cpp index 54323a4fad..f76c53a77a 100644 --- a/indra/newview/llviewercontrol.cpp +++ b/indra/newview/llviewercontrol.cpp @@ -506,6 +506,8 @@ void settings_setup_listeners() gSavedSettings.getControl("RenderTerrainDetail")->getSignal()->connect(boost::bind(&handleTerrainDetailChanged, _2)); gSavedSettings.getControl("RenderUseTriStrips")->getSignal()->connect(boost::bind(&handleResetVertexBuffersChanged, _2)); gSavedSettings.getControl("RenderAnimateTrees")->getSignal()->connect(boost::bind(&handleResetVertexBuffersChanged, _2)); + gSavedSettings.getControl("RenderBakeSunlight")->getSignal()->connect(boost::bind(&handleResetVertexBuffersChanged, _2)); + gSavedSettings.getControl("RenderNoAlpha")->getSignal()->connect(boost::bind(&handleResetVertexBuffersChanged, _2)); gSavedSettings.getControl("RenderAvatarVP")->getSignal()->connect(boost::bind(&handleSetShaderChanged, _2)); gSavedSettings.getControl("VertexShaderEnable")->getSignal()->connect(boost::bind(&handleSetShaderChanged, _2)); gSavedSettings.getControl("RenderUIBuffer")->getSignal()->connect(boost::bind(&handleReleaseGLBufferChanged, _2)); diff --git a/indra/newview/llvovolume.cpp b/indra/newview/llvovolume.cpp index b120f5abf3..c470099594 100644 --- a/indra/newview/llvovolume.cpp +++ b/indra/newview/llvovolume.cpp @@ -3516,6 +3516,8 @@ void LLVolumeGeometryManager::rebuildGeom(LLSpatialGroup* group) vobj->isMesh() && gMeshRepo.getSkinInfo(vobj->getVolume()->getParams().getSculptID()); + bool bake_sunlight = LLPipeline::sBakeSunlight && drawablep->isStatic(); + //for each face for (S32 i = 0; i < drawablep->getNumFaces(); i++) { @@ -3715,7 +3717,7 @@ void LLVolumeGeometryManager::rebuildGeom(LLSpatialGroup* group) bump_faces.push_back(facep); } else if ((te->getShiny() && LLPipeline::sRenderBump) || - !te->getFullbright()) + !(te->getFullbright() || bake_sunlight)) { //needs normal simple_faces.push_back(facep); } @@ -3925,6 +3927,8 @@ void LLVolumeGeometryManager::genDrawInfo(LLSpatialGroup* group, U32 mask, std:: buffer_index = 0; } + bool bake_sunlight = LLPipeline::sBakeSunlight && facep->getDrawable()->isStatic(); + U32 index_count = facep->getIndicesCount(); U32 geom_count = facep->getGeomCount(); @@ -4030,7 +4034,7 @@ void LLVolumeGeometryManager::genDrawInfo(LLSpatialGroup* group, U32 mask, std:: // can we safely treat this as an alpha mask? if (facep->canRenderAsMask()) { - if (te->getFullbright()) + if (te->getFullbright() || LLPipeline::sNoAlpha) { registerFace(group, facep, LLRenderPass::PASS_FULLBRIGHT_ALPHA_MASK); } @@ -4094,7 +4098,7 @@ void LLVolumeGeometryManager::genDrawInfo(LLSpatialGroup* group, U32 mask, std:: { //invisiprim registerFace(group, facep, LLRenderPass::PASS_INVISIBLE); } - else if (fullbright) + else if (fullbright || bake_sunlight) { //fullbright registerFace(group, facep, LLRenderPass::PASS_FULLBRIGHT); if (LLPipeline::sRenderDeferred && LLPipeline::sRenderBump && te->getBumpmap()) diff --git a/indra/newview/pipeline.cpp b/indra/newview/pipeline.cpp index dbd7f86447..f54f29927a 100644 --- a/indra/newview/pipeline.cpp +++ b/indra/newview/pipeline.cpp @@ -275,6 +275,8 @@ BOOL LLPipeline::sAutoMaskAlphaDeferred = TRUE; BOOL LLPipeline::sAutoMaskAlphaNonDeferred = FALSE; BOOL LLPipeline::sDisableShaders = FALSE; BOOL LLPipeline::sRenderBump = TRUE; +BOOL LLPipeline::sBakeSunlight = FALSE; +BOOL LLPipeline::sNoAlpha = FALSE; BOOL LLPipeline::sUseTriStrips = TRUE; BOOL LLPipeline::sUseFarClip = TRUE; BOOL LLPipeline::sShadowRender = FALSE; @@ -5487,6 +5489,8 @@ void LLPipeline::resetVertexBuffers() sRenderBump = gSavedSettings.getBOOL("RenderObjectBump"); sUseTriStrips = gSavedSettings.getBOOL("RenderUseTriStrips"); LLVertexBuffer::sUseStreamDraw = gSavedSettings.getBOOL("RenderUseStreamVBO"); + sBakeSunlight = gSavedSettings.getBOOL("RenderBakeSunlight"); + sNoAlpha = gSavedSettings.getBOOL("RenderNoAlpha"); for (LLWorld::region_list_t::const_iterator iter = LLWorld::getInstance()->getRegionList().begin(); iter != LLWorld::getInstance()->getRegionList().end(); ++iter) diff --git a/indra/newview/pipeline.h b/indra/newview/pipeline.h index 2c14b94847..d8bac8b601 100644 --- a/indra/newview/pipeline.h +++ b/indra/newview/pipeline.h @@ -472,6 +472,8 @@ public: static BOOL sAutoMaskAlphaNonDeferred; static BOOL sDisableShaders; // if TRUE, rendering will be done without shaders static BOOL sRenderBump; + static BOOL sBakeSunlight; + static BOOL sNoAlpha; static BOOL sUseTriStrips; static BOOL sUseFarClip; static BOOL sShadowRender; -- cgit v1.3 From 4d57cb3c0975ff0bcea0d6fb3498f2d90962ff16 Mon Sep 17 00:00:00 2001 From: Dave Parks Date: Sat, 22 May 2010 04:35:02 -0500 Subject: Vectorize/memory align buffers in llvolumeface WIP --- indra/llmath/llvolume.cpp | 720 ++++++++++++++++---------------- indra/llmath/llvolume.h | 35 +- indra/newview/lldrawpoolavatar.cpp | 6 +- indra/newview/llface.cpp | 55 +-- indra/newview/llfloaterimagepreview.cpp | 6 +- indra/newview/llhudicon.cpp | 2 + indra/newview/llhudtext.cpp | 5 + indra/newview/llviewercamera.cpp | 5 +- indra/newview/llvograss.cpp | 2 + indra/newview/llvotextbubble.cpp | 33 +- indra/newview/llvovolume.cpp | 29 +- 11 files changed, 474 insertions(+), 424 deletions(-) (limited to 'indra/newview/llvovolume.cpp') diff --git a/indra/llmath/llvolume.cpp b/indra/llmath/llvolume.cpp index 4e342b0b48..01fe2be371 100644 --- a/indra/llmath/llvolume.cpp +++ b/indra/llmath/llvolume.cpp @@ -1,4 +1,5 @@ /** + * @file llvolume.cpp * * $LicenseInfo:firstyear=2002&license=viewergpl$ @@ -89,8 +90,6 @@ const F32 SKEW_MAX = 0.95f; const F32 SCULPT_MIN_AREA = 0.002f; const S32 SCULPT_MIN_AREA_DETAIL = 1; -#define GEN_TRI_STRIP 0 - BOOL check_same_clock_dir( const LLVector3& pt1, const LLVector3& pt2, const LLVector3& pt3, const LLVector3& norm) { LLVector3 test = (pt2-pt1)%(pt3-pt2); @@ -134,21 +133,25 @@ BOOL LLLineSegmentBoxIntersect(const LLVector3& start, const LLVector3& end, con // and returns the intersection point along dir in intersection_t. // Moller-Trumbore algorithm -BOOL LLTriangleRayIntersect(const LLVector3& vert0, const LLVector3& vert1, const LLVector3& vert2, const LLVector3& orig, const LLVector3& dir, +BOOL LLTriangleRayIntersect(const LLVector4a& vert0, const LLVector4a& vert1, const LLVector4a& vert2, const LLVector4a& orig, const LLVector4a& dir, F32* intersection_a, F32* intersection_b, F32* intersection_t, BOOL two_sided) { F32 u, v, t; /* find vectors for two edges sharing vert0 */ - LLVector3 edge1 = vert1 - vert0; + LLVector4a edge1; + edge1.setSub(vert1, vert0); - LLVector3 edge2 = vert2 - vert0;; + + LLVector4a edge2; + edge2.setSub(vert2, vert0); /* begin calculating determinant - also used to calculate U parameter */ - LLVector3 pvec = dir % edge2; - + LLVector4a pvec; + pvec.setCross3(dir, edge2); + /* if determinant is near zero, ray lies in plane of triangle */ - F32 det = edge1 * pvec; + F32 det = edge1.dot3(pvec); if (!two_sided) { @@ -158,10 +161,11 @@ BOOL LLTriangleRayIntersect(const LLVector3& vert0, const LLVector3& vert1, cons } /* calculate distance from vert0 to ray origin */ - LLVector3 tvec = orig - vert0; + LLVector4a tvec; + tvec.setSub(orig, vert0); /* calculate U parameter and test bounds */ - u = tvec * pvec; + u = tvec.dot3(pvec); if (u < 0.f || u > det) { @@ -169,17 +173,18 @@ BOOL LLTriangleRayIntersect(const LLVector3& vert0, const LLVector3& vert1, cons } /* prepare to test V parameter */ - LLVector3 qvec = tvec % edge1; + LLVector4a qvec; + qvec.setCross3(tvec, edge1); /* calculate V parameter and test bounds */ - v = dir * qvec; + v = dir.dot3(qvec); if (v < 0.f || u + v > det) { return FALSE; } /* calculate t, scale parameters, ray intersects triangle */ - t = edge2 * qvec; + t = edge2.dot3(qvec); F32 inv_det = 1.0 / det; t *= inv_det; u *= inv_det; @@ -195,20 +200,22 @@ BOOL LLTriangleRayIntersect(const LLVector3& vert0, const LLVector3& vert1, cons F32 inv_det = 1.0 / det; /* calculate distance from vert0 to ray origin */ - LLVector3 tvec = orig - vert0; + LLVector4a tvec; + tvec.setSub(orig, vert0); /* calculate U parameter and test bounds */ - u = (tvec * pvec) * inv_det; + u = (tvec.dot3(pvec)) * inv_det; if (u < 0.f || u > 1.f) { return FALSE; } /* prepare to test V parameter */ - LLVector3 qvec = tvec - edge1; + LLVector4a qvec; + qvec.setSub(tvec, edge1); /* calculate V parameter and test bounds */ - v = (dir * qvec) * inv_det; + v = (dir.dot3(qvec)) * inv_det; if (v < 0.f || u + v > 1.f) { @@ -216,7 +223,7 @@ BOOL LLTriangleRayIntersect(const LLVector3& vert0, const LLVector3& vert1, cons } /* calculate t, ray intersects triangle */ - t = (edge2 * qvec) * inv_det; + t = (edge2.dot3(qvec)) * inv_det; } if (intersection_a != NULL) @@ -4120,13 +4127,14 @@ void LLVolume::generateSilhouetteVertices(std::vector &vertices, norm_mat.rotate(n[v1], t); t.normalize3Fast(); - LLVector3 norm1 = LLVector3(t[0], t[1], t[2]); - normals.push_back(norm1); + normals.push_back(LLVector3(t[0], t[1], t[2])); - vertices.push_back(face.mVertices[v2].mPosition*mat); - LLVector3 norm2 = face.mVertices[v2].mNormal * norm_mat; - norm2.normVec(); - normals.push_back(norm2); + mat.affineTransform(v[v2], t); + vertices.push_back(LLVector3(t[0], t[1], t[2])); + + norm_mat.rotate(n[v2], t); + t.normalize3Fast(); + normals.push_back(LLVector3(t[0], t[1], t[2])); segments.push_back(vertices.size()); } @@ -4177,6 +4185,10 @@ S32 LLVolume::lineSegmentIntersect(const LLVector3& start, const LLVector3& end, genBinormals(i); } + LLVector4a starta, dira; + + LLVector4a* p = (LLVector4a*) face.mPositions; + for (U32 tri = 0; tri < face.mNumIndices/3; tri++) { S32 index1 = face.mIndices[tri*3+0]; @@ -4185,15 +4197,15 @@ S32 LLVolume::lineSegmentIntersect(const LLVector3& start, const LLVector3& end, F32 a, b, t; - if (LLTriangleRayIntersect(face.mVertices[index1].mPosition, - face.mVertices[index2].mPosition, - face.mVertices[index3].mPosition, - start, dir, &a, &b, &t, FALSE)) + if (LLTriangleRayIntersect(p[index1], + p[index2], + p[index3], + starta, dira, &a, &b, &t, FALSE)) { if ((t >= 0.f) && // if hit is after start (t <= 1.f) && // and before end (t < closest_t)) // and this hit is closer - { + { closest_t = t; hit_face = i; @@ -4201,27 +4213,35 @@ S32 LLVolume::lineSegmentIntersect(const LLVector3& start, const LLVector3& end, { *intersection = start + dir * closest_t; } - + + if (tex_coord != NULL) - { - *tex_coord = ((1.f - a - b) * face.mVertices[index1].mTexCoord + - a * face.mVertices[index2].mTexCoord + - b * face.mVertices[index3].mTexCoord); + { + LLVector2* tc = (LLVector2*) face.mTexCoords; + *tex_coord = ((1.f - a - b) * tc[index1] + + a * tc[index2] + + b * tc[index3]); } if (normal != NULL) - { - *normal = ((1.f - a - b) * face.mVertices[index1].mNormal + - a * face.mVertices[index2].mNormal + - b * face.mVertices[index3].mNormal); + { + LLVector4* norm = (LLVector4*) face.mNormals; + + *normal = ((1.f - a - b) * LLVector3(norm[index1]) + + a * LLVector3(norm[index2]) + + b * LLVector3(norm[index3])); } if (bi_normal != NULL) - { - *bi_normal = ((1.f - a - b) * face.mVertices[index1].mBinormal + - a * face.mVertices[index2].mBinormal + - b * face.mVertices[index3].mBinormal); + { + LLVector4* binormal = (LLVector4*) face.mBinormals; + if (binormal) + { + *bi_normal = ((1.f - a - b) * LLVector3(binormal[index1]) + + a * LLVector3(binormal[index2]) + + b * LLVector3(binormal[index3])); + } } } @@ -4992,6 +5012,14 @@ std::ostream& operator<<(std::ostream &s, const LLVolume *volumep) return s; } +LLVolumeFace::~LLVolumeFace() +{ + _mm_free(mPositions); + _mm_free(mNormals); + _mm_free(mTexCoords); + _mm_free(mIndices); + _mm_free(mBinormals); +} BOOL LLVolumeFace::create(LLVolume* volume, BOOL partial_build) { @@ -5012,6 +5040,7 @@ BOOL LLVolumeFace::create(LLVolume* volume, BOOL partial_build) void LLVolumeFace::optimize(F32 angle_cutoff) { +#if 0 //disabling until a vectorized version is available LLVolumeFace new_face; VertexMapData::PointMap point_map; @@ -5063,6 +5092,7 @@ void LLVolumeFace::optimize(F32 angle_cutoff) mVertices = new_face.mVertices; mIndices = new_face.mIndices; +#endif } void LerpPlanarVertex(LLVolumeFace::VertexData& v0, @@ -5127,20 +5157,22 @@ BOOL LLVolumeFace::createUnCutCubeCap(LLVolume* volume, BOOL partial_build) corners[1].mTexCoord=corners[2].mTexCoord; corners[2].mTexCoord=swap; } - baseVert.mBinormal = calc_binormal_from_triangle( + + LLVector4a binormal; + + calc_binormal_from_triangle( binormal, corners[0].mPosition, corners[0].mTexCoord, corners[1].mPosition, corners[1].mTexCoord, corners[2].mPosition, corners[2].mTexCoord); - for(int t = 0; t < 4; t++){ - corners[t].mBinormal = baseVert.mBinormal; - corners[t].mNormal = baseVert.mNormal; - } - mHasBinormals = TRUE; - if (partial_build) - { - mVertices.clear(); - } + S32 size = (grid_size+1)*(grid_size+1); + resizeVertices(size); + allocateBinormals(size); + + LLVector4a* pos = (LLVector4a*) mPositions; + LLVector4a* norm = (LLVector4a*) mNormals; + LLVector4a* binorm = (LLVector4a*) mBinormals; + LLVector2* tc = (LLVector2*) mTexCoords; S32 vtop = mNumVertices; for(int gx = 0;gx=0;i--) { - mIndices.push_back(vtop+(gy*(grid_size+1))+gx+idxs[i]); - } - -#if GEN_TRI_STRIP - if (gy == 0) - { - mTriStrip.push_back((gx+1)*(grid_size+1)); - mTriStrip.push_back((gx+1)*(grid_size+1)); - mTriStrip.push_back(gx*(grid_size+1)); - } - - mTriStrip.push_back(gy+1+(gx+1)*(grid_size+1)); - mTriStrip.push_back(gy+1+gx*(grid_size+1)); - - - if (gy == grid_size-1) - { - mTriStrip.push_back(gy+1+gx*(grid_size+1)); - } -#endif + *out++ = (vtop+(gy*(grid_size+1))+gx+idxs[i]); + } } else { for(S32 i=0;i<6;i++) { - mIndices.push_back(vtop+(gy*(grid_size+1))+gx+idxs[i]); - } - -#if GEN_TRI_STRIP - if (gy == 0) - { - mTriStrip.push_back(gx*(grid_size+1)); - mTriStrip.push_back(gx*(grid_size+1)); - mTriStrip.push_back((gx+1)*(grid_size+1)); + *out++ = (vtop+(gy*(grid_size+1))+gx+idxs[i]); } - - mTriStrip.push_back(gy+1+gx*(grid_size+1)); - mTriStrip.push_back(gy+1+(gx+1)*(grid_size+1)); - - if (gy == grid_size-1) - { - mTriStrip.push_back(gy+1+(gx+1)*(grid_size+1)); - } -#endif } } } - -#if GEN_TRI_STRIP - if (mTriStrip.size()%2 == 1) - { - mTriStrip.push_back(mTriStrip[mTriStrip.size()-1]); - } -#endif } return TRUE; @@ -5267,11 +5262,25 @@ BOOL LLVolumeFace::createCap(LLVolume* volume, BOOL partial_build) num_vertices = profile.size(); num_indices = (profile.size() - 2)*3; - mVertices.resize(num_vertices); + if (!(mTypeMask & HOLLOW_MASK) && !(mTypeMask & OPEN_MASK)) + { + resizeVertices(num_vertices+1); + allocateBinormals(num_vertices+1); - if (!partial_build) + if (!partial_build) + { + resizeIndices(num_indices+3); + } + } + else { - mIndices.resize(num_indices); + resizeVertices(num_vertices); + allocateBinormals(num_vertices); + + if (!partial_build) + { + resizeIndices(num_indices); + } } S32 max_s = volume->getProfile().getTotal(); @@ -5298,79 +5307,87 @@ BOOL LLVolumeFace::createCap(LLVolume* volume, BOOL partial_build) LLVector3& min = mExtents[0]; LLVector3& max = mExtents[1]; + LLVector2* tc = (LLVector2*) mTexCoords; + LLVector4a* pos = (LLVector4a*) mPositions; + LLVector4a* norm = (LLVector4a*) mNormals; + LLVector4a* binorm = (LLVector4a*) mBinormals; + // Copy the vertices into the array for (S32 i = 0; i < num_vertices; i++) { if (mTypeMask & TOP_MASK) { - mVertices[i].mTexCoord.mV[0] = profile[i].mV[0]+0.5f; - mVertices[i].mTexCoord.mV[1] = profile[i].mV[1]+0.5f; + tc[i].mV[0] = profile[i].mV[0]+0.5f; + tc[i].mV[1] = profile[i].mV[1]+0.5f; } else { // Mirror for underside. - mVertices[i].mTexCoord.mV[0] = profile[i].mV[0]+0.5f; - mVertices[i].mTexCoord.mV[1] = 0.5f - profile[i].mV[1]; + tc[i].mV[0] = profile[i].mV[0]+0.5f; + tc[i].mV[1] = 0.5f - profile[i].mV[1]; } - mVertices[i].mPosition = mesh[i + offset].mPos; + pos[i].load3(mesh[i + offset].mPos.mV); if (i == 0) { - min = max = mVertices[i].mPosition; - min_uv = max_uv = mVertices[i].mTexCoord; + min = max = mesh[i+offset].mPos; + min_uv = max_uv = tc[i]; } else { - update_min_max(min,max, mVertices[i].mPosition); - update_min_max(min_uv, max_uv, mVertices[i].mTexCoord); + update_min_max(min,max, mesh[i+offset].mPos); + update_min_max(min_uv, max_uv, tc[i]); } } mCenter = (min+max)*0.5f; cuv = (min_uv + max_uv)*0.5f; - LLVector3 binormal = calc_binormal_from_triangle( + LLVector4a binormal; + calc_binormal_from_triangle(binormal, mCenter, cuv, - mVertices[0].mPosition, mVertices[0].mTexCoord, - mVertices[1].mPosition, mVertices[1].mTexCoord); - binormal.normVec(); + mesh[0+offset].mPos, tc[0], + mesh[1+offset].mPos, tc[1]); + binormal.normalize3Fast(); + + LLVector4a normal; + LLVector4a d0, d1; + LLVector4a center; + + center.load3(mCenter.mV); - LLVector3 d0; - LLVector3 d1; - LLVector3 normal; + d0.setSub(center, pos[0]); + d1.setSub(center, pos[1]); - d0 = mCenter-mVertices[0].mPosition; - d1 = mCenter-mVertices[1].mPosition; + if (mTypeMask & TOP_MASK) + { + normal.setCross3(d0, d1); + } + else + { + normal.setCross3(d1, d0); + } - normal = (mTypeMask & TOP_MASK) ? (d0%d1) : (d1%d0); - normal.normVec(); + normal.normalize3Fast(); VertexData vd; vd.mPosition = mCenter; - vd.mNormal = normal; - vd.mBinormal = binormal; vd.mTexCoord = cuv; if (!(mTypeMask & HOLLOW_MASK) && !(mTypeMask & OPEN_MASK)) { - mVertices.push_back(vd); + pos[num_vertices].load4a((F32*) ¢er.mQ); + tc[num_vertices] = cuv; num_vertices++; - if (!partial_build) - { - vector_append(mIndices, 3); - } } - for (S32 i = 0; i < num_vertices; i++) { - mVertices[i].mBinormal = binormal; - mVertices[i].mNormal = normal; + binorm[i].load4a((F32*) &binormal.mQ); + norm[i].load4a((F32*) &normal.mQ); } - mHasBinormals = TRUE; - if (partial_build) { return TRUE; @@ -5478,8 +5495,6 @@ BOOL LLVolumeFace::createCap(LLVolume* volume, BOOL partial_build) pt2--; } } - - makeTriStrip(); } else { @@ -5584,8 +5599,6 @@ BOOL LLVolumeFace::createCap(LLVolume* volume, BOOL partial_build) pt2--; } } - - makeTriStrip(); } } else @@ -5607,131 +5620,63 @@ BOOL LLVolumeFace::createCap(LLVolume* volume, BOOL partial_build) mIndices[3*i+v2] = i + 1; } -#if GEN_TRI_STRIP - //make tri strip - if (mTypeMask & OPEN_MASK) - { - makeTriStrip(); - } - else - { - S32 j = num_vertices-2; - if (mTypeMask & TOP_MASK) - { - mTriStrip.push_back(0); - for (S32 i = 0; i <= j; ++i) - { - mTriStrip.push_back(i); - if (i != j) - { - mTriStrip.push_back(j); - } - --j; - } - } - else - { - mTriStrip.push_back(j); - for (S32 i = 0; i <= j; ++i) - { - if (i != j) - { - mTriStrip.push_back(j); - } - mTriStrip.push_back(i); - --j; - } - } - - mTriStrip.push_back(mTriStrip[mTriStrip.size()-1]); - if (mTriStrip.size()%2 == 1) - { - mTriStrip.push_back(mTriStrip[mTriStrip.size()-1]); - } - } -#endif } return TRUE; } -void LLVolumeFace::makeTriStrip() -{ -#if GEN_TRI_STRIP - for (U32 i = 0; i < mNumIndices; i+=3) - { - U16 i0 = mIndices[i]; - U16 i1 = mIndices[i+1]; - U16 i2 = mIndices[i+2]; - - if ((i/3)%2 == 1) - { - mTriStrip.push_back(i0); - mTriStrip.push_back(i0); - mTriStrip.push_back(i1); - mTriStrip.push_back(i2); - mTriStrip.push_back(i2); - } - else - { - mTriStrip.push_back(i2); - mTriStrip.push_back(i2); - mTriStrip.push_back(i1); - mTriStrip.push_back(i0); - mTriStrip.push_back(i0); - } - } - - if (mTriStrip.size()%2 == 1) - { - mTriStrip.push_back(mTriStrip[mTriStrip.size()-1]); - } -#endif -} - void LLVolumeFace::createBinormals() { LLMemType m1(LLMemType::MTYPE_VOLUME); - if (!mHasBinormals) + if (!mBinormals) { + allocateBinormals(mNumVertices); + //generate binormals + LLStrider pos; + pos = (LLVector3*) mPositions; + pos.setStride(16); + + LLVector2* tc = (LLVector2*) mTexCoords; + LLVector4a* binorm = (LLVector4a*) mBinormals; + for (U32 i = 0; i < mNumIndices/3; i++) { //for each triangle - const VertexData& v0 = mVertices[mIndices[i*3+0]]; - const VertexData& v1 = mVertices[mIndices[i*3+1]]; - const VertexData& v2 = mVertices[mIndices[i*3+2]]; + const U16& i0 = mIndices[i*3+0]; + const U16& i1 = mIndices[i*3+1]; + const U16& i2 = mIndices[i*3+2]; //calculate binormal - LLVector3 binorm = calc_binormal_from_triangle(v0.mPosition, v0.mTexCoord, - v1.mPosition, v1.mTexCoord, - v2.mPosition, v2.mTexCoord); + LLVector4a binormal; + calc_binormal_from_triangle(binormal, + pos[i0], tc[i0], + pos[i1], tc[i1], + pos[i2], tc[i2]); - for (U32 j = 0; j < 3; j++) - { //add triangle normal to vertices - mVertices[mIndices[i*3+j]].mBinormal += binorm; // * (weight_sum - d[j])/weight_sum; - } + + //add triangle normal to vertices + binorm[i0].add(binormal); + binorm[i1].add(binormal); + binorm[i2].add(binormal); //even out quad contributions if (i % 2 == 0) { - mVertices[mIndices[i*3+2]].mBinormal += binorm; + binorm[i2].add(binormal); } else { - mVertices[mIndices[i*3+1]].mBinormal += binorm; + binorm[i1].add(binormal); } } //normalize binormals for (U32 i = 0; i < mNumVertices; i++) { - mVertices[i].mBinormal.normVec(); - mVertices[i].mNormal.normVec(); + binorm[i].normalize3Fast(); } - - mHasBinormals = TRUE; } } @@ -5754,6 +5699,13 @@ void LLVolumeFace::resizeVertices(S32 num_verts) mNumVertices = num_verts; } +void LLVolumeFace::allocateBinormals(S32 num_verts) +{ + _mm_free(mBinormals); + mBinormals = (F32*) _mm_malloc(num_verts*16, 16); +} + + void LLVolumeFace::resizeIndices(S32 num_indices) { _mm_free(mIndices); @@ -5761,44 +5713,107 @@ void LLVolumeFace::resizeIndices(S32 num_indices) //pad index block end to allow for QWORD reads S32 size = ((num_indices*2) + 0xF) & ~0xF; - mIndices = (U16*) _mm_malloc(size); + mIndices = (U16*) _mm_malloc(size,16); mNumIndices = num_indices; } -void LLVolumeFace::appendFace(const LLVolumeFace& face, LLMatrix4& mat, LLMatrix4& norm_mat) +void LLVolumeFace::fillFromLegacyData(std::vector& v, std::vector& idx) +{ + resizeVertices(v.size()); + resizeIndices(idx.size()); + + for (U32 i = 0; i < v.size(); ++i) + { + for (U32 j = 0; j < 3; ++j) + { + mPositions[i*4+j] = v[i].mPosition[j]; + mNormals[i*4+j] = v[i].mNormal[j]; + } + + mTexCoords[i*2+0] = v[i].mTexCoord.mV[0]; + mTexCoords[i*2+1] = v[i].mTexCoord.mV[1]; + } + + for (U32 i = 0; i < idx.size(); ++i) + { + mIndices[i] = idx[i]; + } +} + +void LLVolumeFace::appendFace(const LLVolumeFace& face, LLMatrix4& mat_in, LLMatrix4& norm_mat_in) { U16 offset = mNumVertices; - if (face.mNumVertices + mNumVertices > 65536) + S32 new_count = face.mNumVertices + mNumVertices; + + if (new_count > 65536) { llerrs << "Cannot append face -- 16-bit overflow will occur." << llendl; } + + F32* new_pos = (F32*) _mm_malloc(new_count*16, 16); + F32* new_norm = (F32*) _mm_malloc(new_count*16, 16); + F32* new_tc = (F32*) _mm_malloc((new_count*8+0xF) & ~0xF, 16); + + LLVector4a::memcpyNonAliased16(new_pos, mPositions, new_count*4); + LLVector4a::memcpyNonAliased16(new_norm, mNormals, new_count*4); + LLVector4a::memcpyNonAliased16(new_tc, mTexCoords, new_count*2); + + _mm_free(mPositions); + _mm_free(mNormals); + _mm_free(mTexCoords); + + mPositions = new_pos; + mNormals = new_norm; + mTexCoords = new_tc; + + mNumVertices = new_count; + + LLVector4a* dst_pos = (LLVector4a*) mPositions+offset; + LLVector2* dst_tc = (LLVector2*) mTexCoords+offset; + LLVector4a* dst_norm = (LLVector4a*) mNormals+offset; + + LLVector4a* src_pos = (LLVector4a*) face.mPositions; + LLVector2* src_tc = (LLVector2*) face.mTexCoords; + LLVector4a* src_norm = (LLVector4a*) face.mNormals; + + LLMatrix4a mat, norm_mat; + mat.loadu(mat_in); + norm_mat.loadu(norm_mat_in); + for (U32 i = 0; i < face.mNumVertices; ++i) { - VertexData v = face.mVertices[i]; - v.mPosition = v.mPosition*mat; - v.mNormal = v.mNormal * norm_mat; + mat.affineTransform(src_pos[i], dst_pos[i]); + norm_mat.rotate(src_norm[i], dst_norm[i]); + dst_norm[i].normalize3Fast(); - v.mNormal.normalize(); - - mVertices.push_back(v); + dst_tc[i] = src_tc[i]; if (offset == 0 && i == 0) { - mExtents[0] = mExtents[1] = v.mPosition; + mExtents[0] = mExtents[1] = LLVector3((F32*) &(dst_pos[i].mQ)); } else { - update_min_max(mExtents[0], mExtents[1], v.mPosition); + update_min_max(mExtents[0], mExtents[1], (F32*) &(dst_pos[i].mQ)); } } - + + new_count = mNumIndices + face.mNumIndices; + U16* new_indices = (U16*) _mm_malloc((new_count*2+0xF) & ~0xF, 16); + LLVector4a::memcpyNonAliased16((F32*) new_indices, (F32*) mIndices, new_count/2); + _mm_free(mIndices); + mIndices = new_indices; + mNumIndices = new_count; + + U16* dst_idx = mIndices+offset; + for (U32 i = 0; i < face.mNumIndices; ++i) { - mIndices.push_back(face.mIndices[i]+offset); + dst_idx[i] = face.mIndices[i]+offset; } } @@ -5828,21 +5843,20 @@ BOOL LLVolumeFace::createSide(LLVolume* volume, BOOL partial_build) num_vertices = mNumS*mNumT; num_indices = (mNumS-1)*(mNumT-1)*6; - mVertices.resize(num_vertices); - if (!partial_build) { - mIndices.resize(num_indices); + resizeVertices(num_vertices); + resizeIndices(num_indices); if ((volume->getParams().getSculptType() & LL_SCULPT_TYPE_MASK) != LL_SCULPT_TYPE_MESH) { mEdge.resize(num_indices); } } - else - { - mHasBinormals = FALSE; - } + + LLVector4a* pos = (LLVector4a*) mPositions; + LLVector4a* norm = (LLVector4a*) mNormals; + LLVector2* tc = (LLVector2*) mTexCoords; S32 begin_stex = llfloor( profile[mBeginS].mV[2] ); S32 num_s = ((mTypeMask & INNER_MASK) && (mTypeMask & FLAT_MASK) && mNumS > 2) ? mNumS/2 : mNumS; @@ -5894,21 +5908,21 @@ BOOL LLVolumeFace::createSide(LLVolume* volume, BOOL partial_build) i = mBeginS + s + max_s*t; } - mVertices[cur_vertex].mPosition = mesh[i].mPos; - mVertices[cur_vertex].mTexCoord = LLVector2(ss,tt); + pos[cur_vertex].load3(mesh[i].mPos.mV); + tc[cur_vertex] = LLVector2(ss,tt); - mVertices[cur_vertex].mNormal = LLVector3(0,0,0); - mVertices[cur_vertex].mBinormal = LLVector3(0,0,0); + norm[cur_vertex].clear(); cur_vertex++; if ((mTypeMask & INNER_MASK) && (mTypeMask & FLAT_MASK) && mNumS > 2 && s > 0) { - mVertices[cur_vertex].mPosition = mesh[i].mPos; - mVertices[cur_vertex].mTexCoord = LLVector2(ss,tt); + + pos[cur_vertex].load3(mesh[i].mPos.mV); + tc[cur_vertex] = LLVector2(ss,tt); - mVertices[cur_vertex].mNormal = LLVector3(0,0,0); - mVertices[cur_vertex].mBinormal = LLVector3(0,0,0); + norm[cur_vertex].clear(); + cur_vertex++; } } @@ -5926,12 +5940,10 @@ BOOL LLVolumeFace::createSide(LLVolume* volume, BOOL partial_build) i = mBeginS + s + max_s*t; ss = profile[mBeginS + s].mV[2] - begin_stex; - mVertices[cur_vertex].mPosition = mesh[i].mPos; - mVertices[cur_vertex].mTexCoord = LLVector2(ss,tt); - - mVertices[cur_vertex].mNormal = LLVector3(0,0,0); - mVertices[cur_vertex].mBinormal = LLVector3(0,0,0); - + pos[cur_vertex].load3(mesh[i].mPos.mV); + tc[cur_vertex] = LLVector2(ss,tt); + norm[cur_vertex].clear(); + cur_vertex++; } } @@ -5942,10 +5954,11 @@ BOOL LLVolumeFace::createSide(LLVolume* volume, BOOL partial_build) LLVector3& face_max = mExtents[1]; mCenter.clearVec(); - face_min = face_max = mVertices[0].mPosition; + face_min = face_max = LLVector3((F32*) &(pos[i].mQ)); + for (U32 i = 1; i < mNumVertices; ++i) { - update_min_max(face_min, face_max, mVertices[i].mPosition); + update_min_max(face_min, face_max, (F32*) &(pos[i].mQ)); } mCenter = (face_min + face_max) * 0.5f; @@ -5956,18 +5969,9 @@ BOOL LLVolumeFace::createSide(LLVolume* volume, BOOL partial_build) if (!partial_build) { -#if GEN_TRI_STRIP - mTriStrip.clear(); -#endif - // Now we generate the indices. for (t = 0; t < (mNumT-1); t++) { -#if GEN_TRI_STRIP - //prepend terminating index to strip - mTriStrip.push_back(mNumS*t); -#endif - for (s = 0; s < (mNumS-1); s++) { mIndices[cur_index++] = s + mNumS*t; //bottom left @@ -5977,16 +5981,6 @@ BOOL LLVolumeFace::createSide(LLVolume* volume, BOOL partial_build) mIndices[cur_index++] = s+1 + mNumS*t; //bottom right mIndices[cur_index++] = s+1 + mNumS*(t+1); //top right -#if GEN_TRI_STRIP - if (s == 0) - { - mTriStrip.push_back(s+mNumS*t); - mTriStrip.push_back(s+mNumS*(t+1)); - } - mTriStrip.push_back(s+1+mNumS*t); - mTriStrip.push_back(s+1+mNumS*(t+1)); -#endif - mEdge[cur_edge++] = (mNumS-1)*2*t+s*2+1; //bottom left/top right neighbor face if (t < mNumT-2) { //top right/top left neighbor face mEdge[cur_edge++] = (mNumS-1)*2*(t+1)+s*2+1; @@ -6027,52 +6021,55 @@ BOOL LLVolumeFace::createSide(LLVolume* volume, BOOL partial_build) } mEdge[cur_edge++] = (mNumS-1)*2*t+s*2; //top right/bottom left neighbor face } -#if GEN_TRI_STRIP - //append terminating vertex to strip - mTriStrip.push_back(mNumS-1+mNumS*(t+1)); -#endif } - -#if GEN_TRI_STRIP - if (mTriStrip.size()%2 == 1) - { - mTriStrip.push_back(mTriStrip[mTriStrip.size()-1]); - } -#endif } //generate normals for (U32 i = 0; i < mNumIndices/3; i++) //for each triangle { const U16* idx = &(mIndices[i*3]); - - VertexData* v[] = - { &mVertices[idx[0]], &mVertices[idx[1]], &mVertices[idx[2]] }; - - //calculate triangle normal - LLVector3 norm = (v[0]->mPosition-v[1]->mPosition) % (v[0]->mPosition-v[2]->mPosition); + - v[0]->mNormal += norm; - v[1]->mNormal += norm; - v[2]->mNormal += norm; + LLVector4a* v[] = + { pos+idx[0], pos+idx[1], pos+idx[2] }; + + LLVector4a* n[] = + { norm+idx[0], norm+idx[1], norm+idx[2] }; + + //calculate triangle normal + LLVector4a a, b, c; + + a.setSub(*v[0], *v[1]); + b.setSub(*v[0], *v[2]); + c.setCross3(a,b); + n[0]->add(c); + n[1]->add(c); + n[2]->add(c); + //even out quad contributions - v[i%2+1]->mNormal += norm; + n[i%2+1]->add(c); } // adjust normals based on wrapping and stitching - BOOL s_bottom_converges = ((mVertices[0].mPosition - mVertices[mNumS*(mNumT-2)].mPosition).magVecSquared() < 0.000001f); - BOOL s_top_converges = ((mVertices[mNumS-1].mPosition - mVertices[mNumS*(mNumT-2)+mNumS-1].mPosition).magVecSquared() < 0.000001f); + LLVector4a top; + top.setSub(pos[0], pos[mNumS*(mNumT-2)]); + BOOL s_bottom_converges = (top.dot3(top) < 0.000001f); + + top.setSub(pos[mNumS-1], pos[mNumS*(mNumT-2)+mNumS-1]); + BOOL s_top_converges = (top.dot3(top) < 0.000001f); + if (sculpt_stitching == LL_SCULPT_TYPE_NONE) // logic for non-sculpt volumes { if (volume->getPath().isOpen() == FALSE) { //wrap normals on T for (S32 i = 0; i < mNumS; i++) { - LLVector3 norm = mVertices[i].mNormal + mVertices[mNumS*(mNumT-1)+i].mNormal; - mVertices[i].mNormal = norm; - mVertices[mNumS*(mNumT-1)+i].mNormal = norm; + LLVector4a n; + n.setAdd(norm[i], norm[mNumS*(mNumT-1)+i]); + norm[i] = n; + norm[mNumS*(mNumT-1)+i] = n; } } @@ -6080,9 +6077,10 @@ BOOL LLVolumeFace::createSide(LLVolume* volume, BOOL partial_build) { //wrap normals on S for (S32 i = 0; i < mNumT; i++) { - LLVector3 norm = mVertices[mNumS*i].mNormal + mVertices[mNumS*i+mNumS-1].mNormal; - mVertices[mNumS * i].mNormal = norm; - mVertices[mNumS * i+mNumS-1].mNormal = norm; + LLVector4a n; + n.setAdd(norm[mNumS*i], norm[mNumS*i+mNumS-1]); + norm[mNumS * i] = n; + norm[mNumS * i+mNumS-1] = n; } } @@ -6093,7 +6091,7 @@ BOOL LLVolumeFace::createSide(LLVolume* volume, BOOL partial_build) { //all lower S have same normal for (S32 i = 0; i < mNumT; i++) { - mVertices[mNumS*i].mNormal = LLVector3(1,0,0); + norm[mNumS*i].set(1,0,0); } } @@ -6101,7 +6099,7 @@ BOOL LLVolumeFace::createSide(LLVolume* volume, BOOL partial_build) { //all upper S have same normal for (S32 i = 0; i < mNumT; i++) { - mVertices[mNumS*i+mNumS-1].mNormal = LLVector3(-1,0,0); + norm[mNumS*i+mNumS-1].set(-1,0,0); } } } @@ -6129,30 +6127,33 @@ BOOL LLVolumeFace::createSide(LLVolume* volume, BOOL partial_build) { // average normals for north pole - LLVector3 average(0.0, 0.0, 0.0); + LLVector4a average; + average.clear(); + for (S32 i = 0; i < mNumS; i++) { - average += mVertices[i].mNormal; + average.add(norm[i]); } // set average for (S32 i = 0; i < mNumS; i++) { - mVertices[i].mNormal = average; + norm[i] = average; } // average normals for south pole - average = LLVector3(0.0, 0.0, 0.0); + average.clear(); + for (S32 i = 0; i < mNumS; i++) { - average += mVertices[i + mNumS * (mNumT - 1)].mNormal; + average.add(norm[i + mNumS * (mNumT - 1)]); } // set average for (S32 i = 0; i < mNumS; i++) { - mVertices[i + mNumS * (mNumT - 1)].mNormal = average; + norm[i + mNumS * (mNumT - 1)] = average; } } @@ -6162,23 +6163,22 @@ BOOL LLVolumeFace::createSide(LLVolume* volume, BOOL partial_build) { for (S32 i = 0; i < mNumT; i++) { - LLVector3 norm = mVertices[mNumS*i].mNormal + mVertices[mNumS*i+mNumS-1].mNormal; - mVertices[mNumS * i].mNormal = norm; - mVertices[mNumS * i+mNumS-1].mNormal = norm; + LLVector4a n; + n.setAdd(norm[mNumS*i], norm[mNumS*i+mNumS-1]); + norm[mNumS * i] = n; + norm[mNumS * i+mNumS-1] = n; } } - - if (wrap_t) { for (S32 i = 0; i < mNumS; i++) { - LLVector3 norm = mVertices[i].mNormal + mVertices[mNumS*(mNumT-1)+i].mNormal; - mVertices[i].mNormal = norm; - mVertices[mNumS*(mNumT-1)+i].mNormal = norm; + LLVector4a n; + n.setAdd(norm[i], norm[mNumS*(mNumT-1)+i]); + norm[i] = n; + norm[mNumS*(mNumT-1)+i] = n; } - } } @@ -6188,7 +6188,8 @@ BOOL LLVolumeFace::createSide(LLVolume* volume, BOOL partial_build) // Finds binormal based on three vertices with texture coordinates. // Fills in dummy values if the triangle has degenerate texture coordinates. -LLVector3 calc_binormal_from_triangle( +void calc_binormal_from_triangle(LLVector4a& binormal, + const LLVector3& pos0, const LLVector2& tex0, const LLVector3& pos1, @@ -6196,33 +6197,42 @@ LLVector3 calc_binormal_from_triangle( const LLVector3& pos2, const LLVector2& tex2) { - LLVector3 rx0( pos0.mV[VX], tex0.mV[VX], tex0.mV[VY] ); - LLVector3 rx1( pos1.mV[VX], tex1.mV[VX], tex1.mV[VY] ); - LLVector3 rx2( pos2.mV[VX], tex2.mV[VX], tex2.mV[VY] ); + LLVector4a rx0; rx0.set( pos0.mV[VX], tex0.mV[VX], tex0.mV[VY] ); + LLVector4a rx1; rx1.set( pos1.mV[VX], tex1.mV[VX], tex1.mV[VY] ); + LLVector4a rx2; rx2.set( pos2.mV[VX], tex2.mV[VX], tex2.mV[VY] ); - LLVector3 ry0( pos0.mV[VY], tex0.mV[VX], tex0.mV[VY] ); - LLVector3 ry1( pos1.mV[VY], tex1.mV[VX], tex1.mV[VY] ); - LLVector3 ry2( pos2.mV[VY], tex2.mV[VX], tex2.mV[VY] ); + LLVector4a ry0; ry0.set( pos0.mV[VY], tex0.mV[VX], tex0.mV[VY] ); + LLVector4a ry1; ry1.set( pos1.mV[VY], tex1.mV[VX], tex1.mV[VY] ); + LLVector4a ry2; ry2.set( pos2.mV[VY], tex2.mV[VX], tex2.mV[VY] ); - LLVector3 rz0( pos0.mV[VZ], tex0.mV[VX], tex0.mV[VY] ); - LLVector3 rz1( pos1.mV[VZ], tex1.mV[VX], tex1.mV[VY] ); - LLVector3 rz2( pos2.mV[VZ], tex2.mV[VX], tex2.mV[VY] ); + LLVector4a rz0; rz0.set( pos0.mV[VZ], tex0.mV[VX], tex0.mV[VY] ); + LLVector4a rz1; rz1.set( pos1.mV[VZ], tex1.mV[VX], tex1.mV[VY] ); + LLVector4a rz2; rz2.set( pos2.mV[VZ], tex2.mV[VX], tex2.mV[VY] ); - LLVector3 r0 = (rx0 - rx1) % (rx0 - rx2); - LLVector3 r1 = (ry0 - ry1) % (ry0 - ry2); - LLVector3 r2 = (rz0 - rz1) % (rz0 - rz2); + LLVector4a lhs, rhs; + + LLVector4a r0; + lhs.setSub(rx0, rx1); rhs.setSub(rx0, rx2); + r0.setCross3(lhs, rhs); + + LLVector4a r1; + lhs.setSub(ry0, ry1); rhs.setSub(ry0, ry2); + r1.setCross3(lhs, rhs); + + LLVector4a r2; + lhs.setSub(rz0, rz1); rhs.setSub(rz0, rz2); + r2.setCross3(lhs, rhs); - if( r0.mV[VX] && r1.mV[VX] && r2.mV[VX] ) + if( r0[VX] && r1[VX] && r2[VX] ) { - LLVector3 binormal( - -r0.mV[VZ] / r0.mV[VX], - -r1.mV[VZ] / r1.mV[VX], - -r2.mV[VZ] / r2.mV[VX]); + binormal.set( + -r0[VZ] / r0[VX], + -r1[VZ] / r1[VX], + -r2[VZ] / r2[VX]); // binormal.normVec(); - return binormal; } else { - return LLVector3( 0, 1 , 0 ); + binormal.set( 0, 1 , 0 ); } } diff --git a/indra/llmath/llvolume.h b/indra/llmath/llvolume.h index 17be642d4a..911db6f94b 100644 --- a/indra/llmath/llvolume.h +++ b/indra/llmath/llvolume.h @@ -40,6 +40,7 @@ class LLPathParams; class LLVolumeParams; class LLProfile; class LLPath; +class LLVector4a; class LLVolumeFace; class LLVolume; @@ -791,6 +792,19 @@ public: class LLVolumeFace { public: + class VertexData + { + public: + LLVector3 mPosition; + LLVector3 mNormal; + LLVector3 mBinormal; + LLVector2 mTexCoord; + + bool operator<(const VertexData& rhs) const; + bool operator==(const VertexData& rhs) const; + bool compareNormal(const VertexData& rhs, F32 angle_cutoff) const; + }; + LLVolumeFace() : mID(0), mTypeMask(0), @@ -808,26 +822,18 @@ public: { } + ~LLVolumeFace(); + BOOL create(LLVolume* volume, BOOL partial_build = FALSE); void createBinormals(); void appendFace(const LLVolumeFace& face, LLMatrix4& transform, LLMatrix4& normal_tranform); void resizeVertices(S32 num_verts); + void allocateBinormals(S32 num_verts); void resizeIndices(S32 num_indices); + void fillFromLegacyData(std::vector& v, std::vector& idx); - class VertexData - { - public: - LLVector3 mPosition; - LLVector3 mNormal; - LLVector3 mBinormal; - LLVector2 mTexCoord; - - bool operator<(const VertexData& rhs) const; - bool operator==(const VertexData& rhs) const; - bool compareNormal(const VertexData& rhs, F32 angle_cutoff) const; - }; class VertexMapData : public LLVolumeFace::VertexData { @@ -1051,7 +1057,8 @@ public: std::ostream& operator<<(std::ostream &s, const LLVolumeParams &volume_params); -LLVector3 calc_binormal_from_triangle( +void calc_binormal_from_triangle( + LLVector4a& binormal, const LLVector3& pos0, const LLVector2& tex0, const LLVector3& pos1, @@ -1060,7 +1067,7 @@ LLVector3 calc_binormal_from_triangle( const LLVector2& tex2); BOOL LLLineSegmentBoxIntersect(const LLVector3& start, const LLVector3& end, const LLVector3& center, const LLVector3& size); -BOOL LLTriangleRayIntersect(const LLVector3& vert0, const LLVector3& vert1, const LLVector3& vert2, const LLVector3& orig, const LLVector3& dir, +BOOL LLTriangleRayIntersect(const LLVector4a& vert0, const LLVector4a& vert1, const LLVector4a& vert2, const LLVector4a& orig, const LLVector4a& dir, F32* intersection_a, F32* intersection_b, F32* intersection_t, BOOL two_sided); diff --git a/indra/newview/lldrawpoolavatar.cpp b/indra/newview/lldrawpoolavatar.cpp index 4fb8f5266e..0fa0e80cb7 100644 --- a/indra/newview/lldrawpoolavatar.cpp +++ b/indra/newview/lldrawpoolavatar.cpp @@ -1201,14 +1201,14 @@ void LLDrawPoolAvatar::updateRiggedFaceVertexBuffer(LLFace* face, const LLMeshSk if (!buff || buff->getTypeMask() != data_mask || - buff->getRequestedVerts() != vol_face.mVertices.size()) + buff->getRequestedVerts() != vol_face.mNumVertices) { face->setGeomIndex(0); face->setIndicesIndex(0); - face->setSize(vol_face.mVertices.size(), vol_face.mIndices.size()); + face->setSize(vol_face.mNumVertices, vol_face.mNumIndices); face->mVertexBuffer = new LLVertexBuffer(data_mask, 0); - face->mVertexBuffer->allocateBuffer(vol_face.mVertices.size(), vol_face.mIndices.size(), true); + face->mVertexBuffer->allocateBuffer(vol_face.mNumVertices, vol_face.mNumIndices, true); U16 offset = 0; diff --git a/indra/newview/llface.cpp b/indra/newview/llface.cpp index 53dc335c16..77e8a6fdf9 100644 --- a/indra/newview/llface.cpp +++ b/indra/newview/llface.cpp @@ -928,8 +928,8 @@ BOOL LLFace::getGeometryVolume(const LLVolume& volume, { LLFastTimer t(FTM_FACE_GET_GEOM); const LLVolumeFace &vf = volume.getVolumeFace(f); - S32 num_vertices = (S32)vf.mVertices.size(); - S32 num_indices = LLPipeline::sUseTriStrips ? (S32)vf.mTriStrip.size() : (S32) vf.mIndices.size(); + S32 num_vertices = (S32)vf.mNumVertices; + S32 num_indices = (S32) vf.mNumIndices; if (mVertexBuffer.notNull()) { @@ -1128,19 +1128,9 @@ BOOL LLFace::getGeometryVolume(const LLVolume& volume, if (full_rebuild) { mVertexBuffer->getIndexStrider(indicesp, mIndicesIndex); - if (LLPipeline::sUseTriStrips) + for (U32 i = 0; i < (U32) num_indices; i++) { - for (U32 i = 0; i < (U32) num_indices; i++) - { - *indicesp++ = vf.mTriStrip[i] + index_offset; - } - } - else - { - for (U32 i = 0; i < (U32) num_indices; i++) - { - *indicesp++ = vf.mIndices[i] + index_offset; - } + *indicesp++ = vf.mIndices[i] + index_offset; } } @@ -1214,28 +1204,41 @@ BOOL LLFace::getGeometryVolume(const LLVolume& volume, bool bake_sunlight = !getTextureEntry()->getFullbright() && !mVertexBuffer->hasDataType(LLVertexBuffer::TYPE_NORMAL); + //VECTORIZE THIS for (S32 i = 0; i < num_vertices; i++) { + LLVector3 vf_binormal; + if (vf.mBinormals) + { + vf_binormal.set(vf.mBinormals+i*4); + } + + LLVector3 vf_normal; + vf_normal.set(vf.mNormals+i*4); + + LLVector3 vf_position; + vf_position.set(vf.mPositions+i*4); + if (rebuild_tcoord) { - LLVector2 tc = vf.mVertices[i].mTexCoord; + LLVector2 tc(vf.mTexCoords+i*2); if (texgen != LLTextureEntry::TEX_GEN_DEFAULT) { - LLVector3 vec = vf.mVertices[i].mPosition; + LLVector3 vec = vf_position; vec.scaleVec(scale); switch (texgen) { case LLTextureEntry::TEX_GEN_PLANAR: - planarProjection(tc, vf.mVertices[i].mNormal, vf.mCenter, vec); + planarProjection(tc, vf_normal, vf.mCenter, vec); break; case LLTextureEntry::TEX_GEN_SPHERICAL: - sphericalProjection(tc, vf.mVertices[i].mNormal, vf.mCenter, vec); + sphericalProjection(tc, vf_normal, vf.mCenter, vec); break; case LLTextureEntry::TEX_GEN_CYLINDRICAL: - cylindricalProjection(tc, vf.mVertices[i].mNormal, vf.mCenter, vec); + cylindricalProjection(tc, vf_normal, vf.mCenter, vec); break; default: break; @@ -1345,10 +1348,10 @@ BOOL LLFace::getGeometryVolume(const LLVolume& volume, if (bump_code && mVertexBuffer->hasDataType(LLVertexBuffer::TYPE_TEXCOORD1)) { - LLVector3 tangent = vf.mVertices[i].mBinormal % vf.mVertices[i].mNormal; + LLVector3 tangent = vf_binormal % vf_normal; LLMatrix3 tangent_to_object; - tangent_to_object.setRows(tangent, vf.mVertices[i].mBinormal, vf.mVertices[i].mNormal); + tangent_to_object.setRows(tangent, vf_binormal, vf_normal); LLVector3 binormal = binormal_dir * tangent_to_object; binormal = binormal * mat_normal; @@ -1366,12 +1369,12 @@ BOOL LLFace::getGeometryVolume(const LLVolume& volume, if (rebuild_pos) { - *vertices++ = vf.mVertices[i].mPosition * mat_vert; + *vertices++ = vf_position * mat_vert; } if (rebuild_normal) { - LLVector3 normal = vf.mVertices[i].mNormal * mat_normal; + LLVector3 normal = vf_normal * mat_normal; normal.normVec(); *normals++ = normal; @@ -1379,21 +1382,21 @@ BOOL LLFace::getGeometryVolume(const LLVolume& volume, if (rebuild_binormal) { - LLVector3 binormal = vf.mVertices[i].mBinormal * mat_normal; + LLVector3 binormal = vf_binormal * mat_normal; binormal.normVec(); *binormals++ = binormal; } if (rebuild_weights && vf.mWeights.size() > i) { - *weights++ = vf.mWeights[i]; + (*weights++) = vf.mWeights[i]; } if (rebuild_color) { if (bake_sunlight) { - LLVector3 normal = vf.mVertices[i].mNormal * mat_normal; + LLVector3 normal = vf_normal * mat_normal; normal.normVec(); F32 da = normal * gPipeline.mSunDir; diff --git a/indra/newview/llfloaterimagepreview.cpp b/indra/newview/llfloaterimagepreview.cpp index 159e4b41ca..dae301ae29 100644 --- a/indra/newview/llfloaterimagepreview.cpp +++ b/indra/newview/llfloaterimagepreview.cpp @@ -852,6 +852,7 @@ S8 LLImagePreviewSculpted::getType() const void LLImagePreviewSculpted::setPreviewTarget(LLImageRaw* imagep, F32 distance) { +#if 0 //VECTORIZE THIS mCameraDistance = distance; mCameraZoom = 1.f; mCameraPitch = 0.f; @@ -892,6 +893,7 @@ void LLImagePreviewSculpted::setPreviewTarget(LLImageRaw* imagep, F32 distance) { *(index_strider++) = vf.mIndices[i]; } +#endif } @@ -901,7 +903,7 @@ void LLImagePreviewSculpted::setPreviewTarget(LLImageRaw* imagep, F32 distance) BOOL LLImagePreviewSculpted::render() { mNeedsUpdate = FALSE; - +#if 0 //VECTORIZE THIS LLGLSUIDefault def; LLGLDisable no_blend(GL_BLEND); LLGLEnable cull(GL_CULL_FACE); @@ -959,7 +961,7 @@ BOOL LLImagePreviewSculpted::render() mVertexBuffer->draw(LLRender::TRIANGLES, num_indices, 0); gGL.popMatrix(); - +#endif return TRUE; } diff --git a/indra/newview/llhudicon.cpp b/indra/newview/llhudicon.cpp index 28b0e7356a..c7ad0bde7e 100644 --- a/indra/newview/llhudicon.cpp +++ b/indra/newview/llhudicon.cpp @@ -271,6 +271,7 @@ BOOL LLHUDIcon::lineSegmentIntersect(const LLVector3& start, const LLVector3& en LLVector3 upper_left = icon_position - (x_scale * 0.5f) + y_scale; LLVector3 upper_right = icon_position + (x_scale * 0.5f) + y_scale; +#if 0 //VECTORIZE THIS F32 t = 0.f; LLVector3 dir = end-start; @@ -284,6 +285,7 @@ BOOL LLHUDIcon::lineSegmentIntersect(const LLVector3& start, const LLVector3& en } return TRUE; } +#endif return FALSE; } diff --git a/indra/newview/llhudtext.cpp b/indra/newview/llhudtext.cpp index 9ed5d13831..64c01d0937 100644 --- a/indra/newview/llhudtext.cpp +++ b/indra/newview/llhudtext.cpp @@ -206,6 +206,8 @@ BOOL LLHUDText::lineSegmentIntersect(const LLVector3& start, const LLVector3& en gGL.end(); } +#if 0 //VECTORIZE THIS + LLVector3 dir = end-start; F32 t = 0.f; @@ -218,6 +220,9 @@ BOOL LLHUDText::lineSegmentIntersect(const LLVector3& start, const LLVector3& en return TRUE; } } + +#endif + } return FALSE; diff --git a/indra/newview/llviewercamera.cpp b/indra/newview/llviewercamera.cpp index aa82c216d9..7eca276358 100644 --- a/indra/newview/llviewercamera.cpp +++ b/indra/newview/llviewercamera.cpp @@ -788,13 +788,14 @@ BOOL LLViewerCamera::areVertsVisible(LLViewerObject* volumep, BOOL all_verts) LLMatrix4 render_mat(vo_volume->getRenderRotation(), LLVector4(vo_volume->getRenderPosition())); num_faces = volume->getNumVolumeFaces(); + //VECTORIZE THIS for (i = 0; i < num_faces; i++) { const LLVolumeFace& face = volume->getVolumeFace(i); - for (U32 v = 0; v < face.mVertices.size(); v++) + for (U32 v = 0; v < face.mNumVertices; v++) { - LLVector4 vec = LLVector4(face.mVertices[v].mPosition) * mat; + LLVector4 vec = LLVector4(face.mPositions+v*4) * mat; if (drawablep->isActive()) { diff --git a/indra/newview/llvograss.cpp b/indra/newview/llvograss.cpp index a82afbeb76..edb97eef24 100644 --- a/indra/newview/llvograss.cpp +++ b/indra/newview/llvograss.cpp @@ -640,6 +640,7 @@ BOOL LLVOGrass::lineSegmentIntersect(const LLVector3& start, const LLVector3& en position.mV[2] += blade_height; v[3] = v1 = position + mRegionp->getOriginAgent(); +#if 0 //VECTORIZE THIS F32 a,b,t; @@ -703,6 +704,7 @@ BOOL LLVOGrass::lineSegmentIntersect(const LLVector3& start, const LLVector3& en } } } +#endif } return ret; diff --git a/indra/newview/llvotextbubble.cpp b/indra/newview/llvotextbubble.cpp index 428ef20006..339da3c0bf 100644 --- a/indra/newview/llvotextbubble.cpp +++ b/indra/newview/llvotextbubble.cpp @@ -45,6 +45,7 @@ #include "llviewertexturelist.h" #include "llvolume.h" #include "pipeline.h" +#include "llvector4a.h" #include "llviewerregion.h" LLVOTextBubble::LLVOTextBubble(const LLUUID &id, const LLPCode pcode, LLViewerRegion *regionp) @@ -217,7 +218,7 @@ void LLVOTextBubble::updateFaceSize(S32 idx) else { const LLVolumeFace& vol_face = getVolume()->getVolumeFace(idx); - face->setSize(vol_face.mVertices.size(), vol_face.mIndices.size()); + face->setSize(vol_face.mNumVertices, vol_face.mNumIndices); } } @@ -235,19 +236,37 @@ void LLVOTextBubble::getGeometry(S32 idx, const LLVolumeFace& face = getVolume()->getVolumeFace(idx); - LLVector3 pos = getPositionAgent(); + LLVector4a pos; + pos.load3(getPositionAgent().mV); + + LLVector4a scale; + scale.load3(getScale().mV); + LLColor4U color = LLColor4U(getTE(idx)->getColor()); U32 offset = mDrawable->getFace(idx)->getGeomIndex(); - for (U32 i = 0; i < face.mVertices.size(); i++) + LLVector4a* dst_pos = (LLVector4a*) verticesp.get(); + LLVector4a* src_pos = (LLVector4a*) face.mPositions; + + LLVector4a* dst_norm = (LLVector4a*) normalsp.get(); + LLVector4a* src_norm = (LLVector4a*) face.mNormals; + + LLVector2* dst_tc = (LLVector2*) texcoordsp.get(); + LLVector2* src_tc = (LLVector2*) face.mTexCoords; + + LLVector4a::memcpyNonAliased16((F32*) dst_norm, (F32*) src_norm, face.mNumVertices*4); + LLVector4a::memcpyNonAliased16((F32*) dst_tc, (F32*) src_tc, face.mNumVertices*2); + + + for (U32 i = 0; i < face.mNumVertices; i++) { - *verticesp++ = face.mVertices[i].mPosition.scaledVec(getScale()) + pos; - *normalsp++ = face.mVertices[i].mNormal; - *texcoordsp++ = face.mVertices[i].mTexCoord; + LLVector4a t; + t.setMul(src_pos[i], scale); + dst_pos[i].setAdd(t, pos); *colorsp++ = color; } - for (U32 i = 0; i < face.mIndices.size(); i++) + for (U32 i = 0; i < face.mNumIndices; i++) { *indicesp++ = face.mIndices[i] + offset; } diff --git a/indra/newview/llvovolume.cpp b/indra/newview/llvovolume.cpp index cea964a100..7d80a66041 100644 --- a/indra/newview/llvovolume.cpp +++ b/indra/newview/llvovolume.cpp @@ -60,6 +60,7 @@ #include "llflexibleobject.h" #include "llsky.h" #include "lltexturefetch.h" +#include "llvector4a.h" #include "llviewercamera.h" #include "llviewertexturelist.h" #include "llviewerobjectlist.h" @@ -1601,14 +1602,8 @@ void LLVOVolume::updateFaceSize(S32 idx) else { const LLVolumeFace& vol_face = getVolume()->getVolumeFace(idx); - if (LLPipeline::sUseTriStrips) - { - facep->setSize(vol_face.mVertices.size(), vol_face.mTriStrip.size()); - } - else - { - facep->setSize(vol_face.mVertices.size(), vol_face.mIndices.size()); - } + facep->setSize(vol_face.mNumVertices, vol_face.mNumIndices); + } } @@ -1863,21 +1858,25 @@ bool LLVOVolume::hasMedia() const LLVector3 LLVOVolume::getApproximateFaceNormal(U8 face_id) { LLVolume* volume = getVolume(); - LLVector3 result; + LLVector4a result; + result.clear(); + + LLVector3 ret; if (volume && face_id < volume->getNumVolumeFaces()) { const LLVolumeFace& face = volume->getVolumeFace(face_id); - for (S32 i = 0; i < (S32)face.mVertices.size(); ++i) + for (S32 i = 0; i < (S32)face.mNumVertices; ++i) { - result += face.mVertices[i].mNormal; + result.add(*((LLVector4a*) face.mNormals+i*4)); } - result = volumeDirectionToAgent(result); - result.normVec(); + LLVector3 ret((F32*) &result.mQ); + ret = volumeDirectionToAgent(ret); + ret.normVec(); } - return result; + return ret; } void LLVOVolume::requestMediaDataUpdate(bool isNew) @@ -3032,7 +3031,7 @@ F32 LLVOVolume::getBinRadius() for (S32 i = 0; i < volume->getNumVolumeFaces(); ++i) { const LLVolumeFace& face = volume->getVolumeFace(i); - vert_count += face.mVertices.size(); + vert_count += face.mNumVertices; } scale = 1.f/llmax(vert_count/1024.f, 1.f); -- cgit v1.3 From 7eba473723a260a1025b5a865715573b2369298e Mon Sep 17 00:00:00 2001 From: Dave Parks Date: Mon, 24 May 2010 14:01:18 -0500 Subject: Fix for bad indexes on cube faces. Extra validation on vertex buffers. --- indra/llmath/llvolume.cpp | 11 +++--- indra/llrender/llvertexbuffer.cpp | 34 +++++++++++-------- indra/llrender/llvertexbuffer.h | 5 +++ indra/newview/llface.cpp | 7 ++-- indra/newview/llspatialpartition.cpp | 66 +++++------------------------------- indra/newview/llspatialpartition.h | 4 +-- indra/newview/llvovolume.cpp | 5 +-- 7 files changed, 51 insertions(+), 81 deletions(-) (limited to 'indra/newview/llvovolume.cpp') diff --git a/indra/llmath/llvolume.cpp b/indra/llmath/llvolume.cpp index 88969af4bd..31544016db 100644 --- a/indra/llmath/llvolume.cpp +++ b/indra/llmath/llvolume.cpp @@ -4173,7 +4173,7 @@ S32 LLVolume::lineSegmentIntersect(const LLVector3& start, const LLVector3& end, for (S32 i = start_face; i <= end_face; i++) { - const LLVolumeFace &face = getVolumeFace((U32)i); + LLVolumeFace &face = mVolumeFaces[i]; LLVector3 box_center = (face.mExtents[0] + face.mExtents[1]) / 2.f; LLVector3 box_size = face.mExtents[1] - face.mExtents[0]; @@ -4235,6 +4235,10 @@ S32 LLVolume::lineSegmentIntersect(const LLVector3& start, const LLVector3& end, if (bi_normal != NULL) { + if (!face.mBinormals) + { + face.createBinormals(); + } LLVector4* binormal = (LLVector4*) face.mBinormals; if (binormal) { @@ -5174,7 +5178,6 @@ BOOL LLVolumeFace::createUnCutCubeCap(LLVolume* volume, BOOL partial_build) LLVector4a* binorm = (LLVector4a*) mBinormals; LLVector2* tc = (LLVector2*) mTexCoords; - S32 vtop = mNumVertices; for(int gx = 0;gx=0;i--) { - *out++ = (vtop+(gy*(grid_size+1))+gx+idxs[i]); + *out++ = ((gy*(grid_size+1))+gx+idxs[i]); } } else { for(S32 i=0;i<6;i++) { - *out++ = (vtop+(gy*(grid_size+1))+gx+idxs[i]); + *out++ = ((gy*(grid_size+1))+gx+idxs[i]); } } } diff --git a/indra/llrender/llvertexbuffer.cpp b/indra/llrender/llvertexbuffer.cpp index 0f3c900d2b..e6943cc551 100644 --- a/indra/llrender/llvertexbuffer.cpp +++ b/indra/llrender/llvertexbuffer.cpp @@ -261,10 +261,8 @@ void LLVertexBuffer::setupClientArrays(U32 data_mask) } } -void LLVertexBuffer::drawRange(U32 mode, U32 start, U32 end, U32 count, U32 indices_offset) const +void LLVertexBuffer::validateRange(U32 start, U32 end, U32 count, U32 indices_offset) const { - llassert(mRequestedNumVerts >= 0); - if (start >= (U32) mRequestedNumVerts || end >= (U32) mRequestedNumVerts) { @@ -279,6 +277,25 @@ void LLVertexBuffer::drawRange(U32 mode, U32 start, U32 end, U32 count, U32 indi llerrs << "Bad index buffer draw range: [" << indices_offset << ", " << indices_offset+count << "]" << llendl; } + if (gDebugGL && !useVBOs()) + { + U16* idx = ((U16*) getIndicesPointer())+indices_offset; + for (U32 i = 0; i < count; ++i) + { + if (idx[i] < start || idx[i] > end) + { + llerrs << "Index out of range: " << idx[i] << " not in [" << start << ", " << end << "]" << llendl; + } + } + } +} + +void LLVertexBuffer::drawRange(U32 mode, U32 start, U32 end, U32 count, U32 indices_offset) const +{ + validateRange(start, end, count, indices_offset); + + llassert(mRequestedNumVerts >= 0); + if (mGLIndices != sGLRenderIndices) { llerrs << "Wrong index buffer bound." << llendl; @@ -297,17 +314,6 @@ void LLVertexBuffer::drawRange(U32 mode, U32 start, U32 end, U32 count, U32 indi U16* idx = ((U16*) getIndicesPointer())+indices_offset; - if (gDebugGL && !useVBOs()) - { - for (U32 i = 0; i < count; ++i) - { - if (idx[i] < start || idx[i] > end) - { - llerrs << "Index out of range: " << idx[i] << " not in [" << start << ", " << end << "]" << llendl; - } - } - } - stop_glerror(); glDrawRangeElements(sGLMode[mode], start, end, count, GL_UNSIGNED_SHORT, idx); diff --git a/indra/llrender/llvertexbuffer.h b/indra/llrender/llvertexbuffer.h index c6fd0a9e3c..47146a5ec4 100644 --- a/indra/llrender/llvertexbuffer.h +++ b/indra/llrender/llvertexbuffer.h @@ -207,6 +207,11 @@ public: void drawArrays(U32 mode, U32 offset, U32 count) const; void drawRange(U32 mode, U32 start, U32 end, U32 count, U32 indices_offset) const; + //for debugging, validate data in given range is valid + void validateRange(U32 start, U32 end, U32 count, U32 offset) const; + + + protected: S32 mNumVerts; // Number of vertices allocated S32 mNumIndices; // Number of indices allocated diff --git a/indra/newview/llface.cpp b/indra/newview/llface.cpp index 77e8a6fdf9..a65ee52fa1 100644 --- a/indra/newview/llface.cpp +++ b/indra/newview/llface.cpp @@ -368,10 +368,13 @@ void LLFace::setDrawable(LLDrawable *drawable) void LLFace::setSize(const S32 num_vertices, const S32 num_indices) { - if (mGeomCount != num_vertices || + //allocate vertices in blocks of 4 for alignment + S32 num_verts = (num_vertices + 0x3) & ~0x3; + + if (mGeomCount != num_verts || mIndicesCount != num_indices) { - mGeomCount = num_vertices; + mGeomCount = num_verts; mIndicesCount = num_indices; mVertexBuffer = NULL; mLastVertexBuffer = NULL; diff --git a/indra/newview/llspatialpartition.cpp b/indra/newview/llspatialpartition.cpp index 77c38798d1..470c332b42 100644 --- a/indra/newview/llspatialpartition.cpp +++ b/indra/newview/llspatialpartition.cpp @@ -512,50 +512,6 @@ void LLSpatialGroup::checkStates() #endif } -void validate_draw_info(LLDrawInfo& params) -{ -#if LL_OCTREE_PARANOIA_CHECK - if (params.mVertexBuffer.isNull()) - { - llerrs << "Draw batch has no vertex buffer." << llendl; - } - - //bad range - if (params.mStart >= params.mEnd) - { - llerrs << "Draw batch has invalid range." << llendl; - } - - if (params.mEnd >= (U32) params.mVertexBuffer->getNumVerts()) - { - llerrs << "Draw batch has buffer overrun error." << llendl; - } - - if (params.mOffset + params.mCount > (U32) params.mVertexBuffer->getNumIndices()) - { - llerrs << "Draw batch has index buffer ovverrun error." << llendl; - } - - //bad indices - U16* indicesp = (U16*) params.mVertexBuffer->getIndicesPointer(); - if (indicesp) - { - for (U32 i = params.mOffset; i < params.mOffset+params.mCount; i++) - { - if (indicesp[i] < (U16)params.mStart) - { - llerrs << "Draw batch has vertex buffer index out of range error (index too low)." << llendl; - } - - if (indicesp[i] > (U16)params.mEnd) - { - llerrs << "Draw batch has vertex buffer index out of range error (index too high)." << llendl; - } - } - } -#endif -} - void LLSpatialGroup::validateDrawMap() { #if LL_OCTREE_PARANOIA_CHECK @@ -565,8 +521,8 @@ void LLSpatialGroup::validateDrawMap() for (drawmap_elem_t::iterator j = draw_vec.begin(); j != draw_vec.end(); ++j) { LLDrawInfo& params = **j; - - validate_draw_info(params); + + params.validate(); } } #endif @@ -3379,18 +3335,9 @@ LLDrawInfo::LLDrawInfo(U16 start, U16 end, U32 count, U32 offset, mDistance(0.f), mDrawMode(LLRender::TRIANGLES) { - mDebugColor = (rand() << 16) + rand(); - if (mStart >= mVertexBuffer->getRequestedVerts() || - mEnd >= mVertexBuffer->getRequestedVerts()) - { - llerrs << "Invalid draw info vertex range." << llendl; - } + mVertexBuffer->validateRange(mStart, mEnd, mCount, mOffset); - if (mOffset >= (U32) mVertexBuffer->getRequestedIndices() || - mOffset + mCount > (U32) mVertexBuffer->getRequestedIndices()) - { - llerrs << "Invalid draw info index range." << llendl; - } + mDebugColor = (rand() << 16) + rand(); } LLDrawInfo::~LLDrawInfo() @@ -3406,6 +3353,11 @@ LLDrawInfo::~LLDrawInfo() } } +void LLDrawInfo::validate() +{ + mVertexBuffer->validateRange(mStart, mEnd, mCount, mOffset); +} + LLVertexBuffer* LLGeometryManager::createVertexBuffer(U32 type_mask, U32 usage) { return new LLVertexBuffer(type_mask, usage); diff --git a/indra/newview/llspatialpartition.h b/indra/newview/llspatialpartition.h index 67c33d5b0f..9b252d1035 100644 --- a/indra/newview/llspatialpartition.h +++ b/indra/newview/llspatialpartition.h @@ -75,6 +75,8 @@ public: BOOL fullbright = FALSE, U8 bump = 0, BOOL particle = FALSE, F32 part_size = 0); + void validate(); + LLPointer mVertexBuffer; LLPointer mTexture; LLColor4U mGlowColor; @@ -676,8 +678,6 @@ public: virtual void shift(const LLVector3 &offset); }; -void validate_draw_info(LLDrawInfo& params); - extern const F32 SG_BOX_SIDE; extern const F32 SG_BOX_OFFSET; extern const F32 SG_BOX_RAD; diff --git a/indra/newview/llvovolume.cpp b/indra/newview/llvovolume.cpp index 7d80a66041..d57a535050 100644 --- a/indra/newview/llvovolume.cpp +++ b/indra/newview/llvovolume.cpp @@ -3382,7 +3382,7 @@ void LLVolumeGeometryManager::registerFace(LLSpatialGroup* group, LLFace* facep, draw_vec[idx]->mCount += facep->getIndicesCount(); draw_vec[idx]->mEnd += facep->getGeomCount(); draw_vec[idx]->mVSize = llmax(draw_vec[idx]->mVSize, facep->getVirtualSize()); - validate_draw_info(*draw_vec[idx]); + draw_vec[idx]->validate(); update_min_max(draw_vec[idx]->mExtents[0], draw_vec[idx]->mExtents[1], facep->mExtents[0]); update_min_max(draw_vec[idx]->mExtents[0], draw_vec[idx]->mExtents[1], facep->mExtents[1]); } @@ -3406,12 +3406,13 @@ void LLVolumeGeometryManager::registerFace(LLSpatialGroup* group, LLFace* facep, } draw_info->mExtents[0] = facep->mExtents[0]; draw_info->mExtents[1] = facep->mExtents[1]; - validate_draw_info(*draw_info); if (LLPipeline::sUseTriStrips) { draw_info->mDrawMode = LLRender::TRIANGLE_STRIP; } + + draw_info->validate(); } } -- cgit v1.3 From e6fe3b1f1aa888e4594c89154ef895b3cf5498e9 Mon Sep 17 00:00:00 2001 From: Dave Parks Date: Tue, 25 May 2010 03:55:01 -0500 Subject: Better vectorization of various things. Turn off debug gl by default. --- indra/llcommon/lldefs.h | 8 + indra/llmath/llvolume.cpp | 383 ++++++++++++++++++++------------ indra/llmath/llvolume.h | 110 +++++++-- indra/newview/app_settings/settings.xml | 4 +- indra/newview/llface.cpp | 8 +- indra/newview/llfloaterimagepreview.cpp | 20 +- indra/newview/llhudicon.cpp | 43 +++- indra/newview/llhudtext.cpp | 5 - indra/newview/llviewercamera.cpp | 16 +- indra/newview/llvograss.cpp | 7 +- indra/newview/llvovolume.cpp | 2 +- 11 files changed, 404 insertions(+), 202 deletions(-) (limited to 'indra/newview/llvovolume.cpp') diff --git a/indra/llcommon/lldefs.h b/indra/llcommon/lldefs.h index f3b5ca361f..10e6cb34bf 100644 --- a/indra/llcommon/lldefs.h +++ b/indra/llcommon/lldefs.h @@ -242,5 +242,13 @@ inline LLDATATYPE llclampb(const LLDATATYPE& a) return llmin(llmax(a, (LLDATATYPE)0), (LLDATATYPE)255); } +template +inline void llswap(LLDATATYPE& lhs, LLDATATYPE& rhs) +{ + LLDATATYPE tmp = lhs; + lhs = rhs; + rhs = tmp; +} + #endif // LL_LLDEFS_H diff --git a/indra/llmath/llvolume.cpp b/indra/llmath/llvolume.cpp index f05e6eb9d9..d8fbc081fa 100644 --- a/indra/llmath/llvolume.cpp +++ b/indra/llmath/llvolume.cpp @@ -237,6 +237,21 @@ BOOL LLTriangleRayIntersect(const LLVector4a& vert0, const LLVector4a& vert1, co return TRUE; } +//helper for non-aligned vectors +BOOL LLTriangleRayIntersect(const LLVector3& vert0, const LLVector3& vert1, const LLVector3& vert2, const LLVector3& orig, const LLVector3& dir, + F32* intersection_a, F32* intersection_b, F32* intersection_t, BOOL two_sided) +{ + LLVector4a vert0a, vert1a, vert2a, origa, dira; + vert0a.load3(vert0.mV); + vert1a.load3(vert1.mV); + vert2a.load3(vert2.mV); + origa.load3(orig.mV); + dira.load3(dir.mV); + + return LLTriangleRayIntersect(vert0a, vert1a, vert2a, origa, dira, + intersection_a, intersection_b, intersection_t, two_sided); +} + //------------------------------------------------------------------- // statics @@ -1889,15 +1904,15 @@ bool LLVolumeFace::VertexData::operator==(const LLVolumeFace::VertexData& rhs)co bool LLVolumeFace::VertexData::compareNormal(const LLVolumeFace::VertexData& rhs, F32 angle_cutoff) const { bool retval = false; - if (rhs.mPosition == mPosition && rhs.mTexCoord == mTexCoord) + if (rhs.mData[POSITION].equal3(mData[POSITION]) && rhs.mTexCoord == mTexCoord) { if (angle_cutoff > 1.f) { - retval = (mNormal == rhs.mNormal); + retval = (mData[NORMAL].equal3(rhs.mData[NORMAL])); } else { - F32 cur_angle = rhs.mNormal*mNormal; + F32 cur_angle = rhs.mData[NORMAL].dot3(mData[NORMAL]); retval = cur_angle > angle_cutoff; } } @@ -2081,9 +2096,9 @@ bool LLVolume::unpackVolumeFaces(std::istream& is, S32 size) min = max = LLVector3(0,0,0); - F32* pos_out = face.mPositions; - F32* norm_out = face.mNormals; - F32* tc_out = face.mTexCoords; + F32* pos_out = (F32*) face.mPositions; + F32* norm_out = (F32*) face.mNormals; + F32* tc_out = (F32*) face.mTexCoords; for (U32 j = 0; j < num_verts; ++j) { @@ -2188,13 +2203,15 @@ bool LLVolume::unpackVolumeFaces(std::istream& is, S32 size) void tetrahedron_set_normal(LLVolumeFace::VertexData* cv) { - LLVector3 nrm = (cv[1].mPosition-cv[0].mPosition)%(cv[2].mPosition-cv[0].mPosition); - - nrm.normVec(); - - cv[0].mNormal = nrm; - cv[1].mNormal = nrm; - cv[2].mNormal = nrm; + LLVector4a v0; + v0.setSub(cv[1].getPosition(), cv[0].getNormal()); + LLVector4a v1; + v1.setSub(cv[2].getNormal(), cv[0].getPosition()); + + cv[0].getNormal().setCross3(v0,v1); + cv[0].getNormal().normalize3fast(); + cv[1].setNormal(cv[0].getNormal()); + cv[2].setNormal(cv[1].getNormal()); } BOOL LLVolume::isTetrahedron() @@ -2209,12 +2226,12 @@ void LLVolume::makeTetrahedron() LLVolumeFace face; F32 x = 0.25f; - LLVector3 p[] = + LLVector4a p[] = { //unit tetrahedron corners - LLVector3(x,x,x), - LLVector3(-x,-x,x), - LLVector3(-x,x,-x), - LLVector3(x,-x,-x) + LLVector4a(x,x,x), + LLVector4a(-x,-x,x), + LLVector4a(-x,x,-x), + LLVector4a(x,-x,-x) }; face.mExtents[0].setVec(-x,-x,-x); @@ -2229,9 +2246,9 @@ void LLVolume::makeTetrahedron() //side 1 - cv[0].mPosition = p[1]; - cv[1].mPosition = p[0]; - cv[2].mPosition = p[2]; + cv[0].setPosition(p[1]); + cv[1].setPosition(p[0]); + cv[2].setPosition(p[2]); tetrahedron_set_normal(cv); @@ -2242,14 +2259,14 @@ void LLVolume::makeTetrahedron() LLVector4a* n = (LLVector4a*) face.mNormals; LLVector2* tc = (LLVector2*) face.mTexCoords; - v[0].load3(cv[0].mPosition.mV); - v[1].load3(cv[1].mPosition.mV); - v[2].load3(cv[2].mPosition.mV); + v[0] = cv[0].getPosition(); + v[1] = cv[1].getPosition(); + v[2] = cv[2].getPosition(); v += 3; - n[0].load3(cv[0].mNormal.mV); - n[1].load3(cv[1].mNormal.mV); - n[2].load3(cv[2].mNormal.mV); + n[0] = cv[0].getNormal(); + n[1] = cv[1].getNormal(); + n[2] = cv[2].getNormal(); n += 3; tc[0] = cv[0].mTexCoord; @@ -2259,20 +2276,20 @@ void LLVolume::makeTetrahedron() //side 2 - cv[0].mPosition = p[3]; - cv[1].mPosition = p[0]; - cv[2].mPosition = p[1]; + cv[0].setPosition(p[3]); + cv[1].setPosition(p[0]); + cv[2].setPosition(p[1]); tetrahedron_set_normal(cv); - v[0].load3(cv[0].mPosition.mV); - v[1].load3(cv[1].mPosition.mV); - v[2].load3(cv[2].mPosition.mV); + v[0] = cv[0].getPosition(); + v[1] = cv[1].getPosition(); + v[2] = cv[2].getPosition(); v += 3; - n[0].load3(cv[0].mNormal.mV); - n[1].load3(cv[1].mNormal.mV); - n[2].load3(cv[2].mNormal.mV); + n[0] = cv[0].getNormal(); + n[1] = cv[1].getNormal(); + n[2] = cv[2].getNormal(); n += 3; tc[0] = cv[0].mTexCoord; @@ -2281,20 +2298,20 @@ void LLVolume::makeTetrahedron() tc += 3; //side 3 - cv[0].mPosition = p[3]; - cv[1].mPosition = p[1]; - cv[2].mPosition = p[2]; + cv[0].setPosition(p[3]); + cv[1].setPosition(p[1]); + cv[2].setPosition(p[2]); tetrahedron_set_normal(cv); - v[0].load3(cv[0].mPosition.mV); - v[1].load3(cv[1].mPosition.mV); - v[2].load3(cv[2].mPosition.mV); + v[0] = cv[0].getPosition(); + v[1] = cv[1].getPosition(); + v[2] = cv[2].getPosition(); v += 3; - n[0].load3(cv[0].mNormal.mV); - n[1].load3(cv[1].mNormal.mV); - n[2].load3(cv[2].mNormal.mV); + n[0] = cv[0].getNormal(); + n[1] = cv[1].getNormal(); + n[2] = cv[2].getNormal(); n += 3; tc[0] = cv[0].mTexCoord; @@ -2303,20 +2320,20 @@ void LLVolume::makeTetrahedron() tc += 3; //side 4 - cv[0].mPosition = p[2]; - cv[1].mPosition = p[0]; - cv[2].mPosition = p[3]; + cv[0].setPosition(p[2]); + cv[1].setPosition(p[0]); + cv[2].setPosition(p[3]); tetrahedron_set_normal(cv); - v[0].load3(cv[0].mPosition.mV); - v[1].load3(cv[1].mPosition.mV); - v[2].load3(cv[2].mPosition.mV); + v[0] = cv[0].getPosition(); + v[1] = cv[1].getPosition(); + v[2] = cv[2].getPosition(); v += 3; - n[0].load3(cv[0].mNormal.mV); - n[1].load3(cv[1].mNormal.mV); - n[2].load3(cv[2].mNormal.mV); + n[0] = cv[0].getNormal(); + n[1] = cv[1].getNormal(); + n[2] = cv[2].getNormal(); n += 3; tc[0] = cv[0].mTexCoord; @@ -3974,9 +3991,9 @@ void LLVolume::generateSilhouetteVertices(std::vector &vertices, S32 v3 = face.mIndices[j*3+2]; //get current face center - LLVector3 cCenter = (face.mVertices[v1].mPosition + - face.mVertices[v2].mPosition + - face.mVertices[v3].mPosition) / 3.0f; + LLVector3 cCenter = (face.mVertices[v1].getPosition() + + face.mVertices[v2].getPosition() + + face.mVertices[v3].getPosition()) / 3.0f; //for each edge for (S32 k = 0; k < 3; k++) { @@ -3994,9 +4011,9 @@ void LLVolume::generateSilhouetteVertices(std::vector &vertices, v3 = face.mIndices[nIndex*3+2]; //get neighbor face center - LLVector3 nCenter = (face.mVertices[v1].mPosition + - face.mVertices[v2].mPosition + - face.mVertices[v3].mPosition) / 3.0f; + LLVector3 nCenter = (face.mVertices[v1].getPosition() + + face.mVertices[v2].getPosition() + + face.mVertices[v3].getPosition()) / 3.0f; //draw line vertices.push_back(cCenter); @@ -4020,14 +4037,14 @@ void LLVolume::generateSilhouetteVertices(std::vector &vertices, //for each vertex for (U32 j = 0; j < face.mNumVertices; j++) { - vertices.push_back(face.mVertices[j].mPosition); - vertices.push_back(face.mVertices[j].mPosition + face.mVertices[j].mNormal*0.1f); + vertices.push_back(face.mVertices[j].getPosition()); + vertices.push_back(face.mVertices[j].getPosition() + face.mVertices[j].getNormal()*0.1f); normals.push_back(LLVector3(0,0,1)); normals.push_back(LLVector3(0,0,1)); segments.push_back(vertices.size()); #if DEBUG_SILHOUETTE_BINORMALS - vertices.push_back(face.mVertices[j].mPosition); - vertices.push_back(face.mVertices[j].mPosition + face.mVertices[j].mBinormal*0.1f); + vertices.push_back(face.mVertices[j].getPosition()); + vertices.push_back(face.mVertices[j].getPosition() + face.mVertices[j].mBinormal*0.1f); normals.push_back(LLVector3(0,0,1)); normals.push_back(LLVector3(0,0,1)); segments.push_back(vertices.size()); @@ -5038,9 +5055,15 @@ BOOL LLVolumeFace::create(LLVolume* volume, BOOL partial_build) } } +void LLVolumeFace::getVertexData(U16 index, LLVolumeFace::VertexData& cv) +{ + cv.setPosition(mPositions[index]); + cv.setNormal(mNormals[index]); + cv.mTexCoord = mTexCoords[index]; +} + void LLVolumeFace::optimize(F32 angle_cutoff) { -#if 0 //disabling until a vectorized version is available LLVolumeFace new_face; VertexMapData::PointMap point_map; @@ -5050,10 +5073,11 @@ void LLVolumeFace::optimize(F32 angle_cutoff) { U16 index = mIndices[i]; - LLVolumeFace::VertexData cv = mVertices[index]; - + LLVolumeFace::VertexData cv; + getVertexData(index, cv); + BOOL found = FALSE; - VertexMapData::PointMap::iterator point_iter = point_map.find(cv.mPosition); + VertexMapData::PointMap::iterator point_iter = point_map.find(cv.getPosition()); if (point_iter != point_map.end()) { //duplicate point might exist for (U32 j = 0; j < point_iter->second.size(); ++j) @@ -5062,7 +5086,7 @@ void LLVolumeFace::optimize(F32 angle_cutoff) if (tv.compareNormal(cv, angle_cutoff)) { found = TRUE; - new_face.mIndices.push_back((point_iter->second)[j].mIndex); + new_face.pushIndex((point_iter->second)[j].mIndex); break; } } @@ -5070,14 +5094,14 @@ void LLVolumeFace::optimize(F32 angle_cutoff) if (!found) { - new_face.mVertices.push_back(cv); + new_face.pushVertex(cv); U16 index = (U16) new_face.mNumVertices-1; - new_face.mIndices.push_back(index); + new_face.pushIndex(index); VertexMapData d; - d.mPosition = cv.mPosition; + d.setPosition(cv.getPosition()); d.mTexCoord = cv.mTexCoord; - d.mNormal = cv.mNormal; + d.setNormal(cv.getNormal()); d.mIndex = index; if (point_iter != point_map.end()) { @@ -5085,14 +5109,23 @@ void LLVolumeFace::optimize(F32 angle_cutoff) } else { - point_map[d.mPosition].push_back(d); + point_map[d.getPosition()].push_back(d); } } } - mVertices = new_face.mVertices; - mIndices = new_face.mIndices; -#endif + swapData(new_face); +} + +void LLVolumeFace::swapData(LLVolumeFace& rhs) +{ + llswap(rhs.mPositions, mPositions); + llswap(rhs.mNormals, mNormals); + llswap(rhs.mBinormals, mBinormals); + llswap(rhs.mTexCoords, mTexCoords); + llswap(rhs.mIndices,mIndices); + llswap(rhs.mNumVertices, mNumVertices); + llswap(rhs.mNumIndices, mNumIndices); } void LerpPlanarVertex(LLVolumeFace::VertexData& v0, @@ -5102,10 +5135,21 @@ void LerpPlanarVertex(LLVolumeFace::VertexData& v0, F32 coef01, F32 coef02) { - vout.mPosition = v0.mPosition + ((v1.mPosition-v0.mPosition)*coef01)+((v2.mPosition-v0.mPosition)*coef02); + + LLVector4a lhs; + lhs.setSub(v1.getPosition(), v0.getPosition()); + lhs.mul(coef01); + LLVector4a rhs; + rhs.setSub(v2.getPosition(), v0.getPosition()); + rhs.mul(coef02); + + rhs.add(lhs); + rhs.add(v0.getPosition()); + + vout.setPosition(rhs); + vout.mTexCoord = v0.mTexCoord + ((v1.mTexCoord-v0.mTexCoord)*coef01)+((v2.mTexCoord-v0.mTexCoord)*coef02); - vout.mNormal = v0.mNormal; - vout.mBinormal = v0.mBinormal; + vout.setNormal(v0.getNormal()); } BOOL LLVolumeFace::createUnCutCubeCap(LLVolume* volume, BOOL partial_build) @@ -5137,16 +5181,22 @@ BOOL LLVolumeFace::createUnCutCubeCap(LLVolume* volume, BOOL partial_build) VertexData corners[4]; VertexData baseVert; for(int t = 0; t < 4; t++){ - corners[t].mPosition = mesh[offset + (grid_size*t)].mPos; + corners[t].getPosition().load3( mesh[offset + (grid_size*t)].mPos.mV); corners[t].mTexCoord.mV[0] = profile[grid_size*t].mV[0]+0.5f; corners[t].mTexCoord.mV[1] = 0.5f - profile[grid_size*t].mV[1]; } - baseVert.mNormal = - ((corners[1].mPosition-corners[0].mPosition) % - (corners[2].mPosition-corners[1].mPosition)); - baseVert.mNormal.normVec(); + + { + LLVector4a lhs; + lhs.setSub(corners[1].getPosition(), corners[0].getPosition()); + LLVector4a rhs; + rhs.setSub(corners[2].getPosition(), corners[1].getPosition()); + baseVert.getNormal().setCross3(lhs, rhs); + baseVert.getNormal().normalize3fast(); + } + if(!(mTypeMask & TOP_MASK)){ - baseVert.mNormal *= -1.0f; + baseVert.getNormal().mul(-1.0f); }else{ //Swap the UVs on the U(X) axis for top face LLVector2 swap; @@ -5161,9 +5211,9 @@ BOOL LLVolumeFace::createUnCutCubeCap(LLVolume* volume, BOOL partial_build) LLVector4a binormal; calc_binormal_from_triangle( binormal, - corners[0].mPosition, corners[0].mTexCoord, - corners[1].mPosition, corners[1].mTexCoord, - corners[2].mPosition, corners[2].mTexCoord); + corners[0].getPosition(), corners[0].mTexCoord, + corners[1].getPosition(), corners[1].mTexCoord, + corners[2].getPosition(), corners[2].mTexCoord); S32 size = (grid_size+1)*(grid_size+1); resizeVertices(size); @@ -5185,18 +5235,18 @@ BOOL LLVolumeFace::createUnCutCubeCap(LLVolume* volume, BOOL partial_build) (F32)gx/(F32)grid_size, (F32)gy/(F32)grid_size); - (*pos++).load3(newVert.mPosition.mV); - (*norm++).load3(baseVert.mNormal.mV); - (*tc++) = newVert.mTexCoord; - (*binorm++).load4a((F32*) &binormal.mQ); + *pos++ = newVert.getPosition(); + *norm++ = baseVert.getNormal(); + *tc++ = newVert.mTexCoord; + *binorm++ = binormal; if (gx == 0 && gy == 0) { - min = max = newVert.mPosition; + min = max = LLVector3(newVert.getPosition().getF32()); } else { - update_min_max(min,max,newVert.mPosition); + update_min_max(min,max,newVert.getPosition().getF32()); } } } @@ -5343,18 +5393,19 @@ BOOL LLVolumeFace::createCap(LLVolume* volume, BOOL partial_build) mCenter = (min+max)*0.5f; cuv = (min_uv + max_uv)*0.5f; + LLVector4a center; + center.load3(mCenter.mV); + LLVector4a binormal; calc_binormal_from_triangle(binormal, - mCenter, cuv, - mesh[0+offset].mPos, tc[0], - mesh[1+offset].mPos, tc[1]); + center, cuv, + pos[0], tc[0], + pos[1], tc[1]); binormal.normalize3fast(); LLVector4a normal; LLVector4a d0, d1; - LLVector4a center; - - center.load3(mCenter.mV); + d0.setSub(center, pos[0]); d1.setSub(center, pos[1]); @@ -5371,7 +5422,7 @@ BOOL LLVolumeFace::createCap(LLVolume* volume, BOOL partial_build) normal.normalize3fast(); VertexData vd; - vd.mPosition = mCenter; + vd.getPosition().load3(mCenter.mV); vd.mTexCoord = cuv; if (!(mTypeMask & HOLLOW_MASK) && !(mTypeMask & OPEN_MASK)) @@ -5634,10 +5685,7 @@ void LLVolumeFace::createBinormals() allocateBinormals(mNumVertices); //generate binormals - LLStrider pos; - pos = (LLVector3*) mPositions; - pos.setStride(16); - + LLVector4a* pos = mPositions; LLVector2* tc = (LLVector2*) mTexCoords; LLVector4a* binorm = (LLVector4a*) mBinormals; @@ -5690,12 +5738,12 @@ void LLVolumeFace::resizeVertices(S32 num_verts) if (num_verts) { - mPositions = (F32*) _mm_malloc(num_verts*16, 16); - mNormals = (F32*) _mm_malloc(num_verts*16, 16); + mPositions = (LLVector4a*) _mm_malloc(num_verts*16, 16); + mNormals = (LLVector4a*) _mm_malloc(num_verts*16, 16); //pad texture coordinate block end to allow for QWORD reads S32 size = ((num_verts*8) + 0xF) & ~0xF; - mTexCoords = (F32*) _mm_malloc(size, 16); + mTexCoords = (LLVector2*) _mm_malloc(size, 16); } else { @@ -5707,10 +5755,61 @@ void LLVolumeFace::resizeVertices(S32 num_verts) mNumVertices = num_verts; } +void LLVolumeFace::pushVertex(const LLVolumeFace::VertexData& cv) +{ + pushVertex(cv.getPosition(), cv.getNormal(), cv.mTexCoord); +} + +void LLVolumeFace::pushVertex(const LLVector4a& pos, const LLVector4a& norm, const LLVector2& tc) +{ + S32 new_verts = mNumVertices+1; + S32 new_size = new_verts*16; + + //positions + LLVector4a* dst = (LLVector4a*) _mm_malloc(new_size, 16); + if (mPositions) + { + LLVector4a::memcpyNonAliased16((F32*) dst, (F32*) mPositions, new_size/4); + _mm_free(mPositions); + } + mPositions = dst; + + //normals + dst = (LLVector4a*) _mm_malloc(new_size, 16); + if (mNormals) + { + LLVector4a::memcpyNonAliased16((F32*) dst, (F32*) mNormals, new_size/4); + _mm_free(mNormals); + } + mNormals = dst; + + //tex coords + new_size = ((new_verts*8)+0xF) & ~0xF; + + { + LLVector2* dst = (LLVector2*) _mm_malloc(new_size, 16); + if (mTexCoords) + { + LLVector4a::memcpyNonAliased16((F32*) dst, (F32*) mTexCoords, new_size/4); + _mm_free(mTexCoords); + } + } + + //just clear binormals + _mm_free(mBinormals); + mBinormals = NULL; + + mPositions[mNumVertices] = pos; + mNormals[mNumVertices] = norm; + mTexCoords[mNumVertices] = tc; + + mNumVertices++; +} + void LLVolumeFace::allocateBinormals(S32 num_verts) { _mm_free(mBinormals); - mBinormals = (F32*) _mm_malloc(num_verts*16, 16); + mBinormals = (LLVector4a*) _mm_malloc(num_verts*16, 16); } @@ -5733,6 +5832,23 @@ void LLVolumeFace::resizeIndices(S32 num_indices) mNumIndices = num_indices; } +void LLVolumeFace::pushIndex(const U16& idx) +{ + S32 new_count = mNumIndices + 1; + S32 new_size = ((new_count*2)+0xF) & ~0xF; + + S32 old_size = (mNumIndices+0xF) & ~0xF; + if (new_size != old_size) + { + U16* dst = (U16*) _mm_malloc(new_size, 16); + LLVector4a::memcpyNonAliased16((F32*) dst, (F32*) mIndices, new_size/4); + _mm_free(mIndices); + mIndices = dst; + } + + mIndices[mNumIndices++] = idx; +} + void LLVolumeFace::fillFromLegacyData(std::vector& v, std::vector& idx) { resizeVertices(v.size()); @@ -5740,14 +5856,9 @@ void LLVolumeFace::fillFromLegacyData(std::vector& v, for (U32 i = 0; i < v.size(); ++i) { - for (U32 j = 0; j < 3; ++j) - { - mPositions[i*4+j] = v[i].mPosition[j]; - mNormals[i*4+j] = v[i].mNormal[j]; - } - - mTexCoords[i*2+0] = v[i].mTexCoord.mV[0]; - mTexCoords[i*2+1] = v[i].mTexCoord.mV[1]; + mPositions[i] = v[i].getPosition(); + mNormals[i] = v[i].getNormal(); + mTexCoords[i] = v[i].mTexCoord; } for (U32 i = 0; i < idx.size(); ++i) @@ -5768,13 +5879,13 @@ void LLVolumeFace::appendFace(const LLVolumeFace& face, LLMatrix4& mat_in, LLMat } - F32* new_pos = (F32*) _mm_malloc(new_count*16, 16); - F32* new_norm = (F32*) _mm_malloc(new_count*16, 16); - F32* new_tc = (F32*) _mm_malloc((new_count*8+0xF) & ~0xF, 16); + LLVector4a* new_pos = (LLVector4a*) _mm_malloc(new_count*16, 16); + LLVector4a* new_norm = (LLVector4a*) _mm_malloc(new_count*16, 16); + LLVector2* new_tc = (LLVector2*) _mm_malloc((new_count*8+0xF) & ~0xF, 16); - LLVector4a::memcpyNonAliased16(new_pos, mPositions, new_count*4); - LLVector4a::memcpyNonAliased16(new_norm, mNormals, new_count*4); - LLVector4a::memcpyNonAliased16(new_tc, mTexCoords, new_count*2); + LLVector4a::memcpyNonAliased16((F32*) new_pos, (F32*) mPositions, new_count*4); + LLVector4a::memcpyNonAliased16((F32*) new_norm, (F32*) mNormals, new_count*4); + LLVector4a::memcpyNonAliased16((F32*) new_tc, (F32*) mTexCoords, new_count*2); _mm_free(mPositions); _mm_free(mNormals); @@ -6205,24 +6316,24 @@ BOOL LLVolumeFace::createSide(LLVolume* volume, BOOL partial_build) // Fills in dummy values if the triangle has degenerate texture coordinates. void calc_binormal_from_triangle(LLVector4a& binormal, - const LLVector3& pos0, + const LLVector4a& pos0, const LLVector2& tex0, - const LLVector3& pos1, + const LLVector4a& pos1, const LLVector2& tex1, - const LLVector3& pos2, + const LLVector4a& pos2, const LLVector2& tex2) { - LLVector4a rx0; rx0.set( pos0.mV[VX], tex0.mV[VX], tex0.mV[VY] ); - LLVector4a rx1; rx1.set( pos1.mV[VX], tex1.mV[VX], tex1.mV[VY] ); - LLVector4a rx2; rx2.set( pos2.mV[VX], tex2.mV[VX], tex2.mV[VY] ); + LLVector4a rx0( pos0[VX], tex0.mV[VX], tex0.mV[VY] ); + LLVector4a rx1( pos1[VX], tex1.mV[VX], tex1.mV[VY] ); + LLVector4a rx2( pos2[VX], tex2.mV[VX], tex2.mV[VY] ); - LLVector4a ry0; ry0.set( pos0.mV[VY], tex0.mV[VX], tex0.mV[VY] ); - LLVector4a ry1; ry1.set( pos1.mV[VY], tex1.mV[VX], tex1.mV[VY] ); - LLVector4a ry2; ry2.set( pos2.mV[VY], tex2.mV[VX], tex2.mV[VY] ); + LLVector4a ry0( pos0[VY], tex0.mV[VX], tex0.mV[VY] ); + LLVector4a ry1( pos1[VY], tex1.mV[VX], tex1.mV[VY] ); + LLVector4a ry2( pos2[VY], tex2.mV[VX], tex2.mV[VY] ); - LLVector4a rz0; rz0.set( pos0.mV[VZ], tex0.mV[VX], tex0.mV[VY] ); - LLVector4a rz1; rz1.set( pos1.mV[VZ], tex1.mV[VX], tex1.mV[VY] ); - LLVector4a rz2; rz2.set( pos2.mV[VZ], tex2.mV[VX], tex2.mV[VY] ); + LLVector4a rz0( pos0[VZ], tex0.mV[VX], tex0.mV[VY] ); + LLVector4a rz1( pos1[VZ], tex1.mV[VX], tex1.mV[VY] ); + LLVector4a rz2( pos2[VZ], tex2.mV[VX], tex2.mV[VY] ); LLVector4a lhs, rhs; diff --git a/indra/llmath/llvolume.h b/indra/llmath/llvolume.h index 911db6f94b..aa58d6d114 100644 --- a/indra/llmath/llvolume.h +++ b/indra/llmath/llvolume.h @@ -40,7 +40,6 @@ class LLPathParams; class LLVolumeParams; class LLProfile; class LLPath; -class LLVector4a; class LLVolumeFace; class LLVolume; @@ -56,6 +55,7 @@ class LLVolume; #include "v4coloru.h" #include "llrefcount.h" #include "llfile.h" +#include "llvector4a.h" //============================================================================ @@ -794,15 +794,74 @@ class LLVolumeFace public: class VertexData { + enum + { + POSITION = 0, + NORMAL = 1 + }; + + private: + void init() + { + mData = (LLVector4a*) _mm_malloc(32, 16); + } public: - LLVector3 mPosition; - LLVector3 mNormal; - LLVector3 mBinormal; + VertexData() + { + init(); + } + + VertexData(const VertexData& rhs) + { + init(); + LLVector4a::memcpyNonAliased16((F32*) mData, (F32*) rhs.mData, 8); + mTexCoord = rhs.mTexCoord; + } + + ~VertexData() + { + _mm_free(mData); + } + + LLVector4a& getPosition() + { + return mData[POSITION]; + } + + LLVector4a& getNormal() + { + return mData[NORMAL]; + } + + const LLVector4a& getPosition() const + { + return mData[POSITION]; + } + + const LLVector4a& getNormal() const + { + return mData[NORMAL]; + } + + + void setPosition(const LLVector4a& pos) + { + mData[POSITION] = pos; + } + + void setNormal(const LLVector4a& norm) + { + mData[NORMAL] = norm; + } + LLVector2 mTexCoord; bool operator<(const VertexData& rhs) const; bool operator==(const VertexData& rhs) const; bool compareNormal(const VertexData& rhs, F32 angle_cutoff) const; + + private: + LLVector4a* mData; }; LLVolumeFace() : @@ -834,6 +893,13 @@ public: void resizeIndices(S32 num_indices); void fillFromLegacyData(std::vector& v, std::vector& idx); + void pushVertex(const VertexData& cv); + void pushVertex(const LLVector4a& pos, const LLVector4a& norm, const LLVector2& tc); + void pushIndex(const U16& idx); + + void swapData(LLVolumeFace& rhs); + + void getVertexData(U16 indx, LLVolumeFace::VertexData& cv); class VertexMapData : public LLVolumeFace::VertexData { @@ -842,28 +908,20 @@ public: bool operator==(const LLVolumeFace::VertexData& rhs) const { - return mPosition == rhs.mPosition && + return getPosition().equal3(rhs.getPosition()) && mTexCoord == rhs.mTexCoord && - mNormal == rhs.mNormal; + getNormal().equal3(rhs.getNormal()); } struct ComparePosition { - bool operator()(const LLVector3& a, const LLVector3& b) const + bool operator()(const LLVector4a& a, const LLVector4a& b) const { - if (a.mV[0] != b.mV[0]) - { - return a.mV[0] < b.mV[0]; - } - if (a.mV[1] != b.mV[1]) - { - return a.mV[1] < b.mV[1]; - } - return a.mV[2] < b.mV[2]; + return a.less3(b); } }; - typedef std::map, VertexMapData::ComparePosition > PointMap; + typedef std::map, VertexMapData::ComparePosition > PointMap; }; void optimize(F32 angle_cutoff = 2.f); @@ -899,10 +957,10 @@ public: S32 mNumVertices; S32 mNumIndices; - F32* mPositions; - F32* mNormals; - F32* mBinormals; - F32* mTexCoords; + LLVector4a* mPositions; + LLVector4a* mNormals; + LLVector4a* mBinormals; + LLVector2* mTexCoords; U16* mIndices; std::vector mEdge; @@ -1059,14 +1117,18 @@ std::ostream& operator<<(std::ostream &s, const LLVolumeParams &volume_params); void calc_binormal_from_triangle( LLVector4a& binormal, - const LLVector3& pos0, + const LLVector4a& pos0, const LLVector2& tex0, - const LLVector3& pos1, + const LLVector4a& pos1, const LLVector2& tex1, - const LLVector3& pos2, + const LLVector4a& pos2, const LLVector2& tex2); BOOL LLLineSegmentBoxIntersect(const LLVector3& start, const LLVector3& end, const LLVector3& center, const LLVector3& size); + +BOOL LLTriangleRayIntersect(const LLVector3& vert0, const LLVector3& vert1, const LLVector3& vert2, const LLVector3& orig, const LLVector3& dir, + F32* intersection_a, F32* intersection_b, F32* intersection_t, BOOL two_sided); + BOOL LLTriangleRayIntersect(const LLVector4a& vert0, const LLVector4a& vert1, const LLVector4a& vert2, const LLVector4a& orig, const LLVector4a& dir, F32* intersection_a, F32* intersection_b, F32* intersection_t, BOOL two_sided); diff --git a/indra/newview/app_settings/settings.xml b/indra/newview/app_settings/settings.xml index b0a4c02a43..9b7cc04120 100644 --- a/indra/newview/app_settings/settings.xml +++ b/indra/newview/app_settings/settings.xml @@ -6278,7 +6278,7 @@ Type Boolean Value - 1 + 0 RenderDebugPipeline @@ -7735,7 +7735,7 @@ Type Boolean Value - 0 + 1 RenderUseStreamVBO diff --git a/indra/newview/llface.cpp b/indra/newview/llface.cpp index 679875e6bd..db3c5cca33 100644 --- a/indra/newview/llface.cpp +++ b/indra/newview/llface.cpp @@ -1216,18 +1216,18 @@ BOOL LLFace::getGeometryVolume(const LLVolume& volume, LLVector3 vf_binormal; if (vf.mBinormals) { - vf_binormal.set(vf.mBinormals+i*4); + vf_binormal.setVec(vf.mBinormals[i].getF32()); } LLVector3 vf_normal; - vf_normal.set(vf.mNormals+i*4); + vf_normal.set(vf.mNormals[i].getF32()); LLVector3 vf_position; - vf_position.set(vf.mPositions+i*4); + vf_position.set(vf.mPositions[i].getF32()); if (rebuild_tcoord) { - LLVector2 tc(vf.mTexCoords+i*2); + LLVector2 tc(vf.mTexCoords[i]); if (texgen != LLTextureEntry::TEX_GEN_DEFAULT) { diff --git a/indra/newview/llfloaterimagepreview.cpp b/indra/newview/llfloaterimagepreview.cpp index dae301ae29..28fe2a14b7 100644 --- a/indra/newview/llfloaterimagepreview.cpp +++ b/indra/newview/llfloaterimagepreview.cpp @@ -852,7 +852,6 @@ S8 LLImagePreviewSculpted::getType() const void LLImagePreviewSculpted::setPreviewTarget(LLImageRaw* imagep, F32 distance) { -#if 0 //VECTORIZE THIS mCameraDistance = distance; mCameraZoom = 1.f; mCameraPitch = 0.f; @@ -865,8 +864,8 @@ void LLImagePreviewSculpted::setPreviewTarget(LLImageRaw* imagep, F32 distance) } const LLVolumeFace &vf = mVolume->getVolumeFace(0); - U32 num_indices = vf.mIndices.size(); - U32 num_vertices = vf.mVertices.size(); + U32 num_indices = vf.mNumIndices; + U32 num_vertices = vf.mNumVertices; mVertexBuffer = new LLVertexBuffer(LLVertexBuffer::MAP_VERTEX | LLVertexBuffer::MAP_NORMAL, 0); mVertexBuffer->allocateBuffer(num_vertices, num_indices, TRUE); @@ -880,10 +879,16 @@ void LLImagePreviewSculpted::setPreviewTarget(LLImageRaw* imagep, F32 distance) mVertexBuffer->getIndexStrider(index_strider); // build vertices and normals + LLStrider pos; + pos = (LLVector3*) vf.mPositions; pos.setStride(16); + LLStrider norm; + norm = (LLVector3*) vf.mNormals; norm.setStride(16); + + for (U32 i = 0; i < num_vertices; i++) { - *(vertex_strider++) = vf.mVertices[i].mPosition; - LLVector3 normal = vf.mVertices[i].mNormal; + *(vertex_strider++) = *pos++; + LLVector3 normal = *norm++; normal.normalize(); *(normal_strider++) = normal; } @@ -893,7 +898,6 @@ void LLImagePreviewSculpted::setPreviewTarget(LLImageRaw* imagep, F32 distance) { *(index_strider++) = vf.mIndices[i]; } -#endif } @@ -903,7 +907,6 @@ void LLImagePreviewSculpted::setPreviewTarget(LLImageRaw* imagep, F32 distance) BOOL LLImagePreviewSculpted::render() { mNeedsUpdate = FALSE; -#if 0 //VECTORIZE THIS LLGLSUIDefault def; LLGLDisable no_blend(GL_BLEND); LLGLEnable cull(GL_CULL_FACE); @@ -948,7 +951,7 @@ BOOL LLImagePreviewSculpted::render() LLViewerCamera::getInstance()->setPerspective(FALSE, mOrigin.mX, mOrigin.mY, mFullWidth, mFullHeight, FALSE); const LLVolumeFace &vf = mVolume->getVolumeFace(0); - U32 num_indices = vf.mIndices.size(); + U32 num_indices = vf.mNumIndices; mVertexBuffer->setBuffer(LLVertexBuffer::MAP_VERTEX | LLVertexBuffer::MAP_NORMAL); @@ -961,7 +964,6 @@ BOOL LLImagePreviewSculpted::render() mVertexBuffer->draw(LLRender::TRIANGLES, num_indices, 0); gGL.popMatrix(); -#endif return TRUE; } diff --git a/indra/newview/llhudicon.cpp b/indra/newview/llhudicon.cpp index c7ad0bde7e..3c5a4de7f8 100644 --- a/indra/newview/llhudicon.cpp +++ b/indra/newview/llhudicon.cpp @@ -39,6 +39,7 @@ #include "llviewerobject.h" #include "lldrawable.h" +#include "llvector4a.h" #include "llviewercamera.h" #include "llviewertexture.h" #include "llviewerwindow.h" @@ -266,26 +267,44 @@ BOOL LLHUDIcon::lineSegmentIntersect(const LLVector3& start, const LLVector3& en LLVector3 x_scale = image_aspect * (F32)gViewerWindow->getWindowHeightScaled() * mScale * scale_factor * x_pixel_vec; LLVector3 y_scale = (F32)gViewerWindow->getWindowHeightScaled() * mScale * scale_factor * y_pixel_vec; - LLVector3 lower_left = icon_position - (x_scale * 0.5f); - LLVector3 lower_right = icon_position + (x_scale * 0.5f); - LLVector3 upper_left = icon_position - (x_scale * 0.5f) + y_scale; - LLVector3 upper_right = icon_position + (x_scale * 0.5f) + y_scale; + LLVector4a x_scalea; + LLVector4a icon_positiona; + LLVector4a y_scalea; -#if 0 //VECTORIZE THIS - - F32 t = 0.f; - LLVector3 dir = end-start; + x_scalea.load3(x_scale.mV); + x_scalea.mul(0.5f); + y_scalea.load3(y_scale.mV); + + icon_positiona.load3(icon_position.mV); - if (LLTriangleRayIntersect(upper_right, upper_left, lower_right, start, dir, NULL, NULL, &t, FALSE) || - LLTriangleRayIntersect(upper_left, lower_left, lower_right, start, dir, NULL, NULL, &t, FALSE)) + LLVector4a lower_left; + lower_left.setSub(icon_positiona, x_scalea); + LLVector4a lower_right; + lower_right.setAdd(icon_positiona, x_scalea); + LLVector4a upper_left; + upper_left.setAdd(lower_left, y_scalea); + LLVector4a upper_right; + upper_right.setAdd(lower_right, y_scalea); + + F32 t = 0.f; + LLVector4a enda; + enda.load3(end.mV); + LLVector4a starta; + starta.load3(start.mV); + LLVector4a dir; + dir.setSub(enda, starta); + + if (LLTriangleRayIntersect(upper_right, upper_left, lower_right, starta, dir, NULL, NULL, &t, FALSE) || + LLTriangleRayIntersect(upper_left, lower_left, lower_right, starta, dir, NULL, NULL, &t, FALSE)) { if (intersection) { - *intersection = start + dir*t; + dir.mul(t); + starta.add(dir); + *intersection = LLVector3((F32*) &starta.mQ); } return TRUE; } -#endif return FALSE; } diff --git a/indra/newview/llhudtext.cpp b/indra/newview/llhudtext.cpp index 64c01d0937..9ed5d13831 100644 --- a/indra/newview/llhudtext.cpp +++ b/indra/newview/llhudtext.cpp @@ -206,8 +206,6 @@ BOOL LLHUDText::lineSegmentIntersect(const LLVector3& start, const LLVector3& en gGL.end(); } -#if 0 //VECTORIZE THIS - LLVector3 dir = end-start; F32 t = 0.f; @@ -220,9 +218,6 @@ BOOL LLHUDText::lineSegmentIntersect(const LLVector3& start, const LLVector3& en return TRUE; } } - -#endif - } return FALSE; diff --git a/indra/newview/llviewercamera.cpp b/indra/newview/llviewercamera.cpp index 7eca276358..cef7c4abbb 100644 --- a/indra/newview/llviewercamera.cpp +++ b/indra/newview/llviewercamera.cpp @@ -38,6 +38,7 @@ // Viewer includes #include "llagent.h" #include "llagentcamera.h" +#include "llmatrix4a.h" #include "llviewercontrol.h" #include "llviewerobjectlist.h" #include "llviewerregion.h" @@ -787,22 +788,29 @@ BOOL LLViewerCamera::areVertsVisible(LLViewerObject* volumep, BOOL all_verts) LLMatrix4 render_mat(vo_volume->getRenderRotation(), LLVector4(vo_volume->getRenderPosition())); + LLMatrix4a render_mata; + render_mata.loadu(render_mat); + LLMatrix4a mata; + mata.loadu(mat); + num_faces = volume->getNumVolumeFaces(); - //VECTORIZE THIS for (i = 0; i < num_faces; i++) { const LLVolumeFace& face = volume->getVolumeFace(i); for (U32 v = 0; v < face.mNumVertices; v++) { - LLVector4 vec = LLVector4(face.mPositions+v*4) * mat; + const LLVector4a& src_vec = face.mPositions[v]; + LLVector4a vec; + mata.affineTransform(src_vec, vec); if (drawablep->isActive()) { - vec = vec * render_mat; + LLVector4a t = vec; + render_mata.affineTransform(t, vec); } - BOOL in_frustum = pointInFrustum(LLVector3(vec)) > 0; + BOOL in_frustum = pointInFrustum(LLVector3(vec.getF32())) > 0; if (( !in_frustum && all_verts) || (in_frustum && !all_verts)) diff --git a/indra/newview/llvograss.cpp b/indra/newview/llvograss.cpp index b40e6af496..d2842fd62c 100644 --- a/indra/newview/llvograss.cpp +++ b/indra/newview/llvograss.cpp @@ -594,9 +594,9 @@ BOOL LLVOGrass::lineSegmentIntersect(const LLVector3& start, const LLVector3& en LLVector2 tc[4]; LLVector3 v[4]; - // LLVector3 n[4]; // unused! + LLVector3 n[4]; - // F32 closest_t = 1.f; // unused! + F32 closest_t = 1.f; for (S32 i = 0; i < mNumBlades; i++) { @@ -640,8 +640,6 @@ BOOL LLVOGrass::lineSegmentIntersect(const LLVector3& start, const LLVector3& en position.mV[2] += blade_height; v[3] = v1 = position + mRegionp->getOriginAgent(); -#if 0 //VECTORIZE THIS - F32 a,b,t; BOOL hit = FALSE; @@ -704,7 +702,6 @@ BOOL LLVOGrass::lineSegmentIntersect(const LLVector3& start, const LLVector3& en } } } -#endif } return ret; diff --git a/indra/newview/llvovolume.cpp b/indra/newview/llvovolume.cpp index d57a535050..8022f81f19 100644 --- a/indra/newview/llvovolume.cpp +++ b/indra/newview/llvovolume.cpp @@ -1868,7 +1868,7 @@ LLVector3 LLVOVolume::getApproximateFaceNormal(U8 face_id) const LLVolumeFace& face = volume->getVolumeFace(face_id); for (S32 i = 0; i < (S32)face.mNumVertices; ++i) { - result.add(*((LLVector4a*) face.mNormals+i*4)); + result.add(face.mNormals[i]); } LLVector3 ret((F32*) &result.mQ); -- cgit v1.3 From c98b1b3fd9341834978aff0e841714e206d28c0a Mon Sep 17 00:00:00 2001 From: Dave Parks Date: Wed, 26 May 2010 03:29:19 -0500 Subject: Fully aligned llvolume --- indra/llmath/llvolume.cpp | 353 ++++++++++++++++++++++------- indra/llmath/llvolume.h | 99 +++----- indra/llmath/v3math.h | 1 - indra/newview/llface.cpp | 272 +++++++++++++--------- indra/newview/llpanelprimmediacontrols.cpp | 4 +- indra/newview/llvovolume.cpp | 3 +- 6 files changed, 468 insertions(+), 264 deletions(-) (limited to 'indra/newview/llvovolume.cpp') diff --git a/indra/llmath/llvolume.cpp b/indra/llmath/llvolume.cpp index d8fbc081fa..9b6e2488e6 100644 --- a/indra/llmath/llvolume.cpp +++ b/indra/llmath/llvolume.cpp @@ -107,22 +107,27 @@ BOOL check_same_clock_dir( const LLVector3& pt1, const LLVector3& pt2, const LLV BOOL LLLineSegmentBoxIntersect(const LLVector3& start, const LLVector3& end, const LLVector3& center, const LLVector3& size) { - float fAWdU[3]; - LLVector3 dir; - LLVector3 diff; + return LLLineSegmentBoxIntersect(start.mV, end.mV, center.mV, size.mV); +} + +BOOL LLLineSegmentBoxIntersect(const F32* start, const F32* end, const F32* center, const F32* size) +{ + F32 fAWdU[3]; + F32 dir[3]; + F32 diff[3]; for (U32 i = 0; i < 3; i++) { - dir.mV[i] = 0.5f * (end.mV[i] - start.mV[i]); - diff.mV[i] = (0.5f * (end.mV[i] + start.mV[i])) - center.mV[i]; - fAWdU[i] = fabsf(dir.mV[i]); - if(fabsf(diff.mV[i])>size.mV[i] + fAWdU[i]) return false; + dir[i] = 0.5f * (end[i] - start[i]); + diff[i] = (0.5f * (end[i] + start[i])) - center[i]; + fAWdU[i] = fabsf(dir[i]); + if(fabsf(diff[i])>size[i] + fAWdU[i]) return false; } float f; - f = dir.mV[1] * diff.mV[2] - dir.mV[2] * diff.mV[1]; if(fabsf(f)>size.mV[1]*fAWdU[2] + size.mV[2]*fAWdU[1]) return false; - f = dir.mV[2] * diff.mV[0] - dir.mV[0] * diff.mV[2]; if(fabsf(f)>size.mV[0]*fAWdU[2] + size.mV[2]*fAWdU[0]) return false; - f = dir.mV[0] * diff.mV[1] - dir.mV[1] * diff.mV[0]; if(fabsf(f)>size.mV[0]*fAWdU[1] + size.mV[1]*fAWdU[0]) return false; + f = dir[1] * diff[2] - dir[2] * diff[1]; if(fabsf(f)>size[1]*fAWdU[2] + size[2]*fAWdU[1]) return false; + f = dir[2] * diff[0] - dir[0] * diff[2]; if(fabsf(f)>size[0]*fAWdU[2] + size[2]*fAWdU[0]) return false; + f = dir[0] * diff[1] - dir[1] * diff[0]; if(fabsf(f)>size[0]*fAWdU[1] + size[1]*fAWdU[0]) return false; return true; } @@ -1869,6 +1874,59 @@ BOOL LLVolume::generate() return FALSE; } +void LLVolumeFace::VertexData::init() +{ + mData = (LLVector4a*) _mm_malloc(32, 16); +} + +LLVolumeFace::VertexData::VertexData() +{ + init(); +} + +LLVolumeFace::VertexData::VertexData(const VertexData& rhs) +{ + init(); + LLVector4a::memcpyNonAliased16((F32*) mData, (F32*) rhs.mData, 8); + mTexCoord = rhs.mTexCoord; +} + +LLVolumeFace::VertexData::~VertexData() +{ + _mm_free(mData); +} + +LLVector4a& LLVolumeFace::VertexData::getPosition() +{ + return mData[POSITION]; +} + +LLVector4a& LLVolumeFace::VertexData::getNormal() +{ + return mData[NORMAL]; +} + +const LLVector4a& LLVolumeFace::VertexData::getPosition() const +{ + return mData[POSITION]; +} + +const LLVector4a& LLVolumeFace::VertexData::getNormal() const +{ + return mData[NORMAL]; +} + + +void LLVolumeFace::VertexData::setPosition(const LLVector4a& pos) +{ + mData[POSITION] = pos; +} + +void LLVolumeFace::VertexData::setNormal(const LLVector4a& norm) +{ + mData[NORMAL] = norm; +} + bool LLVolumeFace::VertexData::operator<(const LLVolumeFace::VertexData& rhs)const { const U8* l = (const U8*) this; @@ -2037,7 +2095,7 @@ bool LLVolume::unpackVolumeFaces(std::istream& is, S32 size) if (mdl[i].has("Weights")) { - face.mWeights.resize(num_verts); + face.allocateWeights(num_verts); LLSD::Binary weights = mdl[i]["Weights"]; @@ -2050,13 +2108,15 @@ bool LLVolume::unpackVolumeFaces(std::istream& is, S32 size) U8 joint = weights[idx++]; U32 cur_influence = 0; + LLVector4 wght(0,0,0,0); + while (joint != END_INFLUENCES) { U16 influence = weights[idx++]; influence |= ((U16) weights[idx++] << 8); F32 w = llmin((F32) influence / 65535.f, 0.99999f); - face.mWeights[cur_vertex].mV[cur_influence++] = (F32) joint + w; + wght.mV[cur_influence++] = (F32) joint + w; if (cur_influence >= 4) { @@ -2068,6 +2128,8 @@ bool LLVolume::unpackVolumeFaces(std::istream& is, S32 size) } } + face.mWeights[cur_vertex].loadua(wght.mV); + cur_vertex++; } @@ -2078,62 +2140,70 @@ bool LLVolume::unpackVolumeFaces(std::istream& is, S32 size) } - LLVector3 min_pos; - LLVector3 max_pos; + LLVector3 minp; + LLVector3 maxp; LLVector2 min_tc; LLVector2 max_tc; - min_pos.setValue(mdl[i]["PositionDomain"]["Min"]); - max_pos.setValue(mdl[i]["PositionDomain"]["Max"]); + minp.setValue(mdl[i]["PositionDomain"]["Min"]); + maxp.setValue(mdl[i]["PositionDomain"]["Max"]); + LLVector4a min_pos, max_pos; + min_pos.load3(minp.mV); + max_pos.load3(maxp.mV); + min_tc.setValue(mdl[i]["TexCoord0Domain"]["Min"]); max_tc.setValue(mdl[i]["TexCoord0Domain"]["Max"]); - LLVector3 pos_range = max_pos - min_pos; + LLVector4a pos_range; + pos_range.setSub(max_pos, min_pos); LLVector2 tc_range = max_tc - min_tc; - LLVector3& min = face.mExtents[0]; - LLVector3& max = face.mExtents[1]; - - min = max = LLVector3(0,0,0); + LLVector4a& min = face.mExtents[0]; + LLVector4a& max = face.mExtents[1]; - F32* pos_out = (F32*) face.mPositions; - F32* norm_out = (F32*) face.mNormals; - F32* tc_out = (F32*) face.mTexCoords; + min.clear(); + max.clear(); + + LLVector4a* pos_out = face.mPositions; + LLVector4a* norm_out = face.mNormals; + LLVector2* tc_out = face.mTexCoords; for (U32 j = 0; j < num_verts; ++j) { U16* v = (U16*) &(pos[j*3*2]); - pos_out[0] = (F32) v[0] / 65535.f * pos_range.mV[0] + min_pos.mV[0]; - pos_out[1] = (F32) v[1] / 65535.f * pos_range.mV[1] + min_pos.mV[1]; - pos_out[2] = (F32) v[2] / 65535.f * pos_range.mV[2] + min_pos.mV[2]; - + pos_out->set((F32) v[0], (F32) v[1], (F32) v[2]); + pos_out->div(65535.f); + pos_out->mul(pos_range); + pos_out->add(min_pos); if (j == 0) { - min = max = LLVector3(pos_out); + min = *pos_out; + max = min; } else { - update_min_max(min,max,pos_out); + min.setMin(*pos_out); + max.setMax(*pos_out); } - pos_out += 4; + pos_out++; U16* n = (U16*) &(norm[j*3*2]); - - norm_out[0] = (F32) n[0] / 65535.f * 2.f - 1.f; - norm_out[1] = (F32) n[1] / 65535.f * 2.f - 1.f; - norm_out[2] = (F32) n[2] / 65535.f * 2.f - 1.f; - norm_out += 4; + norm_out->set((F32) n[0], (F32) n[1], (F32) n[2]); + norm_out->div(65535.f); + norm_out->mul(2.f); + norm_out->sub(1.f); + norm_out++; U16* t = (U16*) &(tc[j*2*2]); - tc_out[0] = (F32) t[0] / 65535.f * tc_range.mV[0] + min_tc.mV[0]; - tc_out[1] = (F32) t[1] / 65535.f * tc_range.mV[1] + min_tc.mV[1]; + tc_out->mV[0] = (F32) t[0] / 65535.f * tc_range.mV[0] + min_tc.mV[0]; + tc_out->mV[1] = (F32) t[1] / 65535.f * tc_range.mV[1] + min_tc.mV[1]; - tc_out += 8; + tc_out++; } @@ -2234,8 +2304,8 @@ void LLVolume::makeTetrahedron() LLVector4a(x,-x,-x) }; - face.mExtents[0].setVec(-x,-x,-x); - face.mExtents[1].setVec(x,x,x); + face.mExtents[0].splat(-x); + face.mExtents[1].splat(x); LLVolumeFace::VertexData cv[3]; @@ -4165,6 +4235,18 @@ void LLVolume::generateSilhouetteVertices(std::vector &vertices, S32 LLVolume::lineSegmentIntersect(const LLVector3& start, const LLVector3& end, S32 face, LLVector3* intersection,LLVector2* tex_coord, LLVector3* normal, LLVector3* bi_normal) +{ + LLVector4a starta, enda; + starta.load3(start.mV); + enda.load3(end.mV); + + return lineSegmentIntersect(starta, enda, face, intersection, tex_coord, normal, bi_normal); + +} + +S32 LLVolume::lineSegmentIntersect(const LLVector4a& start, const LLVector4a& end, + S32 face, + LLVector3* intersection,LLVector2* tex_coord, LLVector3* normal, LLVector3* bi_normal) { S32 hit_face = -1; @@ -4182,7 +4264,8 @@ S32 LLVolume::lineSegmentIntersect(const LLVector3& start, const LLVector3& end, end_face = face; } - LLVector3 dir = end - start; + LLVector4a dir; + dir.setSub(end, start); F32 closest_t = 2.f; // must be larger than 1 @@ -4192,21 +4275,20 @@ S32 LLVolume::lineSegmentIntersect(const LLVector3& start, const LLVector3& end, { LLVolumeFace &face = mVolumeFaces[i]; - LLVector3 box_center = (face.mExtents[0] + face.mExtents[1]) / 2.f; - LLVector3 box_size = face.mExtents[1] - face.mExtents[0]; + LLVector4a box_center; + box_center.setAdd(face.mExtents[0], face.mExtents[1]); + box_center.mul(0.5f); + + LLVector4a box_size; + box_size.setSub(face.mExtents[1], face.mExtents[0]); - if (LLLineSegmentBoxIntersect(start, end, box_center, box_size)) + if (LLLineSegmentBoxIntersect(start.getF32(), end.getF32(), box_center.getF32(), box_size.getF32())) { if (bi_normal != NULL) // if the caller wants binormals, we may need to generate them { genBinormals(i); } - LLVector4a starta, dira; - - starta.load3(start.mV); - dira.load3(dir.mV); - LLVector4a* p = (LLVector4a*) face.mPositions; for (U32 tri = 0; tri < face.mNumIndices/3; tri++) @@ -4220,7 +4302,7 @@ S32 LLVolume::lineSegmentIntersect(const LLVector3& start, const LLVector3& end, if (LLTriangleRayIntersect(p[index1], p[index2], p[index3], - starta, dira, &a, &b, &t, FALSE)) + start, dir, &a, &b, &t, FALSE)) { if ((t >= 0.f) && // if hit is after start (t <= 1.f) && // and before end @@ -4231,7 +4313,10 @@ S32 LLVolume::lineSegmentIntersect(const LLVector3& start, const LLVector3& end, if (intersection != NULL) { - *intersection = start + dir * closest_t; + LLVector4a intersect = dir; + intersect.mul(closest_t); + intersect.add(start); + intersection->set(intersect.getF32()); } @@ -5029,6 +5114,107 @@ std::ostream& operator<<(std::ostream &s, const LLVolume *volumep) return s; } +LLVolumeFace::LLVolumeFace() : + mID(0), + mTypeMask(0), + mBeginS(0), + mBeginT(0), + mNumS(0), + mNumT(0), + mNumVertices(0), + mNumIndices(0), + mPositions(NULL), + mNormals(NULL), + mBinormals(NULL), + mTexCoords(NULL), + mIndices(NULL), + mWeights(NULL) +{ + mExtents = (LLVector4a*) _mm_malloc(48, 16); + mCenter = mExtents+2; +} + +LLVolumeFace::LLVolumeFace(const LLVolumeFace& src) +{ + mExtents = (LLVector4a*) _mm_malloc(48, 16); + mCenter = mExtents+2; + *this = src; +} + +LLVolumeFace& LLVolumeFace::operator=(const LLVolumeFace& src) +{ + if (&src == this) + { //self assignment, do nothing + return *this; + } + + mID = src.mID; + mTypeMask = src.mTypeMask; + mBeginS = src.mBeginS; + mBeginT = src.mBeginT; + mNumS = src.mNumS; + mNumT = src.mNumT; + + mNumVertices = 0; + mNumIndices = 0; + mPositions = NULL; + mNormals = NULL; + mBinormals = NULL; + mTexCoords = NULL; + mWeights = NULL; + mIndices = NULL; + + LLVector4a::memcpyNonAliased16((F32*) mExtents, (F32*) src.mExtents, 12); + + resizeVertices(src.mNumVertices); + resizeIndices(src.mNumIndices); + + if (mNumVertices) + { + S32 vert_size = mNumVertices*4; + S32 tc_size = (mNumVertices*8+0xF) & ~0xF; + tc_size /= 4; + + LLVector4a::memcpyNonAliased16((F32*) mPositions, (F32*) src.mPositions, vert_size); + LLVector4a::memcpyNonAliased16((F32*) mNormals, (F32*) src.mNormals, vert_size); + LLVector4a::memcpyNonAliased16((F32*) mTexCoords, (F32*) src.mTexCoords, vert_size); + + if (src.mBinormals) + { + allocateBinormals(src.mNumVertices); + LLVector4a::memcpyNonAliased16((F32*) mBinormals, (F32*) src.mBinormals, vert_size); + } + else + { + _mm_free(mBinormals); + mBinormals = NULL; + } + + if (src.mWeights) + { + allocateWeights(src.mNumVertices); + LLVector4a::memcpyNonAliased16((F32*) mWeights, (F32*) src.mWeights, vert_size); + } + else + { + _mm_free(mWeights); + mWeights = NULL; + } + } + + if (mNumIndices) + { + S32 idx_size = (mNumIndices*2+0xF) & ~0xF; + idx_size /= 4; + + LLVector4a::memcpyNonAliased16((F32*) mIndices, (F32*) src.mIndices, idx_size); + } + + + //delete + return *this; +} + LLVolumeFace::~LLVolumeFace() { _mm_free(mPositions); @@ -5036,6 +5222,8 @@ LLVolumeFace::~LLVolumeFace() _mm_free(mTexCoords); _mm_free(mIndices); _mm_free(mBinormals); + _mm_free(mWeights); + _mm_free(mExtents); } BOOL LLVolumeFace::create(LLVolume* volume, BOOL partial_build) @@ -5169,8 +5357,8 @@ BOOL LLVolumeFace::createUnCutCubeCap(LLVolume* volume, BOOL partial_build) num_vertices = (grid_size+1)*(grid_size+1); num_indices = quad_count * 4; - LLVector3& min = mExtents[0]; - LLVector3& max = mExtents[1]; + LLVector4a& min = mExtents[0]; + LLVector4a& max = mExtents[1]; S32 offset = 0; if (mTypeMask & TOP_MASK) @@ -5242,16 +5430,18 @@ BOOL LLVolumeFace::createUnCutCubeCap(LLVolume* volume, BOOL partial_build) if (gx == 0 && gy == 0) { - min = max = LLVector3(newVert.getPosition().getF32()); + min = max = newVert.getPosition(); } else { - update_min_max(min,max,newVert.getPosition().getF32()); + min.setMin(newVert.getPosition()); + max.setMax(newVert.getPosition()); } } } - mCenter = (min + max) * 0.5f; + mCenter->setAdd(min, max); + mCenter->mul(0.5f); if (!partial_build) { @@ -5335,7 +5525,7 @@ BOOL LLVolumeFace::createCap(LLVolume* volume, BOOL partial_build) S32 max_s = volume->getProfile().getTotal(); S32 max_t = volume->getPath().mPath.size(); - mCenter.clearVec(); + mCenter->clear(); S32 offset = 0; if (mTypeMask & TOP_MASK) @@ -5353,8 +5543,8 @@ BOOL LLVolumeFace::createCap(LLVolume* volume, BOOL partial_build) LLVector2 cuv; LLVector2 min_uv, max_uv; - LLVector3& min = mExtents[0]; - LLVector3& max = mExtents[1]; + LLVector4a& min = mExtents[0]; + LLVector4a& max = mExtents[1]; LLVector2* tc = (LLVector2*) mTexCoords; LLVector4a* pos = (LLVector4a*) mPositions; @@ -5380,25 +5570,24 @@ BOOL LLVolumeFace::createCap(LLVolume* volume, BOOL partial_build) if (i == 0) { - min = max = mesh[i+offset].mPos; + min = max = pos[i]; min_uv = max_uv = tc[i]; } else { - update_min_max(min,max, mesh[i+offset].mPos); + update_min_max(min,max,pos[i]); update_min_max(min_uv, max_uv, tc[i]); } } - mCenter = (min+max)*0.5f; - cuv = (min_uv + max_uv)*0.5f; + mCenter->setAdd(min, max); + mCenter->mul(0.5f); - LLVector4a center; - center.load3(mCenter.mV); + cuv = (min_uv + max_uv)*0.5f; LLVector4a binormal; calc_binormal_from_triangle(binormal, - center, cuv, + *mCenter, cuv, pos[0], tc[0], pos[1], tc[1]); binormal.normalize3fast(); @@ -5407,8 +5596,8 @@ BOOL LLVolumeFace::createCap(LLVolume* volume, BOOL partial_build) LLVector4a d0, d1; - d0.setSub(center, pos[0]); - d1.setSub(center, pos[1]); + d0.setSub(*mCenter, pos[0]); + d1.setSub(*mCenter, pos[1]); if (mTypeMask & TOP_MASK) { @@ -5422,12 +5611,12 @@ BOOL LLVolumeFace::createCap(LLVolume* volume, BOOL partial_build) normal.normalize3fast(); VertexData vd; - vd.getPosition().load3(mCenter.mV); + vd.setPosition(*mCenter); vd.mTexCoord = cuv; if (!(mTypeMask & HOLLOW_MASK) && !(mTypeMask & OPEN_MASK)) { - pos[num_vertices].load4a((F32*) ¢er.mQ); + pos[num_vertices] = *mCenter; tc[num_vertices] = cuv; num_vertices++; } @@ -5812,6 +6001,11 @@ void LLVolumeFace::allocateBinormals(S32 num_verts) mBinormals = (LLVector4a*) _mm_malloc(num_verts*16, 16); } +void LLVolumeFace::allocateWeights(S32 num_verts) +{ + _mm_free(mWeights); + mWeights = (LLVector4a*) _mm_malloc(num_verts*16, 16); +} void LLVolumeFace::resizeIndices(S32 num_indices) { @@ -5919,11 +6113,11 @@ void LLVolumeFace::appendFace(const LLVolumeFace& face, LLMatrix4& mat_in, LLMat if (offset == 0 && i == 0) { - mExtents[0] = mExtents[1] = LLVector3((F32*) &(dst_pos[i].mQ)); + mExtents[0] = mExtents[1] = dst_pos[i]; } else { - update_min_max(mExtents[0], mExtents[1], (F32*) &(dst_pos[i].mQ)); + update_min_max(mExtents[0], mExtents[1], dst_pos[i]); } } @@ -6076,18 +6270,19 @@ BOOL LLVolumeFace::createSide(LLVolume* volume, BOOL partial_build) //get bounding box for this side - LLVector3& face_min = mExtents[0]; - LLVector3& face_max = mExtents[1]; - mCenter.clearVec(); + LLVector4a& face_min = mExtents[0]; + LLVector4a& face_max = mExtents[1]; + mCenter->clear(); - face_min = face_max = LLVector3((F32*) &(pos[0].mQ)); + face_min = face_max = pos[0]; for (U32 i = 1; i < mNumVertices; ++i) { - update_min_max(face_min, face_max, (F32*) &(pos[i].mQ)); + update_min_max(face_min, face_max, pos[i]); } - mCenter = (face_min + face_max) * 0.5f; + mCenter->setAdd(face_min, face_max); + mCenter->mul(0.5f); S32 cur_index = 0; S32 cur_edge = 0; diff --git a/indra/llmath/llvolume.h b/indra/llmath/llvolume.h index aa58d6d114..7c63266aab 100644 --- a/indra/llmath/llvolume.h +++ b/indra/llmath/llvolume.h @@ -50,12 +50,13 @@ class LLVolume; #include "v2math.h" #include "v3math.h" #include "v4math.h" +#include "llvector4a.h" #include "llquaternion.h" #include "llstrider.h" #include "v4coloru.h" #include "llrefcount.h" #include "llfile.h" -#include "llvector4a.h" + //============================================================================ @@ -801,59 +802,19 @@ public: }; private: - void init() - { - mData = (LLVector4a*) _mm_malloc(32, 16); - } + void init(); public: - VertexData() - { - init(); - } - - VertexData(const VertexData& rhs) - { - init(); - LLVector4a::memcpyNonAliased16((F32*) mData, (F32*) rhs.mData, 8); - mTexCoord = rhs.mTexCoord; - } - - ~VertexData() - { - _mm_free(mData); - } - - LLVector4a& getPosition() - { - return mData[POSITION]; - } - - LLVector4a& getNormal() - { - return mData[NORMAL]; - } - - const LLVector4a& getPosition() const - { - return mData[POSITION]; - } - - const LLVector4a& getNormal() const - { - return mData[NORMAL]; - } + VertexData(); + VertexData(const VertexData& rhs); + ~VertexData(); + LLVector4a& getPosition(); + LLVector4a& getNormal(); + const LLVector4a& getPosition() const; + const LLVector4a& getNormal() const; + void setPosition(const LLVector4a& pos); + void setNormal(const LLVector4a& norm); - void setPosition(const LLVector4a& pos) - { - mData[POSITION] = pos; - } - - void setNormal(const LLVector4a& norm) - { - mData[NORMAL] = norm; - } - LLVector2 mTexCoord; bool operator<(const VertexData& rhs) const; @@ -864,22 +825,9 @@ public: LLVector4a* mData; }; - LLVolumeFace() : - mID(0), - mTypeMask(0), - mBeginS(0), - mBeginT(0), - mNumS(0), - mNumT(0), - mNumVertices(0), - mNumIndices(0), - mPositions(NULL), - mNormals(NULL), - mBinormals(NULL), - mTexCoords(NULL), - mIndices(NULL) - { - } + LLVolumeFace(); + LLVolumeFace(const LLVolumeFace& src); + LLVolumeFace& operator=(const LLVolumeFace& rhs); ~LLVolumeFace(); @@ -890,6 +838,7 @@ public: void resizeVertices(S32 num_verts); void allocateBinormals(S32 num_verts); + void allocateWeights(S32 num_verts); void resizeIndices(S32 num_indices); void fillFromLegacyData(std::vector& v, std::vector& idx); @@ -944,15 +893,15 @@ public: public: S32 mID; U32 mTypeMask; - LLVector3 mCenter; - + // Only used for INNER/OUTER faces S32 mBeginS; S32 mBeginT; S32 mNumS; S32 mNumT; - LLVector3 mExtents[2]; //minimum and maximum point of face + LLVector4a* mExtents; //minimum and maximum point of face + LLVector4a* mCenter; S32 mNumVertices; S32 mNumIndices; @@ -968,7 +917,7 @@ public: //list of skin weights for rigged volumes // format is mWeights[vertex_index].mV[influence] = . // mWeights.size() should be empty or match mVertices.size() - std::vector mWeights; + LLVector4a* mWeights; private: BOOL createUnCutCubeCap(LLVolume* volume, BOOL partial_build = FALSE); @@ -1051,6 +1000,13 @@ public: LLVector3* normal = NULL, // return the surface normal at the intersection point LLVector3* bi_normal = NULL // return the surface bi-normal at the intersection point ); + + S32 lineSegmentIntersect(const LLVector4a& start, const LLVector4a& end, + S32 face = 1, + LLVector3* intersection = NULL, + LLVector2* tex_coord = NULL, + LLVector3* normal = NULL, + LLVector3* bi_normal = NULL); // The following cleans up vertices and triangles, // getting rid of degenerate triangles and duplicate vertices, @@ -1124,6 +1080,7 @@ void calc_binormal_from_triangle( const LLVector4a& pos2, const LLVector2& tex2); +BOOL LLLineSegmentBoxIntersect(const F32* start, const F32* end, const F32* center, const F32* size); BOOL LLLineSegmentBoxIntersect(const LLVector3& start, const LLVector3& end, const LLVector3& center, const LLVector3& size); BOOL LLTriangleRayIntersect(const LLVector3& vert0, const LLVector3& vert1, const LLVector3& vert2, const LLVector3& orig, const LLVector3& dir, diff --git a/indra/llmath/v3math.h b/indra/llmath/v3math.h index 0e7d72e958..75c860a91e 100644 --- a/indra/llmath/v3math.h +++ b/indra/llmath/v3math.h @@ -36,7 +36,6 @@ #include "llerror.h" #include "llmath.h" - #include "llsd.h" class LLVector2; class LLVector4; diff --git a/indra/newview/llface.cpp b/indra/newview/llface.cpp index db3c5cca33..5cdfdbacde 100644 --- a/indra/newview/llface.cpp +++ b/indra/newview/llface.cpp @@ -39,6 +39,7 @@ #include "llviewercontrol.h" #include "llvolume.h" #include "m3math.h" +#include "llmatrix4a.h" #include "v3color.h" #include "lldrawpoolavatar.h" @@ -74,35 +75,43 @@ The resulting texture coordinate is: u = 2(B dot P) v = 2(T dot P) */ -void planarProjection(LLVector2 &tc, const LLVector3& normal, - const LLVector3 &mCenter, const LLVector3& vec) -{ //DONE! - LLVector3 binormal; - float d = normal * LLVector3(1,0,0); +void planarProjection(LLVector2 &tc, const LLVector4a& normal, + const LLVector4a ¢er, const LLVector4a& vec) +{ + LLVector4a binormal; + F32 d = normal[0]; + if (d >= 0.5f || d <= -0.5f) { - binormal = LLVector3(0,1,0); - if (normal.mV[0] < 0) + if (d < 0) { - binormal = -binormal; + binormal.set(0,-1,0); + } + else + { + binormal.set(0, 1, 0); } } else { - binormal = LLVector3(1,0,0); - if (normal.mV[1] > 0) + if (normal[1] > 0) + { + binormal.set(-1,0,0); + } + else { - binormal = -binormal; + binormal.set(1,0,0); } } - LLVector3 tangent = binormal % normal; + LLVector4a tangent; + tangent.setCross3(binormal,normal); - tc.mV[1] = -((tangent*vec)*2 - 0.5f); - tc.mV[0] = 1.0f+((binormal*vec)*2 - 0.5f); + tc.mV[1] = -((tangent.dot3(vec))*2 - 0.5f); + tc.mV[0] = 1.0f+((binormal.dot3(vec))*2 - 0.5f); } -void sphericalProjection(LLVector2 &tc, const LLVector3& normal, - const LLVector3 &mCenter, const LLVector3& vec) +void sphericalProjection(LLVector2 &tc, const LLVector4a& normal, + const LLVector4a &mCenter, const LLVector4a& vec) { //BROKEN /*tc.mV[0] = acosf(vd.mNormal * LLVector3(1,0,0))/3.14159f; @@ -113,7 +122,7 @@ void sphericalProjection(LLVector2 &tc, const LLVector3& normal, }*/ } -void cylindricalProjection(LLVector2 &tc, const LLVector3& normal, const LLVector3 &mCenter, const LLVector3& vec) +void cylindricalProjection(LLVector2 &tc, const LLVector4a& normal, const LLVector4a &mCenter, const LLVector4a& vec) { //BROKEN /*LLVector3 binormal; float d = vd.mNormal * LLVector3(1,0,0); @@ -371,7 +380,14 @@ void LLFace::setSize(S32 num_vertices, S32 num_indices, bool align) if (align) { //allocate vertices in blocks of 4 for alignment - S32 num_vertices = (num_vertices + 0x3) & ~0x3; + num_vertices = (num_vertices + 0x3) & ~0x3; + } + else + { + if (mDrawablep->getVOVolume()) + { + llerrs << "WTF?" << llendl; + } } if (mGeomCount != num_vertices || @@ -722,6 +738,7 @@ BOOL LLFace::genVolumeBBoxes(const LLVolume &volume, S32 f, // mLastVertexBuffer = NULL; //} + //VECTORIZE THIS LLVector3 min,max; if (f >= volume.getNumVolumeFaces()) @@ -732,8 +749,8 @@ BOOL LLFace::genVolumeBBoxes(const LLVolume &volume, S32 f, else { const LLVolumeFace &face = volume.getVolumeFace(f); - min = face.mExtents[0]; - max = face.mExtents[1]; + min.set(face.mExtents[0].getF32()); + max.set(face.mExtents[1].getF32()); } //min, max are in volume space, convert to drawable render space @@ -824,18 +841,26 @@ LLVector2 LLFace::surfaceToTexture(LLVector2 surface_coord, LLVector3 position, return surface_coord; } + //VECTORIZE THIS // see if we have a non-default mapping U8 texgen = getTextureEntry()->getTexGen(); if (texgen != LLTextureEntry::TEX_GEN_DEFAULT) { - LLVector3 center = mDrawablep->getVOVolume()->getVolume()->getVolumeFace(mTEOffset).mCenter; + LLVector4a& center = *(mDrawablep->getVOVolume()->getVolume()->getVolumeFace(mTEOffset).mCenter); - LLVector3 scale = (mDrawablep->getVOVolume()->isVolumeGlobal()) ? LLVector3(1,1,1) : mVObjp->getScale(); - LLVector3 volume_position = mDrawablep->getVOVolume()->agentPositionToVolume(position); - volume_position.scaleVec(scale); + LLVector4a volume_position; + volume_position.load3(mDrawablep->getVOVolume()->agentPositionToVolume(position).mV); - LLVector3 volume_normal = mDrawablep->getVOVolume()->agentDirectionToVolume(normal); - volume_normal.normalize(); + if (!mDrawablep->getVOVolume()->isVolumeGlobal()) + { + LLVector4a scale; + scale.load3(mVObjp->getScale().mV); + volume_position.mul(scale); + } + + LLVector4a volume_normal; + volume_normal.load3(mDrawablep->getVOVolume()->agentDirectionToVolume(normal).mV); + volume_normal.normalize3fast(); switch (texgen) { @@ -928,7 +953,7 @@ static LLFastTimer::DeclareTimer FTM_FACE_GET_GEOM("Face Geom"); BOOL LLFace::getGeometryVolume(const LLVolume& volume, const S32 &f, - const LLMatrix4& mat_vert, const LLMatrix3& mat_normal, + const LLMatrix4& mat_vert_in, const LLMatrix3& mat_norm_in, const U16 &index_offset, bool force_rebuild) { @@ -960,14 +985,18 @@ BOOL LLFace::getGeometryVolume(const LLVolume& volume, } } - LLStrider vertices; + LLStrider vert; + LLVector4a* vertices = NULL; LLStrider tex_coords; LLStrider tex_coords2; - LLStrider normals; + LLVector4a* normals = NULL; + LLStrider norm; LLStrider colors; - LLStrider binormals; + LLVector4a* binormals = NULL; + LLStrider binorm; LLStrider indicesp; - LLStrider weights; + LLVector4a* weights = NULL; + LLStrider wght; BOOL full_rebuild = force_rebuild || mDrawablep->isState(LLDrawable::REBUILD_VOLUME); @@ -982,11 +1011,11 @@ BOOL LLFace::getGeometryVolume(const LLVolume& volume, scale = mVObjp->getScale(); } - BOOL rebuild_pos = full_rebuild || mDrawablep->isState(LLDrawable::REBUILD_POSITION); - BOOL rebuild_color = full_rebuild || mDrawablep->isState(LLDrawable::REBUILD_COLOR); - BOOL rebuild_tcoord = full_rebuild || mDrawablep->isState(LLDrawable::REBUILD_TCOORD); - BOOL rebuild_normal = rebuild_pos && mVertexBuffer->hasDataType(LLVertexBuffer::TYPE_NORMAL); - BOOL rebuild_binormal = rebuild_pos && mVertexBuffer->hasDataType(LLVertexBuffer::TYPE_BINORMAL); + bool rebuild_pos = full_rebuild || mDrawablep->isState(LLDrawable::REBUILD_POSITION); + bool rebuild_color = full_rebuild || mDrawablep->isState(LLDrawable::REBUILD_COLOR); + bool rebuild_tcoord = full_rebuild || mDrawablep->isState(LLDrawable::REBUILD_TCOORD); + bool rebuild_normal = rebuild_pos && mVertexBuffer->hasDataType(LLVertexBuffer::TYPE_NORMAL); + bool rebuild_binormal = rebuild_pos && mVertexBuffer->hasDataType(LLVertexBuffer::TYPE_BINORMAL); bool rebuild_weights = rebuild_pos && mVertexBuffer->hasDataType(LLVertexBuffer::TYPE_WEIGHT4); const LLTextureEntry *tep = mVObjp->getTE(f); @@ -994,19 +1023,23 @@ BOOL LLFace::getGeometryVolume(const LLVolume& volume, if (rebuild_pos) { - mVertexBuffer->getVertexStrider(vertices, mGeomIndex); + mVertexBuffer->getVertexStrider(vert, mGeomIndex); + vertices = (LLVector4a*) vert.get(); } if (rebuild_normal) { - mVertexBuffer->getNormalStrider(normals, mGeomIndex); + mVertexBuffer->getNormalStrider(norm, mGeomIndex); + normals = (LLVector4a*) norm.get(); } if (rebuild_binormal) { - mVertexBuffer->getBinormalStrider(binormals, mGeomIndex); + mVertexBuffer->getBinormalStrider(binorm, mGeomIndex); + binormals = (LLVector4a*) binorm.get(); } if (rebuild_weights) { - mVertexBuffer->getWeight4Strider(weights, mGeomIndex); + mVertexBuffer->getWeight4Strider(wght, mGeomIndex); + weights = (LLVector4a*) wght.get(); } F32 tcoord_xoffset = 0.f ; @@ -1058,8 +1091,6 @@ BOOL LLFace::getGeometryVolume(const LLVolume& volume, LLVector2 tmin, tmax; - - if (rebuild_tcoord) { if (tep) @@ -1142,9 +1173,9 @@ BOOL LLFace::getGeometryVolume(const LLVolume& volume, //bump setup - LLVector3 binormal_dir( -sin_ang, cos_ang, 0 ); - LLVector3 bump_s_primary_light_ray; - LLVector3 bump_t_primary_light_ray; + LLVector4a binormal_dir( -sin_ang, cos_ang, 0 ); + LLVector4a bump_s_primary_light_ray; + LLVector4a bump_t_primary_light_ray; LLQuaternion bump_quat; if (mDrawablep->isActive()) @@ -1196,8 +1227,9 @@ BOOL LLFace::getGeometryVolume(const LLVolume& volume, LLVector3 moon_ray = gSky.getMoonDirection(); LLVector3& primary_light_ray = (sun_ray.mV[VZ] > 0) ? sun_ray : moon_ray; - bump_s_primary_light_ray = offset_multiple * s_scale * primary_light_ray; - bump_t_primary_light_ray = offset_multiple * t_scale * primary_light_ray; + bump_s_primary_light_ray; + bump_s_primary_light_ray.load3((offset_multiple * s_scale * primary_light_ray).mV); + bump_t_primary_light_ray.load3((offset_multiple * t_scale * primary_light_ray).mV); } U8 texgen = getTextureEntry()->getTexGen(); @@ -1206,45 +1238,47 @@ BOOL LLFace::getGeometryVolume(const LLVolume& volume, mVObjp->getVolume()->genBinormals(f); } + LLMatrix4a mat_normal; + + if (rebuild_normal || rebuild_binormal || rebuild_tcoord) + { + mat_normal.loadu(mat_norm_in); + } + //if it's not fullbright and has no normals, bake sunlight based on face normal bool bake_sunlight = !getTextureEntry()->getFullbright() && !mVertexBuffer->hasDataType(LLVertexBuffer::TYPE_NORMAL); - //VECTORIZE THIS - for (S32 i = 0; i < num_vertices; i++) + if (rebuild_tcoord) { - LLVector3 vf_binormal; - if (vf.mBinormals) - { - vf_binormal.setVec(vf.mBinormals[i].getF32()); - } + LLVector4a scalea; + scalea.load3(scale.mV); - LLVector3 vf_normal; - vf_normal.set(vf.mNormals[i].getF32()); - LLVector3 vf_position; - vf_position.set(vf.mPositions[i].getF32()); - - if (rebuild_tcoord) - { + for (S32 i = 0; i < num_vertices; i++) + { LLVector2 tc(vf.mTexCoords[i]); + LLVector4a& norm = vf.mNormals[i]; + + LLVector4a& center = *(vf.mCenter); + if (texgen != LLTextureEntry::TEX_GEN_DEFAULT) { - LLVector3 vec = vf_position; + LLVector4a vec = vf.mPositions[i]; - vec.scaleVec(scale); + vec.mul(scalea); switch (texgen) { case LLTextureEntry::TEX_GEN_PLANAR: - planarProjection(tc, vf_normal, vf.mCenter, vec); + planarProjection(tc, norm, center, vec); break; case LLTextureEntry::TEX_GEN_SPHERICAL: - sphericalProjection(tc, vf_normal, vf.mCenter, vec); + sphericalProjection(tc, norm, center, vec); break; case LLTextureEntry::TEX_GEN_CYLINDRICAL: - cylindricalProjection(tc, vf_normal, vf.mCenter, vec); + cylindricalProjection(tc, norm, center, vec); break; default: break; @@ -1351,68 +1385,84 @@ BOOL LLFace::getGeometryVolume(const LLVolume& volume, *tex_coords++ = tc; - + if (bump_code && mVertexBuffer->hasDataType(LLVertexBuffer::TYPE_TEXCOORD1)) { - LLVector3 tangent = vf_binormal % vf_normal; - - LLMatrix3 tangent_to_object; - tangent_to_object.setRows(tangent, vf_binormal, vf_normal); - LLVector3 binormal = binormal_dir * tangent_to_object; - binormal = binormal * mat_normal; - + LLVector4a tangent; + tangent.setCross3(vf.mBinormals[i], vf.mNormals[i]); + + LLMatrix4a tangent_to_object; + tangent_to_object.setRows(tangent, vf.mBinormals[i], vf.mNormals[i]); + LLVector4a t; + tangent_to_object.rotate(binormal_dir, t); + LLVector4a binormal; + mat_normal.rotate(t, binormal); + + //VECTORIZE THIS if (mDrawablep->isActive()) { - binormal *= bump_quat; + LLVector3 t; + t.set(binormal.getF32()); + t *= bump_quat; + binormal.load3(t.mV); } - binormal.normVec(); - tc += LLVector2( bump_s_primary_light_ray * tangent, bump_t_primary_light_ray * binormal ); + binormal.normalize3fast(); + tc += LLVector2( bump_s_primary_light_ray.dot3(tangent), bump_t_primary_light_ray.dot3(binormal) ); *tex_coords2++ = tc; } } - - if (rebuild_pos) - { - *vertices++ = vf_position * mat_vert; - } - - if (rebuild_normal) - { - LLVector3 normal = vf_normal * mat_normal; - normal.normVec(); - - *normals++ = normal; + } + + if (rebuild_pos) + { + LLMatrix4a mat_vert; + mat_vert.loadu(mat_vert_in); + + for (S32 i = 0; i < num_vertices; i++) + { + mat_vert.affineTransform(vf.mPositions[i], vertices[i]); } + } - if (rebuild_binormal) - { - LLVector3 binormal = vf_binormal * mat_normal; - binormal.normVec(); - *binormals++ = binormal; + if (rebuild_normal) + { + for (S32 i = 0; i < num_vertices; i++) + { + LLVector4a normal; + mat_normal.rotate(vf.mNormals[i], normal); + normal.normalize3fast(); + normals[i] = normal; } + } - if (rebuild_weights && vf.mWeights.size() > i) - { - (*weights++) = vf.mWeights[i]; + if (rebuild_binormal) + { + for (S32 i = 0; i < num_vertices; i++) + { + LLVector4a binormal; + mat_normal.rotate(vf.mBinormals[i], binormal); + binormal.normalize3fast(); + binormals[i] = binormal; } + } + + if (rebuild_weights && vf.mWeights) + { + LLVector4a::memcpyNonAliased16((F32*) weights, (F32*) vf.mWeights, num_vertices/4); + } - if (rebuild_color) - { - if (bake_sunlight) - { - LLVector3 normal = vf_normal * mat_normal; - normal.normVec(); - - F32 da = normal * gPipeline.mSunDir; + if (rebuild_color) + { + LLVector4a src; - *colors++ = LLColor4U(U8(color.mV[0]*da), U8(color.mV[1]*da), U8(color.mV[2]*da), color.mV[3]); - } - else - { - *colors++ = color; - } + src.splat(reinterpret_cast(color.mAll)); + + F32* dst = (F32*) colors.get(); + for (S32 i = 0; i < num_vertices; i+=4) + { + LLVector4a::copy4a(dst+i, (F32*) &src); } } diff --git a/indra/newview/llpanelprimmediacontrols.cpp b/indra/newview/llpanelprimmediacontrols.cpp index 0648d99685..a5804aa04e 100644 --- a/indra/newview/llpanelprimmediacontrols.cpp +++ b/indra/newview/llpanelprimmediacontrols.cpp @@ -574,7 +574,9 @@ void LLPanelPrimMediaControls::updateShape() { const LLVolumeFace& vf = volume->getVolumeFace(mTargetObjectFace); - const LLVector3* ext = vf.mExtents; + LLVector3 ext[2]; + ext[0].set(vf.mExtents[0].getF32()); + ext[1].set(vf.mExtents[1].getF32()); LLVector3 center = (ext[0]+ext[1])*0.5f; LLVector3 size = (ext[1]-ext[0])*0.5f; diff --git a/indra/newview/llvovolume.cpp b/indra/newview/llvovolume.cpp index 8022f81f19..c03ec67c79 100644 --- a/indra/newview/llvovolume.cpp +++ b/indra/newview/llvovolume.cpp @@ -1602,7 +1602,8 @@ void LLVOVolume::updateFaceSize(S32 idx) else { const LLVolumeFace& vol_face = getVolume()->getVolumeFace(idx); - facep->setSize(vol_face.mNumVertices, vol_face.mNumIndices); + facep->setSize(vol_face.mNumVertices, vol_face.mNumIndices, + true); // <--- volume faces should be padded for 16-byte alignment } } -- cgit v1.3 From bac3623e26d62f6e1eb3c6be1f5fcfcf69a0e714 Mon Sep 17 00:00:00 2001 From: Dave Parks Date: Wed, 26 May 2010 14:55:29 -0500 Subject: Faster index buffer copying. --- indra/newview/llface.cpp | 16 ++++++++++++++-- indra/newview/llvovolume.cpp | 1 - 2 files changed, 14 insertions(+), 3 deletions(-) (limited to 'indra/newview/llvovolume.cpp') diff --git a/indra/newview/llface.cpp b/indra/newview/llface.cpp index 5cdfdbacde..26554997c0 100644 --- a/indra/newview/llface.cpp +++ b/indra/newview/llface.cpp @@ -1165,9 +1165,21 @@ BOOL LLFace::getGeometryVolume(const LLVolume& volume, if (full_rebuild) { mVertexBuffer->getIndexStrider(indicesp, mIndicesIndex); - for (U32 i = 0; i < (U32) num_indices; i++) + __m128i* dst = (__m128i*) indicesp.get(); + __m128i* src = (__m128i*) vf.mIndices; + __m128i offset = _mm_set1_epi16(index_offset); + + S32 end = num_indices/8; + + for (S32 i = 0; i < end; i++) + { + __m128i res = _mm_add_epi16(src[i], offset); + _mm_storeu_si128(dst+i, res); + } + + for (S32 i = end*8; i < num_indices; ++i) { - *indicesp++ = vf.mIndices[i] + index_offset; + indicesp[i] = vf.mIndices[i]+index_offset; } } diff --git a/indra/newview/llvovolume.cpp b/indra/newview/llvovolume.cpp index c03ec67c79..a9f3abeef8 100644 --- a/indra/newview/llvovolume.cpp +++ b/indra/newview/llvovolume.cpp @@ -3354,7 +3354,6 @@ void LLVolumeGeometryManager::registerFace(LLSpatialGroup* group, LLFace* facep, model_mat = &(drawable->getRegion()->mRenderMatrix); } - U8 bump = (type == LLRenderPass::PASS_BUMP || type == LLRenderPass::PASS_POST_BUMP) ? facep->getTextureEntry()->getBumpmap() : 0; LLViewerTexture* tex = facep->getTexture(); -- cgit v1.3 From 26ba00b5554d20ee958693ced87b36fa7f6e3d99 Mon Sep 17 00:00:00 2001 From: Dave Parks Date: Thu, 3 Jun 2010 12:52:28 -0500 Subject: Vectorized octree and much of llspatialpartition and lldrawable. Octree driven raycast. --- indra/llmath/llcamera.cpp | 216 ++++++--------- indra/llmath/llcamera.h | 29 +- indra/llmath/lloctree.h | 239 ++++++++-------- indra/llmath/llvolume.cpp | 31 +-- indra/llmath/llvolume.h | 16 +- indra/llrender/llvertexbuffer.h | 11 + indra/llui/lllineeditor.cpp | 9 +- indra/newview/lldrawable.cpp | 163 ++++++----- indra/newview/lldrawable.h | 34 ++- indra/newview/llface.cpp | 164 +++++++---- indra/newview/llface.h | 15 +- indra/newview/llflexibleobject.cpp | 6 +- indra/newview/llflexibleobject.h | 2 +- indra/newview/llselectmgr.cpp | 16 +- indra/newview/llspatialpartition.cpp | 520 ++++++++++++++++++++++------------- indra/newview/llspatialpartition.h | 80 ++++-- indra/newview/llsurfacepatch.cpp | 6 +- indra/newview/llviewerdisplay.cpp | 7 +- indra/newview/llviewerobject.cpp | 40 ++- indra/newview/llviewerobject.h | 4 +- indra/newview/llviewerpartsim.cpp | 4 +- indra/newview/llvoavatar.cpp | 100 ++++--- indra/newview/llvoavatar.h | 10 +- indra/newview/llvopartgroup.cpp | 10 +- indra/newview/llvopartgroup.h | 2 +- indra/newview/llvosurfacepatch.cpp | 19 +- indra/newview/llvosurfacepatch.h | 2 +- indra/newview/llvotree.cpp | 17 +- indra/newview/llvotree.h | 2 +- indra/newview/llvovolume.cpp | 35 +-- indra/newview/llvovolume.h | 6 +- indra/newview/llvowater.cpp | 20 +- indra/newview/llvowater.h | 2 +- indra/newview/pipeline.cpp | 257 +++++++++++++++-- indra/newview/pipeline.h | 9 + 35 files changed, 1333 insertions(+), 770 deletions(-) (limited to 'indra/newview/llvovolume.cpp') diff --git a/indra/llmath/llcamera.cpp b/indra/llmath/llcamera.cpp index 487ed6451f..6b56e4870e 100644 --- a/indra/llmath/llcamera.cpp +++ b/indra/llmath/llcamera.cpp @@ -48,10 +48,10 @@ LLCamera::LLCamera() : mPlaneCount(6), mFrustumCornerDist(0.f) { + alignPlanes(); calculateFrustumPlanes(); } - LLCamera::LLCamera(F32 vertical_fov_rads, F32 aspect_ratio, S32 view_height_in_pixels, F32 near_plane, F32 far_plane) : LLCoordFrame(), mViewHeightInPixels(view_height_in_pixels), @@ -59,6 +59,7 @@ LLCamera::LLCamera(F32 vertical_fov_rads, F32 aspect_ratio, S32 view_height_in_p mPlaneCount(6), mFrustumCornerDist(0.f) { + alignPlanes(); mAspect = llclamp(aspect_ratio, MIN_ASPECT_RATIO, MAX_ASPECT_RATIO); mNearPlane = llclamp(near_plane, MIN_NEAR_PLANE, MAX_NEAR_PLANE); if(far_plane < 0) far_plane = DEFAULT_FAR_PLANE; @@ -67,6 +68,23 @@ LLCamera::LLCamera(F32 vertical_fov_rads, F32 aspect_ratio, S32 view_height_in_p setView(vertical_fov_rads); } +LLCamera::~LLCamera() +{ + +} + +const LLCamera& LLCamera::operator=(const LLCamera& rhs) +{ + memcpy(this, &rhs, sizeof(LLCamera)); + alignPlanes(); + LLVector4a::memcpyNonAliased16((F32*) mAgentPlanes, (F32*) rhs.mAgentPlanes, 4*7); + return *this; +} + +void LLCamera::alignPlanes() +{ + mAgentPlanes = (LLPlane*) LL_NEXT_ALIGNED_ADDRESS(mAgentPlaneBuffer); +} // ---------------- LLCamera::getFoo() member functions ---------------- @@ -91,8 +109,8 @@ F32 LLCamera::getMaxView() const void LLCamera::setUserClipPlane(LLPlane plane) { mPlaneCount = 7; - mAgentPlanes[6].p = plane; - mAgentPlanes[6].mask = calcPlaneMask(plane); + mAgentPlanes[6] = plane; + mPlaneMask[6] = calcPlaneMask(plane); } void LLCamera::disableUserClipPlane() @@ -164,129 +182,66 @@ size_t LLCamera::readFrustumFromBuffer(const char *buffer) // ---------------- test methods ---------------- -S32 LLCamera::AABBInFrustum(const LLVector3 ¢er, const LLVector3& radius) -{ - static const LLVector3 scaler[] = { - LLVector3(-1,-1,-1), - LLVector3( 1,-1,-1), - LLVector3(-1, 1,-1), - LLVector3( 1, 1,-1), - LLVector3(-1,-1, 1), - LLVector3( 1,-1, 1), - LLVector3(-1, 1, 1), - LLVector3( 1, 1, 1) +S32 LLCamera::AABBInFrustum(const LLVector4a ¢er, const LLVector4a& radius) +{ + static const LLVector4a scaler[] = { + LLVector4a(-1,-1,-1), + LLVector4a( 1,-1,-1), + LLVector4a(-1, 1,-1), + LLVector4a( 1, 1,-1), + LLVector4a(-1,-1, 1), + LLVector4a( 1,-1, 1), + LLVector4a(-1, 1, 1), + LLVector4a( 1, 1, 1) }; U8 mask = 0; S32 result = 2; - /*if (mFrustumCornerDist > 0.f && radius.magVecSquared() > mFrustumCornerDist * mFrustumCornerDist) - { //box is larger than frustum, check frustum quads against box planes - - static const LLVector3 dir[] = - { - LLVector3(1, 0, 0), - LLVector3(-1, 0, 0), - LLVector3(0, 1, 0), - LLVector3(0, -1, 0), - LLVector3(0, 0, 1), - LLVector3(0, 0, -1) - }; - - U32 quads[] = + for (U32 i = 0; i < mPlaneCount; i++) + { + mask = mPlaneMask[i]; + if (mask == 0xff) { - 0, 1, 2, 3, - 0, 1, 5, 4, - 2, 3, 7, 6, - 3, 0, 7, 4, - 1, 2, 6, 4, - 4, 5, 6, 7 - }; - - result = 0; - - BOOL total_inside = TRUE; - for (U32 i = 0; i < 6; i++) - { - LLVector3 p = center + radius.scaledVec(dir[i]); - F32 d = -p*dir[i]; - - for (U32 j = 0; j < 6; j++) - { //for each quad - F32 dist = mAgentFrustum[quads[j*4+0]]*dir[i] + d; - if (dist > 0) - { //at least one frustum point is outside the AABB - total_inside = FALSE; - for (U32 k = 1; k < 4; k++) - { //for each other point on quad - if ( mAgentFrustum[quads[j*4+k]]*dir[i]+d <= 0.f) - { //quad is straddling some plane of AABB - return 1; - } - } - } - else - { - for (U32 k = 1; k < 4; k++) - { - if (mAgentFrustum[quads[j*4+k]]*dir[i]+d > 0.f) - { - return 1; - } - } - } - } + continue; } - if (total_inside) + const LLPlane& p = mAgentPlanes[i]; + const LLVector4a& n = reinterpret_cast(p); + float d = p.mV[3]; + LLVector4a rscale; + rscale.setMul(radius, scaler[mask]); + + LLVector4a minp, maxp; + minp.setSub(center, rscale); + maxp.setAdd(center, rscale); + + if (n.dot3(minp) > -d) { - result = 1; + return 0; } - } - else*/ - { - for (U32 i = 0; i < mPlaneCount; i++) + + if (n.dot3(maxp) > -d) { - mask = mAgentPlanes[i].mask; - if (mask == 0xff) - { - continue; - } - LLPlane p = mAgentPlanes[i].p; - LLVector3 n = LLVector3(p); - float d = p.mV[3]; - LLVector3 rscale = radius.scaledVec(scaler[mask]); - - LLVector3 minp = center - rscale; - LLVector3 maxp = center + rscale; - - if (n * minp > -d) - { - return 0; - } - - if (n * maxp > -d) - { - result = 1; - } + result = 1; } } - return result; } -S32 LLCamera::AABBInFrustumNoFarClip(const LLVector3 ¢er, const LLVector3& radius) -{ - static const LLVector3 scaler[] = { - LLVector3(-1,-1,-1), - LLVector3( 1,-1,-1), - LLVector3(-1, 1,-1), - LLVector3( 1, 1,-1), - LLVector3(-1,-1, 1), - LLVector3( 1,-1, 1), - LLVector3(-1, 1, 1), - LLVector3( 1, 1, 1) + +S32 LLCamera::AABBInFrustumNoFarClip(const LLVector4a& center, const LLVector4a& radius) +{ + static const LLVector4a scaler[] = { + LLVector4a(-1,-1,-1), + LLVector4a( 1,-1,-1), + LLVector4a(-1, 1,-1), + LLVector4a( 1, 1,-1), + LLVector4a(-1,-1, 1), + LLVector4a( 1,-1, 1), + LLVector4a(-1, 1, 1), + LLVector4a( 1, 1, 1) }; U8 mask = 0; @@ -299,25 +254,28 @@ S32 LLCamera::AABBInFrustumNoFarClip(const LLVector3 ¢er, const LLVector3& r continue; } - mask = mAgentPlanes[i].mask; + mask = mPlaneMask[i]; if (mask == 0xff) { continue; } - LLPlane p = mAgentPlanes[i].p; - LLVector3 n = LLVector3(p); + + const LLPlane& p = mAgentPlanes[i]; + const LLVector4a& n = reinterpret_cast(p); float d = p.mV[3]; - LLVector3 rscale = radius.scaledVec(scaler[mask]); + LLVector4a rscale; + rscale.setMul(radius, scaler[mask]); - LLVector3 minp = center - rscale; - LLVector3 maxp = center + rscale; + LLVector4a minp, maxp; + minp.setSub(center, rscale); + maxp.setAdd(center, rscale); - if (n * minp > -d) + if (n.dot3(minp) > -d) { return 0; } - if (n * maxp > -d) + if (n.dot3(maxp) > -d) { result = 1; } @@ -447,12 +405,12 @@ int LLCamera::sphereInFrustum(const LLVector3 &sphere_center, const F32 radius) int res = 2; for (int i = 0; i < 6; i++) { - if (mAgentPlanes[i].mask == 0xff) + if (mPlaneMask[i] == 0xff) { continue; } - float d = mAgentPlanes[i].p.dist(sphere_center); + float d = mAgentPlanes[i].dist(sphere_center); if (d > radius) { @@ -644,12 +602,14 @@ void LLCamera::ignoreAgentFrustumPlane(S32 idx) return; } - mAgentPlanes[idx].mask = 0xff; - mAgentPlanes[idx].p.clearVec(); + mPlaneMask[idx] = 0xff; + mAgentPlanes[idx].clearVec(); } void LLCamera::calcAgentFrustumPlanes(LLVector3* frust) { + alignPlanes(); + for (int i = 0; i < 8; i++) { mAgentFrustum[i] = frust[i]; @@ -662,27 +622,27 @@ void LLCamera::calcAgentFrustumPlanes(LLVector3* frust) //order of planes is important, keep most likely to fail in the front of the list //near - frust[0], frust[1], frust[2] - mAgentPlanes[2].p = planeFromPoints(frust[0], frust[1], frust[2]); + mAgentPlanes[2] = planeFromPoints(frust[0], frust[1], frust[2]); //far - mAgentPlanes[5].p = planeFromPoints(frust[5], frust[4], frust[6]); + mAgentPlanes[5] = planeFromPoints(frust[5], frust[4], frust[6]); //left - mAgentPlanes[0].p = planeFromPoints(frust[4], frust[0], frust[7]); + mAgentPlanes[0] = planeFromPoints(frust[4], frust[0], frust[7]); //right - mAgentPlanes[1].p = planeFromPoints(frust[1], frust[5], frust[6]); + mAgentPlanes[1] = planeFromPoints(frust[1], frust[5], frust[6]); //top - mAgentPlanes[4].p = planeFromPoints(frust[3], frust[2], frust[6]); + mAgentPlanes[4] = planeFromPoints(frust[3], frust[2], frust[6]); //bottom - mAgentPlanes[3].p = planeFromPoints(frust[1], frust[0], frust[4]); + mAgentPlanes[3] = planeFromPoints(frust[1], frust[0], frust[4]); //cache plane octant facing mask for use in AABBInFrustum for (U32 i = 0; i < mPlaneCount; i++) { - mAgentPlanes[i].mask = calcPlaneMask(mAgentPlanes[i].p); + mPlaneMask[i] = calcPlaneMask(mAgentPlanes[i]); } } diff --git a/indra/llmath/llcamera.h b/indra/llmath/llcamera.h index d6c5f7bbb1..c40e819dcf 100644 --- a/indra/llmath/llcamera.h +++ b/indra/llmath/llcamera.h @@ -37,6 +37,7 @@ #include "llmath.h" #include "llcoordframe.h" #include "llplane.h" +#include "llvector4a.h" const F32 DEFAULT_FIELD_OF_VIEW = 60.f * DEG_TO_RAD; const F32 DEFAULT_ASPECT_RATIO = 640.f / 480.f; @@ -79,6 +80,14 @@ class LLCamera : public LLCoordFrame { public: + + LLCamera(const LLCamera& rhs) + { + *this = rhs; + } + + const LLCamera& operator=(const LLCamera& rhs); + enum { PLANE_LEFT = 0, PLANE_RIGHT = 1, @@ -129,13 +138,9 @@ private: LLPlane mWorldPlanes[PLANE_NUM]; LLPlane mHorizPlanes[HORIZ_PLANE_NUM]; - struct frustum_plane - { - frustum_plane() : mask(0) {} - LLPlane p; - U8 mask; - }; - frustum_plane mAgentPlanes[7]; //frustum planes in agent space a la gluUnproject (I'm a bastard, I know) - DaveP + LLPlane* mAgentPlanes; //frustum planes in agent space a la gluUnproject (I'm a bastard, I know) - DaveP + U8 mAgentPlaneBuffer[sizeof(LLPlane)*8]; + U8 mPlaneMask[7]; U32 mPlaneCount; //defaults to 6, if setUserClipPlane is called, uses user supplied clip plane in @@ -143,12 +148,14 @@ private: public: LLVector3 mAgentFrustum[8]; //8 corners of 6-plane frustum F32 mFrustumCornerDist; //distance to corner of frustum against far clip plane - LLPlane& getAgentPlane(U32 idx) { return mAgentPlanes[idx].p; } + LLPlane& getAgentPlane(U32 idx) { return mAgentPlanes[idx]; } public: LLCamera(); LLCamera(F32 vertical_fov_rads, F32 aspect_ratio, S32 view_height_in_pixels, F32 near_plane, F32 far_plane); - virtual ~LLCamera(){} // no-op virtual destructor + virtual ~LLCamera(); + + void alignPlanes(); void setUserClipPlane(LLPlane plane); void disableUserClipPlane(); @@ -199,8 +206,8 @@ public: S32 sphereInFrustum(const LLVector3 ¢er, const F32 radius) const; S32 pointInFrustum(const LLVector3 &point) const { return sphereInFrustum(point, 0.0f); } S32 sphereInFrustumFull(const LLVector3 ¢er, const F32 radius) const { return sphereInFrustum(center, radius); } - S32 AABBInFrustum(const LLVector3 ¢er, const LLVector3& radius); - S32 AABBInFrustumNoFarClip(const LLVector3 ¢er, const LLVector3& radius); + S32 AABBInFrustum(const LLVector4a& center, const LLVector4a& radius); + S32 AABBInFrustumNoFarClip(const LLVector4a& center, const LLVector4a& radius); //does a quick 'n dirty sphere-sphere check S32 sphereInFrustumQuick(const LLVector3 &sphere_center, const F32 radius); diff --git a/indra/llmath/lloctree.h b/indra/llmath/lloctree.h index 8bba12783f..ae2259dba0 100644 --- a/indra/llmath/lloctree.h +++ b/indra/llmath/lloctree.h @@ -35,6 +35,7 @@ #include "lltreenode.h" #include "v3math.h" +#include "llvector4a.h" #include #include @@ -44,7 +45,7 @@ #define OCT_ERRS LL_WARNS("OctreeErrors") #endif -#define LL_OCTREE_PARANOIA_CHECK 0 +#define LL_OCTREE_PARANOIA_CHECK 1 #if LL_DARWIN #define LL_OCTREE_MAX_CAPACITY 32 #else @@ -94,23 +95,22 @@ public: typedef LLOctreeNode oct_node; typedef LLOctreeListener oct_listener; - static const U8 OCTANT_POSITIVE_X = 0x01; - static const U8 OCTANT_POSITIVE_Y = 0x02; - static const U8 OCTANT_POSITIVE_Z = 0x04; - - LLOctreeNode( LLVector3d center, - LLVector3d size, + LLOctreeNode( const LLVector4a& center, + const LLVector4a& size, BaseType* parent, - U8 octant = 255) + S32 octant = -1) : mParent((oct_node*)parent), - mCenter(center), - mSize(size), mOctant(octant) { + mD = (LLVector4a*) _mm_malloc(sizeof(LLVector4a)*4, 16); + + mD[CENTER] = center; + mD[SIZE] = size; + updateMinMax(); - if ((mOctant == 255) && mParent) + if ((mOctant == -1) && mParent) { - mOctant = ((oct_node*) mParent)->getOctant(mCenter.mdV); + mOctant = ((oct_node*) mParent)->getOctant(mD[CENTER]); } clearChildren(); @@ -124,43 +124,30 @@ public: { delete getChild(i); } + + _mm_free(mD); } inline const BaseType* getParent() const { return mParent; } - inline void setParent(BaseType* parent) { mParent = (oct_node*) parent; } - inline const LLVector3d& getCenter() const { return mCenter; } - inline const LLVector3d& getSize() const { return mSize; } - inline void setCenter(LLVector3d center) { mCenter = center; } - inline void setSize(LLVector3d size) { mSize = size; } - inline oct_node* getNodeAt(T* data) { return getNodeAt(data->getPositionGroup(), data->getBinRadius()); } - inline U8 getOctant() const { return mOctant; } - inline void setOctant(U8 octant) { mOctant = octant; } + inline void setParent(BaseType* parent) { mParent = (oct_node*) parent; } + inline const LLVector4a& getCenter() const { return mD[CENTER]; } + inline const LLVector4a& getSize() const { return mD[SIZE]; } + inline void setCenter(const LLVector4a& center) { mD[CENTER] = center; } + inline void setSize(const LLVector4a& size) { mD[SIZE] = size; } + inline oct_node* getNodeAt(T* data) { return getNodeAt(data->getPositionGroup(), data->getBinRadius()); } + inline S32 getOctant() const { return mOctant; } + inline void setOctant(S32 octant) { mOctant = octant; } inline const oct_node* getOctParent() const { return (const oct_node*) getParent(); } inline oct_node* getOctParent() { return (oct_node*) getParent(); } - U8 getOctant(const F64 pos[]) const //get the octant pos is in + S32 getOctant(const LLVector4a& pos) const //get the octant pos is in { - U8 ret = 0; - - if (pos[0] > mCenter.mdV[0]) - { - ret |= OCTANT_POSITIVE_X; - } - if (pos[1] > mCenter.mdV[1]) - { - ret |= OCTANT_POSITIVE_Y; - } - if (pos[2] > mCenter.mdV[2]) - { - ret |= OCTANT_POSITIVE_Z; - } - - return ret; + return pos.greaterThan4(mD[CENTER]).getComparisonMask() & 0x7; } - inline bool isInside(const LLVector3d& pos, const F64& rad) const + inline bool isInside(const LLVector4a& pos, const F32& rad) const { - return rad <= mSize.mdV[0]*2.0 && isInside(pos); + return rad <= mD[SIZE][0]*2.f && isInside(pos); } inline bool isInside(T* data) const @@ -168,29 +155,27 @@ public: return isInside(data->getPositionGroup(), data->getBinRadius()); } - bool isInside(const LLVector3d& pos) const + bool isInside(const LLVector4a& pos) const { - const F64& x = pos.mdV[0]; - const F64& y = pos.mdV[1]; - const F64& z = pos.mdV[2]; - - if (x > mMax.mdV[0] || x <= mMin.mdV[0] || - y > mMax.mdV[1] || y <= mMin.mdV[1] || - z > mMax.mdV[2] || z <= mMin.mdV[2]) + S32 gt = pos.greaterThan4(mD[MAX]).getComparisonMask() & 0x7; + if (gt) { return false; } - + + S32 lt = pos.lessEqual4(mD[MIN]).getComparisonMask() & 0x7; + if (lt) + { + return false; + } + return true; } void updateMinMax() { - for (U32 i = 0; i < 3; i++) - { - mMax.mdV[i] = mCenter.mdV[i] + mSize.mdV[i]; - mMin.mdV[i] = mCenter.mdV[i] - mSize.mdV[i]; - } + mD[MAX].setAdd(mD[CENTER], mD[SIZE]); + mD[MIN].setSub(mD[CENTER], mD[SIZE]); } inline oct_listener* getOctListener(U32 index) @@ -203,34 +188,34 @@ public: return contains(xform->getBinRadius()); } - bool contains(F64 radius) + bool contains(F32 radius) { if (mParent == NULL) { //root node contains nothing return false; } - F64 size = mSize.mdV[0]; - F64 p_size = size * 2.0; + F32 size = mD[SIZE][0]; + F32 p_size = size * 2.f; - return (radius <= 0.001 && size <= 0.001) || + return (radius <= 0.001f && size <= 0.001f) || (radius <= p_size && radius > size); } - static void pushCenter(LLVector3d ¢er, const LLVector3d &size, const T* data) + static void pushCenter(LLVector4a ¢er, const LLVector4a &size, const T* data) { - const LLVector3d& pos = data->getPositionGroup(); - for (U32 i = 0; i < 3; i++) - { - if (pos.mdV[i] > center.mdV[i]) - { - center.mdV[i] += size.mdV[i]; - } - else - { - center.mdV[i] -= size.mdV[i]; - } - } + const LLVector4a& pos = data->getPositionGroup(); + + LLVector4a gt = pos.greaterThan4(center); + + LLVector4a up; + up.mQ = _mm_and_ps(size.mQ, gt.mQ); + + LLVector4a down; + down.mQ = _mm_andnot_ps(gt.mQ, size.mQ); + + center.add(up); + center.sub(down); } void accept(oct_traveler* visitor) { visitor->visit(this); } @@ -249,21 +234,21 @@ public: void accept(tree_traveler* visitor) const { visitor->visit(this); } void accept(oct_traveler* visitor) const { visitor->visit(this); } - oct_node* getNodeAt(const LLVector3d& pos, const F64& rad) + oct_node* getNodeAt(const LLVector4a& pos, const F32& rad) { LLOctreeNode* node = this; if (node->isInside(pos, rad)) { //do a quick search by octant - U8 octant = node->getOctant(pos.mdV); + S32 octant = node->getOctant(pos); BOOL keep_going = TRUE; //traverse the tree until we find a node that has no node //at the appropriate octant or is smaller than the object. //by definition, that node is the smallest node that contains // the data - while (keep_going && node->getSize().mdV[0] >= rad) + while (keep_going && node->getSize()[0] >= rad) { keep_going = FALSE; for (U32 i = 0; i < node->getChildCount() && !keep_going; i++) @@ -271,7 +256,7 @@ public: if (node->getChild(i)->getOctant() == octant) { node = node->getChild(i); - octant = node->getOctant(pos.mdV); + octant = node->getOctant(pos); keep_going = TRUE; } } @@ -289,7 +274,7 @@ public: { if (data == NULL) { - //OCT_ERRS << "!!! INVALID ELEMENT ADDED TO OCTREE BRANCH !!!" << llendl; + OCT_ERRS << "!!! INVALID ELEMENT ADDED TO OCTREE BRANCH !!!" << llendl; return false; } LLOctreeNode* parent = getOctParent(); @@ -299,7 +284,7 @@ public: { if (getElementCount() < LL_OCTREE_MAX_CAPACITY && (contains(data->getBinRadius()) || - (data->getBinRadius() > getSize().mdV[0] && + (data->getBinRadius() > getSize()[0] && parent && parent->getElementCount() >= LL_OCTREE_MAX_CAPACITY))) { //it belongs here #if LL_OCTREE_PARANOIA_CHECK @@ -330,16 +315,22 @@ public: } //it's here, but no kids are in the right place, make a new kid - LLVector3d center(getCenter()); - LLVector3d size(getSize()*0.5); + LLVector4a center = getCenter(); + LLVector4a size = getSize(); + size.mul(0.5f); //push center in direction of data LLOctreeNode::pushCenter(center, size, data); // handle case where floating point number gets too small - if( llabs(center.mdV[0] - getCenter().mdV[0]) < F_APPROXIMATELY_ZERO && - llabs(center.mdV[1] - getCenter().mdV[1]) < F_APPROXIMATELY_ZERO && - llabs(center.mdV[2] - getCenter().mdV[2]) < F_APPROXIMATELY_ZERO) + LLVector4a val; + val.setSub(center, getCenter()); + val.setAbs(val); + LLVector4a app_zero; + app_zero.mQ = F_APPROXIMATELY_ZERO_4A; + S32 lt = val.lessThan4(app_zero).getComparisonMask() & 0x7; + + if( lt == 0x7 ) { mData.insert(data); BaseType::insert(data); @@ -357,7 +348,7 @@ public: //make sure no existing node matches this position for (U32 i = 0; i < getChildCount(); i++) { - if (mChild[i]->getCenter() == center) + if (mChild[i]->getCenter().equal3(center)) { OCT_ERRS << "Octree detected duplicate child center and gave up." << llendl; return false; @@ -375,7 +366,7 @@ public: else { //it's not in here, give it to the root - //OCT_ERRS << "Octree insertion failed, starting over from root!" << llendl; + OCT_ERRS << "Octree insertion failed, starting over from root!" << llendl; oct_node* node = this; @@ -482,13 +473,19 @@ public: void addChild(oct_node* child, BOOL silent = FALSE) { #if LL_OCTREE_PARANOIA_CHECK + + if (child->getSize().equal3(getSize())) + { + OCT_ERRS << "Child size is same as parent size!" << llendl; + } + for (U32 i = 0; i < getChildCount(); i++) { - if(mChild[i]->getSize() != child->getSize()) + if(!mChild[i]->getSize().equal3(child->getSize())) { OCT_ERRS <<"Invalid octree child size." << llendl; } - if (mChild[i]->getCenter() == child->getCenter()) + if (mChild[i]->getCenter().equal3(child->getCenter())) { OCT_ERRS <<"Duplicate octree child position." << llendl; } @@ -513,7 +510,7 @@ public: } } - void removeChild(U8 index, BOOL destroy = FALSE) + void removeChild(S32 index, BOOL destroy = FALSE) { for (U32 i = 0; i < this->getListenerCount(); i++) { @@ -554,18 +551,26 @@ public: } } - //OCT_ERRS << "Octree failed to delete requested child." << llendl; + OCT_ERRS << "Octree failed to delete requested child." << llendl; } protected: + typedef enum + { + CENTER = 0, + SIZE = 1, + MAX = 2, + MIN = 3 + } eDName; + + LLVector4a* mD; + + oct_node* mParent; + S32 mOctant; + child_list mChild; element_list mData; - oct_node* mParent; - LLVector3d mCenter; - LLVector3d mSize; - LLVector3d mMax; - LLVector3d mMin; - U8 mOctant; + }; //just like a regular node, except it might expand on insert and compress on balance @@ -576,9 +581,9 @@ public: typedef LLOctreeNode BaseType; typedef LLOctreeNode oct_node; - LLOctreeRoot( LLVector3d center, - LLVector3d size, - BaseType* parent) + LLOctreeRoot(const LLVector4a& center, + const LLVector4a& size, + BaseType* parent) : BaseType(center, size, parent) { } @@ -619,28 +624,33 @@ public: { if (data == NULL) { - //OCT_ERRS << "!!! INVALID ELEMENT ADDED TO OCTREE ROOT !!!" << llendl; + OCT_ERRS << "!!! INVALID ELEMENT ADDED TO OCTREE ROOT !!!" << llendl; return false; } if (data->getBinRadius() > 4096.0) { - //OCT_ERRS << "!!! ELEMENT EXCEEDS MAXIMUM SIZE IN OCTREE ROOT !!!" << llendl; + OCT_ERRS << "!!! ELEMENT EXCEEDS MAXIMUM SIZE IN OCTREE ROOT !!!" << llendl; return false; } - const F64 MAX_MAG = 1024.0*1024.0; + LLVector4a MAX_MAG; + MAX_MAG.splat(1024.f*1024.f); + + const LLVector4a& v = data->getPositionGroup(); + + LLVector4a val; + val.setSub(v, mD[CENTER]); + val.setAbs(val); + S32 lt = val.lessThan4(MAX_MAG).getComparisonMask() & 0x7; - const LLVector3d& v = data->getPositionGroup(); - if (!(fabs(v.mdV[0]-this->mCenter.mdV[0]) < MAX_MAG && - fabs(v.mdV[1]-this->mCenter.mdV[1]) < MAX_MAG && - fabs(v.mdV[2]-this->mCenter.mdV[2]) < MAX_MAG)) + if (lt != 0x7) { - //OCT_ERRS << "!!! ELEMENT EXCEEDS RANGE OF SPATIAL PARTITION !!!" << llendl; + OCT_ERRS << "!!! ELEMENT EXCEEDS RANGE OF SPATIAL PARTITION !!!" << llendl; return false; } - if (this->getSize().mdV[0] > data->getBinRadius() && isInside(data->getPositionGroup())) + if (this->getSize()[0] > data->getBinRadius() && isInside(data->getPositionGroup())) { //we got it, just act like a branch oct_node* node = getNodeAt(data); @@ -656,31 +666,34 @@ public: else if (this->getChildCount() == 0) { //first object being added, just wrap it up - while (!(this->getSize().mdV[0] > data->getBinRadius() && isInside(data->getPositionGroup()))) + while (!(this->getSize()[0] > data->getBinRadius() && isInside(data->getPositionGroup()))) { - LLVector3d center, size; + LLVector4a center, size; center = this->getCenter(); size = this->getSize(); LLOctreeNode::pushCenter(center, size, data); this->setCenter(center); - this->setSize(size*2); + size.mul(2.f); + this->setSize(size); this->updateMinMax(); } LLOctreeNode::insert(data); } else { - while (!(this->getSize().mdV[0] > data->getBinRadius() && isInside(data->getPositionGroup()))) + while (!(this->getSize()[0] > data->getBinRadius() && isInside(data->getPositionGroup()))) { //the data is outside the root node, we need to grow - LLVector3d center(this->getCenter()); - LLVector3d size(this->getSize()); + LLVector4a center(this->getCenter()); + LLVector4a size(this->getSize()); //expand this node - LLVector3d newcenter(center); + LLVector4a newcenter(center); LLOctreeNode::pushCenter(newcenter, size, data); this->setCenter(newcenter); - this->setSize(size*2); + LLVector4a size2 = size; + size2.mul(2.f); + this->setSize(size2); this->updateMinMax(); //copy our children to a new branch diff --git a/indra/llmath/llvolume.cpp b/indra/llmath/llvolume.cpp index c4172de651..72833c019f 100644 --- a/indra/llmath/llvolume.cpp +++ b/indra/llmath/llvolume.cpp @@ -295,7 +295,7 @@ BOOL LLTriangleRayIntersect(const LLVector3& vert0, const LLVector3& vert1, cons } } -class LLVolumeOctreeRebound : public LLOctreeTravelerDepthFirst +class LLVolumeOctreeRebound : public LLOctreeTravelerDepthFirst { public: const LLVolumeFace* mFace; @@ -305,7 +305,7 @@ public: mFace = face; } - virtual void visit(const LLOctreeNode* branch) + virtual void visit(const LLOctreeNode* branch) { LLVolumeOctreeListener* node = (LLVolumeOctreeListener*) branch->getListener(0); @@ -314,12 +314,12 @@ public: if (branch->getElementCount() != 0) { - const LLVolumeFace::Triangle* tri = *(branch->getData().begin()); + const LLVolumeTriangle* tri = *(branch->getData().begin()); min = *(tri->mV[0]); max = *(tri->mV[0]); - for (LLOctreeNode::const_element_iter iter = + for (LLOctreeNode::const_element_iter iter = branch->getData().begin(); iter != branch->getData().end(); ++iter) { //stretch by triangles in node @@ -4394,7 +4394,7 @@ S32 LLVolume::lineSegmentIntersect(const LLVector4a& start, const LLVector4a& en LLVector4a box_size; box_size.setSub(face.mExtents[1], face.mExtents[0]); - if (LLLineSegmentBoxIntersect(start.getF32(), end.getF32(), box_center.getF32(), box_size.getF32())) + if (LLLineSegmentBoxIntersect(start, end, box_center, box_size)) { if (bi_normal != NULL) // if the caller wants binormals, we may need to generate them { @@ -5418,12 +5418,17 @@ void LLVolumeFace::optimize(F32 angle_cutoff) void LLVolumeFace::createOctree() { - mOctree = new LLOctreeRoot(LLVector3d(0,0,0), LLVector3d(1,1,1), NULL); + LLVector4a center; + LLVector4a size; + center.splat(0.f); + size.splat(1.f); + + mOctree = new LLOctreeRoot(center, size, NULL); new LLVolumeOctreeListener(mOctree); for (U32 i = 0; i < mNumIndices; i+= 3) { - Triangle* tri = new Triangle(); + LLPointer tri = new LLVolumeTriangle(); const LLVector4a& v0 = mPositions[mIndices[i]]; const LLVector4a& v1 = mPositions[mIndices[i+1]]; @@ -5449,8 +5454,7 @@ void LLVolumeFace::createOctree() center.setAdd(min, max); center.mul(0.5f); - - tri->mPositionGroup.setVec(center[0], center[1], center[2]); + *tri->mPositionGroup = center; LLVector4a size; size.setSub(max,min); @@ -5464,15 +5468,6 @@ void LLVolumeFace::createOctree() rebound.traverse(mOctree); } -const LLVector3d& LLVolumeFace::Triangle::getPositionGroup() const -{ - return mPositionGroup; -} - -const F64& LLVolumeFace::Triangle::getBinRadius() const -{ - return mRadius; -} void LLVolumeFace::swapData(LLVolumeFace& rhs) { diff --git a/indra/llmath/llvolume.h b/indra/llmath/llvolume.h index 0ae8aa19ca..c49d1c650d 100644 --- a/indra/llmath/llvolume.h +++ b/indra/llmath/llvolume.h @@ -46,6 +46,7 @@ template class LLOctreeNode; class LLVector4a; class LLVolumeFace; class LLVolume; +class LLVolumeTriangle; #include "lldarray.h" #include "lluuid.h" @@ -918,20 +919,7 @@ public: // mWeights.size() should be empty or match mVertices.size() LLVector4a* mWeights; - class Triangle : public LLRefCount - { - public: - const LLVector4a* mV[3]; - U16 mIndex[3]; - - LLVector3d mPositionGroup; - F64 mRadius; - - virtual const LLVector3d& getPositionGroup() const; - virtual const F64& getBinRadius() const; - }; - - LLOctreeNode* mOctree; + LLOctreeNode* mOctree; private: BOOL createUnCutCubeCap(LLVolume* volume, BOOL partial_build = FALSE); diff --git a/indra/llrender/llvertexbuffer.h b/indra/llrender/llvertexbuffer.h index 47146a5ec4..715309b64a 100644 --- a/indra/llrender/llvertexbuffer.h +++ b/indra/llrender/llvertexbuffer.h @@ -79,6 +79,17 @@ protected: class LLVertexBuffer : public LLRefCount { public: + LLVertexBuffer(const LLVertexBuffer& rhs) + { + *this = rhs; + } + + const LLVertexBuffer& operator=(const LLVertexBuffer& rhs) + { + llerrs << "Illegal operation!" << llendl; + return *this; + } + static LLVBOPool sStreamVBOPool; static LLVBOPool sDynamicVBOPool; static LLVBOPool sStreamIBOPool; diff --git a/indra/llui/lllineeditor.cpp b/indra/llui/lllineeditor.cpp index 45f9de8e8d..c0cc294c02 100644 --- a/indra/llui/lllineeditor.cpp +++ b/indra/llui/lllineeditor.cpp @@ -377,7 +377,14 @@ void LLLineEditor::setText(const LLStringExplicit &new_text) setCursor(llmin((S32)mText.length(), getCursor())); // Set current history line to end of history. - mCurrentHistoryLine = mLineHistory.end() - 1; + if (mLineHistory.empty()) + { + mCurrentHistoryLine = mLineHistory.end(); + } + else + { + mCurrentHistoryLine = mLineHistory.end() - 1; + } mPrevText = mText; } diff --git a/indra/newview/lldrawable.cpp b/indra/newview/lldrawable.cpp index 03eee12707..04e433dcfd 100644 --- a/indra/newview/lldrawable.cpp +++ b/indra/newview/lldrawable.cpp @@ -41,6 +41,7 @@ #include "llcriticaldamp.h" #include "llface.h" #include "lllightconstants.h" +#include "llmatrix4a.h" #include "llsky.h" #include "llsurfacepatch.h" #include "llviewercamera.h" @@ -91,8 +92,12 @@ void LLDrawable::incrementVisible() sCurVisible++; sCurPixelAngle = (F32) gViewerWindow->getWindowHeightRaw()/LLViewerCamera::getInstance()->getView(); } + void LLDrawable::init() { + mExtents = (LLVector4a*) _mm_malloc(sizeof(LLVector4a)*3, 32); + mPositionGroup = mExtents + 2; + // mXform mParent = NULL; mRenderType = 0; @@ -121,6 +126,11 @@ void LLDrawable::initClass() void LLDrawable::destroy() { + if (gDebugGL) + { + gPipeline.checkReferences(this); + } + if (isDead()) { sNumZombieDrawables--; @@ -139,6 +149,9 @@ void LLDrawable::destroy() { llinfos << "- Zombie drawables: " << sNumZombieDrawables << llendl; }*/ + + _mm_free(mExtents); + mExtents = mPositionGroup = NULL; } void LLDrawable::markDead() @@ -714,12 +727,14 @@ void LLDrawable::updateDistance(LLCamera& camera, bool force_update) LLFace* facep = getFace(i); if (force_update || facep->getPoolType() == LLDrawPool::POOL_ALPHA) { - LLVector3 box = (facep->mExtents[1] - facep->mExtents[0]) * 0.25f; + LLVector4a box; + box.setSub(facep->mExtents[1], facep->mExtents[0]); + box.mul(0.25f); LLVector3 v = (facep->mCenterLocal-camera.getOrigin()); const LLVector3& at = camera.getAtAxis(); for (U32 j = 0; j < 3; j++) { - v.mV[j] -= box.mV[j] * at.mV[j]; + v.mV[j] -= box[j] * at.mV[j]; } facep->mDistance = v * camera.getAtAxis(); } @@ -728,7 +743,7 @@ void LLDrawable::updateDistance(LLCamera& camera, bool force_update) } else { - pos = LLVector3(getPositionGroup()); + pos = LLVector3(getPositionGroup().getF32()); } pos -= camera.getOrigin(); @@ -777,7 +792,7 @@ BOOL LLDrawable::updateGeometry(BOOL priority) return res; } -void LLDrawable::shiftPos(const LLVector3 &shift_vector) +void LLDrawable::shiftPos(const LLVector4a &shift_vector) { if (isDead()) { @@ -809,9 +824,9 @@ void LLDrawable::shiftPos(const LLVector3 &shift_vector) for (S32 i = 0; i < getNumFaces(); i++) { LLFace *facep = getFace(i); - facep->mCenterAgent += shift_vector; - facep->mExtents[0] += shift_vector; - facep->mExtents[1] += shift_vector; + facep->mCenterAgent += LLVector3(shift_vector.getF32()); + facep->mExtents[0].add(shift_vector); + facep->mExtents[1].add(shift_vector); if (!volume && facep->hasGeometry()) { @@ -820,9 +835,9 @@ void LLDrawable::shiftPos(const LLVector3 &shift_vector) } } - mExtents[0] += shift_vector; - mExtents[1] += shift_vector; - mPositionGroup += LLVector3d(shift_vector); + mExtents[0].add(shift_vector); + mExtents[1].add(shift_vector); + mPositionGroup->add(shift_vector); } else if (mSpatialBridge) { @@ -830,9 +845,9 @@ void LLDrawable::shiftPos(const LLVector3 &shift_vector) } else if (isAvatar()) { - mExtents[0] += shift_vector; - mExtents[1] += shift_vector; - mPositionGroup += LLVector3d(shift_vector); + mExtents[0].add(shift_vector); + mExtents[1].add(shift_vector); + mPositionGroup->add(shift_vector); } mVObjp->onShift(shift_vector); @@ -844,21 +859,26 @@ const LLVector3& LLDrawable::getBounds(LLVector3& min, LLVector3& max) const return mXform.getPositionW(); } -const LLVector3* LLDrawable::getSpatialExtents() const +const LLVector4a* LLDrawable::getSpatialExtents() const { return mExtents; } -void LLDrawable::setSpatialExtents(LLVector3 min, LLVector3 max) +void LLDrawable::setSpatialExtents(const LLVector3& min, const LLVector3& max) +{ + mExtents[0].load3(min.mV); + mExtents[1].load3(max.mV); +} + +void LLDrawable::setSpatialExtents(const LLVector4a& min, const LLVector4a& max) { - LLVector3 size = max - min; mExtents[0] = min; - mExtents[1] = max; + mExtents[1] = max; } -void LLDrawable::setPositionGroup(const LLVector3d& pos) +void LLDrawable::setPositionGroup(const LLVector4a& pos) { - mPositionGroup.setVec(pos); + *mPositionGroup = pos; } void LLDrawable::updateSpatialExtents() @@ -872,7 +892,7 @@ void LLDrawable::updateSpatialExtents() if (mSpatialBridge.notNull()) { - mPositionGroup.setVec(0,0,0); + mPositionGroup->splat(0.f); } } @@ -1083,59 +1103,72 @@ void LLSpatialBridge::updateSpatialExtents() root->rebound(); } - LLXformMatrix* mat = mDrawable->getXform(); - - LLVector3 offset = root->mBounds[0]; - LLVector3 size = root->mBounds[1]; + LLVector4a offset; + LLVector4a size = root->mBounds[1]; - LLVector3 center = LLVector3(0,0,0) * mat->getWorldMatrix(); - LLQuaternion rotation = LLQuaternion(mat->getWorldMatrix()); + //VECTORIZE THIS + LLMatrix4a mat; + mat.loadu(mDrawable->getXform()->getWorldMatrix()); + + LLVector4a t; + t.splat(0.f); + + LLVector4a center; + mat.affineTransform(t, center); - offset *= rotation; - center += offset; + mat.rotate(root->mBounds[0], offset); + center.add(offset); - LLVector3 v[4]; + LLVector4a v[4]; + //get 4 corners of bounding box - v[0] = (size * rotation); - v[1] = (LLVector3(-size.mV[0], -size.mV[1], size.mV[2]) * rotation); - v[2] = (LLVector3(size.mV[0], -size.mV[1], -size.mV[2]) * rotation); - v[3] = (LLVector3(-size.mV[0], size.mV[1], -size.mV[2]) * rotation); + mat.rotate(size,v[0]); - LLVector3& newMin = mExtents[0]; - LLVector3& newMax = mExtents[1]; + LLVector4a scale; + + scale.set(-1.f, -1.f, 1.f); + scale.mul(size); + mat.rotate(scale, v[1]); + + scale.set(1.f, -1.f, -1.f); + scale.mul(size); + mat.rotate(scale, v[2]); + + scale.set(-1.f, 1.f, -1.f); + scale.mul(size); + mat.rotate(scale, v[3]); + + + LLVector4a& newMin = mExtents[0]; + LLVector4a& newMax = mExtents[1]; newMin = newMax = center; for (U32 i = 0; i < 4; i++) { - for (U32 j = 0; j < 3; j++) - { - F32 delta = fabsf(v[i].mV[j]); - F32 min = center.mV[j] - delta; - F32 max = center.mV[j] + delta; - - if (min < newMin.mV[j]) - { - newMin.mV[j] = min; - } - - if (max > newMax.mV[j]) - { - newMax.mV[j] = max; - } - } - } + LLVector4a delta; + delta.setAbs(v[i]); + LLVector4a min; + min.setSub(center, delta); + LLVector4a max; + max.setAdd(center, delta); - LLVector3 diagonal = newMax - newMin; - mRadius = diagonal.magVec() * 0.5f; + newMin.setMin(min); + newMax.setMax(max); + } + + LLVector4a diagonal; + diagonal.setSub(newMax, newMin); + mRadius = diagonal.length3() * 0.5f; - mPositionGroup.setVec((newMin + newMax) * 0.5f); + mPositionGroup->setAdd(newMin,newMax); + mPositionGroup->mul(0.5f); updateBinRadius(); } void LLSpatialBridge::updateBinRadius() { - mBinRadius = llmin((F32) mOctree->getSize().mdV[0]*0.5f, 256.f); + mBinRadius = llmin( mOctree->getSize()[0]*0.5f, 256.f); } LLCamera LLSpatialBridge::transformCamera(LLCamera& camera) @@ -1276,8 +1309,12 @@ void LLSpatialBridge::setVisible(LLCamera& camera_in, std::vector* LLSpatialGroup* group = (LLSpatialGroup*) mOctree->getListener(0); group->rebound(); - LLVector3 center = (mExtents[0] + mExtents[1]) * 0.5f; - LLVector3 size = (mExtents[1]-mExtents[0]) * 0.5f; + LLVector4a center; + center.setAdd(mExtents[0], mExtents[1]); + center.mul(0.5f); + LLVector4a size; + size.setSub(mExtents[1], mExtents[0]); + size.mul(0.5f); if ((LLPipeline::sShadowRender && camera_in.AABBInFrustum(center, size)) || LLPipeline::sImpostorRender || @@ -1389,11 +1426,11 @@ BOOL LLSpatialBridge::updateMove() return TRUE; } -void LLSpatialBridge::shiftPos(const LLVector3& vec) +void LLSpatialBridge::shiftPos(const LLVector4a& vec) { - mExtents[0] += vec; - mExtents[1] += vec; - mPositionGroup += LLVector3d(vec); + mExtents[0].add(vec); + mExtents[1].add(vec); + mPositionGroup->add(vec); } void LLSpatialBridge::cleanupReferences() @@ -1511,7 +1548,7 @@ F32 LLHUDBridge::calcPixelArea(LLSpatialGroup* group, LLCamera& camera) } -void LLHUDBridge::shiftPos(const LLVector3& vec) +void LLHUDBridge::shiftPos(const LLVector4a& vec) { //don't shift hud bridges on region crossing } diff --git a/indra/newview/lldrawable.h b/indra/newview/lldrawable.h index c3c6cbe12f..811ff1801b 100644 --- a/indra/newview/lldrawable.h +++ b/indra/newview/lldrawable.h @@ -41,6 +41,7 @@ #include "v4math.h" #include "m4math.h" #include "v4coloru.h" +#include "llvector4a.h" #include "llquaternion.h" #include "xform.h" #include "llmemtype.h" @@ -66,6 +67,17 @@ const U32 SILHOUETTE_HIGHLIGHT = 0; class LLDrawable : public LLRefCount { public: + LLDrawable(const LLDrawable& rhs) + { + *this = rhs; + } + + const LLDrawable& operator=(const LLDrawable& rhs) + { + llerrs << "Illegal operation!" << llendl; + return *this; + } + static void initClass(); LLDrawable() { init(); } @@ -94,14 +106,14 @@ public: const LLVector3& getPosition() const { return mXform.getPosition(); } const LLVector3& getWorldPosition() const { return mXform.getPositionW(); } const LLVector3 getPositionAgent() const; - const LLVector3d& getPositionGroup() const { return mPositionGroup; } + const LLVector4a& getPositionGroup() const { return *mPositionGroup; } const LLVector3& getScale() const { return mCurrentScale; } void setScale(const LLVector3& scale) { mCurrentScale = scale; } const LLQuaternion& getWorldRotation() const { return mXform.getWorldRotation(); } const LLQuaternion& getRotation() const { return mXform.getRotation(); } F32 getIntensity() const { return llmin(mXform.getScale().mV[0], 4.f); } S32 getLOD() const { return mVObjp ? mVObjp->getLOD() : 1; } - F64 getBinRadius() const { return mBinRadius; } + F32 getBinRadius() const { return mBinRadius; } void getMinMax(LLVector3& min,LLVector3& max) const { mXform.getMinMax(min,max); } LLXformMatrix* getXform() { return &mXform; } @@ -155,7 +167,7 @@ public: void updateSpecialHoverCursor(BOOL enabled); - virtual void shiftPos(const LLVector3 &shift_vector); + virtual void shiftPos(const LLVector4a &shift_vector); S32 getGeneration() const { return mGeneration; } @@ -173,11 +185,12 @@ public: const LLVector3& getBounds(LLVector3& min, LLVector3& max) const; virtual void updateSpatialExtents(); virtual void updateBinRadius(); - const LLVector3* getSpatialExtents() const; - void setSpatialExtents(LLVector3 min, LLVector3 max); - void setPositionGroup(const LLVector3d& pos); - void setPositionGroup(const LLVector3& pos) { setPositionGroup(LLVector3d(pos)); } + const LLVector4a* getSpatialExtents() const; + void setSpatialExtents(const LLVector3& min, const LLVector3& max); + void setSpatialExtents(const LLVector4a& min, const LLVector4a& max); + void setPositionGroup(const LLVector4a& pos); + void setRenderType(S32 type) { mRenderType = type; } BOOL isRenderType(S32 type) { return mRenderType == type; } S32 getRenderType() { return mRenderType; } @@ -288,6 +301,9 @@ public: private: typedef std::vector face_list_t; + LLVector4a* mExtents; + LLVector4a* mPositionGroup; + U32 mState; S32 mRenderType; LLPointer mVObjp; @@ -297,9 +313,7 @@ private: mutable U32 mVisible; F32 mRadius; - LLVector3 mExtents[2]; - LLVector3d mPositionGroup; - F64 mBinRadius; + F32 mBinRadius; S32 mGeneration; LLVector3 mCurrentScale; diff --git a/indra/newview/llface.cpp b/indra/newview/llface.cpp index 98a50ca4e7..b8407a6f5b 100644 --- a/indra/newview/llface.cpp +++ b/indra/newview/llface.cpp @@ -152,6 +152,8 @@ void cylindricalProjection(LLVector2 &tc, const LLVector4a& normal, const LLVect void LLFace::init(LLDrawable* drawablep, LLViewerObject* objp) { + mExtents = (LLVector4a*) _mm_malloc(sizeof(LLVector4a)*2, 16); + mLastUpdateTime = gFrameTimeSeconds; mLastMoveTime = 0.f; mVSize = 0.f; @@ -206,6 +208,12 @@ static LLFastTimer::DeclareTimer FTM_FACE_DEREF("Deref"); void LLFace::destroy() { LLFastTimer t(FTM_DESTROY_FACE); + + if (gDebugGL) + { + gPipeline.checkReferences(this); + } + if(mTexture.notNull()) { LLFastTimer t(FTM_DESTROY_TEXTURE); @@ -260,6 +268,9 @@ void LLFace::destroy() mDrawablep = NULL; mVObjp = NULL; } + + _mm_free(mExtents); + mExtents = NULL; } @@ -725,13 +736,20 @@ static void xform(LLVector2 &tex_coord, F32 cosAng, F32 sinAng, F32 offS, F32 of BOOL LLFace::genVolumeBBoxes(const LLVolume &volume, S32 f, - const LLMatrix4& mat_vert, const LLMatrix3& mat_normal, BOOL global_volume) + const LLMatrix4& mat_vert_in, const LLMatrix3& mat_normal_in, BOOL global_volume) { LLMemType mt1(LLMemType::MTYPE_DRAWABLE); //get bounding box if (mDrawablep->isState(LLDrawable::REBUILD_VOLUME | LLDrawable::REBUILD_POSITION)) { + //VECTORIZE THIS + LLMatrix4a mat_vert; + mat_vert.loadu(mat_vert_in); + + LLMatrix4a mat_normal; + mat_normal.loadu(mat_normal_in); + //if (mDrawablep->isState(LLDrawable::REBUILD_VOLUME)) //{ //vertex buffer no longer valid // mVertexBuffer = NULL; @@ -739,82 +757,96 @@ BOOL LLFace::genVolumeBBoxes(const LLVolume &volume, S32 f, //} //VECTORIZE THIS - LLVector3 min,max; + LLVector4a min,max; if (f >= volume.getNumVolumeFaces()) { - min = LLVector3(-1,-1,-1); - max = LLVector3(1,1,1); - } - else - { - const LLVolumeFace &face = volume.getVolumeFace(f); - min.set(face.mExtents[0].getF32()); - max.set(face.mExtents[1].getF32()); + llwarns << "Generating bounding box for invalid face index!" << llendl; + f = 0; } + const LLVolumeFace &face = volume.getVolumeFace(f); + min = face.mExtents[0]; + max = face.mExtents[1]; + + //min, max are in volume space, convert to drawable render space - LLVector3 center = ((min + max) * 0.5f)*mat_vert; - LLVector3 size = ((max-min) * 0.5f); + LLVector4a center; + LLVector4a t; + t.setAdd(min, max); + t.mul(0.5f); + mat_vert.affineTransform(t, center); + LLVector4a size; + size.setSub(max, min); + size.mul(0.5f); + if (!global_volume) { - size.scaleVec(mDrawablep->getVObj()->getScale()); + //VECTORIZE THIS + LLVector4a scale; + scale.load3(mDrawablep->getVObj()->getScale().mV); + size.mul(scale); } - LLMatrix3 mat = mat_normal; - LLVector3 x = mat.getFwdRow(); - LLVector3 y = mat.getLeftRow(); - LLVector3 z = mat.getUpRow(); - x.normVec(); - y.normVec(); - z.normVec(); + mat_normal.mMatrix[0].normalize3fast(); + mat_normal.mMatrix[1].normalize3fast(); + mat_normal.mMatrix[2].normalize3fast(); + + LLVector4a v[4]; - mat.setRows(x,y,z); + //get 4 corners of bounding box + mat_normal.rotate(size,v[0]); - LLQuaternion rotation = LLQuaternion(mat); + //VECTORIZE THIS + LLVector4a scale; - LLVector3 v[4]; - //get 4 corners of bounding box - v[0] = (size * rotation); - v[1] = (LLVector3(-size.mV[0], -size.mV[1], size.mV[2]) * rotation); - v[2] = (LLVector3(size.mV[0], -size.mV[1], -size.mV[2]) * rotation); - v[3] = (LLVector3(-size.mV[0], size.mV[1], -size.mV[2]) * rotation); + scale.set(-1.f, -1.f, 1.f); + scale.mul(size); + mat_normal.rotate(scale, v[1]); + + scale.set(1.f, -1.f, -1.f); + scale.mul(size); + mat_normal.rotate(scale, v[2]); + + scale.set(-1.f, 1.f, -1.f); + scale.mul(size); + mat_normal.rotate(scale, v[3]); - LLVector3& newMin = mExtents[0]; - LLVector3& newMax = mExtents[1]; + LLVector4a& newMin = mExtents[0]; + LLVector4a& newMax = mExtents[1]; newMin = newMax = center; for (U32 i = 0; i < 4; i++) { - for (U32 j = 0; j < 3; j++) - { - F32 delta = fabsf(v[i].mV[j]); - F32 min = center.mV[j] - delta; - F32 max = center.mV[j] + delta; - - if (min < newMin.mV[j]) - { - newMin.mV[j] = min; - } - - if (max > newMax.mV[j]) - { - newMax.mV[j] = max; - } - } + LLVector4a delta; + delta.setAbs(v[i]); + LLVector4a min; + min.setSub(center, delta); + LLVector4a max; + max.setAdd(center, delta); + + newMin.setMin(min); + newMax.setMax(max); } if (!mDrawablep->isActive()) { - LLVector3 offset = mDrawablep->getRegion()->getOriginAgent(); - newMin += offset; - newMax += offset; + LLVector4a offset; + offset.load3(mDrawablep->getRegion()->getOriginAgent().mV); + newMin.add(offset); + newMax.add(offset); } - mCenterLocal = (newMin+newMax)*0.5f; - LLVector3 tmp = (newMin - newMax) ; - mBoundingSphereRadius = tmp.length() * 0.5f ; + t.setAdd(newMin, newMax); + t.mul(0.5f); + + //VECTORIZE THIS + mCenterLocal.set(t.getF32()); + + t.setSub(newMax,newMin); + t.mul(0.5f); + mBoundingSphereRadius = t.length3(); updateCenterAgent(); } @@ -1647,20 +1679,31 @@ F32 LLFace::getTextureVirtualSize() BOOL LLFace::calcPixelArea(F32& cos_angle_to_view_dir, F32& radius) { + //VECTORIZE THIS //get area of circle around face - LLVector3 center = getPositionAgent(); - LLVector3 size = (mExtents[1] - mExtents[0]) * 0.5f; + LLVector4a center; + center.load3(getPositionAgent().mV); + LLVector4a size; + size.setSub(mExtents[1], mExtents[0]); + size.mul(0.5f); + LLViewerCamera* camera = LLViewerCamera::getInstance(); - F32 size_squared = size.lengthSquared() ; - LLVector3 lookAt = center - camera->getOrigin(); - F32 dist = lookAt.normVec() ; + F32 size_squared = size.dot3(size); + LLVector4a lookAt; + LLVector4a t; + t.load3(camera->getOrigin().mV); + lookAt.setSub(center, t); + F32 dist = lookAt.length3(); + lookAt.normalize3fast() ; //get area of circle around node F32 app_angle = atanf(fsqrtf(size_squared) / dist); radius = app_angle*LLDrawable::sCurPixelAngle; mPixelArea = radius*radius * 3.14159f; - cos_angle_to_view_dir = lookAt * camera->getXAxis() ; + LLVector4a x_axis; + x_axis.load3(camera->getXAxis().mV); + cos_angle_to_view_dir = lookAt.dot3(x_axis); //if has media, check if the face is out of the view frustum. if(hasMedia()) @@ -1676,7 +1719,10 @@ BOOL LLFace::calcPixelArea(F32& cos_angle_to_view_dir, F32& radius) } else { - if(dist * dist * (lookAt - camera->getXAxis()).lengthSquared() < size_squared) + LLVector4a d; + d.setSub(lookAt, x_axis); + + if(dist * dist * d.dot3(d) < size_squared) { cos_angle_to_view_dir = 1.0f ; } diff --git a/indra/newview/llface.h b/indra/newview/llface.h index 48909d7895..0cd472a2fd 100644 --- a/indra/newview/llface.h +++ b/indra/newview/llface.h @@ -65,6 +65,17 @@ class LLFace { public: + LLFace(const LLFace& rhs) + { + *this = rhs; + } + + const LLFace& operator=(const LLFace& rhs) + { + llerrs << "Illegal operation!" << llendl; + return *this; + } + enum EMasks { LIGHT = 0x0001, @@ -221,7 +232,9 @@ public: LLVector3 mCenterLocal; LLVector3 mCenterAgent; - LLVector3 mExtents[2]; + + LLVector4a* mExtents; + LLVector2 mTexExtents[2]; F32 mDistance; LLPointer mVertexBuffer; diff --git a/indra/newview/llflexibleobject.cpp b/indra/newview/llflexibleobject.cpp index 561965d021..8be4e34748 100644 --- a/indra/newview/llflexibleobject.cpp +++ b/indra/newview/llflexibleobject.cpp @@ -97,11 +97,13 @@ void LLVolumeImplFlexible::onParameterChanged(U16 param_type, LLNetworkData *dat } } -void LLVolumeImplFlexible::onShift(const LLVector3 &shift_vector) +void LLVolumeImplFlexible::onShift(const LLVector4a &shift_vector) { + //VECTORIZE THIS + LLVector3 shift(shift_vector.getF32()); for (int section = 0; section < (1<getRenderRotation(); LLVector3 first_grid_obj_pos = first_grid_object->getRenderPosition(); - LLVector3 min_extents(F32_MAX, F32_MAX, F32_MAX); - LLVector3 max_extents(-F32_MAX, -F32_MAX, -F32_MAX); + LLVector4a min_extents(F32_MAX); + LLVector4a max_extents(-F32_MAX); BOOL grid_changed = FALSE; for (LLObjectSelection::iterator iter = mGridObjects.begin(); iter != mGridObjects.end(); ++iter) @@ -1110,7 +1110,7 @@ void LLSelectMgr::getGrid(LLVector3& origin, LLQuaternion &rotation, LLVector3 & LLDrawable* drawable = object->mDrawable; if (drawable) { - const LLVector3* ext = drawable->getSpatialExtents(); + const LLVector4a* ext = drawable->getSpatialExtents(); update_min_max(min_extents, max_extents, ext[0]); update_min_max(min_extents, max_extents, ext[1]); grid_changed = TRUE; @@ -1118,13 +1118,19 @@ void LLSelectMgr::getGrid(LLVector3& origin, LLQuaternion &rotation, LLVector3 & } if (grid_changed) { - mGridOrigin = lerp(min_extents, max_extents, 0.5f); + LLVector4a center, size; + center.setAdd(min_extents, max_extents); + center.mul(0.5f); + size.setSub(max_extents, min_extents); + size.mul(0.5f); + + mGridOrigin.set(center.getF32()); LLDrawable* drawable = first_grid_object->mDrawable; if (drawable && drawable->isActive()) { mGridOrigin = mGridOrigin * first_grid_object->getRenderMatrix(); } - mGridScale = (max_extents - min_extents) * 0.5f; + mGridScale.set(size.getF32()); } } else // GRID_MODE_WORLD or just plain default diff --git a/indra/newview/llspatialpartition.cpp b/indra/newview/llspatialpartition.cpp index 355173772b..3cf0138303 100644 --- a/indra/newview/llspatialpartition.cpp +++ b/indra/newview/llspatialpartition.cpp @@ -102,23 +102,6 @@ void sg_assert(BOOL expr) #endif } -#if LL_DEBUG -void validate_drawable(LLDrawable* drawablep) -{ - F64 rad = drawablep->getBinRadius(); - const LLVector3* ext = drawablep->getSpatialExtents(); - - if (rad < 0 || rad > 4096 || - (ext[1]-ext[0]).magVec() > 4096) - { - llwarns << "Invalid drawable found in octree." << llendl; - } -} -#else -#define validate_drawable(x) -#endif - - S32 AABBSphereIntersect(const LLVector3& min, const LLVector3& max, const LLVector3 &origin, const F32 &rad) { return AABBSphereIntersectR2(min, max, origin, rad*rad); @@ -158,6 +141,55 @@ S32 AABBSphereIntersectR2(const LLVector3& min, const LLVector3& max, const LLVe } +S32 AABBSphereIntersect(const LLVector4a& min, const LLVector4a& max, const LLVector3 &origin, const F32 &rad) +{ + return AABBSphereIntersectR2(min, max, origin, rad*rad); +} + +S32 AABBSphereIntersectR2(const LLVector4a& min, const LLVector4a& max, const LLVector3 &origin, const F32 &r) +{ + F32 d = 0.f; + F32 t; + + LLVector4a origina; + origina.load3(origin.mV); + + LLVector4a v; + v.setSub(min, origina); + + if (v.dot3(v) < r) + { + v.setSub(max, origina); + if (v.dot3(v) < r) + { + return 2; + } + } + + + for (U32 i = 0; i < 3; i++) + { + if (origin.mV[i] < min[i]) + { + t = min[i] - origin.mV[i]; + d += t*t; + } + else if (origin.mV[i] > max[i]) + { + t = origin.mV[i] - max[i]; + d += t*t; + } + + if (d > r) + { + return 0; + } + } + + return 1; +} + + typedef enum { b000 = 0x00, @@ -193,24 +225,13 @@ static U8 sOcclusionIndices[] = b000, b110, b100, b101, b001, b011, b010, b110, }; -U8* get_box_fan_indices(LLCamera* camera, const LLVector3& center) +U8* get_box_fan_indices(LLCamera* camera, const LLVector4a& center) { - LLVector3 d = center - camera->getOrigin(); - - U8 cypher = 0; - if (d.mV[0] > 0) - { - cypher |= b100; - } - if (d.mV[1] > 0) - { - cypher |= b010; - } - if (d.mV[2] > 0) - { - cypher |= b001; - } + LLVector4a origin; + origin.load3(camera->getOrigin().mV); + S32 cypher = center.greaterThan4(origin).getComparisonMask() & 0x7; + return sOcclusionIndices+cypher*8; } @@ -218,33 +239,49 @@ void LLSpatialGroup::buildOcclusion() { if (!mOcclusionVerts) { - mOcclusionVerts = new F32[8*3]; + mOcclusionVerts = (LLVector4a*) _mm_malloc(sizeof(LLVector4a)*8, 16); } - LLVector3 r = mBounds[1] + LLVector3(SG_OCCLUSION_FUDGE, SG_OCCLUSION_FUDGE, SG_OCCLUSION_FUDGE); + LLVector4a fudge; + fudge.splat(SG_OCCLUSION_FUDGE); - for (U32 k = 0; k < 3; k++) - { - r.mV[k] = llmin(mBounds[1].mV[k]+0.25f, r.mV[k]); - } + LLVector4a r; + r.setAdd(mBounds[1], fudge); - F32* v = mOcclusionVerts; - F32* c = mBounds[0].mV; - F32* s = r.mV; + LLVector4a r2; + r2.splat(0.25f); + r2.add(mBounds[1]); + + r.setMin(r2); + + LLVector4a* v = mOcclusionVerts; + const LLVector4a& c = mBounds[0]; + const LLVector4a& s = r; + static const LLVector4a octant[] = + { + LLVector4a(-1.f, -1.f, -1.f), + LLVector4a(-1.f, -1.f, 1.f), + LLVector4a(-1.f, 1.f, -1.f), + LLVector4a(-1.f, 1.f, 1.f), + + LLVector4a(1.f, -1.f, -1.f), + LLVector4a(1.f, -1.f, 1.f), + LLVector4a(1.f, 1.f, -1.f), + LLVector4a(1.f, 1.f, 1.f), + }; + //vertex positions are encoded so the 3 bits of their vertex index //correspond to their axis facing, with bit position 3,2,1 matching //axis facing x,y,z, bit set meaning positive facing, bit clear //meaning negative facing - v[0] = c[0]-s[0]; v[1] = c[1]-s[1]; v[2] = c[2]-s[2]; // 0 - 0000 - v[3] = c[0]-s[0]; v[4] = c[1]-s[1]; v[5] = c[2]+s[2]; // 1 - 0001 - v[6] = c[0]-s[0]; v[7] = c[1]+s[1]; v[8] = c[2]-s[2]; // 2 - 0010 - v[9] = c[0]-s[0]; v[10] = c[1]+s[1]; v[11] = c[2]+s[2]; // 3 - 0011 - - v[12] = c[0]+s[0]; v[13] = c[1]-s[1]; v[14] = c[2]-s[2]; // 4 - 0100 - v[15] = c[0]+s[0]; v[16] = c[1]-s[1]; v[17] = c[2]+s[2]; // 5 - 0101 - v[18] = c[0]+s[0]; v[19] = c[1]+s[1]; v[20] = c[2]-s[2]; // 6 - 0110 - v[21] = c[0]+s[0]; v[22] = c[1]+s[1]; v[23] = c[2]+s[2]; // 7 - 0111 + + for (S32 i = 0; i < 8; ++i) + { + v[i] = s; + v[i].mul(octant[i]); + v[i].add(c); + } clearState(LLSpatialGroup::OCCLUSION_DIRTY); } @@ -288,6 +325,11 @@ LLSpatialGroup::~LLSpatialGroup() llerrs << "Illegal deletion of LLSpatialGroup!" << llendl; }*/ + if (gDebugGL) + { + gPipeline.checkReferences(this); + } + if (isState(DEAD)) { sZombieGroups--; @@ -300,11 +342,13 @@ LLSpatialGroup::~LLSpatialGroup() sQueryPool.release(mOcclusionQuery[LLViewerCamera::sCurCameraID]); } - delete [] mOcclusionVerts; + _mm_free(mOcclusionVerts); LLMemType mt(LLMemType::MTYPE_SPACE_PARTITION); clearDrawMap(); clearAtlasList() ; + + _mm_free(mBounds); } BOOL LLSpatialGroup::hasAtlas(LLTextureAtlas* atlasp) @@ -456,8 +500,10 @@ void LLSpatialGroup::validate() sg_assert(!isState(DIRTY)); sg_assert(!isDead()); - LLVector3 myMin = mBounds[0] - mBounds[1]; - LLVector3 myMax = mBounds[0] + mBounds[1]; + LLVector4a myMin; + myMin.setSub(mBounds[0], mBounds[1]); + LLVector4a myMax; + myMax.setAdd(mBounds[0], mBounds[1]); validateDrawMap(); @@ -489,16 +535,18 @@ void LLSpatialGroup::validate() group->validate(); //ensure all children are enclosed in this node - LLVector3 center = group->mBounds[0]; - LLVector3 size = group->mBounds[1]; + LLVector4a center = group->mBounds[0]; + LLVector4a size = group->mBounds[1]; - LLVector3 min = center - size; - LLVector3 max = center + size; + LLVector4a min; + min.setSub(center, size); + LLVector4a max; + max.setAdd(center, size); for (U32 j = 0; j < 3; j++) { - sg_assert(min.mV[j] >= myMin.mV[j]-0.02f); - sg_assert(max.mV[j] <= myMax.mV[j]+0.02f); + sg_assert(min[j] >= myMin[j]-0.02f); + sg_assert(max[j] <= myMax[j]+0.02f); } } @@ -508,8 +556,8 @@ void LLSpatialGroup::validate() void LLSpatialGroup::checkStates() { #if LL_OCTREE_PARANOIA_CHECK - LLOctreeStateCheck checker; - checker.traverse(mOctreeNode); + //LLOctreeStateCheck checker; + //checker.traverse(mOctreeNode); #endif } @@ -534,19 +582,17 @@ BOOL LLSpatialGroup::updateInGroup(LLDrawable *drawablep, BOOL immediate) LLMemType mt(LLMemType::MTYPE_SPACE_PARTITION); drawablep->updateSpatialExtents(); - validate_drawable(drawablep); OctreeNode* parent = mOctreeNode->getOctParent(); if (mOctreeNode->isInside(drawablep->getPositionGroup()) && (mOctreeNode->contains(drawablep) || - (drawablep->getBinRadius() > mOctreeNode->getSize().mdV[0] && + (drawablep->getBinRadius() > mOctreeNode->getSize()[0] && parent && parent->getElementCount() >= LL_OCTREE_MAX_CAPACITY))) { unbound(); setState(OBJECT_DIRTY); //setState(GEOM_DIRTY); - validate_drawable(drawablep); return TRUE; } @@ -564,7 +610,6 @@ BOOL LLSpatialGroup::addObject(LLDrawable *drawablep, BOOL add_all, BOOL from_oc else { drawablep->setSpatialGroup(this); - validate_drawable(drawablep); setState(OBJECT_DIRTY | GEOM_DIRTY); setOcclusionState(LLSpatialGroup::DISCARD_QUERY, LLSpatialGroup::STATE_MODE_ALL_CAMERAS); gPipeline.markRebuild(this, TRUE); @@ -665,7 +710,7 @@ void LLSpatialPartition::rebuildMesh(LLSpatialGroup* group) } -BOOL LLSpatialGroup::boundObjects(BOOL empty, LLVector3& minOut, LLVector3& maxOut) +BOOL LLSpatialGroup::boundObjects(BOOL empty, LLVector4a& minOut, LLVector4a& maxOut) { const OctreeNode* node = mOctreeNode; @@ -678,8 +723,8 @@ BOOL LLSpatialGroup::boundObjects(BOOL empty, LLVector3& minOut, LLVector3& maxO return FALSE; } - LLVector3& newMin = mObjectExtents[0]; - LLVector3& newMax = mObjectExtents[1]; + LLVector4a& newMin = mObjectExtents[0]; + LLVector4a& newMax = mObjectExtents[1]; if (isState(OBJECT_DIRTY)) { //calculate new bounding box @@ -688,10 +733,10 @@ BOOL LLSpatialGroup::boundObjects(BOOL empty, LLVector3& minOut, LLVector3& maxO //initialize bounding box to first element OctreeNode::const_element_iter i = node->getData().begin(); LLDrawable* drawablep = *i; - const LLVector3* minMax = drawablep->getSpatialExtents(); + const LLVector4a* minMax = drawablep->getSpatialExtents(); - newMin.setVec(minMax[0]); - newMax.setVec(minMax[1]); + newMin = minMax[0]; + newMax = minMax[1]; for (++i; i != node->getData().end(); ++i) { @@ -715,8 +760,10 @@ BOOL LLSpatialGroup::boundObjects(BOOL empty, LLVector3& minOut, LLVector3& maxO }*/ } - mObjectBounds[0] = (newMin + newMax) * 0.5f; - mObjectBounds[1] = (newMax - newMin) * 0.5f; + mObjectBounds[0].setAdd(newMin, newMax); + mObjectBounds[0].mul(0.5f); + mObjectBounds[1].setSub(newMax, newMin); + mObjectBounds[1].mul(0.5f); } if (empty) @@ -726,17 +773,8 @@ BOOL LLSpatialGroup::boundObjects(BOOL empty, LLVector3& minOut, LLVector3& maxO } else { - for (U32 i = 0; i < 3; i++) - { - if (newMin.mV[i] < minOut.mV[i]) - { - minOut.mV[i] = newMin.mV[i]; - } - if (newMax.mV[i] > maxOut.mV[i]) - { - maxOut.mV[i] = newMax.mV[i]; - } - } + minOut.setMin(newMin); + maxOut.setMax(newMax); } return TRUE; @@ -827,18 +865,19 @@ BOOL LLSpatialGroup::removeObject(LLDrawable *drawablep, BOOL from_octree) return TRUE; } -void LLSpatialGroup::shift(const LLVector3 &offset) +void LLSpatialGroup::shift(const LLVector4a &offset) { LLMemType mt(LLMemType::MTYPE_SPACE_PARTITION); - LLVector3d offsetd(offset); - mOctreeNode->setCenter(mOctreeNode->getCenter()+offsetd); + LLVector4a t = mOctreeNode->getCenter(); + t.add(offset); + mOctreeNode->setCenter(t); mOctreeNode->updateMinMax(); - mBounds[0] += offset; - mExtents[0] += offset; - mExtents[1] += offset; - mObjectBounds[0] += offset; - mObjectExtents[0] += offset; - mObjectExtents[1] += offset; + mBounds[0].add(offset); + mExtents[0].add(offset); + mExtents[1].add(offset); + mObjectBounds[0].add(offset); + mObjectExtents[0].add(offset); + mObjectExtents[1].add(offset); //if (!mSpatialPartition->mRenderByGroup) { @@ -850,10 +889,7 @@ void LLSpatialGroup::shift(const LLVector3 &offset) { for (U32 i = 0; i < 8; i++) { - F32* v = mOcclusionVerts+i*3; - v[0] += offset.mV[0]; - v[1] += offset.mV[1]; - v[2] += offset.mV[2]; + mOcclusionVerts[i].add(offset); } } } @@ -1119,8 +1155,6 @@ LLSpatialGroup::LLSpatialGroup(OctreeNode* node, LLSpatialPartition* part) : mDepth(0.f), mLastUpdateDistance(-1.f), mLastUpdateTime(gFrameTimeSeconds), - mViewAngle(0.f), - mLastUpdateViewAngle(-1.f), mAtlasList(4), mCurUpdatingTime(0), mCurUpdatingSlotp(NULL), @@ -1129,13 +1163,25 @@ LLSpatialGroup::LLSpatialGroup(OctreeNode* node, LLSpatialPartition* part) : sNodeCount++; LLMemType mt(LLMemType::MTYPE_SPACE_PARTITION); + mBounds = (LLVector4a*) _mm_malloc(sizeof(LLVector4a) * V4_COUNT, 16); + mExtents = mBounds + EXTENTS; + mObjectBounds = mBounds + OBJECT_BOUNDS; + mObjectExtents = mBounds + OBJECT_EXTENTS; + mViewAngle = mBounds+VIEW_ANGLE; + mLastUpdateViewAngle = mBounds+LAST_VIEW_ANGLE; + + mViewAngle->splat(0.f); + mLastUpdateViewAngle->splat(-1.f); + mExtents[0] = mExtents[1] = mObjectBounds[0] = mObjectBounds[0] = mObjectBounds[1] = + mObjectExtents[0] = mObjectExtents[1] = *mViewAngle; + sg_assert(mOctreeNode->getListenerCount() == 0); mOctreeNode->addListener(this); setState(SG_INITIAL_STATE_MASK); gPipeline.markRebuild(this, TRUE); - mBounds[0] = LLVector3(node->getCenter()); - mBounds[1] = LLVector3(node->getSize()); + mBounds[0] = node->getCenter(); + mBounds[1] = node->getSize(); part->mLODSeed = (part->mLODSeed+1)%part->mLODPeriod; mLODHash = part->mLODSeed; @@ -1172,8 +1218,8 @@ void LLSpatialGroup::updateDistance(LLCamera &camera) #endif if (!getData().empty()) { - mRadius = mSpatialPartition->mRenderByGroup ? mObjectBounds[1].magVec() : - (F32) mOctreeNode->getSize().magVec(); + mRadius = mSpatialPartition->mRenderByGroup ? mObjectBounds[1].length3() : + (F32) mOctreeNode->getSize().length3(); mDistance = mSpatialPartition->calcDistance(this, camera); mPixelArea = mSpatialPartition->calcPixelArea(this, camera); } @@ -1181,27 +1227,34 @@ void LLSpatialGroup::updateDistance(LLCamera &camera) F32 LLSpatialPartition::calcDistance(LLSpatialGroup* group, LLCamera& camera) { - LLVector3 eye = group->mObjectBounds[0] - camera.getOrigin(); + LLVector4a eye; + LLVector4a origin; + origin.load3(camera.getOrigin().mV); + + eye.setSub(group->mObjectBounds[0], origin); F32 dist = 0.f; if (group->mDrawMap.find(LLRenderPass::PASS_ALPHA) != group->mDrawMap.end()) { - LLVector3 v = eye; - dist = eye.normVec(); + LLVector4a v = eye; + + dist = eye.length3(); + eye.normalize3fast(); if (!group->isState(LLSpatialGroup::ALPHA_DIRTY)) { if (!group->mSpatialPartition->isBridge()) { - LLVector3 view_angle = LLVector3(eye * LLVector3(1,0,0), - eye * LLVector3(0,1,0), - eye * LLVector3(0,0,1)); + LLVector4a view_angle = eye; - if ((view_angle-group->mLastUpdateViewAngle).magVec() > 0.64f) + LLVector4a diff; + diff.setSub(view_angle, *group->mLastUpdateViewAngle); + + if (diff.length3() > 0.64f) { - group->mViewAngle = view_angle; - group->mLastUpdateViewAngle = view_angle; + *group->mViewAngle = view_angle; + *group->mLastUpdateViewAngle = view_angle; //for occasional alpha sorting within the group //NOTE: If there is a trivial way to detect that alpha sorting here would not change the render order, //not setting this node to dirty would be a very good thing @@ -1215,17 +1268,20 @@ F32 LLSpatialPartition::calcDistance(LLSpatialGroup* group, LLCamera& camera) LLVector3 at = camera.getAtAxis(); - //front of bounding box - for (U32 i = 0; i < 3; i++) - { - v.mV[i] -= group->mObjectBounds[1].mV[i]*0.25f * at.mV[i]; - } + LLVector4a ata; + ata.load3(at.mV); - group->mDepth = v * at; + LLVector4a t = ata; + //front of bounding box + t.mul(0.25f); + t.mul(group->mObjectBounds[1]); + v.sub(t); + + group->mDepth = v.dot3(ata); } else { - dist = eye.magVec(); + dist = eye.length3(); } if (dist < 16.f) @@ -1378,7 +1434,7 @@ void LLSpatialGroup::destroyGL() } } - delete [] mOcclusionVerts; + _mm_free(mOcclusionVerts); mOcclusionVerts = NULL; for (LLSpatialGroup::element_iter i = getData().begin(); i != getData().end(); ++i) @@ -1421,8 +1477,8 @@ BOOL LLSpatialGroup::rebound() } else { - LLVector3& newMin = mExtents[0]; - LLVector3& newMax = mExtents[1]; + LLVector4a& newMin = mExtents[0]; + LLVector4a& newMax = mExtents[1]; LLSpatialGroup* group = (LLSpatialGroup*) mOctreeNode->getChild(0)->getListener(0); group->clearState(SKIP_FRUSTUM_CHECK); group->rebound(); @@ -1436,26 +1492,19 @@ BOOL LLSpatialGroup::rebound() group = (LLSpatialGroup*) mOctreeNode->getChild(i)->getListener(0); group->clearState(SKIP_FRUSTUM_CHECK); group->rebound(); - const LLVector3& max = group->mExtents[1]; - const LLVector3& min = group->mExtents[0]; + const LLVector4a& max = group->mExtents[1]; + const LLVector4a& min = group->mExtents[0]; - for (U32 j = 0; j < 3; j++) - { - if (max.mV[j] > newMax.mV[j]) - { - newMax.mV[j] = max.mV[j]; - } - if (min.mV[j] < newMin.mV[j]) - { - newMin.mV[j] = min.mV[j]; - } - } + newMax.setMax(max); + newMin.setMin(min); } boundObjects(FALSE, newMin, newMax); - mBounds[0] = (newMin + newMax)*0.5f; - mBounds[1] = (newMax - newMin)*0.5f; + mBounds[0].setAdd(newMin, newMax); + mBounds[0].mul(0.5f); + mBounds[1].setSub(newMax, newMin); + mBounds[1].mul(0.5f); } setState(OCCLUSION_DIRTY); @@ -1540,7 +1589,7 @@ void LLSpatialGroup::doOcclusion(LLCamera* camera) } glBeginQueryARB(GL_SAMPLES_PASSED_ARB, mOcclusionQuery[LLViewerCamera::sCurCameraID]); - glVertexPointer(3, GL_FLOAT, 0, mOcclusionVerts); + glVertexPointer(3, GL_FLOAT, 16, mOcclusionVerts); if (camera->getOrigin().isExactlyZero()) { //origin is invalid, draw entire box glDrawRangeElements(GL_TRIANGLE_FAN, 0, 7, 8, @@ -1581,8 +1630,11 @@ LLSpatialPartition::LLSpatialPartition(U32 data_mask, BOOL render_by_group, U32 LLGLNamePool::registerPool(&sQueryPool); - mOctree = new LLSpatialGroup::OctreeRoot(LLVector3d(0,0,0), - LLVector3d(1,1,1), + LLVector4a center, size; + center.splat(0.f); + size.splat(1.f); + + mOctree = new LLSpatialGroup::OctreeRoot(center,size, NULL); new LLSpatialGroup(mOctree, this); } @@ -1602,7 +1654,6 @@ LLSpatialGroup *LLSpatialPartition::put(LLDrawable *drawablep, BOOL was_visible) LLMemType mt(LLMemType::MTYPE_SPACE_PARTITION); drawablep->updateSpatialExtents(); - validate_drawable(drawablep); //keep drawable from being garbage collected LLPointer ptr = drawablep; @@ -1686,16 +1737,16 @@ void LLSpatialPartition::move(LLDrawable *drawablep, LLSpatialGroup *curp, BOOL class LLSpatialShift : public LLSpatialGroup::OctreeTraveler { public: - LLSpatialShift(LLVector3 offset) : mOffset(offset) { } + const LLVector4a& mOffset; + + LLSpatialShift(const LLVector4a& offset) : mOffset(offset) { } virtual void visit(const LLSpatialGroup::OctreeNode* branch) { ((LLSpatialGroup*) branch->getListener(0))->shift(mOffset); } - - LLVector3 mOffset; }; -void LLSpatialPartition::shift(const LLVector3 &offset) +void LLSpatialPartition::shift(const LLVector4a &offset) { //shift octree node bounding boxes by offset LLMemType mt(LLMemType::MTYPE_SPACE_PARTITION); LLSpatialShift shifter(offset); @@ -1857,7 +1908,7 @@ public: class LLOctreeCullVisExtents: public LLOctreeCullShadow { public: - LLOctreeCullVisExtents(LLCamera* camera, LLVector3& min, LLVector3& max) + LLOctreeCullVisExtents(LLCamera* camera, LLVector4a& min, LLVector4a& max) : LLOctreeCullShadow(camera), mMin(min), mMax(max), mEmpty(TRUE) { } virtual bool earlyFail(LLSpatialGroup* group) @@ -1924,8 +1975,8 @@ public: } BOOL mEmpty; - LLVector3& mMin; - LLVector3& mMax; + LLVector4a& mMin; + LLVector4a& mMax; }; class LLOctreeCullDetectVisible: public LLOctreeCullShadow @@ -2029,6 +2080,11 @@ void drawBox(const LLVector3& c, const LLVector3& r) gGL.end(); } +void drawBox(const LLVector4a& c, const LLVector4a& r) +{ + drawBox(reinterpret_cast(c), reinterpret_cast(r)); +} + void drawBoxOutline(const LLVector3& pos, const LLVector3& size) { LLVector3 v1 = size.scaledVec(LLVector3( 1, 1,1)); @@ -2075,6 +2131,11 @@ void drawBoxOutline(const LLVector3& pos, const LLVector3& size) gGL.end(); } +void drawBoxOutline(const LLVector4a& pos, const LLVector4a& size) +{ + drawBoxOutline(reinterpret_cast(pos), reinterpret_cast(size)); +} + class LLOctreeDirty : public LLOctreeTraveler { public: @@ -2118,14 +2179,21 @@ BOOL LLSpatialPartition::isOcclusionEnabled() BOOL LLSpatialPartition::getVisibleExtents(LLCamera& camera, LLVector3& visMin, LLVector3& visMax) { + LLVector4a visMina, visMaxa; + visMina.load3(visMin.mV); + visMaxa.load3(visMax.mV); + { LLFastTimer ftm(FTM_CULL_REBOUND); LLSpatialGroup* group = (LLSpatialGroup*) mOctree->getListener(0); group->rebound(); } - LLOctreeCullVisExtents vis(&camera, visMin, visMax); + LLOctreeCullVisExtents vis(&camera, visMina, visMaxa); vis.traverse(mOctree); + + visMin.set(visMina.getF32()); + visMax.set(visMina.getF32()); return vis.mEmpty; } @@ -2188,25 +2256,36 @@ BOOL earlyFail(LLCamera* camera, LLSpatialGroup* group) } const F32 vel = SG_OCCLUSION_FUDGE*2.f; - LLVector3 c = group->mBounds[0]; - LLVector3 r = group->mBounds[1] + LLVector3(vel,vel,vel); - + LLVector4a fudge; + fudge.splat(vel); + + const LLVector4a& c = group->mBounds[0]; + LLVector4a r; + r.setAdd(group->mBounds[1], fudge); + /*if (r.magVecSquared() > 1024.0*1024.0) { return TRUE; }*/ - LLVector3 e = camera->getOrigin(); + LLVector4a e; + e.load3(camera->getOrigin().mV); - LLVector3 min = c - r; - LLVector3 max = c + r; + LLVector4a min; + min.setSub(c,r); + LLVector4a max; + max.setAdd(c,r); - for (U32 j = 0; j < 3; j++) + S32 lt = e.lessThan4(min).getComparisonMask() & 0x7; + if (lt) { - if (e.mV[j] < min.mV[j] || e.mV[j] > max.mV[j]) - { - return FALSE; - } + return FALSE; + } + + S32 gt = e.greaterThan4(max).getComparisonMask() & 0x7; + if (gt) + { + return FALSE; } return TRUE; @@ -2411,7 +2490,13 @@ void renderOctree(LLSpatialGroup* group) } gGL.color4fv(col.mV); - drawBox(group->mObjectBounds[0], group->mObjectBounds[1]*1.01f+LLVector3(0.001f, 0.001f, 0.001f)); + LLVector4a fudge; + fudge.splat(0.001f); + LLVector4a size = group->mObjectBounds[1]; + size.mul(1.01f); + size.add(fudge); + + drawBox(group->mObjectBounds[0], fudge); gGL.setSceneBlendType(LLRender::BT_ALPHA); @@ -2442,8 +2527,12 @@ void renderOctree(LLSpatialGroup* group) for (LLSpatialGroup::drawmap_elem_t::iterator j = i->second.begin(); j != i->second.end(); ++j) { LLDrawInfo* draw_info = *j; - LLVector3 center = (draw_info->mExtents[1] + draw_info->mExtents[0])*0.5f; - LLVector3 size = (draw_info->mExtents[1] - draw_info->mExtents[0])*0.5f; + LLVector4a center; + center.setAdd(draw_info->mExtents[1], draw_info->mExtents[0]); + center.mul(0.5f); + LLVector4a size; + size.setSub(draw_info->mExtents[1], draw_info->mExtents[0]); + size.mul(0.5f); drawBoxOutline(center, size); } } @@ -2493,7 +2582,7 @@ void renderVisibility(LLSpatialGroup* group, LLCamera* camera) else if (camera && group->mOcclusionVerts) { LLVertexBuffer::unbind(); - glVertexPointer(3, GL_FLOAT, 0, group->mOcclusionVerts); + glVertexPointer(3, GL_FLOAT, 16, group->mOcclusionVerts); glColor4f(1.0f, 0.f, 0.f, 0.5f); glDrawRangeElements(GL_TRIANGLE_FAN, 0, 7, 8, GL_UNSIGNED_BYTE, get_box_fan_indices(camera, group->mBounds[0])); @@ -2572,8 +2661,8 @@ void renderBoundingBox(LLDrawable* drawable, BOOL set_color = TRUE) } } - const LLVector3* ext; - LLVector3 pos, size; + const LLVector4a* ext; + LLVector4a pos, size; //render face bounding boxes for (S32 i = 0; i < drawable->getNumFaces(); i++) @@ -2582,20 +2671,21 @@ void renderBoundingBox(LLDrawable* drawable, BOOL set_color = TRUE) ext = facep->mExtents; - if (ext[0].isExactlyZero() && ext[1].isExactlyZero()) - { - continue; - } - pos = (ext[0] + ext[1]) * 0.5f; - size = (ext[1] - ext[0]) * 0.5f; + pos.setAdd(ext[0], ext[1]); + pos.mul(0.5f); + size.setSub(ext[1], ext[0]); + size.mul(0.5f); + drawBoxOutline(pos,size); } //render drawable bounding box ext = drawable->getSpatialExtents(); - pos = (ext[0] + ext[1]) * 0.5f; - size = (ext[1] - ext[0]) * 0.5f; + pos.setAdd(ext[0], ext[1]); + pos.mul(0.5f); + size.setSub(ext[1], ext[0]); + size.mul(0.5f); LLViewerObject* vobj = drawable->getVObj(); if (vobj && vobj->onActiveList()) @@ -2651,8 +2741,13 @@ void renderTexturePriority(LLDrawable* drawable) // gGL.color4f(1,0,1,1); //} - LLVector3 center = (facep->mExtents[1]+facep->mExtents[0])*0.5f; - LLVector3 size = (facep->mExtents[1]-facep->mExtents[0])*0.5f + LLVector3(0.01f, 0.01f, 0.01f); + LLVector4a center; + center.setAdd(facep->mExtents[1],facep->mExtents[0]); + center.mul(0.5f); + LLVector4a size; + size.setSub(facep->mExtents[1],facep->mExtents[0]); + size.mul(0.5f); + size.add(LLVector4a(0.01f)); drawBox(center, size); /*S32 boost = imagep->getBoostLevel(); @@ -2676,7 +2771,6 @@ void renderPoints(LLDrawable* drawablep) { gGL.begin(LLRender::POINTS); gGL.color3f(1,1,1); - LLVector3 center(drawablep->getPositionGroup()); for (S32 i = 0; i < drawablep->getNumFaces(); i++) { gGL.vertex3fv(drawablep->getFace(i)->mCenterLocal.mV); @@ -2708,8 +2802,12 @@ void renderShadowFrusta(LLDrawInfo* params) LLGLEnable blend(GL_BLEND); gGL.setSceneBlendType(LLRender::BT_ADD); - LLVector3 center = (params->mExtents[1]+params->mExtents[0])*0.5f; - LLVector3 size = (params->mExtents[1]-params->mExtents[0])*0.5f; + LLVector4a center; + center.setAdd(params->mExtents[1], params->mExtents[0]); + center.mul(0.5f); + LLVector4a size; + size.setSub(params->mExtents[1],params->mExtents[0]); + size.mul(0.5f); if (gPipeline.mShadowCamera[4].AABBInFrustum(center, size)) { @@ -2753,10 +2851,14 @@ void renderLights(LLDrawable* drawablep) pushVerts(drawablep->getFace(i), LLVertexBuffer::MAP_VERTEX); } - const LLVector3* ext = drawablep->getSpatialExtents(); + const LLVector4a* ext = drawablep->getSpatialExtents(); - LLVector3 pos = (ext[0] + ext[1]) * 0.5f; - LLVector3 size = (ext[1] - ext[0]) * 0.5f; + LLVector4a pos; + pos.setAdd(ext[0], ext[1]); + pos.mul(0.5f); + LLVector4a size; + size.setSub(ext[1], ext[0]); + size.mul(0.5f); { LLGLDepthTest depth(GL_FALSE, GL_TRUE); @@ -2766,7 +2868,7 @@ void renderLights(LLDrawable* drawablep) gGL.color4f(1,1,0,1); F32 rad = drawablep->getVOVolume()->getLightRadius(); - drawBoxOutline(pos, LLVector3(rad,rad,rad)); + drawBoxOutline(pos, LLVector4a(rad)); } } @@ -2781,7 +2883,7 @@ public: mDir.setSub(mEnd, mStart); } - void visit(const LLOctreeNode* branch) + void visit(const LLOctreeNode* branch) { LLVolumeOctreeListener* vl = (LLVolumeOctreeListener*) branch->getListener(0); @@ -2859,10 +2961,14 @@ void renderRaycast(LLDrawable* drawablep) glPopMatrix(); // draw bounding box of prim - const LLVector3* ext = drawablep->getSpatialExtents(); + const LLVector4a* ext = drawablep->getSpatialExtents(); - LLVector3 pos = (ext[0] + ext[1]) * 0.5f; - LLVector3 size = (ext[1] - ext[0]) * 0.5f; + LLVector4a pos; + pos.setAdd(ext[0], ext[1]); + pos.mul(0.5f); + LLVector4a size; + size.setSub(ext[1], ext[0]); + size.mul(0.5f); LLGLDepthTest depth(GL_FALSE, GL_TRUE); gGL.color4f(0,0.5f,0.5f,1); @@ -2949,8 +3055,8 @@ public: return; } - LLVector3 nodeCenter = group->mBounds[0]; - LLVector3 octCenter = LLVector3(group->mOctreeNode->getCenter()); + LLVector4a nodeCenter = group->mBounds[0]; + LLVector4a octCenter = group->mOctreeNode->getCenter(); group->rebuildGeom(); group->rebuildMesh(); @@ -2979,8 +3085,14 @@ public: if (drawable->isState(LLDrawable::IN_REBUILD_Q2)) { gGL.color4f(0.6f, 0.6f, 0.1f, 1.f); - const LLVector3* ext = drawable->getSpatialExtents(); - drawBoxOutline((ext[0]+ext[1])*0.5f, (ext[1]-ext[0])*0.5f); + const LLVector4a* ext = drawable->getSpatialExtents(); + LLVector4a center; + center.setAdd(ext[0], ext[1]); + center.mul(0.5f); + LLVector4a size; + size.setSub(ext[1], ext[0]); + size.mul(0.5f); + drawBoxOutline(center, size); } } @@ -3211,7 +3323,11 @@ void LLSpatialPartition::renderDebug() void LLSpatialGroup::drawObjectBox(LLColor4 col) { gGL.color4fv(col.mV); - drawBox(mObjectBounds[0], mObjectBounds[1]*1.01f+LLVector3(0.001f, 0.001f, 0.001f)); + LLVector4a size; + size = mObjectBounds[0]; + size.mul(1.01f); + size.add(LLVector4a(0.001f)); + drawBox(mObjectBounds[0], size); } @@ -3271,8 +3387,8 @@ public: LLSpatialGroup* group = (LLSpatialGroup*) child->getListener(0); - LLVector3 size; - LLVector3 center; + LLVector4a size; + LLVector4a center; size = group->mBounds[1]; center = group->mBounds[0]; @@ -3289,7 +3405,11 @@ public: local_end = mEnd * local_matrix; } - if (LLLineSegmentBoxIntersect(local_start, local_end, center, size)) + LLVector4a start, end; + start.load3(local_start.mV); + end.load3(local_end.mV); + + if (LLLineSegmentBoxIntersect(start, end, center, size)) { check(child); } @@ -3380,6 +3500,7 @@ LLDrawInfo::LLDrawInfo(U16 start, U16 end, U32 count, U32 offset, mDrawMode(LLRender::TRIANGLES) { mVertexBuffer->validateRange(mStart, mEnd, mCount, mOffset); + mExtents = (LLVector4a*) _mm_malloc(sizeof(LLVector4a)*2, 16); mDebugColor = (rand() << 16) + rand(); } @@ -3395,6 +3516,13 @@ LLDrawInfo::~LLDrawInfo() { mFace->setDrawInfo(NULL); } + + if (gDebugGL) + { + gPipeline.checkReferences(this); + } + + _mm_free(mExtents); } void LLDrawInfo::validate() diff --git a/indra/newview/llspatialpartition.h b/indra/newview/llspatialpartition.h index 9b252d1035..119945113a 100644 --- a/indra/newview/llspatialpartition.h +++ b/indra/newview/llspatialpartition.h @@ -45,7 +45,7 @@ #include "lldrawpool.h" #include "llface.h" #include "llviewercamera.h" - +#include "llvector4a.h" #include #define SG_STATE_INHERIT_MASK (OCCLUDED) @@ -57,12 +57,15 @@ class LLSpatialGroup; class LLTextureAtlas; class LLTextureAtlasSlot; +S32 AABBSphereIntersect(const LLVector4a& min, const LLVector4a& max, const LLVector3 &origin, const F32 &rad); +S32 AABBSphereIntersectR2(const LLVector4a& min, const LLVector4a& max, const LLVector3 &origin, const F32 &radius_squared); + S32 AABBSphereIntersect(const LLVector3& min, const LLVector3& max, const LLVector3 &origin, const F32 &rad); S32 AABBSphereIntersectR2(const LLVector3& min, const LLVector3& max, const LLVector3 &origin, const F32 &radius_squared); void pushVerts(LLFace* face, U32 mask); // get index buffer for binary encoded axis vertex buffer given a box at center being viewed by given camera -U8* get_box_fan_indices(LLCamera* camera, const LLVector3& center); +U8* get_box_fan_indices(LLCamera* camera, const LLVector4a& center); class LLDrawInfo : public LLRefCount { @@ -70,6 +73,18 @@ protected: ~LLDrawInfo(); public: + + LLDrawInfo(const LLDrawInfo& rhs) + { + *this = rhs; + } + + const LLDrawInfo& operator=(const LLDrawInfo& rhs) + { + llerrs << "Illegal operation!" << llendl; + return *this; + } + LLDrawInfo(U16 start, U16 end, U32 count, U32 offset, LLViewerTexture* image, LLVertexBuffer* buffer, BOOL fullbright = FALSE, U8 bump = 0, BOOL particle = FALSE, F32 part_size = 0); @@ -77,6 +92,8 @@ public: void validate(); + LLVector4a* mExtents; + LLPointer mVertexBuffer; LLPointer mTexture; LLColor4U mGlowColor; @@ -95,7 +112,6 @@ public: LLSpatialGroup* mGroup; LLFace* mFace; //associated face F32 mDistance; - LLVector3 mExtents[2]; U32 mDrawMode; struct CompareTexture @@ -158,11 +174,24 @@ public: }; }; +LL_ALIGN_PREFIX(64) class LLSpatialGroup : public LLOctreeListener { friend class LLSpatialPartition; friend class LLOctreeStateCheck; public: + + LLSpatialGroup(const LLSpatialGroup& rhs) + { + *this = rhs; + } + + const LLSpatialGroup& operator=(const LLSpatialGroup& rhs) + { + llerrs << "Illegal operation!" << llendl; + return *this; + } + static U32 sNodeCount; static BOOL sNoDelete; //deletion of spatial groups and draw info not allowed if TRUE @@ -273,8 +302,8 @@ public: BOOL isVisible() const; BOOL isRecentlyVisible() const; void setVisible(); - void shift(const LLVector3 &offset); - BOOL boundObjects(BOOL empty, LLVector3& newMin, LLVector3& newMax); + void shift(const LLVector4a &offset); + BOOL boundObjects(BOOL empty, LLVector4a& newMin, LLVector4a& newMax); void unbound(); BOOL rebound(); void buildOcclusion(); //rebuild mOcclusionVerts @@ -322,6 +351,27 @@ public: void addAtlas(LLTextureAtlas* atlasp, S8 recursive_level = 3) ; void removeAtlas(LLTextureAtlas* atlasp, BOOL remove_group = TRUE, S8 recursive_level = 3) ; void clearAtlasList() ; + +public: + + typedef enum + { + BOUNDS = 0, + EXTENTS = 2, + OBJECT_BOUNDS = 4, + OBJECT_EXTENTS = 6, + VIEW_ANGLE = 8, + LAST_VIEW_ANGLE = 9, + V4_COUNT = 10 + } eV4Index; + + LLVector4a* mBounds; // bounding box (center, size) of this node and all its children (tight fit to objects) + LLVector4a* mExtents; // extents (min, max) of this node and all its children + LLVector4a* mObjectExtents; // extents (min, max) of objects in this node + LLVector4a* mObjectBounds; // bounding box (center, size) of objects in this node + LLVector4a* mViewAngle; + LLVector4a* mLastUpdateViewAngle; + private: U32 mCurUpdatingTime ; //do not make the below two to use LLPointer @@ -349,14 +399,9 @@ public: F32 mBuilt; OctreeNode* mOctreeNode; LLSpatialPartition* mSpatialPartition; - LLVector3 mBounds[2]; // bounding box (center, size) of this node and all its children (tight fit to objects) - LLVector3 mExtents[2]; // extents (min, max) of this node and all its children - LLVector3 mObjectExtents[2]; // extents (min, max) of objects in this node - LLVector3 mObjectBounds[2]; // bounding box (center, size) of objects in this node - LLPointer mVertexBuffer; - F32* mOcclusionVerts; + LLVector4a* mOcclusionVerts; GLuint mOcclusionQuery[LLViewerCamera::NUM_CAMERAS]; U32 mBufferUsage; @@ -367,13 +412,10 @@ public: F32 mDepth; F32 mLastUpdateDistance; F32 mLastUpdateTime; - - LLVector3 mViewAngle; - LLVector3 mLastUpdateViewAngle; F32 mPixelArea; F32 mRadius; -}; +} LL_ALIGN_POSTFIX(64); class LLGeometryManager { @@ -409,7 +451,7 @@ public: // If the drawable moves, move it here. virtual void move(LLDrawable *drawablep, LLSpatialGroup *curp, BOOL immediate = FALSE); - virtual void shift(const LLVector3 &offset); + virtual void shift(const LLVector4a &offset); virtual F32 calcDistance(LLSpatialGroup* group, LLCamera& camera); virtual F32 calcPixelArea(LLSpatialGroup* group, LLCamera& camera); @@ -467,7 +509,7 @@ public: virtual void makeActive(); virtual void move(LLDrawable *drawablep, LLSpatialGroup *curp, BOOL immediate = FALSE); virtual BOOL updateMove(); - virtual void shiftPos(const LLVector3& vec); + virtual void shiftPos(const LLVector4a& vec); virtual void cleanupReferences(); virtual LLSpatialPartition* asPartition() { return this; } virtual LLSpatialBridge* asBridge() { return this; } @@ -658,7 +700,7 @@ class LLHUDBridge : public LLVolumeBridge { public: LLHUDBridge(LLDrawable* drawablep); - virtual void shiftPos(const LLVector3& vec); + virtual void shiftPos(const LLVector4a& vec); virtual F32 calcPixelArea(LLSpatialGroup* group, LLCamera& camera); }; @@ -675,7 +717,7 @@ class LLHUDPartition : public LLBridgePartition { public: LLHUDPartition(); - virtual void shift(const LLVector3 &offset); + virtual void shift(const LLVector4a &offset); }; extern const F32 SG_BOX_SIDE; diff --git a/indra/newview/llsurfacepatch.cpp b/indra/newview/llsurfacepatch.cpp index 48e4a6ccc7..06431d428f 100644 --- a/indra/newview/llsurfacepatch.cpp +++ b/indra/newview/llsurfacepatch.cpp @@ -860,8 +860,10 @@ void LLSurfacePatch::updateVisibility() F32 stride_per_distance = DEFAULT_DELTA_ANGLE / mSurfacep->getMetersPerGrid(); U32 grids_per_patch_edge = mSurfacep->getGridsPerPatchEdge(); - LLVector3 center = mCenterRegion + mSurfacep->getOriginAgent(); - LLVector3 radius = LLVector3(mRadius, mRadius, mRadius); + LLVector4a center; + center.load3( (mCenterRegion + mSurfacep->getOriginAgent()).mV); + LLVector4a radius; + radius.splat(mRadius); // sphere in frustum on global coordinates if (LLViewerCamera::getInstance()->AABBInFrustumNoFarClip(center, radius)) diff --git a/indra/newview/llviewerdisplay.cpp b/indra/newview/llviewerdisplay.cpp index 1490f8153c..d31b0f51fd 100644 --- a/indra/newview/llviewerdisplay.cpp +++ b/indra/newview/llviewerdisplay.cpp @@ -900,9 +900,11 @@ void display(BOOL rebuild, F32 zoom_factor, int subfield, BOOL for_snapshot) render_ui(); } - gPipeline.rebuildGroups(); - + LLSpatialGroup::sNoDelete = FALSE; + gPipeline.clearReferences(); + + gPipeline.rebuildGroups(); } LLAppViewer::instance()->pingMainloopTimeout("Display:FrameStats"); @@ -1000,6 +1002,7 @@ void render_hud_attachments() gPipeline.renderGeom(hud_cam); LLSpatialGroup::sNoDelete = FALSE; + gPipeline.clearReferences(); render_hud_elements(); diff --git a/indra/newview/llviewerobject.cpp b/indra/newview/llviewerobject.cpp index 3aecd0175d..0ed2d1da09 100644 --- a/indra/newview/llviewerobject.cpp +++ b/indra/newview/llviewerobject.cpp @@ -2863,21 +2863,26 @@ void LLViewerObject::setScale(const LLVector3 &scale, BOOL damped) } } -void LLViewerObject::updateSpatialExtents(LLVector3& newMin, LLVector3 &newMax) -{ - LLVector3 center = getRenderPosition(); - LLVector3 size = getScale(); - newMin.setVec(center-size); - newMax.setVec(center+size); - mDrawable->setPositionGroup((newMin + newMax) * 0.5f); +void LLViewerObject::updateSpatialExtents(LLVector4a& newMin, LLVector4a &newMax) +{ + LLVector4a center; + center.load3(getRenderPosition().mV); + LLVector4a size; + size.load3(getScale().mV); + newMin.setSub(center, size); + newMax.setAdd(center, size); + + mDrawable->setPositionGroup(center); } F32 LLViewerObject::getBinRadius() { if (mDrawable.notNull()) { - const LLVector3* ext = mDrawable->getSpatialExtents(); - return (ext[1]-ext[0]).magVec(); + const LLVector4a* ext = mDrawable->getSpatialExtents(); + LLVector4a diff; + diff.setSub(ext[1], ext[0]); + return diff.length3(); } return getScale().magVec(); @@ -3469,12 +3474,21 @@ BOOL LLViewerObject::lineSegmentBoundingBox(const LLVector3& start, const LLVect return FALSE; } - const LLVector3* ext = mDrawable->getSpatialExtents(); + const LLVector4a* ext = mDrawable->getSpatialExtents(); + + //VECTORIZE THIS + LLVector4a center; + center.setAdd(ext[1], ext[0]); + center.mul(0.5f); + LLVector4a size; + size.setSub(ext[1], ext[0]); + size.mul(0.5f); - LLVector3 center = (ext[1]+ext[0])*0.5f; - LLVector3 size = (ext[1]-ext[0])*0.5f; + LLVector4a starta, enda; + starta.load3(start.mV); + enda.load3(end.mV); - return LLLineSegmentBoxIntersect(start, end, center, size); + return LLLineSegmentBoxIntersect(starta, enda, center, size); } U8 LLViewerObject::getMediaType() const diff --git a/indra/newview/llviewerobject.h b/indra/newview/llviewerobject.h index 0fd0cbfa60..6ebd1cbe21 100644 --- a/indra/newview/llviewerobject.h +++ b/indra/newview/llviewerobject.h @@ -373,7 +373,7 @@ public: void markForUpdate(BOOL priority); void updateVolume(const LLVolumeParams& volume_params); - virtual void updateSpatialExtents(LLVector3& min, LLVector3& max); + virtual void updateSpatialExtents(LLVector4a& min, LLVector4a& max); virtual F32 getBinRadius(); LLBBox getBoundingBoxAgent() const; @@ -386,7 +386,7 @@ public: void clearDrawableState(U32 state, BOOL recursive = TRUE); // Called when the drawable shifts - virtual void onShift(const LLVector3 &shift_vector) { } + virtual void onShift(const LLVector4a &shift_vector) { } ////////////////////////////////////// // diff --git a/indra/newview/llviewerpartsim.cpp b/indra/newview/llviewerpartsim.cpp index 6b480ccf8e..41848e8b7a 100644 --- a/indra/newview/llviewerpartsim.cpp +++ b/indra/newview/llviewerpartsim.cpp @@ -161,8 +161,8 @@ LLViewerPartGroup::LLViewerPartGroup(const LLVector3 ¢er_agent, const F32 bo if (group != NULL) { - LLVector3 center(group->mOctreeNode->getCenter()); - LLVector3 size(group->mOctreeNode->getSize()); + LLVector3 center(group->mOctreeNode->getCenter().getF32()); + LLVector3 size(group->mOctreeNode->getSize().getF32()); size += LLVector3(0.01f, 0.01f, 0.01f); mMinObjPos = center - size; mMaxObjPos = center + size; diff --git a/indra/newview/llvoavatar.cpp b/indra/newview/llvoavatar.cpp index ac109771dd..b097461822 100644 --- a/indra/newview/llvoavatar.cpp +++ b/indra/newview/llvoavatar.cpp @@ -1286,41 +1286,46 @@ void LLVOAvatar::updateDrawable(BOOL force_damped) clearChanged(SHIFTED); } -void LLVOAvatar::onShift(const LLVector3& shift_vector) +void LLVOAvatar::onShift(const LLVector4a& shift_vector) { - mLastAnimExtents[0] += shift_vector; - mLastAnimExtents[1] += shift_vector; + const LLVector3& shift = reinterpret_cast(shift_vector); + mLastAnimExtents[0] += shift; + mLastAnimExtents[1] += shift; mNeedsImpostorUpdate = TRUE; mNeedsAnimUpdate = TRUE; } -void LLVOAvatar::updateSpatialExtents(LLVector3& newMin, LLVector3 &newMax) +void LLVOAvatar::updateSpatialExtents(LLVector4a& newMin, LLVector4a &newMax) { if (isImpostor() && !needsImpostorUpdate()) { LLVector3 delta = getRenderPosition() - - ((LLVector3(mDrawable->getPositionGroup())-mImpostorOffset)); + ((LLVector3(mDrawable->getPositionGroup().getF32())-mImpostorOffset)); - newMin = mLastAnimExtents[0] + delta; - newMax = mLastAnimExtents[1] + delta; + newMin.load3( (mLastAnimExtents[0] + delta).mV); + newMax.load3( (mLastAnimExtents[1] + delta).mV); } else { getSpatialExtents(newMin,newMax); - mLastAnimExtents[0] = newMin; - mLastAnimExtents[1] = newMax; - LLVector3 pos_group = (newMin+newMax)*0.5f; - mImpostorOffset = pos_group-getRenderPosition(); + mLastAnimExtents[0].set(newMin.getF32()); + mLastAnimExtents[1].set(newMax.getF32()); + LLVector4a pos_group; + pos_group.setAdd(newMin,newMax); + pos_group.mul(0.5f); + mImpostorOffset = LLVector3(pos_group.getF32())-getRenderPosition(); mDrawable->setPositionGroup(pos_group); } } -void LLVOAvatar::getSpatialExtents(LLVector3& newMin, LLVector3& newMax) +void LLVOAvatar::getSpatialExtents(LLVector4a& newMin, LLVector4a& newMax) { - LLVector3 buffer(0.25f, 0.25f, 0.25f); - LLVector3 pos = getRenderPosition(); - newMin = pos - buffer; - newMax = pos + buffer; + LLVector4a buffer(0.25f); + LLVector4a pos; + pos.load3(getRenderPosition().mV); + newMin.setSub(pos, buffer); + newMax.setAdd(pos, buffer); + float max_attachment_span = DEFAULT_MAX_PRIM_SCALE * 5.0f; //stretch bounding box by joint positions @@ -1329,12 +1334,20 @@ void LLVOAvatar::getSpatialExtents(LLVector3& newMin, LLVector3& newMax) LLPolyMesh* mesh = i->second; for (S32 joint_num = 0; joint_num < mesh->mJointRenderData.count(); joint_num++) { - update_min_max(newMin, newMax, - mesh->mJointRenderData[joint_num]->mWorldMatrix->getTranslation()); + LLVector4a trans; + trans.load3( mesh->mJointRenderData[joint_num]->mWorldMatrix->getTranslation().mV); + update_min_max(newMin, newMax, trans); } } - mPixelArea = LLPipeline::calcPixelArea((newMin+newMax)*0.5f, (newMax-newMin)*0.5f, *LLViewerCamera::getInstance()); + LLVector4a center, size; + center.setAdd(newMin, newMax); + center.mul(0.5f); + + size.setSub(newMax,newMin); + size.mul(0.5f); + + mPixelArea = LLPipeline::calcPixelArea(center, size, *LLViewerCamera::getInstance()); //stretch bounding box by attachments for (attachment_map_t::iterator iter = mAttachmentPoints.begin(); @@ -1361,15 +1374,17 @@ void LLVOAvatar::getSpatialExtents(LLVector3& newMin, LLVector3& newMax) LLSpatialBridge* bridge = drawable->getSpatialBridge(); if (bridge) { - const LLVector3* ext = bridge->getSpatialExtents(); - LLVector3 distance = (ext[1] - ext[0]); + const LLVector4a* ext = bridge->getSpatialExtents(); + LLVector4a distance; + distance.setSub(ext[1], ext[0]); + LLVector4a max_span(max_attachment_span); + + S32 lt = distance.lessThan4(max_span).getComparisonMask() & 0x7; // Only add the prim to spatial extents calculations if it isn't a megaprim. // max_attachment_span calculated at the start of the function // (currently 5 times our max prim size) - if (distance.mV[0] < max_attachment_span - && distance.mV[1] < max_attachment_span - && distance.mV[2] < max_attachment_span) + if (lt == 0x7) { update_min_max(newMin,newMax,ext[0]); update_min_max(newMin,newMax,ext[1]); @@ -1381,8 +1396,9 @@ void LLVOAvatar::getSpatialExtents(LLVector3& newMin, LLVector3& newMax) } //pad bounding box - newMin -= buffer; - newMax += buffer; + + newMin.sub(buffer); + newMax.add(buffer); } //----------------------------------------------------------------------------- @@ -2371,7 +2387,7 @@ void LLVOAvatar::idleUpdateMisc(bool detailed_update) if (isImpostor() && !mNeedsImpostorUpdate) { - LLVector3 ext[2]; + LLVector4a ext[2]; F32 distance; LLVector3 angle; @@ -2400,12 +2416,22 @@ void LLVOAvatar::idleUpdateMisc(bool detailed_update) } else { + //VECTORIZE THIS getSpatialExtents(ext[0], ext[1]); - if ((ext[1]-mImpostorExtents[1]).length() > 0.05f || - (ext[0]-mImpostorExtents[0]).length() > 0.05f) + LLVector4a diff; + diff.setSub(ext[1], mImpostorExtents[1]); + if (diff.length3() > 0.05f) { mNeedsImpostorUpdate = TRUE; } + else + { + diff.setSub(ext[0], mImpostorExtents[0]); + if (diff.length3() > 0.05f) + { + mNeedsImpostorUpdate = TRUE; + } + } } } } @@ -5151,9 +5177,13 @@ void LLVOAvatar::setPixelAreaAndAngle(LLAgent &agent) return; } - const LLVector3* ext = mDrawable->getSpatialExtents(); - LLVector3 center = (ext[1] + ext[0]) * 0.5f; - LLVector3 size = (ext[1]-ext[0])*0.5f; + const LLVector4a* ext = mDrawable->getSpatialExtents(); + LLVector4a center; + center.setAdd(ext[1], ext[0]); + center.mul(0.5f); + LLVector4a size; + size.setSub(ext[1], ext[0]); + size.mul(0.5f); mImpostorPixelArea = LLPipeline::calcPixelArea(center, size, *LLViewerCamera::getInstance()); @@ -5165,7 +5195,7 @@ void LLVOAvatar::setPixelAreaAndAngle(LLAgent &agent) } else { - F32 radius = size.length(); + F32 radius = size.length3(); mAppAngle = (F32) atan2( radius, range) * RAD_TO_DEG; } @@ -7546,9 +7576,9 @@ void LLVOAvatar::cacheImpostorValues() getImpostorValues(mImpostorExtents, mImpostorAngle, mImpostorDistance); } -void LLVOAvatar::getImpostorValues(LLVector3* extents, LLVector3& angle, F32& distance) const +void LLVOAvatar::getImpostorValues(LLVector4a* extents, LLVector3& angle, F32& distance) const { - const LLVector3* ext = mDrawable->getSpatialExtents(); + const LLVector4a* ext = mDrawable->getSpatialExtents(); extents[0] = ext[0]; extents[1] = ext[1]; diff --git a/indra/newview/llvoavatar.h b/indra/newview/llvoavatar.h index a851b7a150..71c3ed1cc2 100644 --- a/indra/newview/llvoavatar.h +++ b/indra/newview/llvoavatar.h @@ -127,7 +127,7 @@ public: virtual BOOL isActive() const; // Whether this object needs to do an idleUpdate. virtual void updateTextures(); virtual S32 setTETexture(const U8 te, const LLUUID& uuid); // If setting a baked texture, need to request it from a non-local sim. - virtual void onShift(const LLVector3& shift_vector); + virtual void onShift(const LLVector4a& shift_vector); virtual U32 getPartitionType() const; virtual const LLVector3 getRenderPosition() const; virtual void updateDrawable(BOOL force_damped); @@ -135,8 +135,8 @@ public: virtual BOOL updateGeometry(LLDrawable *drawable); virtual void setPixelAreaAndAngle(LLAgent &agent); virtual void updateRegion(LLViewerRegion *regionp); - virtual void updateSpatialExtents(LLVector3& newMin, LLVector3 &newMax); - virtual void getSpatialExtents(LLVector3& newMin, LLVector3& newMax); + virtual void updateSpatialExtents(LLVector4a& newMin, LLVector4a &newMax); + virtual void getSpatialExtents(LLVector4a& newMin, LLVector4a& newMax); virtual BOOL lineSegmentIntersect(const LLVector3& start, const LLVector3& end, S32 face = -1, // which face to check, -1 = ALL_SIDES BOOL pick_transparent = FALSE, @@ -391,7 +391,7 @@ public: BOOL needsImpostorUpdate() const; const LLVector3& getImpostorOffset() const; const LLVector2& getImpostorDim() const; - void getImpostorValues(LLVector3* extents, LLVector3& angle, F32& distance) const; + void getImpostorValues(LLVector4a* extents, LLVector3& angle, F32& distance) const; void cacheImpostorValues(); void setImpostorDim(const LLVector2& dim); static void resetImpostors(); @@ -402,7 +402,7 @@ private: LLVector3 mImpostorOffset; LLVector2 mImpostorDim; BOOL mNeedsAnimUpdate; - LLVector3 mImpostorExtents[2]; + LL_ALIGN_16(LLVector4a mImpostorExtents[2]); LLVector3 mImpostorAngle; F32 mImpostorDistance; F32 mImpostorPixelArea; diff --git a/indra/newview/llvopartgroup.cpp b/indra/newview/llvopartgroup.cpp index 3ba4ecad0c..b5fd8182c6 100644 --- a/indra/newview/llvopartgroup.cpp +++ b/indra/newview/llvopartgroup.cpp @@ -79,12 +79,14 @@ F32 LLVOPartGroup::getBinRadius() return mScale.mV[0]*2.f; } -void LLVOPartGroup::updateSpatialExtents(LLVector3& newMin, LLVector3& newMax) +void LLVOPartGroup::updateSpatialExtents(LLVector4a& newMin, LLVector4a& newMax) { const LLVector3& pos_agent = getPositionAgent(); - newMin = pos_agent - mScale; - newMax = pos_agent + mScale; - mDrawable->setPositionGroup(pos_agent); + newMin.load3( (pos_agent - mScale).mV); + newMax.load3( (pos_agent + mScale).mV); + LLVector4a pos; + pos.load3(pos_agent.mV); + mDrawable->setPositionGroup(pos); } BOOL LLVOPartGroup::idleUpdate(LLAgent &agent, LLWorld &world, const F64 &time) diff --git a/indra/newview/llvopartgroup.h b/indra/newview/llvopartgroup.h index 18583b4be9..771ae1c1eb 100644 --- a/indra/newview/llvopartgroup.h +++ b/indra/newview/llvopartgroup.h @@ -57,7 +57,7 @@ public: BOOL idleUpdate(LLAgent &agent, LLWorld &world, const F64 &time); virtual F32 getBinRadius(); - virtual void updateSpatialExtents(LLVector3& newMin, LLVector3& newMax); + virtual void updateSpatialExtents(LLVector4a& newMin, LLVector4a& newMax); virtual U32 getPartitionType() const; /*virtual*/ void setPixelAreaAndAngle(LLAgent &agent); diff --git a/indra/newview/llvosurfacepatch.cpp b/indra/newview/llvosurfacepatch.cpp index eef62ddf1a..02e7e7e60f 100644 --- a/indra/newview/llvosurfacepatch.cpp +++ b/indra/newview/llvosurfacepatch.cpp @@ -995,7 +995,13 @@ BOOL LLVOSurfacePatch::lineSegmentIntersect(const LLVector3& start, const LLVect //step one meter at a time until intersection point found - const LLVector3* ext = mDrawable->getSpatialExtents(); + //VECTORIZE THIS + const LLVector4a* exta = mDrawable->getSpatialExtents(); + + LLVector3 ext[2]; + ext[0].set(exta[0].getF32()); + ext[1].set(exta[1].getF32()); + F32 rad = (delta*tdelta).magVecSquared(); F32 t = 0.f; @@ -1057,13 +1063,16 @@ BOOL LLVOSurfacePatch::lineSegmentIntersect(const LLVector3& start, const LLVect return FALSE; } -void LLVOSurfacePatch::updateSpatialExtents(LLVector3& newMin, LLVector3 &newMax) +void LLVOSurfacePatch::updateSpatialExtents(LLVector4a& newMin, LLVector4a &newMax) { LLVector3 posAgent = getPositionAgent(); LLVector3 scale = getScale(); - newMin = posAgent-scale*0.5f; // Changing to 2.f makes the culling a -little- better, but still wrong - newMax = posAgent+scale*0.5f; - mDrawable->setPositionGroup((newMin+newMax)*0.5f); + newMin.load3( (posAgent-scale*0.5f).mV); // Changing to 2.f makes the culling a -little- better, but still wrong + newMax.load3( (posAgent+scale*0.5f).mV); + LLVector4a pos; + pos.setAdd(newMin,newMax); + pos.mul(0.5f); + mDrawable->setPositionGroup(pos); } U32 LLVOSurfacePatch::getPartitionType() const diff --git a/indra/newview/llvosurfacepatch.h b/indra/newview/llvosurfacepatch.h index 10a5888526..15442e1947 100644 --- a/indra/newview/llvosurfacepatch.h +++ b/indra/newview/llvosurfacepatch.h @@ -78,7 +78,7 @@ public: /*virtual*/ void updateTextures(); /*virtual*/ void setPixelAreaAndAngle(LLAgent &agent); // generate accurate apparent angle and area - /*virtual*/ void updateSpatialExtents(LLVector3& newMin, LLVector3& newMax); + /*virtual*/ void updateSpatialExtents(LLVector4a& newMin, LLVector4a& newMax); /*virtual*/ BOOL isActive() const; // Whether this object needs to do an idleUpdate. void setPatch(LLSurfacePatch *patchp); diff --git a/indra/newview/llvotree.cpp b/indra/newview/llvotree.cpp index b89c0cd638..d564643161 100644 --- a/indra/newview/llvotree.cpp +++ b/indra/newview/llvotree.cpp @@ -1238,7 +1238,7 @@ void LLVOTree::updateRadius() mDrawable->setRadius(32.0f); } -void LLVOTree::updateSpatialExtents(LLVector3& newMin, LLVector3& newMax) +void LLVOTree::updateSpatialExtents(LLVector4a& newMin, LLVector4a& newMax) { F32 radius = getScale().length()*0.05f; LLVector3 center = getRenderPosition(); @@ -1248,9 +1248,11 @@ void LLVOTree::updateSpatialExtents(LLVector3& newMin, LLVector3& newMax) center += LLVector3(0, 0, size.mV[2]) * getRotation(); - newMin.set(center-size); - newMax.set(center+size); - mDrawable->setPositionGroup(center); + newMin.load3((center-size).mV); + newMax.load3((center+size).mV); + LLVector4a pos; + pos.load3(center.mV); + mDrawable->setPositionGroup(pos); } BOOL LLVOTree::lineSegmentIntersect(const LLVector3& start, const LLVector3& end, S32 face, BOOL pick_transparent, S32 *face_hitp, @@ -1263,8 +1265,13 @@ BOOL LLVOTree::lineSegmentIntersect(const LLVector3& start, const LLVector3& end return FALSE; } - const LLVector3* ext = mDrawable->getSpatialExtents(); + const LLVector4a* exta = mDrawable->getSpatialExtents(); + //VECTORIZE THIS + LLVector3 ext[2]; + ext[0].set(exta[0].getF32()); + ext[1].set(exta[1].getF32()); + LLVector3 center = (ext[1]+ext[0])*0.5f; LLVector3 size = (ext[1]-ext[0]); diff --git a/indra/newview/llvotree.h b/indra/newview/llvotree.h index feac9e0675..2ce1b03d26 100644 --- a/indra/newview/llvotree.h +++ b/indra/newview/llvotree.h @@ -73,7 +73,7 @@ public: /*virtual*/ LLDrawable* createDrawable(LLPipeline *pipeline); /*virtual*/ BOOL updateGeometry(LLDrawable *drawable); - /*virtual*/ void updateSpatialExtents(LLVector3 &min, LLVector3 &max); + /*virtual*/ void updateSpatialExtents(LLVector4a &min, LLVector4a &max); virtual U32 getPartitionType() const; diff --git a/indra/newview/llvovolume.cpp b/indra/newview/llvovolume.cpp index a9f3abeef8..db9e0b88e1 100644 --- a/indra/newview/llvovolume.cpp +++ b/indra/newview/llvovolume.cpp @@ -701,7 +701,7 @@ void LLVOVolume::updateTextureVirtualSize() const LLTextureEntry *te = face->getTextureEntry(); LLViewerTexture *imagep = face->getTexture(); if (!imagep || !te || - face->mExtents[0] == face->mExtents[1]) + face->mExtents[0].equal3(face->mExtents[1])) { continue; } @@ -1332,7 +1332,7 @@ BOOL LLVOVolume::genBBoxes(BOOL force_global) { BOOL res = TRUE; - LLVector3 min,max; + LLVector4a min,max; BOOL rebuild = mDrawable->isState(LLDrawable::REBUILD_VOLUME | LLDrawable::REBUILD_POSITION); @@ -1356,17 +1356,8 @@ BOOL LLVOVolume::genBBoxes(BOOL force_global) } else { - for (U32 i = 0; i < 3; i++) - { - if (face->mExtents[0].mV[i] < min.mV[i]) - { - min.mV[i] = face->mExtents[0].mV[i]; - } - if (face->mExtents[1].mV[i] > max.mV[i]) - { - max.mV[i] = face->mExtents[1].mV[i]; - } - } + min.setMin(face->mExtents[0]); + max.setMax(face->mExtents[1]); } } } @@ -1374,7 +1365,9 @@ BOOL LLVOVolume::genBBoxes(BOOL force_global) if (rebuild) { mDrawable->setSpatialExtents(min,max); - mDrawable->setPositionGroup((min+max)*0.5f); + min.add(max); + min.mul(0.5f); + mDrawable->setPositionGroup(min); } updateRadius(); @@ -3007,7 +3000,7 @@ void LLVOVolume::setSelected(BOOL sel) } } -void LLVOVolume::updateSpatialExtents(LLVector3& newMin, LLVector3& newMax) +void LLVOVolume::updateSpatialExtents(LLVector4a& newMin, LLVector4a& newMax) { } @@ -3039,7 +3032,7 @@ F32 LLVOVolume::getBinRadius() } } - const LLVector3* ext = mDrawable->getSpatialExtents(); + const LLVector4a* ext = mDrawable->getSpatialExtents(); BOOL shrink_wrap = mDrawable->isAnimating(); BOOL alpha_wrap = FALSE; @@ -3071,7 +3064,10 @@ F32 LLVOVolume::getBinRadius() } else if (shrink_wrap) { - radius = (ext[1]-ext[0]).length()*0.5f; + LLVector4a rad; + rad.setSub(ext[1], ext[0]); + + radius = rad.length3()*0.5f; } else if (mDrawable->isStatic()) { @@ -3107,7 +3103,7 @@ const LLVector3 LLVOVolume::getPivotPositionAgent() const return LLViewerObject::getPivotPositionAgent(); } -void LLVOVolume::onShift(const LLVector3 &shift_vector) +void LLVOVolume::onShift(const LLVector4a &shift_vector) { if (mVolumeImpl) { @@ -3610,7 +3606,6 @@ void LLVolumeGeometryManager::rebuildGeom(LLSpatialGroup* group) } } } - } continue; @@ -4217,7 +4212,7 @@ LLHUDPartition::LLHUDPartition() mLODPeriod = 1; } -void LLHUDPartition::shift(const LLVector3 &offset) +void LLHUDPartition::shift(const LLVector4a &offset) { //HUD objects don't shift with region crossing. That would be silly. } diff --git a/indra/newview/llvovolume.h b/indra/newview/llvovolume.h index 2776988a12..d5606034d0 100644 --- a/indra/newview/llvovolume.h +++ b/indra/newview/llvovolume.h @@ -66,7 +66,7 @@ public: virtual void onSetVolume(const LLVolumeParams &volume_params, const S32 detail) = 0; virtual void onSetScale(const LLVector3 &scale, BOOL damped) = 0; virtual void onParameterChanged(U16 param_type, LLNetworkData *data, BOOL in_use, bool local_origin) = 0; - virtual void onShift(const LLVector3 &shift_vector) = 0; + virtual void onShift(const LLVector4a &shift_vector) = 0; virtual bool isVolumeUnique() const = 0; // Do we need a unique LLVolume instance? virtual bool isVolumeGlobal() const = 0; // Are we in global space? virtual bool isActive() const = 0; // Is this object currently active? @@ -145,7 +145,7 @@ public: void markForUpdate(BOOL priority) { LLViewerObject::markForUpdate(priority); mVolumeChanged = TRUE; } - /*virtual*/ void onShift(const LLVector3 &shift_vector); // Called when the drawable shifts + /*virtual*/ void onShift(const LLVector4a &shift_vector); // Called when the drawable shifts /*virtual*/ void parameterChanged(U16 param_type, bool local_origin); /*virtual*/ void parameterChanged(U16 param_type, LLNetworkData* data, BOOL in_use, bool local_origin); @@ -201,7 +201,7 @@ public: void regenFaces(); BOOL genBBoxes(BOOL force_global); void preRebuild(); - virtual void updateSpatialExtents(LLVector3& min, LLVector3& max); + virtual void updateSpatialExtents(LLVector4a& min, LLVector4a& max); virtual F32 getBinRadius(); virtual U32 getPartitionType() const; diff --git a/indra/newview/llvowater.cpp b/indra/newview/llvowater.cpp index a8c4625f6e..7c1b22d432 100644 --- a/indra/newview/llvowater.cpp +++ b/indra/newview/llvowater.cpp @@ -258,15 +258,21 @@ void LLVOWater::setIsEdgePatch(const BOOL edge_patch) mIsEdgePatch = edge_patch; } -void LLVOWater::updateSpatialExtents(LLVector3 &newMin, LLVector3& newMax) +void LLVOWater::updateSpatialExtents(LLVector4a &newMin, LLVector4a& newMax) { - LLVector3 pos = getPositionAgent(); - LLVector3 scale = getScale(); - - newMin = pos - scale * 0.5f; - newMax = pos + scale * 0.5f; + LLVector4a pos; + pos.load3(getPositionAgent().mV); + LLVector4a scale; + scale.load3(getScale().mV); + scale.mul(0.5f); + + newMin.setSub(pos, scale); + newMax.setAdd(pos, scale); + + pos.setAdd(newMin,newMax); + pos.mul(0.5f); - mDrawable->setPositionGroup((newMin + newMax) * 0.5f); + mDrawable->setPositionGroup(pos); } U32 LLVOWater::getPartitionType() const diff --git a/indra/newview/llvowater.h b/indra/newview/llvowater.h index 3cc031e589..a868afe58b 100644 --- a/indra/newview/llvowater.h +++ b/indra/newview/llvowater.h @@ -66,7 +66,7 @@ public: /*virtual*/ BOOL idleUpdate(LLAgent &agent, LLWorld &world, const F64 &time); /*virtual*/ LLDrawable* createDrawable(LLPipeline *pipeline); /*virtual*/ BOOL updateGeometry(LLDrawable *drawable); - /*virtual*/ void updateSpatialExtents(LLVector3& newMin, LLVector3& newMax); + /*virtual*/ void updateSpatialExtents(LLVector4a& newMin, LLVector4a& newMax); /*virtual*/ void updateTextures(); /*virtual*/ void setPixelAreaAndAngle(LLAgent &agent); // generate accurate apparent angle and area diff --git a/indra/newview/pipeline.cpp b/indra/newview/pipeline.cpp index 91c3805d3b..01027e5be6 100644 --- a/indra/newview/pipeline.cpp +++ b/indra/newview/pipeline.cpp @@ -1508,11 +1508,214 @@ F32 LLPipeline::calcPixelArea(LLVector3 center, LLVector3 size, LLCamera &camera return radius*radius * F_PI; } +//static +F32 LLPipeline::calcPixelArea(const LLVector4a& center, const LLVector4a& size, LLCamera &camera) +{ + LLVector4a origin; + origin.load3(camera.getOrigin().mV); + + LLVector4a lookAt; + lookAt.setSub(center, origin); + F32 dist = lookAt.length3(); + + //ramp down distance for nearby objects + //shrink dist by dist/16. + if (dist < 16.f) + { + dist /= 16.f; + dist *= dist; + dist *= 16.f; + } + + //get area of circle around node + F32 app_angle = atanf(size.length3()/dist); + F32 radius = app_angle*LLDrawable::sCurPixelAngle; + return radius*radius * F_PI; +} + void LLPipeline::grabReferences(LLCullResult& result) { sCull = &result; } +void LLPipeline::clearReferences() +{ + sCull = NULL; +} + +void check_references(LLSpatialGroup* group, LLDrawable* drawable) +{ + for (LLSpatialGroup::element_iter i = group->getData().begin(); i != group->getData().end(); ++i) + { + if (drawable == *i) + { + llerrs << "LLDrawable deleted while actively reference by LLPipeline." << llendl; + } + } +} + +void check_references(LLDrawable* drawable, LLFace* face) +{ + for (S32 i = 0; i < drawable->getNumFaces(); ++i) + { + if (drawable->getFace(i) == face) + { + llerrs << "LLFace deleted while actively referenced by LLPipeline." << llendl; + } + } +} + +void check_references(LLSpatialGroup* group, LLFace* face) +{ + for (LLSpatialGroup::element_iter i = group->getData().begin(); i != group->getData().end(); ++i) + { + LLDrawable* drawable = *i; + check_references(drawable, face); + } +} + +void LLPipeline::checkReferences(LLFace* face) +{ +#if 0 + if (sCull) + { + for (LLCullResult::sg_list_t::iterator iter = sCull->beginVisibleGroups(); iter != sCull->endVisibleGroups(); ++iter) + { + LLSpatialGroup* group = *iter; + check_references(group, face); + } + + for (LLCullResult::sg_list_t::iterator iter = sCull->beginAlphaGroups(); iter != sCull->endAlphaGroups(); ++iter) + { + LLSpatialGroup* group = *iter; + check_references(group, face); + } + + for (LLCullResult::sg_list_t::iterator iter = sCull->beginDrawableGroups(); iter != sCull->endDrawableGroups(); ++iter) + { + LLSpatialGroup* group = *iter; + check_references(group, face); + } + + for (LLCullResult::drawable_list_t::iterator iter = sCull->beginVisibleList(); iter != sCull->endVisibleList(); ++iter) + { + LLDrawable* drawable = *iter; + check_references(drawable, face); + } + } +#endif +} + +void LLPipeline::checkReferences(LLDrawable* drawable) +{ +#if 0 + if (sCull) + { + for (LLCullResult::sg_list_t::iterator iter = sCull->beginVisibleGroups(); iter != sCull->endVisibleGroups(); ++iter) + { + LLSpatialGroup* group = *iter; + check_references(group, drawable); + } + + for (LLCullResult::sg_list_t::iterator iter = sCull->beginAlphaGroups(); iter != sCull->endAlphaGroups(); ++iter) + { + LLSpatialGroup* group = *iter; + check_references(group, drawable); + } + + for (LLCullResult::sg_list_t::iterator iter = sCull->beginDrawableGroups(); iter != sCull->endDrawableGroups(); ++iter) + { + LLSpatialGroup* group = *iter; + check_references(group, drawable); + } + + for (LLCullResult::drawable_list_t::iterator iter = sCull->beginVisibleList(); iter != sCull->endVisibleList(); ++iter) + { + if (drawable == *iter) + { + llerrs << "LLDrawable deleted while actively referenced by LLPipeline." << llendl; + } + } + } +#endif +} + +void check_references(LLSpatialGroup* group, LLDrawInfo* draw_info) +{ + for (LLSpatialGroup::draw_map_t::iterator i = group->mDrawMap.begin(); i != group->mDrawMap.end(); ++i) + { + LLSpatialGroup::drawmap_elem_t& draw_vec = i->second; + for (LLSpatialGroup::drawmap_elem_t::iterator j = draw_vec.begin(); j != draw_vec.end(); ++j) + { + LLDrawInfo* params = *j; + if (params == draw_info) + { + llerrs << "LLDrawInfo deleted while actively referenced by LLPipeline." << llendl; + } + } + } +} + + +void LLPipeline::checkReferences(LLDrawInfo* draw_info) +{ +#if 0 + if (sCull) + { + for (LLCullResult::sg_list_t::iterator iter = sCull->beginVisibleGroups(); iter != sCull->endVisibleGroups(); ++iter) + { + LLSpatialGroup* group = *iter; + check_references(group, draw_info); + } + + for (LLCullResult::sg_list_t::iterator iter = sCull->beginAlphaGroups(); iter != sCull->endAlphaGroups(); ++iter) + { + LLSpatialGroup* group = *iter; + check_references(group, draw_info); + } + + for (LLCullResult::sg_list_t::iterator iter = sCull->beginDrawableGroups(); iter != sCull->endDrawableGroups(); ++iter) + { + LLSpatialGroup* group = *iter; + check_references(group, draw_info); + } + } +#endif +} + +void LLPipeline::checkReferences(LLSpatialGroup* group) +{ +#if 0 + if (sCull) + { + for (LLCullResult::sg_list_t::iterator iter = sCull->beginVisibleGroups(); iter != sCull->endVisibleGroups(); ++iter) + { + if (group == *iter) + { + llerrs << "LLSpatialGroup deleted while actively referenced by LLPipeline." << llendl; + } + } + + for (LLCullResult::sg_list_t::iterator iter = sCull->beginAlphaGroups(); iter != sCull->endAlphaGroups(); ++iter) + { + if (group == *iter) + { + llerrs << "LLSpatialGroup deleted while actively referenced by LLPipeline." << llendl; + } + } + + for (LLCullResult::sg_list_t::iterator iter = sCull->beginDrawableGroups(); iter != sCull->endDrawableGroups(); ++iter) + { + if (group == *iter) + { + llerrs << "LLSpatialGroup deleted while actively referenced by LLPipeline." << llendl; + } + } + } +#endif +} + + BOOL LLPipeline::visibleObjectsInFrustum(LLCamera& camera) { for (LLWorld::region_list_t::const_iterator iter = LLWorld::getInstance()->getRegionList().begin(); @@ -1714,7 +1917,7 @@ void LLPipeline::markNotCulled(LLSpatialGroup* group, LLCamera& camera) } if (sMinRenderSize > 0.f && - llmax(llmax(group->mBounds[1].mV[0], group->mBounds[1].mV[1]), group->mBounds[1].mV[2]) < sMinRenderSize) + llmax(llmax(group->mBounds[1][0], group->mBounds[1][1]), group->mBounds[1][2]) < sMinRenderSize) { return; } @@ -2100,6 +2303,9 @@ void LLPipeline::shiftObjects(const LLVector3 &offset) glClear(GL_DEPTH_BUFFER_BIT); gDepthDirty = TRUE; + LLVector4a offseta; + offseta.load3(offset.mV); + for (LLDrawable::drawable_vector_t::iterator iter = mShiftList.begin(); iter != mShiftList.end(); iter++) { @@ -2108,7 +2314,7 @@ void LLPipeline::shiftObjects(const LLVector3 &offset) { continue; } - drawablep->shiftPos(offset); + drawablep->shiftPos(offseta); drawablep->clearState(LLDrawable::ON_SHIFT_LIST); } mShiftList.resize(0); @@ -2122,7 +2328,7 @@ void LLPipeline::shiftObjects(const LLVector3 &offset) LLSpatialPartition* part = region->getSpatialPartition(i); if (part) { - part->shift(offset); + part->shift(offseta); } } } @@ -2659,8 +2865,10 @@ void LLPipeline::postSort(LLCamera& camera) { if (sMinRenderSize > 0.f) { - LLVector3 bounds = (*k)->mExtents[1]-(*k)->mExtents[0]; - if (llmax(llmax(bounds.mV[0], bounds.mV[1]), bounds.mV[2]) > sMinRenderSize) + LLVector4a bounds; + bounds.setSub((*k)->mExtents[1],(*k)->mExtents[0]); + + if (llmax(llmax(bounds[0], bounds[1]), bounds[2]) > sMinRenderSize) { sCull->pushDrawInfo(j->first, *k); } @@ -6770,8 +6978,9 @@ void LLPipeline::renderDeferredLighting() } - LLVector3 center = drawablep->getPositionAgent(); - F32* c = center.mV; + LLVector4a center; + center.load3(drawablep->getPositionAgent().mV); + const F32* c = center.getF32(); F32 s = volume->getLightRadius()*1.5f; LLColor3 col = volume->getLightColor(); @@ -6787,7 +6996,9 @@ void LLPipeline::renderDeferredLighting() continue; } - if (camera->AABBInFrustumNoFarClip(center, LLVector3(s,s,s)) == 0) + LLVector4a sa; + sa.splat(s); + if (camera->AABBInFrustumNoFarClip(center, sa) == 0) { continue; } @@ -6865,8 +7076,9 @@ void LLPipeline::renderDeferredLighting() LLVOVolume* volume = drawablep->getVOVolume(); - LLVector3 center = drawablep->getPositionAgent(); - F32* c = center.mV; + LLVector4a center; + center.load3(drawablep->getPositionAgent().mV); + const F32* c = center.getF32(); F32 s = volume->getLightRadius()*1.5f; sVisibleLightCount++; @@ -8952,7 +9164,7 @@ void LLPipeline::generateImpostor(LLVOAvatar* avatar) stateSort(*LLViewerCamera::getInstance(), result); - const LLVector3* ext = avatar->mDrawable->getSpatialExtents(); + const LLVector4a* ext = avatar->mDrawable->getSpatialExtents(); LLVector3 pos(avatar->getRenderPosition()+avatar->getImpostorOffset()); LLCamera camera = *viewer_camera; @@ -8961,18 +9173,23 @@ void LLPipeline::generateImpostor(LLVOAvatar* avatar) LLVector2 tdim; - LLVector3 half_height = (ext[1]-ext[0])*0.5f; - LLVector3 left = camera.getLeftAxis(); - left *= left; - left.normalize(); + LLVector4a half_height; + half_height.setSub(ext[1], ext[0]); + half_height.mul(0.5f); + + LLVector4a left; + left.load3(camera.getLeftAxis().mV); + left.mul(left); + left.normalize3fast(); - LLVector3 up = camera.getUpAxis(); - up *= up; - up.normalize(); + LLVector4a up; + up.load3(camera.getUpAxis().mV); + up.mul(up); + up.normalize3fast(); - tdim.mV[0] = fabsf(half_height * left); - tdim.mV[1] = fabsf(half_height * up); + tdim.mV[0] = fabsf(half_height.dot3(left)); + tdim.mV[1] = fabsf(half_height.dot3(up)); glMatrixMode(GL_PROJECTION); glPushMatrix(); diff --git a/indra/newview/pipeline.h b/indra/newview/pipeline.h index c9384f5ba2..52f943cd1d 100644 --- a/indra/newview/pipeline.h +++ b/indra/newview/pipeline.h @@ -217,6 +217,7 @@ public: //calculate pixel area of given box from vantage point of given camera static F32 calcPixelArea(LLVector3 center, LLVector3 size, LLCamera& camera); + static F32 calcPixelArea(const LLVector4a& center, const LLVector4a& size, LLCamera &camera); void stateSort(LLCamera& camera, LLCullResult& result); void stateSort(LLSpatialGroup* group, LLCamera& camera); @@ -229,6 +230,14 @@ public: void renderGroups(LLRenderPass* pass, U32 type, U32 mask, BOOL texture); void grabReferences(LLCullResult& result); + void clearReferences(); + + //check references will assert that there are no references in sCullResult to the provided data + void checkReferences(LLFace* face); + void checkReferences(LLDrawable* drawable); + void checkReferences(LLDrawInfo* draw_info); + void checkReferences(LLSpatialGroup* group); + void renderGeom(LLCamera& camera, BOOL forceVBOUpdate = FALSE); void renderGeomDeferred(LLCamera& camera); -- cgit v1.3 From f9b13d8f8510b1f7f02fcf92685471461b79858e Mon Sep 17 00:00:00 2001 From: Dave Parks Date: Thu, 10 Jun 2010 00:45:48 -0500 Subject: Add "LL_MESH_ENABLED" preprocessor flag for toggling mesh code. Couple of merge fixes. --- indra/llcommon/llassettype.cpp | 3 +- indra/llcommon/llassettype.h | 3 +- indra/llinventory/llinventorytype.cpp | 5 ++- indra/llinventory/llinventorytype.h | 2 + indra/llmath/llvolume.cpp | 13 +++--- indra/llmath/llvolume.h | 7 ++++ indra/llrender/llvertexbuffer.cpp | 4 +- indra/llvfs/llvfs.cpp | 2 + indra/newview/app_settings/settings.xml | 4 +- indra/newview/llappviewer.cpp | 4 ++ indra/newview/llassetuploadresponders.cpp | 5 +++ indra/newview/llassetuploadresponders.h | 3 +- indra/newview/lldrawpoolavatar.cpp | 48 +++++++++++++++++++++- indra/newview/lldrawpoolavatar.h | 40 ++++++++++-------- indra/newview/llface.cpp | 2 + indra/newview/llinventorybridge.cpp | 10 ++++- indra/newview/llinventorybridge.h | 4 +- indra/newview/llpanelgroupnotices.cpp | 2 + indra/newview/llpanelobjectinventory.cpp | 6 +++ indra/newview/llselectmgr.cpp | 2 + indra/newview/lltexturectrl.cpp | 12 +++++- indra/newview/lltooldraganddrop.cpp | 10 +++++ indra/newview/lltooldraganddrop.h | 4 ++ indra/newview/llviewerassettype.cpp | 2 + indra/newview/llviewerfloaterreg.cpp | 4 ++ indra/newview/llviewermenufile.cpp | 9 +++- indra/newview/llviewermenufile.h | 2 + indra/newview/llviewertexteditor.cpp | 2 + indra/newview/llviewerwindow.cpp | 4 ++ indra/newview/llvovolume.cpp | 15 ++++++- indra/newview/pipeline.cpp | 4 ++ indra/newview/pipeline.h | 13 +++--- .../default/xui/en/panel_preferences_graphics1.xml | 10 ++--- 33 files changed, 209 insertions(+), 51 deletions(-) (limited to 'indra/newview/llvovolume.cpp') diff --git a/indra/llcommon/llassettype.cpp b/indra/llcommon/llassettype.cpp index e19f8b0454..476a23ec64 100644 --- a/indra/llcommon/llassettype.cpp +++ b/indra/llcommon/llassettype.cpp @@ -99,8 +99,9 @@ LLAssetDictionary::LLAssetDictionary() addEntry(LLAssetType::AT_LINK, new AssetEntry("LINK", "link", "sym link", false, false, true)); addEntry(LLAssetType::AT_LINK_FOLDER, new AssetEntry("FOLDER_LINK", "link_f", "sym folder link", false, false, true)); - +#if LL_MESH_ENABLED addEntry(LLAssetType::AT_MESH, new AssetEntry("MESH", "mesh", "mesh", false, false, false)); +#endif addEntry(LLAssetType::AT_NONE, new AssetEntry("NONE", "-1", NULL, FALSE, FALSE, FALSE)); diff --git a/indra/llcommon/llassettype.h b/indra/llcommon/llassettype.h index 90cd03c433..ebc43134cb 100644 --- a/indra/llcommon/llassettype.h +++ b/indra/llcommon/llassettype.h @@ -114,9 +114,10 @@ public: AT_LINK_FOLDER = 25, // Inventory folder link - +#if LL_MESH_ENABLED AT_MESH = 49, // Mesh data in our proprietary SLM format +#endif AT_COUNT = 50, diff --git a/indra/llinventory/llinventorytype.cpp b/indra/llinventory/llinventorytype.cpp index 2d4ee604b0..82cd22a832 100644 --- a/indra/llinventory/llinventorytype.cpp +++ b/indra/llinventory/llinventorytype.cpp @@ -89,7 +89,9 @@ LLInventoryDictionary::LLInventoryDictionary() addEntry(LLInventoryType::IT_WEARABLE, new InventoryEntry("wearable", "wearable", 2, LLAssetType::AT_CLOTHING, LLAssetType::AT_BODYPART)); addEntry(LLInventoryType::IT_ANIMATION, new InventoryEntry("animation", "animation", 1, LLAssetType::AT_ANIMATION)); addEntry(LLInventoryType::IT_GESTURE, new InventoryEntry("gesture", "gesture", 1, LLAssetType::AT_GESTURE)); +#if LL_MESH_ENABLED addEntry(LLInventoryType::IT_MESH, new InventoryEntry("mesh", "mesh", 1, LLAssetType::AT_MESH)); +#endif } @@ -150,8 +152,9 @@ DEFAULT_ASSET_FOR_INV_TYPE[LLAssetType::AT_COUNT] = LLInventoryType::IT_NONE, // AT_NONE LLInventoryType::IT_NONE, // AT_NONE LLInventoryType::IT_NONE, // AT_NONE - +#if LL_MESH_ENABLED LLInventoryType::IT_MESH // AT_MESH +#endif }; // static diff --git a/indra/llinventory/llinventorytype.h b/indra/llinventory/llinventorytype.h index 37829f5eae..d2fc67ef64 100644 --- a/indra/llinventory/llinventorytype.h +++ b/indra/llinventory/llinventorytype.h @@ -68,7 +68,9 @@ public: IT_ANIMATION = 19, IT_GESTURE = 20, +#if LL_MESH_ENABLED IT_MESH = 22, +#endif IT_COUNT = 23, IT_NONE = -1 diff --git a/indra/llmath/llvolume.cpp b/indra/llmath/llvolume.cpp index 09ab47b890..05868e3517 100644 --- a/indra/llmath/llvolume.cpp +++ b/indra/llmath/llvolume.cpp @@ -2546,10 +2546,12 @@ S32 LLVolume::getNumFaces() const { U8 sculpt_type = (mParams.getSculptType() & LL_SCULPT_TYPE_MASK); +#if LL_MESH_ENABLED if (sculpt_type == LL_SCULPT_TYPE_MESH) { return LL_SCULPT_MESH_MAX_FACES; } +#endif return (S32)mProfilep->mFaces.size(); } @@ -2922,11 +2924,6 @@ void LLVolume::sculpt(U16 sculpt_width, U16 sculpt_height, S8 sculpt_components, LLMemType m1(LLMemType::MTYPE_VOLUME); U8 sculpt_type = mParams.getSculptType(); - if (sculpt_type & LL_SCULPT_TYPE_MASK == LL_SCULPT_TYPE_MESH) - { - llerrs << "WTF?" << llendl; - } - BOOL data_is_empty = FALSE; if (sculpt_width == 0 || sculpt_height == 0 || sculpt_components < 3 || sculpt_data == NULL) @@ -4135,10 +4132,12 @@ void LLVolume::generateSilhouetteVertices(std::vector &vertices, normals.clear(); segments.clear(); +#if LL_MESH_ENABLED if ((mParams.getSculptType() & LL_SCULPT_TYPE_MASK) == LL_SCULPT_TYPE_MESH) { return; } +#endif S32 cur_index = 0; //for each face @@ -6335,10 +6334,14 @@ BOOL LLVolumeFace::createSide(LLVolume* volume, BOOL partial_build) resizeVertices(num_vertices); resizeIndices(num_indices); +#if LL_MESH_ENABLED if ((volume->getParams().getSculptType() & LL_SCULPT_TYPE_MASK) != LL_SCULPT_TYPE_MESH) { mEdge.resize(num_indices); } +#else + mEdge.resize(num_indices); +#endif } LLVector4a* pos = (LLVector4a*) mPositions; diff --git a/indra/llmath/llvolume.h b/indra/llmath/llvolume.h index c49d1c650d..0782944079 100644 --- a/indra/llmath/llvolume.h +++ b/indra/llmath/llvolume.h @@ -41,6 +41,8 @@ class LLVolumeParams; class LLProfile; class LLPath; +#define LL_MESH_ENABLED 0 + template class LLOctreeNode; class LLVector4a; @@ -190,10 +192,15 @@ const U8 LL_SCULPT_TYPE_SPHERE = 1; const U8 LL_SCULPT_TYPE_TORUS = 2; const U8 LL_SCULPT_TYPE_PLANE = 3; const U8 LL_SCULPT_TYPE_CYLINDER = 4; +#if LL_MESH_ENABLED const U8 LL_SCULPT_TYPE_MESH = 5; const U8 LL_SCULPT_TYPE_MASK = LL_SCULPT_TYPE_SPHERE | LL_SCULPT_TYPE_TORUS | LL_SCULPT_TYPE_PLANE | LL_SCULPT_TYPE_CYLINDER | LL_SCULPT_TYPE_MESH; +#else +const U8 LL_SCULPT_TYPE_MASK = LL_SCULPT_TYPE_SPHERE | LL_SCULPT_TYPE_TORUS | LL_SCULPT_TYPE_PLANE | + LL_SCULPT_TYPE_CYLINDER; +#endif const U8 LL_SCULPT_FLAG_INVERT = 64; const U8 LL_SCULPT_FLAG_MIRROR = 128; diff --git a/indra/llrender/llvertexbuffer.cpp b/indra/llrender/llvertexbuffer.cpp index 48c20b09a8..514ca25aa0 100644 --- a/indra/llrender/llvertexbuffer.cpp +++ b/indra/llrender/llvertexbuffer.cpp @@ -1318,7 +1318,7 @@ void LLVertexBuffer::setBuffer(U32 data_mask) { if (mGLBuffer) { - if (useVBOs() && sVBOActive) + if (sVBOActive) { glBindBufferARB(GL_ARRAY_BUFFER_ARB, 0); sBindCount++; @@ -1330,7 +1330,7 @@ void LLVertexBuffer::setBuffer(U32 data_mask) setup = TRUE; // ... or a client memory pointer changed } } - if (useVBOs() && mGLIndices && sIBOActive) + if (mGLIndices && sIBOActive) { /*if (sMapped) { diff --git a/indra/llvfs/llvfs.cpp b/indra/llvfs/llvfs.cpp index ddb76fb2ba..2b01288fd7 100644 --- a/indra/llvfs/llvfs.cpp +++ b/indra/llvfs/llvfs.cpp @@ -2041,9 +2041,11 @@ std::string get_extension(LLAssetType::EType type) case LLAssetType::AT_ANIMATION: extension = ".lla"; break; +#if LL_MESH_ENABLED case LLAssetType::AT_MESH: extension = ".slm"; break; +#endif default: // Just use the asset server filename extension in most cases extension += "."; diff --git a/indra/newview/app_settings/settings.xml b/indra/newview/app_settings/settings.xml index 79b58bcd7c..df9393c316 100644 --- a/indra/newview/app_settings/settings.xml +++ b/indra/newview/app_settings/settings.xml @@ -4908,7 +4908,7 @@ Type Boolean Value - 1 + 0 MigrateCacheDirectory @@ -6818,7 +6818,7 @@ Type F32 Value - 0.0 + -0.001 RenderSpotShadowOffset diff --git a/indra/newview/llappviewer.cpp b/indra/newview/llappviewer.cpp index 512b1dde7f..6978cdf672 100644 --- a/indra/newview/llappviewer.cpp +++ b/indra/newview/llappviewer.cpp @@ -1302,8 +1302,10 @@ bool LLAppViewer::cleanup() llinfos << "Cleaning Up" << llendflush; +#if LL_MESH_ENABLED // shut down mesh streamer gMeshRepo.shutdown(); +#endif // Must clean up texture references before viewer window is destroyed. LLHUDManager::getInstance()->updateEffects(); @@ -1728,8 +1730,10 @@ bool LLAppViewer::initThreads() mFastTimerLogThread->start(); } +#if LL_MESH_ENABLED // Mesh streaming and caching gMeshRepo.init(); +#endif LLFilePickerThread::initClass(); diff --git a/indra/newview/llassetuploadresponders.cpp b/indra/newview/llassetuploadresponders.cpp index c48dcdb061..afb76735ec 100644 --- a/indra/newview/llassetuploadresponders.cpp +++ b/indra/newview/llassetuploadresponders.cpp @@ -305,6 +305,7 @@ void LLAssetUploadResponder::uploadComplete(const LLSD& content) { } +#if LL_MESH_ENABLED LLNewAgentInventoryResponder::LLNewAgentInventoryResponder( const LLSD& post_data, const LLUUID& vfile_id, @@ -426,6 +427,7 @@ void LLNewAgentInventoryResponder::uploadComplete(const LLSD& content) LLImportColladaAssetCache::getInstance()->assetUploaded(mVFileID, content["new_asset"], TRUE); } +#endif LLSendTexLayerResponder::LLSendTexLayerResponder(const LLSD& post_data, const LLUUID& vfile_id, @@ -675,6 +677,7 @@ void LLUpdateTaskInventoryResponder::uploadComplete(const LLSD& content) } +#if LL_MESH_ENABLED ///////////////////////////////////////////////////// // LLNewAgentInventoryVariablePriceResponder::Impl // ///////////////////////////////////////////////////// @@ -1142,3 +1145,5 @@ void LLNewAgentInventoryVariablePriceResponder::showConfirmationDialog( boost::intrusive_ptr(this))); } } +#endif + diff --git a/indra/newview/llassetuploadresponders.h b/indra/newview/llassetuploadresponders.h index af3b3daf46..9abaccfde0 100644 --- a/indra/newview/llassetuploadresponders.h +++ b/indra/newview/llassetuploadresponders.h @@ -61,7 +61,7 @@ protected: std::string mFileName; }; - +#if LL_MESH_ENABLED // TODO*: Remove this once deprecated class LLNewAgentInventoryResponder : public LLAssetUploadResponder { @@ -116,6 +116,7 @@ private: class Impl; Impl* mImpl; }; +#endif struct LLBakedUploadData; class LLSendTexLayerResponder : public LLAssetUploadResponder diff --git a/indra/newview/lldrawpoolavatar.cpp b/indra/newview/lldrawpoolavatar.cpp index 9268cb1d47..63ca17d62d 100644 --- a/indra/newview/lldrawpoolavatar.cpp +++ b/indra/newview/lldrawpoolavatar.cpp @@ -103,6 +103,7 @@ S32 normal_channel = -1; S32 specular_channel = -1; S32 cube_channel = -1; +#if LL_MESH_ENABLED static const U32 rigged_data_mask[] = { LLDrawPoolAvatar::RIGGED_SIMPLE_MASK, LLDrawPoolAvatar::RIGGED_FULLBRIGHT_MASK, @@ -114,6 +115,7 @@ static const U32 rigged_data_mask[] = { LLDrawPoolAvatar::RIGGED_DEFERRED_BUMP_MASK, LLDrawPoolAvatar::RIGGED_DEFERRED_SIMPLE_MASK, }; +#endif static LLFastTimer::DeclareTimer FTM_SHADOW_AVATAR("Avatar Shadow"); @@ -194,12 +196,14 @@ void LLDrawPoolAvatar::beginDeferredPass(S32 pass) case 2: beginDeferredSkinned(); break; +#if LL_MESH_ENABLED case 3: beginDeferredRiggedSimple(); break; case 4: beginDeferredRiggedBump(); break; +#endif } } @@ -226,12 +230,14 @@ void LLDrawPoolAvatar::endDeferredPass(S32 pass) case 2: endDeferredSkinned(); break; +#if LL_MESH_ENABLED case 3: endDeferredRiggedSimple(); break; case 4: endDeferredRiggedBump(); break; +#endif } } @@ -242,7 +248,11 @@ void LLDrawPoolAvatar::renderDeferred(S32 pass) S32 LLDrawPoolAvatar::getNumPostDeferredPasses() { +#if LL_MESH_ENABLED return 6; +#else + return 1; +#endif } void LLDrawPoolAvatar::beginPostDeferredPass(S32 pass) @@ -252,6 +262,7 @@ void LLDrawPoolAvatar::beginPostDeferredPass(S32 pass) case 0: beginPostDeferredAlpha(); break; +#if LL_MESH_ENABLED case 1: beginRiggedFullbright(); break; @@ -267,6 +278,7 @@ void LLDrawPoolAvatar::beginPostDeferredPass(S32 pass) case 5: beginRiggedGlow(); break; +#endif } } @@ -284,6 +296,7 @@ void LLDrawPoolAvatar::beginPostDeferredAlpha() enable_vertex_weighting(sVertexProgram->mAttribute[LLViewerShaderMgr::AVATAR_WEIGHT]); } +#if LL_MESH_ENABLED void LLDrawPoolAvatar::beginDeferredRiggedAlpha() { sVertexProgram = &gDeferredSkinnedAlphaProgram; @@ -301,6 +314,7 @@ void LLDrawPoolAvatar::endDeferredRiggedAlpha() LLVertexBuffer::sWeight4Loc = -1; sVertexProgram = NULL; } +#endif void LLDrawPoolAvatar::endPostDeferredPass(S32 pass) { @@ -309,6 +323,7 @@ void LLDrawPoolAvatar::endPostDeferredPass(S32 pass) case 0: endPostDeferredAlpha(); break; +#if LL_MESH_ENABLED case 1: endRiggedFullbright(); break; @@ -324,6 +339,7 @@ void LLDrawPoolAvatar::endPostDeferredPass(S32 pass) case 5: endRiggedGlow(); break; +#endif } } @@ -357,7 +373,11 @@ void LLDrawPoolAvatar::renderPostDeferred(S32 pass) S32 LLDrawPoolAvatar::getNumShadowPasses() { +#if LL_MESH_ENABLED return 2; +#else + return 1; +#endif } void LLDrawPoolAvatar::beginShadowPass(S32 pass) @@ -448,6 +468,7 @@ void LLDrawPoolAvatar::renderShadow(S32 pass) avatarp->renderSkinned(AVATAR_RENDER_PASS_SINGLE); } +#if LL_MESH_ENABLED else { renderRigged(avatarp, RIGGED_SIMPLE); @@ -457,10 +478,12 @@ void LLDrawPoolAvatar::renderShadow(S32 pass) renderRigged(avatarp, RIGGED_SHINY); renderRigged(avatarp, RIGGED_FULLBRIGHT_ALPHA); } +#endif } S32 LLDrawPoolAvatar::getNumPasses() { +#if LL_MESH_ENABLED if (LLPipeline::sImpostorRender) { return 8; @@ -473,8 +496,19 @@ S32 LLDrawPoolAvatar::getNumPasses() { return 3; } +#else + if (LLPipeline::sImpostorRender) + { + return 1; + } + else + { + return 3; + } +#endif } + S32 LLDrawPoolAvatar::getNumDeferredPasses() { if (LLPipeline::sImpostorRender) @@ -522,6 +556,7 @@ void LLDrawPoolAvatar::beginRenderPass(S32 pass) case 2: beginSkinned(); break; +#if LL_MESH_ENABLED case 3: beginRiggedSimple(); break; @@ -543,6 +578,7 @@ void LLDrawPoolAvatar::beginRenderPass(S32 pass) case 9: beginRiggedGlow(); break; +#endif } } @@ -566,6 +602,7 @@ void LLDrawPoolAvatar::endRenderPass(S32 pass) case 2: endSkinned(); break; +#if LL_MESH_ENABLED case 3: endRiggedSimple(); break; @@ -587,6 +624,7 @@ void LLDrawPoolAvatar::endRenderPass(S32 pass) case 9: endRiggedGlow(); break; +#endif } } @@ -772,6 +810,7 @@ void LLDrawPoolAvatar::endSkinned() gGL.getTexUnit(0)->activate(); } +#if LL_MESH_ENABLED void LLDrawPoolAvatar::beginRiggedSimple() { sVertexProgram = &gSkinnedObjectSimpleProgram; @@ -924,6 +963,7 @@ void LLDrawPoolAvatar::endDeferredRiggedBump() sDiffuseChannel = 0; sVertexProgram = NULL; } +#endif void LLDrawPoolAvatar::beginDeferredSkinned() { @@ -1069,6 +1109,7 @@ void LLDrawPoolAvatar::renderAvatars(LLVOAvatar* single_avatar, S32 pass) return; } +#if LL_MESH_ENABLED if (pass == 3) { if (is_deferred_render) @@ -1150,7 +1191,7 @@ void LLDrawPoolAvatar::renderAvatars(LLVOAvatar* single_avatar, S32 pass) gGL.setSceneBlendType(LLRender::BT_ALPHA); return; } - +#endif if (sShaderLevel > 0) { @@ -1188,6 +1229,7 @@ void LLDrawPoolAvatar::renderAvatars(LLVOAvatar* single_avatar, S32 pass) } } +#if LL_MESH_ENABLED void LLDrawPoolAvatar::updateRiggedFaceVertexBuffer(LLFace* face, const LLMeshSkinInfo* skin, LLVolume* volume, const LLVolumeFace& vol_face) { U32 data_mask = 0; @@ -1377,7 +1419,7 @@ void LLDrawPoolAvatar::renderRiggedGlow(LLVOAvatar* avatar) { renderRigged(avatar, RIGGED_GLOW, true); } - +#endif @@ -1479,6 +1521,7 @@ LLColor3 LLDrawPoolAvatar::getDebugColor() const return LLColor3(0.f, 1.f, 0.f); } +#if LL_MESH_ENABLED void LLDrawPoolAvatar::addRiggedFace(LLFace* facep, U32 type) { if (facep->mRiggedIndex.empty()) @@ -1533,6 +1576,7 @@ void LLDrawPoolAvatar::removeRiggedFace(LLFace* facep) } } } +#endif LLVertexBufferAvatar::LLVertexBufferAvatar() : LLVertexBuffer(sDataMask, diff --git a/indra/newview/lldrawpoolavatar.h b/indra/newview/lldrawpoolavatar.h index 4a5b009412..46ffc42f04 100644 --- a/indra/newview/lldrawpoolavatar.h +++ b/indra/newview/lldrawpoolavatar.h @@ -96,6 +96,23 @@ public: void beginRigid(); void beginImpostor(); void beginSkinned(); + + void endRigid(); + void endImpostor(); + void endSkinned(); + + void beginDeferredImpostor(); + void beginDeferredRigid(); + void beginDeferredSkinned(); + + void endDeferredImpostor(); + void endDeferredRigid(); + void endDeferredSkinned(); + + void beginPostDeferredAlpha(); + void endPostDeferredAlpha(); + +#if LL_MESH_ENABLED void beginRiggedSimple(); void beginRiggedFullbright(); void beginRiggedFullbrightShiny(); @@ -103,12 +120,8 @@ public: void beginRiggedAlpha(); void beginRiggedFullbrightAlpha(); void beginRiggedGlow(); - void beginPostDeferredAlpha(); void beginDeferredRiggedAlpha(); - void endRigid(); - void endImpostor(); - void endSkinned(); void endRiggedSimple(); void endRiggedFullbright(); void endRiggedFullbrightShiny(); @@ -116,18 +129,11 @@ public: void endRiggedAlpha(); void endRiggedFullbrightAlpha(); void endRiggedGlow(); - void endPostDeferredAlpha(); void endDeferredRiggedAlpha(); - void beginDeferredImpostor(); - void beginDeferredRigid(); - void beginDeferredSkinned(); void beginDeferredRiggedSimple(); void beginDeferredRiggedBump(); - void endDeferredImpostor(); - void endDeferredRigid(); - void endDeferredSkinned(); void endDeferredRiggedSimple(); void endDeferredRiggedBump(); @@ -147,11 +153,6 @@ public: void renderDeferredRiggedSimple(LLVOAvatar* avatar); void renderDeferredRiggedBump(LLVOAvatar* avatar); - /*virtual*/ LLViewerTexture *getDebugTexture(); - /*virtual*/ LLColor3 getDebugColor() const; // For AGP debug display - - void renderAvatars(LLVOAvatar *single_avatar, S32 pass = -1); // renders only one avatar if single_avatar is not null. - typedef enum { RIGGED_SIMPLE = 0, @@ -202,6 +203,13 @@ public: void removeRiggedFace(LLFace* facep); std::vector mRiggedFace[NUM_RIGGED_PASSES]; +#endif + + /*virtual*/ LLViewerTexture *getDebugTexture(); + /*virtual*/ LLColor3 getDebugColor() const; // For AGP debug display + + void renderAvatars(LLVOAvatar *single_avatar, S32 pass = -1); // renders only one avatar if single_avatar is not null. + static BOOL sSkipOpaque; static BOOL sSkipTransparent; diff --git a/indra/newview/llface.cpp b/indra/newview/llface.cpp index e4f9e28c77..c48106863e 100644 --- a/indra/newview/llface.cpp +++ b/indra/newview/llface.cpp @@ -224,11 +224,13 @@ void LLFace::destroy() { LLFastTimer t(FTM_DESTROY_DRAWPOOL); +#if LL_MESH_ENABLED if (this->isState(LLFace::RIGGED) && mDrawPoolp->getType() == LLDrawPool::POOL_AVATAR) { ((LLDrawPoolAvatar*) mDrawPoolp)->removeRiggedFace(this); } else +#endif { mDrawPoolp->removeFace(this); } diff --git a/indra/newview/llinventorybridge.cpp b/indra/newview/llinventorybridge.cpp index 78789eb9ce..3999a8ee88 100644 --- a/indra/newview/llinventorybridge.cpp +++ b/indra/newview/llinventorybridge.cpp @@ -917,6 +917,7 @@ LLInvFVBridge* LLInvFVBridge::createBridge(LLAssetType::EType asset_type, // Only should happen for broken links. new_listener = new LLLinkItemBridge(inventory, root, uuid); break; +#if LL_MESH_ENABLED case LLAssetType::AT_MESH: if(!(inv_type == LLInventoryType::IT_MESH)) { @@ -924,6 +925,7 @@ LLInvFVBridge* LLInvFVBridge::createBridge(LLAssetType::EType asset_type, } new_listener = new LLMeshBridge(inventory, root, uuid); break; +#endif default: llinfos << "Unhandled asset type (llassetstorage.h): " @@ -2765,7 +2767,9 @@ BOOL LLFolderBridge::dragOrDrop(MASK mask, BOOL drop, case DAD_ANIMATION: case DAD_GESTURE: case DAD_LINK: +#if LL_MESH_ENABLED case DAD_MESH: +#endif accept = dragItemIntoFolder((LLInventoryItem*)cargo_data, drop); break; @@ -3635,7 +3639,9 @@ BOOL LLCallingCardBridge::dragOrDrop(MASK mask, BOOL drop, case DAD_BODYPART: case DAD_ANIMATION: case DAD_GESTURE: - case DAD_MESH: +#if LL_MESH_ENABLED + case DAD_MESH: +#endif { LLInventoryItem* inv_item = (LLInventoryItem*)cargo_data; const LLPermissions& perm = inv_item->getPermissions(); @@ -4843,6 +4849,7 @@ void LLLinkItemBridge::buildContextMenu(LLMenuGL& menu, U32 flags) hide_context_entries(menu, items, disabled_items); } +#if LL_MESH_ENABLED // +=================================================+ // | LLMeshBridge | // +=================================================+ @@ -4899,6 +4906,7 @@ void LLMeshBridge::buildContextMenu(LLMenuGL& menu, U32 flags) hide_context_entries(menu, items, disabled_items); } +#endif // +=================================================+ // | LLLinkBridge | diff --git a/indra/newview/llinventorybridge.h b/indra/newview/llinventorybridge.h index d637fc6e36..8de7d63173 100644 --- a/indra/newview/llinventorybridge.h +++ b/indra/newview/llinventorybridge.h @@ -527,7 +527,7 @@ protected: }; - +#if LL_MESH_ENABLED class LLMeshBridge : public LLItemBridge { friend class LLInvFVBridge; @@ -543,7 +543,7 @@ protected: const LLUUID& uuid) : LLItemBridge(inventory, root, uuid) {} }; - +#endif diff --git a/indra/newview/llpanelgroupnotices.cpp b/indra/newview/llpanelgroupnotices.cpp index 7d82a10ca2..d8cfd4fc5d 100644 --- a/indra/newview/llpanelgroupnotices.cpp +++ b/indra/newview/llpanelgroupnotices.cpp @@ -159,7 +159,9 @@ BOOL LLGroupDropTarget::handleDragAndDrop(S32 x, S32 y, MASK mask, BOOL drop, case DAD_ANIMATION: case DAD_GESTURE: case DAD_CALLINGCARD: +#if LL_MESH_ENABLED case DAD_MESH: +#endif { LLViewerInventoryItem* inv_item = (LLViewerInventoryItem*)cargo_data; if(gInventory.getItem(inv_item->getUUID()) diff --git a/indra/newview/llpanelobjectinventory.cpp b/indra/newview/llpanelobjectinventory.cpp index 898cabb03d..58df2746f3 100644 --- a/indra/newview/llpanelobjectinventory.cpp +++ b/indra/newview/llpanelobjectinventory.cpp @@ -819,7 +819,9 @@ BOOL LLTaskCategoryBridge::dragOrDrop(MASK mask, BOOL drop, case DAD_ANIMATION: case DAD_GESTURE: case DAD_CALLINGCARD: +#if LL_MESH_ENABLED case DAD_MESH: +#endif accept = LLToolDragAndDrop::isInventoryDropAcceptable(object, (LLViewerInventoryItem*)cargo_data); if(accept && drop) { @@ -1240,6 +1242,7 @@ LLUIImagePtr LLTaskWearableBridge::getIcon() const return LLInventoryIcon::getIcon(mAssetType, mInventoryType, FALSE, mFlags, FALSE ); } +#if LL_MESH_ENABLED ///---------------------------------------------------------------------------- /// Class LLTaskMeshBridge ///---------------------------------------------------------------------------- @@ -1351,6 +1354,7 @@ void LLTaskMeshBridge::buildContextMenu(LLMenuGL& menu, U32 flags) hide_context_entries(menu, items, disabled_items); } +#endif ///---------------------------------------------------------------------------- /// LLTaskInvFVBridge impl @@ -1431,11 +1435,13 @@ LLTaskInvFVBridge* LLTaskInvFVBridge::createObjectBridge(LLPanelObjectInventory* object->getUUID(), object->getName()); break; +#if LL_MESH_ENABLED case LLAssetType::AT_MESH: new_bridge = new LLTaskMeshBridge(panel, object->getUUID(), object->getName()); break; +#endif default: llinfos << "Unhandled inventory type (llassetstorage.h): " << (S32)type << llendl; diff --git a/indra/newview/llselectmgr.cpp b/indra/newview/llselectmgr.cpp index ceef03d2f8..0ce7ffb077 100644 --- a/indra/newview/llselectmgr.cpp +++ b/indra/newview/llselectmgr.cpp @@ -6165,6 +6165,7 @@ S32 LLObjectSelection::getObjectCount(BOOL mesh_adjust) cleanupNodes(); S32 count = mList.size(); +#if LL_MESH_ENABLED if (mesh_adjust) { for (list_t::iterator iter = mList.begin(); iter != mList.end(); ++iter) @@ -6185,6 +6186,7 @@ S32 LLObjectSelection::getObjectCount(BOOL mesh_adjust) } } +#endif return count; } diff --git a/indra/newview/lltexturectrl.cpp b/indra/newview/lltexturectrl.cpp index 7b57f09e59..6088c31559 100644 --- a/indra/newview/lltexturectrl.cpp +++ b/indra/newview/lltexturectrl.cpp @@ -292,7 +292,11 @@ BOOL LLFloaterTexturePicker::handleDragAndDrop( { BOOL handled = FALSE; - if ((cargo_type == DAD_TEXTURE) || (cargo_type == DAD_MESH)) + bool is_mesh = false; +#if LL_MESH_ENABLED + is_mesh = cargo_type == DAD_MESH; +#endif + if ((cargo_type == DAD_TEXTURE) || is_mesh) { LLInventoryItem *item = (LLInventoryItem *)cargo_data; @@ -1177,8 +1181,12 @@ BOOL LLTextureCtrl::handleDragAndDrop(S32 x, S32 y, MASK mask, // returns true, then the cast was valid, and we can perform // the third test without problems. LLInventoryItem* item = (LLInventoryItem*)cargo_data; + bool is_mesh = false; +#if LL_MESH_ENABLED + is_mesh = cargo_type == DAD_MESH; +#endif if (getEnabled() && - ((cargo_type == DAD_TEXTURE) || (cargo_type == DAD_MESH)) && + ((cargo_type == DAD_TEXTURE) || is_mesh) && allowDrop(item)) { if (drop) diff --git a/indra/newview/lltooldraganddrop.cpp b/indra/newview/lltooldraganddrop.cpp index 15924d67cf..d679ccd3d8 100644 --- a/indra/newview/lltooldraganddrop.cpp +++ b/indra/newview/lltooldraganddrop.cpp @@ -331,7 +331,9 @@ LLToolDragAndDrop::LLDragAndDropDictionary::LLDragAndDropDictionary() addEntry(DAD_ANIMATION, new DragAndDropEntry(&LLToolDragAndDrop::dad3dNULL, &LLToolDragAndDrop::dad3dNULL, &LLToolDragAndDrop::dad3dGiveInventory, &LLToolDragAndDrop::dad3dUpdateInventory, &LLToolDragAndDrop::dad3dNULL)); addEntry(DAD_GESTURE, new DragAndDropEntry(&LLToolDragAndDrop::dad3dNULL, &LLToolDragAndDrop::dad3dActivateGesture, &LLToolDragAndDrop::dad3dGiveInventory, &LLToolDragAndDrop::dad3dUpdateInventory, &LLToolDragAndDrop::dad3dNULL)); addEntry(DAD_LINK, new DragAndDropEntry(&LLToolDragAndDrop::dad3dNULL, &LLToolDragAndDrop::dad3dNULL, &LLToolDragAndDrop::dad3dNULL, &LLToolDragAndDrop::dad3dNULL, &LLToolDragAndDrop::dad3dNULL)); +#if LL_MESH_ENABLED addEntry(DAD_MESH, new DragAndDropEntry(&LLToolDragAndDrop::dad3dNULL, &LLToolDragAndDrop::dad3dNULL, &LLToolDragAndDrop::dad3dGiveInventory, &LLToolDragAndDrop::dad3dMeshObject, &LLToolDragAndDrop::dad3dNULL)); +#endif // TODO: animation on self could play it? edit it? // TODO: gesture on self could play it? edit it? }; @@ -1035,6 +1037,7 @@ void LLToolDragAndDrop::dropTextureAllFaces(LLViewerObject* hit_obj, hit_obj->sendTEUpdate(); } +#if LL_MESH_EANBLED void LLToolDragAndDrop::dropMesh(LLViewerObject* hit_obj, LLInventoryItem* item, LLToolDragAndDrop::ESource source, @@ -1059,6 +1062,7 @@ void LLToolDragAndDrop::dropMesh(LLViewerObject* hit_obj, dialog_refresh_all(); } +#endif /* void LLToolDragAndDrop::dropTextureOneFaceAvatar(LLVOAvatar* avatar, S32 hit_face, LLInventoryItem* item) @@ -1501,7 +1505,9 @@ bool LLToolDragAndDrop::handleGiveDragAndDrop(LLUUID dest_agent, LLUUID session_ case DAD_ANIMATION: case DAD_GESTURE: case DAD_CALLINGCARD: +#if LL_MESH_ENABLED case DAD_MESH: +#endif { LLViewerInventoryItem* inv_item = (LLViewerInventoryItem*)cargo_data; if(gInventory.getItem(inv_item->getUUID()) @@ -1858,10 +1864,12 @@ EAcceptance LLToolDragAndDrop::dad3dApplyToObject( dropTextureOneFace(obj, face, item, mSource, mSourceID); } } +#if LL_MESH_ENABLED else if (cargo_type == DAD_MESH) { dropMesh(obj, item, mSource, mSourceID); } +#endif else { llwarns << "unsupported asset type" << llendl; @@ -1886,11 +1894,13 @@ EAcceptance LLToolDragAndDrop::dad3dTextureObject( return dad3dApplyToObject(obj, face, mask, drop, DAD_TEXTURE); } +#if LL_MESH_ENABLED EAcceptance LLToolDragAndDrop::dad3dMeshObject( LLViewerObject* obj, S32 face, MASK mask, BOOL drop) { return dad3dApplyToObject(obj, face, mask, drop, DAD_MESH); } +#endif diff --git a/indra/newview/lltooldraganddrop.h b/indra/newview/lltooldraganddrop.h index 596cc2ee88..ceeaa8c820 100644 --- a/indra/newview/lltooldraganddrop.h +++ b/indra/newview/lltooldraganddrop.h @@ -154,8 +154,10 @@ protected: MASK mask, BOOL drop); EAcceptance dad3dTextureObject(LLViewerObject* obj, S32 face, MASK mask, BOOL drop); +#if LL_MESH_ENABLED EAcceptance dad3dMeshObject(LLViewerObject* obj, S32 face, MASK mask, BOOL drop); +#endif // EAcceptance dad3dTextureSelf(LLViewerObject* obj, S32 face, // MASK mask, BOOL drop); EAcceptance dad3dWearItem(LLViewerObject* obj, S32 face, @@ -242,10 +244,12 @@ public: LLInventoryItem* item, ESource source, const LLUUID& src_id); +#if LL_MESH_ENABLED static void dropMesh(LLViewerObject* hit_obj, LLInventoryItem* item, ESource source, const LLUUID& src_id); +#endif //static void dropTextureOneFaceAvatar(LLVOAvatar* avatar,S32 hit_face, // LLInventoryItem* item) diff --git a/indra/newview/llviewerassettype.cpp b/indra/newview/llviewerassettype.cpp index 4c088a72b7..370767002a 100644 --- a/indra/newview/llviewerassettype.cpp +++ b/indra/newview/llviewerassettype.cpp @@ -85,7 +85,9 @@ LLViewerAssetDictionary::LLViewerAssetDictionary() addEntry(LLViewerAssetType::AT_LINK, new ViewerAssetEntry(DAD_LINK)); addEntry(LLViewerAssetType::AT_LINK_FOLDER, new ViewerAssetEntry(DAD_LINK)); +#if LL_MESH_ENABLED addEntry(LLViewerAssetType::AT_MESH, new ViewerAssetEntry(DAD_MESH)); +#endif addEntry(LLViewerAssetType::AT_NONE, new ViewerAssetEntry(DAD_NONE)); }; diff --git a/indra/newview/llviewerfloaterreg.cpp b/indra/newview/llviewerfloaterreg.cpp index 525610f983..c953fb23e3 100644 --- a/indra/newview/llviewerfloaterreg.cpp +++ b/indra/newview/llviewerfloaterreg.cpp @@ -177,7 +177,9 @@ void LLViewerFloaterReg::registerFloaters() LLFloaterReg::add("hud", "floater_hud.xml", (LLFloaterBuildFunc)&LLFloaterReg::build); LLFloaterReg::add("impanel", "floater_im_session.xml", (LLFloaterBuildFunc)&LLFloaterReg::build); +#if LL_MESH_ENABLED LLFloaterReg::add("import_collada", "floater_import_collada.xml", (LLFloaterBuildFunc)&LLFloaterReg::build); +#endif LLFloaterReg::add("im_container", "floater_im_container.xml", (LLFloaterBuildFunc)&LLFloaterReg::build); LLFloaterReg::add("im_well_window", "floater_sys_well.xml", (LLFloaterBuildFunc)&LLFloaterReg::build); LLFloaterReg::add("incoming_call", "floater_incoming_call.xml", (LLFloaterBuildFunc)&LLFloaterReg::build); @@ -258,7 +260,9 @@ void LLViewerFloaterReg::registerFloaters() LLFloaterReg::add("upload_anim", "floater_animation_preview.xml", (LLFloaterBuildFunc)&LLFloaterReg::build, "upload"); LLFloaterReg::add("upload_image", "floater_image_preview.xml", (LLFloaterBuildFunc)&LLFloaterReg::build, "upload"); LLFloaterReg::add("upload_sound", "floater_sound_preview.xml", (LLFloaterBuildFunc)&LLFloaterReg::build, "upload"); +#if LL_MESH_ENABLED LLFloaterReg::add("upload_model", "floater_model_preview.xml", (LLFloaterBuildFunc)&LLFloaterReg::build, "upload"); +#endif LLFloaterReg::add("voice_controls", "floater_voice_controls.xml", (LLFloaterBuildFunc)&LLFloaterReg::build); LLFloaterReg::add("voice_effect", "floater_voice_effect.xml", (LLFloaterBuildFunc)&LLFloaterReg::build); diff --git a/indra/newview/llviewermenufile.cpp b/indra/newview/llviewermenufile.cpp index dd65ca2d25..af4549f7e1 100644 --- a/indra/newview/llviewermenufile.cpp +++ b/indra/newview/llviewermenufile.cpp @@ -350,6 +350,7 @@ class LLFileUploadImage : public view_listener_t } }; +#if LL_MESH_ENABLED class LLFileUploadScene : public view_listener_t { bool handleEvent(const LLSD& userdata) @@ -376,6 +377,7 @@ class LLFileUploadModel : public view_listener_t return TRUE; } }; +#endif class LLFileUploadSound : public view_listener_t { @@ -1144,6 +1146,7 @@ void upload_new_resource( lldebugs << "Folder: " << gInventory.findCategoryUUIDForType((destination_folder_type == LLFolderType::FT_NONE) ? LLFolderType::assetTypeToFolderType(asset_type) : destination_folder_type) << llendl; lldebugs << "Asset Type: " << LLAssetType::lookup(asset_type) << llendl; +#if LL_MESH_ENABLED std::string url = gAgent.getRegion()->getCapability( "NewFileAgentInventory"); @@ -1171,8 +1174,8 @@ void upload_new_resource( asset_type)); } else +#endif { - llinfos << "NewAgentInventory capability not found, new agent inventory via asset system." << llendl; // check for adequate funds // TODO: do this check on the sim if (LLAssetType::AT_SOUND == asset_type || @@ -1217,6 +1220,7 @@ void upload_new_resource( } } +#if LL_MESH_ENABLED BOOL upload_new_variable_price_resource( const LLTransactionID &tid, LLAssetType::EType asset_type, @@ -1288,6 +1292,7 @@ BOOL upload_new_variable_price_resource( return FALSE; } } +#endif LLAssetID generate_asset_id_for_new_upload(const LLTransactionID& tid) { @@ -1358,8 +1363,10 @@ void init_menu_file() view_listener_t::addCommit(new LLFileUploadImage(), "File.UploadImage"); view_listener_t::addCommit(new LLFileUploadSound(), "File.UploadSound"); view_listener_t::addCommit(new LLFileUploadAnim(), "File.UploadAnim"); +#if LL_MESH_ENABLED view_listener_t::addCommit(new LLFileUploadModel(), "File.UploadModel"); view_listener_t::addCommit(new LLFileUploadScene(), "File.UploadScene"); +#endif view_listener_t::addCommit(new LLFileUploadBulk(), "File.UploadBulk"); view_listener_t::addCommit(new LLFileCloseWindow(), "File.CloseWindow"); view_listener_t::addCommit(new LLFileCloseAllWindows(), "File.CloseAllWindows"); diff --git a/indra/newview/llviewermenufile.h b/indra/newview/llviewermenufile.h index 08444551a9..bb7cfce862 100644 --- a/indra/newview/llviewermenufile.h +++ b/indra/newview/llviewermenufile.h @@ -74,6 +74,7 @@ void upload_new_resource( S32 expected_upload_cost, void *userdata); +#if LL_MESH_ENABLED // TODO* : Move all uploads to use this new function // since at some point, that upload path will be deprecated and no longer // used @@ -91,6 +92,7 @@ BOOL upload_new_variable_price_resource( U32 everyone_perms, const std::string& display_name, const LLSD& asset_resources); +#endif LLAssetID generate_asset_id_for_new_upload(const LLTransactionID& tid); void increase_new_upload_stats(LLAssetType::EType asset_type); diff --git a/indra/newview/llviewertexteditor.cpp b/indra/newview/llviewertexteditor.cpp index d35be8e1bf..697e8afa65 100644 --- a/indra/newview/llviewertexteditor.cpp +++ b/indra/newview/llviewertexteditor.cpp @@ -535,7 +535,9 @@ LLUIImagePtr LLEmbeddedItems::getItemImage(llwchar ext_char) const case LLAssetType::AT_BODYPART: img_name = "Inv_Skin"; break; case LLAssetType::AT_ANIMATION: img_name = "Inv_Animation"; break; case LLAssetType::AT_GESTURE: img_name = "Inv_Gesture"; break; +#if LL_MESH_ENABLED case LLAssetType::AT_MESH: img_name = "Inv_Mesh"; break; +#endif default: llassert(0); } diff --git a/indra/newview/llviewerwindow.cpp b/indra/newview/llviewerwindow.cpp index fc4b590bf4..9f559331b0 100644 --- a/indra/newview/llviewerwindow.cpp +++ b/indra/newview/llviewerwindow.cpp @@ -591,8 +591,10 @@ public: if (gSavedSettings.getBOOL("DebugShowUploadCost")) { +#if LL_MESH_ENABLED addText(xpos, ypos, llformat(" Meshes: L$%d", gPipeline.mDebugMeshUploadCost)); ypos += y_inc/2; +#endif addText(xpos, ypos, llformat(" Sculpties: L$%d", gPipeline.mDebugSculptUploadCost)); ypos += y_inc/2; addText(xpos, ypos, llformat(" Textures: L$%d", gPipeline.mDebugTextureUploadCost)); @@ -602,6 +604,7 @@ public: ypos += y_inc; } +#if LL_MESH_ENABLED //temporary hack to give feedback on mesh upload progress if (!gMeshRepo.mUploads.empty()) { @@ -631,6 +634,7 @@ public: ypos += y_inc; } +#endif } void draw() diff --git a/indra/newview/llvovolume.cpp b/indra/newview/llvovolume.cpp index db9e0b88e1..d66aa567a8 100644 --- a/indra/newview/llvovolume.cpp +++ b/indra/newview/llvovolume.cpp @@ -926,6 +926,7 @@ BOOL LLVOVolume::setVolume(const LLVolumeParams ¶ms, const S32 detail, bool BOOL is404 = FALSE; +#if LL_MESH_ENABLED if (isSculpted()) { // if it's a mesh @@ -945,6 +946,7 @@ BOOL LLVOVolume::setVolume(const LLVolumeParams ¶ms, const S32 detail, bool } } } +#endif // Check if we need to change implementations bool is_flexible = (volume_params.getPathParams().getCurveType() == LL_PCODE_PATH_FLEXIBLE); @@ -989,10 +991,11 @@ BOOL LLVOVolume::setVolume(const LLVolumeParams ¶ms, const S32 detail, bool updateSculptTexture(); + if (isSculpted()) { updateSculptTexture(); - +#if LL_MESH_ENABLED // if it's a mesh if ((volume_params.getSculptType() & LL_SCULPT_TYPE_MASK) == LL_SCULPT_TYPE_MESH) { @@ -1008,6 +1011,7 @@ BOOL LLVOVolume::setVolume(const LLVolumeParams ¶ms, const S32 detail, bool } } else // otherwise is sculptie +#endif { if (mSculptTexture.notNull()) { @@ -2659,12 +2663,13 @@ BOOL LLVOVolume::isMesh() const { LLSculptParams *sculpt_params = (LLSculptParams *)getParameterEntry(LLNetworkData::PARAMS_SCULPT); U8 sculpt_type = sculpt_params->getSculptType(); - +#if LL_MESH_ENABLED if ((sculpt_type & LL_SCULPT_TYPE_MASK) == LL_SCULPT_TYPE_MESH) // mesh is a mesh { return TRUE; } +#endif } return FALSE; @@ -3010,6 +3015,7 @@ F32 LLVOVolume::getBinRadius() F32 scale = 1.f; +#if LL_MESH_ENABLED if (isSculpted()) { LLSculptParams *sculpt_params = (LLSculptParams *)getParameterEntry(LLNetworkData::PARAMS_SCULPT); @@ -3031,6 +3037,7 @@ F32 LLVOVolume::getBinRadius() scale = 1.f/llmax(vert_count/1024.f, 1.f); } } +#endif const LLVector4a* ext = mDrawable->getSpatialExtents(); @@ -3513,9 +3520,11 @@ void LLVolumeGeometryManager::rebuildGeom(LLSpatialGroup* group) drawablep->clearState(LLDrawable::HAS_ALPHA); +#if LL_MESH_ENABLED bool rigged = vobj->isAttachment() && vobj->isMesh() && gMeshRepo.getSkinInfo(vobj->getVolume()->getParams().getSculptID()); +#endif bool bake_sunlight = LLPipeline::sBakeSunlight && drawablep->isStatic(); @@ -3526,6 +3535,7 @@ void LLVolumeGeometryManager::rebuildGeom(LLSpatialGroup* group) drawablep->updateFaceSize(i); LLFace* facep = drawablep->getFace(i); +#if LL_MESH_ENABLED if (rigged) { if (!facep->isState(LLFace::RIGGED)) @@ -3622,6 +3632,7 @@ void LLVolumeGeometryManager::rebuildGeom(LLSpatialGroup* group) facep->clearState(LLFace::RIGGED); } } +#endif if (cur_total > max_total || facep->getIndicesCount() <= 0 || facep->getGeomCount() <= 0) { diff --git a/indra/newview/pipeline.cpp b/indra/newview/pipeline.cpp index 583b3df0e5..c5c7277211 100644 --- a/indra/newview/pipeline.cpp +++ b/indra/newview/pipeline.cpp @@ -2025,7 +2025,9 @@ void LLPipeline::rebuildPriorityGroups() assertInitialized(); +#if LL_MESH_ENABLED gMeshRepo.notifyLoadedMeshes(); +#endif // Iterate through all drawables on the priority build queue, for (LLSpatialGroup::sg_list_t::iterator iter = mGroupQ1.begin(); @@ -3859,6 +3861,7 @@ void LLPipeline::renderDebug() gPipeline.mDebugTextureUploadCost = textures.size() * 10; gPipeline.mDebugSculptUploadCost = sculpts.size()*10; +#if LL_MESH_ENABLED U32 mesh_cost = 0; for (std::set::iterator iter = meshes.begin(); iter != meshes.end(); ++iter) @@ -3867,6 +3870,7 @@ void LLPipeline::renderDebug() } gPipeline.mDebugMeshUploadCost = mesh_cost; +#endif } for (LLCullResult::bridge_list_t::const_iterator i = sCull->beginVisibleBridge(); i != sCull->endVisibleBridge(); ++i) diff --git a/indra/newview/pipeline.h b/indra/newview/pipeline.h index 4e8760947c..95f951b393 100644 --- a/indra/newview/pipeline.h +++ b/indra/newview/pipeline.h @@ -45,6 +45,7 @@ #include "llgl.h" #include "lldrawable.h" #include "llrendertarget.h" +#include "llmodel.h" //for LL_MESH_ENaBLED #include @@ -61,7 +62,10 @@ class LLCullResult; class LLVOAvatar; class LLGLSLShader; class LLCurlRequest; + +#if LL_MESH_ENABLED class LLMeshResponder; +#endif typedef enum e_avatar_skinning_method { @@ -465,7 +469,9 @@ public: S32 mDebugTextureUploadCost; S32 mDebugSculptUploadCost; +#if LL_MESH_ENABLED S32 mDebugMeshUploadCost; +#endif S32 mLightingChanges; S32 mGeometryChanges; @@ -712,13 +718,6 @@ public: protected: std::vector mSelectedFaces; - - typedef std::map > mesh_load_map; - mesh_load_map mLoadingMeshes[4]; - - typedef std::list mesh_response_list; - mesh_response_list mMeshResponseList; - LLPointer mFaceSelectImagep; U32 mLightMask; diff --git a/indra/newview/skins/default/xui/en/panel_preferences_graphics1.xml b/indra/newview/skins/default/xui/en/panel_preferences_graphics1.xml index 8b431e9c54..b54bad98d0 100644 --- a/indra/newview/skins/default/xui/en/panel_preferences_graphics1.xml +++ b/indra/newview/skins/default/xui/en/panel_preferences_graphics1.xml @@ -160,7 +160,7 @@ layout="topleft" left="5" name="CustomGraphics Panel" - top="101" + top="76" width="485"> Low @@ -625,7 +625,7 @@ follows="left|top" height="12" layout="topleft" - left_delta="-260" + left="200" name="AvatarRenderingText" top_pad="8" width="128"> @@ -673,7 +673,7 @@ left="358" left_pad="-30" name="TerrainDetailText" - top="488" + top="226" width="155"> Terrain detail: -- cgit v1.3 From a7eadcccd02210349a07d10a52669f4678d8fae2 Mon Sep 17 00:00:00 2001 From: Dave Parks Date: Thu, 10 Jun 2010 23:11:21 -0500 Subject: Unused variable/function --- indra/newview/llvovolume.cpp | 9 ++++++--- 1 file changed, 6 insertions(+), 3 deletions(-) (limited to 'indra/newview/llvovolume.cpp') diff --git a/indra/newview/llvovolume.cpp b/indra/newview/llvovolume.cpp index d66aa567a8..38e54342ea 100644 --- a/indra/newview/llvovolume.cpp +++ b/indra/newview/llvovolume.cpp @@ -2659,18 +2659,20 @@ BOOL LLVOVolume::isSculpted() const BOOL LLVOVolume::isMesh() const { +#if LL_MESH_ENABLED if (isSculpted()) { LLSculptParams *sculpt_params = (LLSculptParams *)getParameterEntry(LLNetworkData::PARAMS_SCULPT); U8 sculpt_type = sculpt_params->getSculptType(); -#if LL_MESH_ENABLED + if ((sculpt_type & LL_SCULPT_TYPE_MASK) == LL_SCULPT_TYPE_MESH) // mesh is a mesh { return TRUE; } -#endif } +#endif + return FALSE; } @@ -3427,6 +3429,7 @@ void LLVolumeGeometryManager::getGeometry(LLSpatialGroup* group) static LLFastTimer::DeclareTimer FTM_REBUILD_VOLUME_VB("Volume"); static LLFastTimer::DeclareTimer FTM_REBUILD_VBO("VBO Rebuilt"); +#if LL_MESH_ENABLED static LLDrawPoolAvatar* get_avatar_drawpool(LLViewerObject* vobj) { LLVOAvatar* avatar = vobj->getAvatar(); @@ -3453,7 +3456,7 @@ static LLDrawPoolAvatar* get_avatar_drawpool(LLViewerObject* vobj) return NULL; } - +#endif void LLVolumeGeometryManager::rebuildGeom(LLSpatialGroup* group) { -- cgit v1.3 From 6e37ec08f678451a526f34218cb070d117cdf60a Mon Sep 17 00:00:00 2001 From: Dave Parks Date: Mon, 14 Jun 2010 23:13:10 -0500 Subject: Builds with LLConvexDecompInter as a static lib. --- indra/cmake/CMakeLists.txt | 2 ++ indra/llcommon/llassettype.h | 2 -- indra/llinventory/llinventorytype.h | 3 --- indra/llmath/llvolume.cpp | 8 ++++---- indra/llmath/llvolume.h | 4 ++-- indra/newview/CMakeLists.txt | 3 +++ indra/newview/llassetuploadresponders.h | 1 + indra/newview/lldrawpoolavatar.h | 1 + indra/newview/llinventorybridge.h | 1 + indra/newview/lltooldraganddrop.cpp | 2 +- indra/newview/lltooldraganddrop.h | 1 + indra/newview/llvovolume.cpp | 2 +- indra/newview/skins/default/xui/en/floater_about.xml | 4 +++- 13 files changed, 20 insertions(+), 14 deletions(-) (limited to 'indra/newview/llvovolume.cpp') diff --git a/indra/cmake/CMakeLists.txt b/indra/cmake/CMakeLists.txt index 4fc25dcc24..8612c46376 100644 --- a/indra/cmake/CMakeLists.txt +++ b/indra/cmake/CMakeLists.txt @@ -34,6 +34,7 @@ set(cmake_SOURCE_FILES FindXmlRpcEpi.cmake FMOD.cmake FreeType.cmake + GLOD.cmake GStreamer010Plugin.cmake GooglePerfTools.cmake JPEG.cmake @@ -41,6 +42,7 @@ set(cmake_SOURCE_FILES LLAudio.cmake LLCharacter.cmake LLCommon.cmake + LLConvexDecompInter.cmake LLCrashLogger.cmake LLDatabase.cmake LLImage.cmake diff --git a/indra/llcommon/llassettype.h b/indra/llcommon/llassettype.h index ebc43134cb..27d35e95ff 100644 --- a/indra/llcommon/llassettype.h +++ b/indra/llcommon/llassettype.h @@ -114,10 +114,8 @@ public: AT_LINK_FOLDER = 25, // Inventory folder link -#if LL_MESH_ENABLED AT_MESH = 49, // Mesh data in our proprietary SLM format -#endif AT_COUNT = 50, diff --git a/indra/llinventory/llinventorytype.h b/indra/llinventory/llinventorytype.h index d2fc67ef64..17ed3df951 100644 --- a/indra/llinventory/llinventorytype.h +++ b/indra/llinventory/llinventorytype.h @@ -67,10 +67,7 @@ public: IT_WEARABLE = 18, IT_ANIMATION = 19, IT_GESTURE = 20, - -#if LL_MESH_ENABLED IT_MESH = 22, -#endif IT_COUNT = 23, IT_NONE = -1 diff --git a/indra/llmath/llvolume.cpp b/indra/llmath/llvolume.cpp index 10cef533b0..53f484fb79 100644 --- a/indra/llmath/llvolume.cpp +++ b/indra/llmath/llvolume.cpp @@ -5355,9 +5355,9 @@ bool LLVolumeFace::VertexMapData::operator==(const LLVolumeFace::VertexData& rhs getNormal().equal3(rhs.getNormal()); } -bool LLVolumeFace::VertexMapData::ComparePosition::operator()(const LLVector4a& a, const LLVector4a& b) const +bool LLVolumeFace::VertexMapData::ComparePosition::operator()(const LLVector3& a, const LLVector3& b) const { - return a.less3(b); + return a < b; } void LLVolumeFace::optimize(F32 angle_cutoff) @@ -5375,7 +5375,7 @@ void LLVolumeFace::optimize(F32 angle_cutoff) getVertexData(index, cv); BOOL found = FALSE; - VertexMapData::PointMap::iterator point_iter = point_map.find(cv.getPosition()); + VertexMapData::PointMap::iterator point_iter = point_map.find(LLVector3(cv.getPosition().getF32())); if (point_iter != point_map.end()) { //duplicate point might exist for (U32 j = 0; j < point_iter->second.size(); ++j) @@ -5407,7 +5407,7 @@ void LLVolumeFace::optimize(F32 angle_cutoff) } else { - point_map[d.getPosition()].push_back(d); + point_map[LLVector3(d.getPosition().getF32())].push_back(d); } } } diff --git a/indra/llmath/llvolume.h b/indra/llmath/llvolume.h index 98db7f31c0..4aef3be973 100644 --- a/indra/llmath/llvolume.h +++ b/indra/llmath/llvolume.h @@ -873,10 +873,10 @@ public: struct ComparePosition { - bool operator()(const LLVector4a& a, const LLVector4a& b) const; + bool operator()(const LLVector3& a, const LLVector3& b) const; }; - typedef std::map, VertexMapData::ComparePosition > PointMap; + typedef std::map, VertexMapData::ComparePosition > PointMap; }; void optimize(F32 angle_cutoff = 2.f); diff --git a/indra/newview/CMakeLists.txt b/indra/newview/CMakeLists.txt index d2ae81180b..9de1d4aca5 100644 --- a/indra/newview/CMakeLists.txt +++ b/indra/newview/CMakeLists.txt @@ -16,6 +16,7 @@ include(FindOpenGL) include(LLAudio) include(LLCharacter) include(LLCommon) +include(LLConvexDecompInter) include(LLImage) include(LLImageJ2COJ) include(LLInventory) @@ -49,6 +50,7 @@ include_directories( ${LLAUDIO_INCLUDE_DIRS} ${LLCHARACTER_INCLUDE_DIRS} ${LLCOMMON_INCLUDE_DIRS} + ${LLCONVEXDECOMPINTER_INCLUDE_DIRS} ${LLIMAGE_INCLUDE_DIRS} ${LLINVENTORY_INCLUDE_DIRS} ${LLMATH_INCLUDE_DIRS} @@ -1683,6 +1685,7 @@ target_link_libraries(${VIEWER_BINARY_NAME} ${CRYPTO_LIBRARIES} ${LLLOGIN_LIBRARIES} ${GOOGLE_PERFTOOLS_LIBRARIES} + ${LLCONVEXDECOMPINTER_LIBRARIES} ) build_version(viewer) diff --git a/indra/newview/llassetuploadresponders.h b/indra/newview/llassetuploadresponders.h index 9abaccfde0..5890fbbc08 100644 --- a/indra/newview/llassetuploadresponders.h +++ b/indra/newview/llassetuploadresponders.h @@ -34,6 +34,7 @@ #define LL_LLASSETUPLOADRESPONDER_H #include "llhttpclient.h" +#include "llvolume.h" //for LL_MESH_ENABLED // Abstract class for supporting asset upload // via capabilities diff --git a/indra/newview/lldrawpoolavatar.h b/indra/newview/lldrawpoolavatar.h index 46ffc42f04..b01394534b 100644 --- a/indra/newview/lldrawpoolavatar.h +++ b/indra/newview/lldrawpoolavatar.h @@ -34,6 +34,7 @@ #define LL_LLDRAWPOOLAVATAR_H #include "lldrawpool.h" +#include "llvolume.h" // for LL_MESH_ENABLED class LLVOAvatar; class LLGLSLShader; diff --git a/indra/newview/llinventorybridge.h b/indra/newview/llinventorybridge.h index 8de7d63173..0e3b5bc3c9 100644 --- a/indra/newview/llinventorybridge.h +++ b/indra/newview/llinventorybridge.h @@ -40,6 +40,7 @@ #include "llinventoryobserver.h" #include "llviewercontrol.h" #include "llwearable.h" +#include "llvolume.h" //for LL_MESH_ENABLED class LLInventoryPanel; class LLInventoryModel; diff --git a/indra/newview/lltooldraganddrop.cpp b/indra/newview/lltooldraganddrop.cpp index d679ccd3d8..c4ecba1aa2 100644 --- a/indra/newview/lltooldraganddrop.cpp +++ b/indra/newview/lltooldraganddrop.cpp @@ -1037,7 +1037,7 @@ void LLToolDragAndDrop::dropTextureAllFaces(LLViewerObject* hit_obj, hit_obj->sendTEUpdate(); } -#if LL_MESH_EANBLED +#if LL_MESH_ENABLED void LLToolDragAndDrop::dropMesh(LLViewerObject* hit_obj, LLInventoryItem* item, LLToolDragAndDrop::ESource source, diff --git a/indra/newview/lltooldraganddrop.h b/indra/newview/lltooldraganddrop.h index ceeaa8c820..09da4c2955 100644 --- a/indra/newview/lltooldraganddrop.h +++ b/indra/newview/lltooldraganddrop.h @@ -43,6 +43,7 @@ #include "llpermissions.h" #include "llwindow.h" #include "llviewerinventory.h" +#include "llvolume.h" //for LL_MESH_ENABLED class LLToolDragAndDrop; class LLViewerRegion; diff --git a/indra/newview/llvovolume.cpp b/indra/newview/llvovolume.cpp index d66aa567a8..8f51edc1de 100644 --- a/indra/newview/llvovolume.cpp +++ b/indra/newview/llvovolume.cpp @@ -1594,7 +1594,7 @@ void LLVOVolume::updateFaceSize(S32 idx) LLFace* facep = mDrawable->getFace(idx); if (idx >= getVolume()->getNumVolumeFaces()) { - facep->setSize(0,0); + facep->setSize(0,0, true); } else { diff --git a/indra/newview/skins/default/xui/en/floater_about.xml b/indra/newview/skins/default/xui/en/floater_about.xml index 294a68255d..860cff6664 100644 --- a/indra/newview/skins/default/xui/en/floater_about.xml +++ b/indra/newview/skins/default/xui/en/floater_about.xml @@ -160,10 +160,12 @@ xmlrpc-epi Copyright (C) 2000 Epinions, Inc. zlib Copyright (C) 1995-2002 Jean-loup Gailly and Mark Adler. google-perftools Copyright (c) 2005, Google Inc. +Second Life Viewer uses Havok (TM) Physics. (c)Copyright 1999-2010 Havok.com Inc. (and its Licensors). All Rights Reserved. See www.havok.com for details. + All rights reserved. See licenses.txt for details. Voice chat Audio coding: Polycom(R) Siren14(TM) (ITU-T Rec. G.722.1 Annex C) - + -- cgit v1.3 From 93ea3d2850067c23ff07f0ffb362b73247840e9a Mon Sep 17 00:00:00 2001 From: Dave Parks Date: Thu, 15 Jul 2010 14:17:51 -0500 Subject: Merge cleanup. --- indra/llmath/llvolume.cpp | 22 +--------------------- indra/newview/llappviewer.cpp | 2 -- indra/newview/llfloatersnapshot.cpp | 4 ++-- indra/newview/llpanelobject.cpp | 2 ++ indra/newview/llvovolume.cpp | 5 ----- 5 files changed, 5 insertions(+), 30 deletions(-) (limited to 'indra/newview/llvovolume.cpp') diff --git a/indra/llmath/llvolume.cpp b/indra/llmath/llvolume.cpp index 25db58491e..0976487cbd 100644 --- a/indra/llmath/llvolume.cpp +++ b/indra/llmath/llvolume.cpp @@ -93,8 +93,6 @@ const F32 SKEW_MAX = 0.95f; const F32 SCULPT_MIN_AREA = 0.002f; const S32 SCULPT_MIN_AREA_DETAIL = 1; -#define GEN_TRI_STRIP 0 - BOOL check_same_clock_dir( const LLVector3& pt1, const LLVector3& pt2, const LLVector3& pt3, const LLVector3& norm) { LLVector3 test = (pt2-pt1)%(pt3-pt2); @@ -5663,8 +5661,6 @@ BOOL LLVolumeFace::createUnCutCubeCap(LLVolume* volume, BOOL partial_build) { *out++ = ((gy*(grid_size+1))+gx+idxs[i]); } -#if GEN_TRI_STRIP -#endif } else { @@ -5672,14 +5668,9 @@ BOOL LLVolumeFace::createUnCutCubeCap(LLVolume* volume, BOOL partial_build) { *out++ = ((gy*(grid_size+1))+gx+idxs[i]); } -#if GEN_TRI_STRIP -#endif } - } - -#if GEN_TRI_STRIP + } } -#endif } return TRUE; @@ -6539,13 +6530,9 @@ BOOL LLVolumeFace::createSide(LLVolume* volume, BOOL partial_build) if (!partial_build) { -#if GEN_TRI_STRIP -#endif // Now we generate the indices. for (t = 0; t < (mNumT-1); t++) { -#if GEN_TRI_STRIP -#endif for (s = 0; s < (mNumS-1); s++) { mIndices[cur_index++] = s + mNumS*t; //bottom left @@ -6555,8 +6542,6 @@ BOOL LLVolumeFace::createSide(LLVolume* volume, BOOL partial_build) mIndices[cur_index++] = s+1 + mNumS*t; //bottom right mIndices[cur_index++] = s+1 + mNumS*(t+1); //top right -#if GEN_TRI_STRIP -#endif mEdge[cur_edge++] = (mNumS-1)*2*t+s*2+1; //bottom left/top right neighbor face if (t < mNumT-2) { //top right/top left neighbor face mEdge[cur_edge++] = (mNumS-1)*2*(t+1)+s*2+1; @@ -6597,11 +6582,7 @@ BOOL LLVolumeFace::createSide(LLVolume* volume, BOOL partial_build) } mEdge[cur_edge++] = (mNumS-1)*2*t+s*2; //top right/bottom left neighbor face } -#if GEN_TRI_STRIP -#endif -#if GEN_TRI_STRIP } -#endif } //generate normals @@ -6684,7 +6665,6 @@ BOOL LLVolumeFace::createSide(LLVolume* volume, BOOL partial_build) } } } - else // logic for sculpt volumes { BOOL average_poles = FALSE; diff --git a/indra/newview/llappviewer.cpp b/indra/newview/llappviewer.cpp index 0ef403720b..8ffdd30d04 100644 --- a/indra/newview/llappviewer.cpp +++ b/indra/newview/llappviewer.cpp @@ -3515,8 +3515,6 @@ static LLFastTimer::DeclareTimer FTM_REGION_UPDATE("Update Region"); static LLFastTimer::DeclareTimer FTM_WORLD_UPDATE("Update World"); static LLFastTimer::DeclareTimer FTM_NETWORK("Network"); static LLFastTimer::DeclareTimer FTM_AGENT_NETWORK("Agent Network"); -static LLFastTimer::DeclareTimer FTM_AGENT_AUTOPILOT("Autopilot"); -static LLFastTimer::DeclareTimer FTM_AGENT_UPDATE("Update"); static LLFastTimer::DeclareTimer FTM_VLMANAGER("VL Manager"); /////////////////////////////////////////////////////// diff --git a/indra/newview/llfloatersnapshot.cpp b/indra/newview/llfloatersnapshot.cpp index 187a5e2040..43ea6143b1 100644 --- a/indra/newview/llfloatersnapshot.cpp +++ b/indra/newview/llfloatersnapshot.cpp @@ -999,7 +999,7 @@ void LLSnapshotLivePreview::saveTexture() void *userdata = NULL; upload_new_resource(tid, // tid LLAssetType::AT_TEXTURE, - snapname, + "Snapshot : " + pos_string, "Taken by " + who_took_it + " at " + pos_string, 0, LLFolderType::FT_SNAPSHOT_CATEGORY, @@ -1007,7 +1007,7 @@ void LLSnapshotLivePreview::saveTexture() PERM_ALL, // Note: Snapshots to inventory is a special case of content upload PERM_NONE, // that ignores the user's premissions preferences and continues to PERM_NONE, // always use these fairly permissive hard-coded initial perms. - MG - snapname, + "Snapshot : " + pos_string, callback, expected_upload_cost, userdata); gViewerWindow->playSnapshotAnimAndSound(); } diff --git a/indra/newview/llpanelobject.cpp b/indra/newview/llpanelobject.cpp index 63490d4fa2..1de0496737 100644 --- a/indra/newview/llpanelobject.cpp +++ b/indra/newview/llpanelobject.cpp @@ -1289,6 +1289,8 @@ void LLPanelObject::sendPhysicsParam() body.append(LLSelectMgr::getInstance()->getSelection()->getFirstObject()->getID()); LLHTTPClient::post( url, body, new CostResponder(body[0].asInteger()) ); +} + void LLPanelObject::sendCastShadows() { BOOL value = mCheckCastShadows->get(); diff --git a/indra/newview/llvovolume.cpp b/indra/newview/llvovolume.cpp index 4faef57074..36391a3602 100644 --- a/indra/newview/llvovolume.cpp +++ b/indra/newview/llvovolume.cpp @@ -96,13 +96,8 @@ LLPointer LLVOVolume::sObjectMediaNavigateClient = static LLFastTimer::DeclareTimer FTM_GEN_TRIANGLES("Generate Triangles"); static LLFastTimer::DeclareTimer FTM_GEN_VOLUME("Generate Volumes"); -static LLFastTimer::DeclareTimer FTM_BUILD_MESH("Mesh"); -static LLFastTimer::DeclareTimer FTM_MESH_VFS("VFS"); -static LLFastTimer::DeclareTimer FTM_MESH_STREAM("Stream"); -static LLFastTimer::DeclareTimer FTM_MESH_FACES("Faces"); static LLFastTimer::DeclareTimer FTM_VOLUME_TEXTURES("Volume Textures"); - // Implementation class of LLMediaDataClientObject. See llmediadataclient.h class LLMediaDataClientObjectImpl : public LLMediaDataClientObject { -- cgit v1.3 From a919bda24dfc6fb7fc031546e69ed7f531d828e8 Mon Sep 17 00:00:00 2001 From: Dave Parks Date: Sat, 17 Jul 2010 11:23:26 -0500 Subject: Fix for rigged attachments disappearing when toggling lighting and shadows. --- indra/newview/llvovolume.cpp | 111 ++++++++++++++++++++++--------------------- 1 file changed, 56 insertions(+), 55 deletions(-) (limited to 'indra/newview/llvovolume.cpp') diff --git a/indra/newview/llvovolume.cpp b/indra/newview/llvovolume.cpp index 36391a3602..b69fd3f90b 100644 --- a/indra/newview/llvovolume.cpp +++ b/indra/newview/llvovolume.cpp @@ -3544,77 +3544,78 @@ void LLVolumeGeometryManager::rebuildGeom(LLSpatialGroup* group) { facep->mVertexBuffer = NULL; facep->mLastVertexBuffer = NULL; - facep->setState(LLFace::RIGGED); - - //get drawpool of avatar with rigged face - LLDrawPoolAvatar* pool = get_avatar_drawpool(vobj); - - if (pool) + } + + facep->setState(LLFace::RIGGED); + + //get drawpool of avatar with rigged face + LLDrawPoolAvatar* pool = get_avatar_drawpool(vobj); + + if (pool) + { + const LLTextureEntry* te = facep->getTextureEntry(); + + //remove face from old pool if it exists + LLDrawPool* old_pool = facep->getPool(); + if (old_pool && old_pool->getType() == LLDrawPool::POOL_AVATAR) { - const LLTextureEntry* te = facep->getTextureEntry(); + ((LLDrawPoolAvatar*) old_pool)->removeRiggedFace(facep); + } + + //add face to new pool + LLViewerTexture* tex = facep->getTexture(); + U32 type = gPipeline.getPoolTypeFromTE(te, tex); - //remove face from old pool if it exists - LLDrawPool* old_pool = facep->getPool(); - if (old_pool && old_pool->getType() == LLDrawPool::POOL_AVATAR) + if (type == LLDrawPool::POOL_ALPHA) + { + if (te->getFullbright()) { - ((LLDrawPoolAvatar*) old_pool)->removeRiggedFace(facep); + pool->addRiggedFace(facep, LLDrawPoolAvatar::RIGGED_FULLBRIGHT_ALPHA); } - - //add face to new pool - LLViewerTexture* tex = facep->getTexture(); - U32 type = gPipeline.getPoolTypeFromTE(te, tex); - - if (type == LLDrawPool::POOL_ALPHA) + else { - if (te->getFullbright()) - { - pool->addRiggedFace(facep, LLDrawPoolAvatar::RIGGED_FULLBRIGHT_ALPHA); - } - else - { - pool->addRiggedFace(facep, LLDrawPoolAvatar::RIGGED_ALPHA); - } + pool->addRiggedFace(facep, LLDrawPoolAvatar::RIGGED_ALPHA); } - else if (te->getShiny()) + } + else if (te->getShiny()) + { + if (te->getFullbright()) { - if (te->getFullbright()) - { - pool->addRiggedFace(facep, LLDrawPoolAvatar::RIGGED_FULLBRIGHT_SHINY); - } - else - { - pool->addRiggedFace(facep, LLDrawPoolAvatar::RIGGED_SHINY); - } + pool->addRiggedFace(facep, LLDrawPoolAvatar::RIGGED_FULLBRIGHT_SHINY); } else { - if (te->getFullbright()) - { - pool->addRiggedFace(facep, LLDrawPoolAvatar::RIGGED_FULLBRIGHT); - } - else - { - pool->addRiggedFace(facep, LLDrawPoolAvatar::RIGGED_SIMPLE); - } + pool->addRiggedFace(facep, LLDrawPoolAvatar::RIGGED_SHINY); } - - if (te->getGlow()) + } + else + { + if (te->getFullbright()) { - pool->addRiggedFace(facep, LLDrawPoolAvatar::RIGGED_GLOW); + pool->addRiggedFace(facep, LLDrawPoolAvatar::RIGGED_FULLBRIGHT); } + else + { + pool->addRiggedFace(facep, LLDrawPoolAvatar::RIGGED_SIMPLE); + } + } - if (LLPipeline::sRenderDeferred) + if (te->getGlow()) + { + pool->addRiggedFace(facep, LLDrawPoolAvatar::RIGGED_GLOW); + } + + if (LLPipeline::sRenderDeferred) + { + if (type != LLDrawPool::POOL_ALPHA && !te->getFullbright()) { - if (type != LLDrawPool::POOL_ALPHA && !te->getFullbright()) + if (te->getBumpmap()) + { + pool->addRiggedFace(facep, LLDrawPoolAvatar::RIGGED_DEFERRED_BUMP); + } + else { - if (te->getBumpmap()) - { - pool->addRiggedFace(facep, LLDrawPoolAvatar::RIGGED_DEFERRED_BUMP); - } - else - { - pool->addRiggedFace(facep, LLDrawPoolAvatar::RIGGED_DEFERRED_SIMPLE); - } + pool->addRiggedFace(facep, LLDrawPoolAvatar::RIGGED_DEFERRED_SIMPLE); } } } -- cgit v1.3 From 340a729d55983110a85cfc55f1ce0a952bfe15b4 Mon Sep 17 00:00:00 2001 From: Dave Parks Date: Thu, 12 Aug 2010 17:05:40 -0500 Subject: Vertex buffer cleanup, adding water shader to simple rigged attachments. Reviewed by Nyx. --- indra/llrender/llvertexbuffer.cpp | 232 +++++------------------------------- indra/llrender/llvertexbuffer.h | 20 ++-- indra/newview/lldrawpoolavatar.cpp | 32 +++-- indra/newview/llviewershadermgr.cpp | 25 ++++ indra/newview/llviewershadermgr.h | 5 + indra/newview/llvosurfacepatch.cpp | 22 ++-- indra/newview/llvovolume.cpp | 6 +- indra/newview/llvowlsky.cpp | 2 +- 8 files changed, 105 insertions(+), 239 deletions(-) (limited to 'indra/newview/llvovolume.cpp') diff --git a/indra/llrender/llvertexbuffer.cpp b/indra/llrender/llvertexbuffer.cpp index 9783435e2f..71620feed3 100644 --- a/indra/llrender/llvertexbuffer.cpp +++ b/indra/llrender/llvertexbuffer.cpp @@ -68,11 +68,8 @@ S32 LLVertexBuffer::sWeight4Loc = -1; std::vector LLVertexBuffer::sDeleteList; -#define LL_ALIGNED_VB 1 -#if LL_ALIGNED_VB - -S32 LLVertexBuffer::sTypeOffsets[LLVertexBuffer::TYPE_MAX] = +S32 LLVertexBuffer::sTypeSize[LLVertexBuffer::TYPE_MAX] = { sizeof(LLVector4), // TYPE_VERTEX, sizeof(LLVector4), // TYPE_NORMAL, @@ -87,25 +84,6 @@ S32 LLVertexBuffer::sTypeOffsets[LLVertexBuffer::TYPE_MAX] = sizeof(LLVector4), // TYPE_CLOTHWEIGHT, }; -#else - -S32 LLVertexBuffer::sTypeOffsets[LLVertexBuffer::TYPE_MAX] = -{ - sizeof(LLVector3), // TYPE_VERTEX, - sizeof(LLVector3), // TYPE_NORMAL, - sizeof(LLVector2), // TYPE_TEXCOORD0, - sizeof(LLVector2), // TYPE_TEXCOORD1, - sizeof(LLVector2), // TYPE_TEXCOORD2, - sizeof(LLVector2), // TYPE_TEXCOORD3, - sizeof(LLColor4U), // TYPE_COLOR, - sizeof(LLVector3), // TYPE_BINORMAL, - sizeof(F32), // TYPE_WEIGHT, - sizeof(LLVector4), // TYPE_WEIGHT4, - sizeof(LLVector4), // TYPE_CLOTHWEIGHT, -}; - -#endif - U32 LLVertexBuffer::sGLMode[LLRender::NUM_MODES] = { GL_TRIANGLES, @@ -460,19 +438,22 @@ LLVertexBuffer::LLVertexBuffer(U32 typemask, S32 usage) : mUsage = 0; } - S32 stride = calcStride(typemask, mOffsets); + //zero out offsets + for (U32 i = 0; i < TYPE_MAX; i++) + { + mOffsets[i] = 0; + } mTypeMask = typemask; - mStride = stride; + mSize = 0; mAlignedOffset = 0; mAlignedIndexOffset = 0; sCount++; } -#if LL_ALIGNED_VB //static -S32 LLVertexBuffer::calcStride(const U32& typemask, S32* offsets, S32 num_vertices) +S32 LLVertexBuffer::calcOffsets(const U32& typemask, S32* offsets, S32 num_vertices) { S32 offset = 0; for (S32 i=0; i struct VertexBufferStrider { typedef LLStrider strider_t; @@ -1051,7 +1015,7 @@ template struct VertexBufferStrider } else if (vbo.hasDataType(type)) { - S32 stride = LLVertexBuffer::sTypeOffsets[type]; + S32 stride = LLVertexBuffer::sTypeSize[type]; strider = (T*)(vbo.getMappedData() + vbo.getOffset(type)+index*stride); strider.setStride(stride); return TRUE; @@ -1064,45 +1028,6 @@ template struct VertexBufferStrider } }; -#else - -template struct VertexBufferStrider -{ - typedef LLStrider strider_t; - static bool get(LLVertexBuffer& vbo, - strider_t& strider, - S32 index) - { - if (vbo.mapBuffer() == NULL) - { - llwarns << "mapBuffer failed!" << llendl; - return FALSE; - } - - if (type == LLVertexBuffer::TYPE_INDEX) - { - S32 stride = sizeof(T); - strider = (T*)(vbo.getMappedIndices() + index*stride); - strider.setStride(0); - return TRUE; - } - else if (vbo.hasDataType(type)) - { - S32 stride = vbo.getStride(); - strider = (T*)(vbo.getMappedData() + vbo.getOffset(type) + index*stride); - strider.setStride(stride); - return TRUE; - } - else - { - llerrs << "VertexBufferStrider could not find valid vertex data." << llendl; - } - return FALSE; - } -}; - -#endif - bool LLVertexBuffer::getVertexStrider(LLStrider& strider, S32 index) { return VertexBufferStrider::get(*this, strider, index); @@ -1154,25 +1079,6 @@ bool LLVertexBuffer::getClothWeightStrider(LLStrider& strider, S32 in return VertexBufferStrider::get(*this, strider, index); } -void LLVertexBuffer::setStride(S32 type, S32 new_stride) -{ - LLMemType mt2(LLMemType::MTYPE_VERTEX_SET_STRIDE); - if (mNumVerts) - { - llerrs << "LLVertexBuffer::setOffset called with mNumVerts = " << mNumVerts << llendl; - } - // This code assumes that setStride() will only be called once per VBO per type. - S32 delta = new_stride - sTypeOffsets[type]; - for (S32 i=type+1; i buffer_list_t; static BOOL sEnableVBOs; - static S32 sTypeOffsets[TYPE_MAX]; + static S32 sTypeSize[TYPE_MAX]; static U32 sGLMode[LLRender::NUM_MODES]; static U32 sGLRenderBuffer; static U32 sGLRenderIndices; diff --git a/indra/newview/lldrawpoolavatar.cpp b/indra/newview/lldrawpoolavatar.cpp index ef1d2ec1c5..1788a3da3d 100644 --- a/indra/newview/lldrawpoolavatar.cpp +++ b/indra/newview/lldrawpoolavatar.cpp @@ -814,17 +814,25 @@ void LLDrawPoolAvatar::endSkinned() #if LL_MESH_ENABLED void LLDrawPoolAvatar::beginRiggedSimple() { - sVertexProgram = &gSkinnedObjectSimpleProgram; + if (LLPipeline::sUnderWaterRender) + { + sVertexProgram = &gSkinnedObjectSimpleWaterProgram; + } + else + { + sVertexProgram = &gSkinnedObjectSimpleProgram; + } + sDiffuseChannel = 0; - gSkinnedObjectSimpleProgram.bind(); - LLVertexBuffer::sWeight4Loc = gSkinnedObjectSimpleProgram.getAttribLocation(LLViewerShaderMgr::OBJECT_WEIGHT); + sVertexProgram->bind(); + LLVertexBuffer::sWeight4Loc = sVertexProgram->getAttribLocation(LLViewerShaderMgr::OBJECT_WEIGHT); } void LLDrawPoolAvatar::endRiggedSimple() { - sVertexProgram = NULL; LLVertexBuffer::unbind(); - gSkinnedObjectSimpleProgram.unbind(); + sVertexProgram->unbind(); + sVertexProgram = NULL; LLVertexBuffer::sWeight4Loc = -1; } @@ -1596,25 +1604,25 @@ void LLVertexBufferAvatar::setupVertexBuffer(U32 data_mask) const { if (sRenderingSkinned) { - U8* base = useVBOs() ? NULL : mMappedData; + U8* base = useVBOs() ? (U8*) mAlignedOffset : mMappedData; - glVertexPointer(3,GL_FLOAT, LLVertexBuffer::sTypeOffsets[LLVertexBuffer::TYPE_VERTEX], (void*)(base + 0)); - glNormalPointer(GL_FLOAT, LLVertexBuffer::sTypeOffsets[LLVertexBuffer::TYPE_NORMAL], (void*)(base + mOffsets[TYPE_NORMAL])); - glTexCoordPointer(2,GL_FLOAT, LLVertexBuffer::sTypeOffsets[LLVertexBuffer::TYPE_TEXCOORD0], (void*)(base + mOffsets[TYPE_TEXCOORD0])); + glVertexPointer(3,GL_FLOAT, LLVertexBuffer::sTypeSize[LLVertexBuffer::TYPE_VERTEX], (void*)(base + 0)); + glNormalPointer(GL_FLOAT, LLVertexBuffer::sTypeSize[LLVertexBuffer::TYPE_NORMAL], (void*)(base + mOffsets[TYPE_NORMAL])); + glTexCoordPointer(2,GL_FLOAT, LLVertexBuffer::sTypeSize[LLVertexBuffer::TYPE_TEXCOORD0], (void*)(base + mOffsets[TYPE_TEXCOORD0])); set_vertex_weights(LLDrawPoolAvatar::sVertexProgram->mAttribute[LLViewerShaderMgr::AVATAR_WEIGHT], - LLVertexBuffer::sTypeOffsets[LLVertexBuffer::TYPE_WEIGHT], (F32*)(base + mOffsets[TYPE_WEIGHT])); + LLVertexBuffer::sTypeSize[LLVertexBuffer::TYPE_WEIGHT], (F32*)(base + mOffsets[TYPE_WEIGHT])); if (sShaderLevel >= LLDrawPoolAvatar::SHADER_LEVEL_BUMP) { set_binormals(LLDrawPoolAvatar::sVertexProgram->mAttribute[LLViewerShaderMgr::BINORMAL], - LLVertexBuffer::sTypeOffsets[LLVertexBuffer::TYPE_BINORMAL], (LLVector3*)(base + mOffsets[TYPE_BINORMAL])); + LLVertexBuffer::sTypeSize[LLVertexBuffer::TYPE_BINORMAL], (LLVector3*)(base + mOffsets[TYPE_BINORMAL])); } if (sShaderLevel >= LLDrawPoolAvatar::SHADER_LEVEL_CLOTH) { set_vertex_clothing_weights(LLDrawPoolAvatar::sVertexProgram->mAttribute[LLViewerShaderMgr::AVATAR_CLOTHING], - LLVertexBuffer::sTypeOffsets[LLVertexBuffer::TYPE_CLOTHWEIGHT], (LLVector4*)(base + mOffsets[TYPE_CLOTHWEIGHT])); + LLVertexBuffer::sTypeSize[LLVertexBuffer::TYPE_CLOTHWEIGHT], (LLVector4*)(base + mOffsets[TYPE_CLOTHWEIGHT])); } } else diff --git a/indra/newview/llviewershadermgr.cpp b/indra/newview/llviewershadermgr.cpp index 49eeffe228..e887766492 100644 --- a/indra/newview/llviewershadermgr.cpp +++ b/indra/newview/llviewershadermgr.cpp @@ -83,6 +83,11 @@ LLGLSLShader gSkinnedObjectFullbrightProgram; LLGLSLShader gSkinnedObjectFullbrightShinyProgram; LLGLSLShader gSkinnedObjectShinySimpleProgram; +LLGLSLShader gSkinnedObjectSimpleWaterProgram; +//LLGLSLShader gSkinnedObjectFullbrightProgram; +//LLGLSLShader gSkinnedObjectFullbrightShinyProgram; +//LLGLSLShader gSkinnedObjectShinySimpleProgram; + //environment shaders LLGLSLShader gTerrainProgram; LLGLSLShader gTerrainWaterProgram; @@ -159,6 +164,7 @@ LLViewerShaderMgr::LLViewerShaderMgr() : mShaderList.push_back(&gObjectFullbrightProgram); mShaderList.push_back(&gObjectFullbrightShinyProgram); mShaderList.push_back(&gSkinnedObjectSimpleProgram); + mShaderList.push_back(&gSkinnedObjectSimpleWaterProgram); mShaderList.push_back(&gSkinnedObjectFullbrightProgram); mShaderList.push_back(&gSkinnedObjectFullbrightShinyProgram); mShaderList.push_back(&gSkinnedObjectShinySimpleProgram); @@ -570,6 +576,7 @@ void LLViewerShaderMgr::unloadShaders() gObjectShinyWaterProgram.unload(); gSkinnedObjectSimpleProgram.unload(); + gSkinnedObjectSimpleWaterProgram.unload(); gSkinnedObjectFullbrightProgram.unload(); gSkinnedObjectFullbrightShinyProgram.unload(); gSkinnedObjectShinySimpleProgram.unload(); @@ -1299,6 +1306,7 @@ BOOL LLViewerShaderMgr::loadShadersObject() gObjectFullbrightProgram.unload(); gObjectFullbrightWaterProgram.unload(); gSkinnedObjectSimpleProgram.unload(); + gSkinnedObjectSimpleWaterProgram.unload(); gSkinnedObjectFullbrightProgram.unload(); gSkinnedObjectFullbrightShinyProgram.unload(); gSkinnedObjectShinySimpleProgram.unload(); @@ -1428,6 +1436,23 @@ BOOL LLViewerShaderMgr::loadShadersObject() success = gSkinnedObjectSimpleProgram.createShader(NULL, NULL); } + if (success) + { + gSkinnedObjectSimpleWaterProgram.mName = "Skinned Simple Water Shader"; + gSkinnedObjectSimpleWaterProgram.mFeatures.calculatesLighting = true; + gSkinnedObjectSimpleWaterProgram.mFeatures.calculatesAtmospherics = true; + gSkinnedObjectSimpleWaterProgram.mFeatures.hasGamma = true; + gSkinnedObjectSimpleWaterProgram.mFeatures.hasAtmospherics = true; + gSkinnedObjectSimpleWaterProgram.mFeatures.hasLighting = true; + gSkinnedObjectSimpleWaterProgram.mFeatures.hasWaterFog = true; + gSkinnedObjectSimpleWaterProgram.mFeatures.hasObjectSkinning = true; + gSkinnedObjectSimpleWaterProgram.mShaderFiles.clear(); + gSkinnedObjectSimpleWaterProgram.mShaderFiles.push_back(make_pair("objects/simpleSkinnedV.glsl", GL_VERTEX_SHADER_ARB)); + gSkinnedObjectSimpleWaterProgram.mShaderFiles.push_back(make_pair("objects/simpleWaterF.glsl", GL_FRAGMENT_SHADER_ARB)); + gSkinnedObjectSimpleWaterProgram.mShaderLevel = mVertexShaderLevel[SHADER_OBJECT]; + success = gSkinnedObjectSimpleWaterProgram.createShader(NULL, NULL); + } + if (success) { gSkinnedObjectFullbrightProgram.mName = "Skinned Fullbright Shader"; diff --git a/indra/newview/llviewershadermgr.h b/indra/newview/llviewershadermgr.h index da16a38427..99787f3c0a 100644 --- a/indra/newview/llviewershadermgr.h +++ b/indra/newview/llviewershadermgr.h @@ -319,6 +319,11 @@ extern LLGLSLShader gSkinnedObjectFullbrightProgram; extern LLGLSLShader gSkinnedObjectFullbrightShinyProgram; extern LLGLSLShader gSkinnedObjectShinySimpleProgram; +extern LLGLSLShader gSkinnedObjectSimpleWaterProgram; +//extern LLGLSLShader gSkinnedObjectFullbrightProgram; +//extern LLGLSLShader gSkinnedObjectFullbrightShinyProgram; +//extern LLGLSLShader gSkinnedObjectShinySimpleProgram; + //environment shaders extern LLGLSLShader gTerrainProgram; extern LLGLSLShader gTerrainWaterProgram; diff --git a/indra/newview/llvosurfacepatch.cpp b/indra/newview/llvosurfacepatch.cpp index 02e7e7e60f..63f662c5a1 100644 --- a/indra/newview/llvosurfacepatch.cpp +++ b/indra/newview/llvosurfacepatch.cpp @@ -77,58 +77,58 @@ public: if (data_mask & MAP_NORMAL) { - glNormalPointer(GL_FLOAT, LLVertexBuffer::sTypeOffsets[TYPE_NORMAL], (void*)(base + mOffsets[TYPE_NORMAL])); + glNormalPointer(GL_FLOAT, LLVertexBuffer::sTypeSize[TYPE_NORMAL], (void*)(base + mOffsets[TYPE_NORMAL])); } if (data_mask & MAP_TEXCOORD3) { //substitute tex coord 0 for tex coord 3 glClientActiveTextureARB(GL_TEXTURE3_ARB); - glTexCoordPointer(2,GL_FLOAT, LLVertexBuffer::sTypeOffsets[TYPE_TEXCOORD0], (void*)(base + mOffsets[TYPE_TEXCOORD0])); + glTexCoordPointer(2,GL_FLOAT, LLVertexBuffer::sTypeSize[TYPE_TEXCOORD0], (void*)(base + mOffsets[TYPE_TEXCOORD0])); glClientActiveTextureARB(GL_TEXTURE0_ARB); } if (data_mask & MAP_TEXCOORD2) { //substitute tex coord 0 for tex coord 2 glClientActiveTextureARB(GL_TEXTURE2_ARB); - glTexCoordPointer(2,GL_FLOAT, LLVertexBuffer::sTypeOffsets[TYPE_TEXCOORD0], (void*)(base + mOffsets[TYPE_TEXCOORD0])); + glTexCoordPointer(2,GL_FLOAT, LLVertexBuffer::sTypeSize[TYPE_TEXCOORD0], (void*)(base + mOffsets[TYPE_TEXCOORD0])); glClientActiveTextureARB(GL_TEXTURE0_ARB); } if (data_mask & MAP_TEXCOORD1) { glClientActiveTextureARB(GL_TEXTURE1_ARB); - glTexCoordPointer(2,GL_FLOAT, LLVertexBuffer::sTypeOffsets[TYPE_TEXCOORD1], (void*)(base + mOffsets[TYPE_TEXCOORD1])); + glTexCoordPointer(2,GL_FLOAT, LLVertexBuffer::sTypeSize[TYPE_TEXCOORD1], (void*)(base + mOffsets[TYPE_TEXCOORD1])); glClientActiveTextureARB(GL_TEXTURE0_ARB); } if (data_mask & MAP_BINORMAL) { glClientActiveTextureARB(GL_TEXTURE2_ARB); - glTexCoordPointer(3,GL_FLOAT, LLVertexBuffer::sTypeOffsets[TYPE_BINORMAL], (void*)(base + mOffsets[TYPE_BINORMAL])); + glTexCoordPointer(3,GL_FLOAT, LLVertexBuffer::sTypeSize[TYPE_BINORMAL], (void*)(base + mOffsets[TYPE_BINORMAL])); glClientActiveTextureARB(GL_TEXTURE0_ARB); } if (data_mask & MAP_TEXCOORD0) { - glTexCoordPointer(2,GL_FLOAT, LLVertexBuffer::sTypeOffsets[TYPE_TEXCOORD0], (void*)(base + mOffsets[TYPE_TEXCOORD0])); + glTexCoordPointer(2,GL_FLOAT, LLVertexBuffer::sTypeSize[TYPE_TEXCOORD0], (void*)(base + mOffsets[TYPE_TEXCOORD0])); } if (data_mask & MAP_COLOR) { - glColorPointer(4, GL_UNSIGNED_BYTE, LLVertexBuffer::sTypeOffsets[TYPE_COLOR], (void*)(base + mOffsets[TYPE_COLOR])); + glColorPointer(4, GL_UNSIGNED_BYTE, LLVertexBuffer::sTypeSize[TYPE_COLOR], (void*)(base + mOffsets[TYPE_COLOR])); } if (data_mask & MAP_WEIGHT) { - glVertexAttribPointerARB(1, 1, GL_FLOAT, FALSE, LLVertexBuffer::sTypeOffsets[TYPE_WEIGHT], (void*)(base + mOffsets[TYPE_WEIGHT])); + glVertexAttribPointerARB(1, 1, GL_FLOAT, FALSE, LLVertexBuffer::sTypeSize[TYPE_WEIGHT], (void*)(base + mOffsets[TYPE_WEIGHT])); } if (data_mask & MAP_WEIGHT4 && sWeight4Loc != -1) { - glVertexAttribPointerARB(sWeight4Loc, 4, GL_FLOAT, FALSE, LLVertexBuffer::sTypeOffsets[TYPE_WEIGHT4], (void*)(base+mOffsets[TYPE_WEIGHT4])); + glVertexAttribPointerARB(sWeight4Loc, 4, GL_FLOAT, FALSE, LLVertexBuffer::sTypeSize[TYPE_WEIGHT4], (void*)(base+mOffsets[TYPE_WEIGHT4])); } if (data_mask & MAP_CLOTHWEIGHT) { - glVertexAttribPointerARB(4, 4, GL_FLOAT, TRUE, LLVertexBuffer::sTypeOffsets[TYPE_CLOTHWEIGHT], (void*)(base + mOffsets[TYPE_CLOTHWEIGHT])); + glVertexAttribPointerARB(4, 4, GL_FLOAT, TRUE, LLVertexBuffer::sTypeSize[TYPE_CLOTHWEIGHT], (void*)(base + mOffsets[TYPE_CLOTHWEIGHT])); } if (data_mask & MAP_VERTEX) { - glVertexPointer(3,GL_FLOAT, LLVertexBuffer::sTypeOffsets[TYPE_VERTEX], (void*)(base + 0)); + glVertexPointer(3,GL_FLOAT, LLVertexBuffer::sTypeSize[TYPE_VERTEX], (void*)(base + 0)); } } }; diff --git a/indra/newview/llvovolume.cpp b/indra/newview/llvovolume.cpp index b69fd3f90b..6fa97b6688 100644 --- a/indra/newview/llvovolume.cpp +++ b/indra/newview/llvovolume.cpp @@ -3494,8 +3494,8 @@ void LLVolumeGeometryManager::rebuildGeom(LLSpatialGroup* group) std::vector alpha_faces; U32 useage = group->mSpatialPartition->mBufferUsage; - U32 max_vertices = (gSavedSettings.getS32("RenderMaxVBOSize")*1024)/LLVertexBuffer::calcStride(group->mSpatialPartition->mVertexDataMask); - U32 max_total = (gSavedSettings.getS32("RenderMaxNodeSize")*1024)/LLVertexBuffer::calcStride(group->mSpatialPartition->mVertexDataMask); + U32 max_vertices = (gSavedSettings.getS32("RenderMaxVBOSize")*1024)/LLVertexBuffer::calcVertexSize(group->mSpatialPartition->mVertexDataMask); + U32 max_total = (gSavedSettings.getS32("RenderMaxNodeSize")*1024)/LLVertexBuffer::calcVertexSize(group->mSpatialPartition->mVertexDataMask); max_vertices = llmin(max_vertices, (U32) 65535); U32 cur_total = 0; @@ -3894,7 +3894,7 @@ void LLVolumeGeometryManager::rebuildMesh(LLSpatialGroup* group) void LLVolumeGeometryManager::genDrawInfo(LLSpatialGroup* group, U32 mask, std::vector& faces, BOOL distance_sort) { //calculate maximum number of vertices to store in a single buffer - U32 max_vertices = (gSavedSettings.getS32("RenderMaxVBOSize")*1024)/LLVertexBuffer::calcStride(group->mSpatialPartition->mVertexDataMask); + U32 max_vertices = (gSavedSettings.getS32("RenderMaxVBOSize")*1024)/LLVertexBuffer::calcVertexSize(group->mSpatialPartition->mVertexDataMask); max_vertices = llmin(max_vertices, (U32) 65535); if (!distance_sort) diff --git a/indra/newview/llvowlsky.cpp b/indra/newview/llvowlsky.cpp index 1749ccef94..9398bf425e 100644 --- a/indra/newview/llvowlsky.cpp +++ b/indra/newview/llvowlsky.cpp @@ -338,7 +338,7 @@ BOOL LLVOWLSky::updateGeometry(LLDrawable * drawable) { const U32 max_buffer_bytes = gSavedSettings.getS32("RenderMaxVBOSize")*1024; const U32 data_mask = LLDrawPoolWLSky::SKY_VERTEX_DATA_MASK; - const U32 max_verts = max_buffer_bytes / LLVertexBuffer::calcStride(data_mask); + const U32 max_verts = max_buffer_bytes / LLVertexBuffer::calcVertexSize(data_mask); const U32 total_stacks = getNumStacks(); -- cgit v1.3 From 9d5a3d7a2d830c07dd6b79534fdbe7c782b30afe Mon Sep 17 00:00:00 2001 From: Dave Parks Date: Fri, 13 Aug 2010 19:35:03 -0500 Subject: Better support for curious mixtures fullbright/shiny/water with rigged attachments, deferred rendering, and not. --- indra/llrender/llshadermgr.cpp | 9 ++++++++- indra/newview/lldrawpoolbump.cpp | 2 +- indra/newview/lldrawpoolsimple.cpp | 6 ------ indra/newview/llviewershadermgr.cpp | 26 ++++++++++++++++++++++++-- indra/newview/llviewershadermgr.h | 1 + indra/newview/llvovolume.cpp | 11 +++++++++-- 6 files changed, 43 insertions(+), 12 deletions(-) (limited to 'indra/newview/llvovolume.cpp') diff --git a/indra/llrender/llshadermgr.cpp b/indra/llrender/llshadermgr.cpp index 0f54ebeb69..9eb5668d88 100644 --- a/indra/llrender/llshadermgr.cpp +++ b/indra/llrender/llshadermgr.cpp @@ -234,7 +234,14 @@ BOOL LLShaderMgr::attachShaderFeatures(LLGLSLShader * shader) else if (features->isFullbright) { - if (features->hasWaterFog) + if (features->isShiny && features->hasWaterFog) + { + if (!shader->attachObject("lighting/lightFullbrightShinyWaterF.glsl")) + { + return FALSE; + } + } + else if (features->hasWaterFog) { if (!shader->attachObject("lighting/lightFullbrightWaterF.glsl")) { diff --git a/indra/newview/lldrawpoolbump.cpp b/indra/newview/lldrawpoolbump.cpp index 20097e29fe..0b0297d2b8 100644 --- a/indra/newview/lldrawpoolbump.cpp +++ b/indra/newview/lldrawpoolbump.cpp @@ -501,7 +501,7 @@ void LLDrawPoolBump::beginFullbrightShiny() if (LLPipeline::sUnderWaterRender) { - shader = &gObjectShinyWaterProgram; + shader = &gObjectFullbrightShinyWaterProgram; } else { diff --git a/indra/newview/lldrawpoolsimple.cpp b/indra/newview/lldrawpoolsimple.cpp index 91191287cd..30d1ffc783 100644 --- a/indra/newview/lldrawpoolsimple.cpp +++ b/indra/newview/lldrawpoolsimple.cpp @@ -149,12 +149,6 @@ void LLDrawPoolSimple::render(S32 pass) LLFastTimer t(FTM_RENDER_SIMPLE); gPipeline.enableLightsDynamic(); renderTexture(LLRenderPass::PASS_SIMPLE, getVertexDataMask()); - - if (LLPipeline::sRenderDeferred) - { //if deferred rendering is enabled, bump faces aren't reigstered as simple - //render bump faces here as simple so bump faces will appear under water - renderTexture(LLRenderPass::PASS_BUMP, getVertexDataMask()); - } } } diff --git a/indra/newview/llviewershadermgr.cpp b/indra/newview/llviewershadermgr.cpp index 20e4333209..f0f83ec6e4 100644 --- a/indra/newview/llviewershadermgr.cpp +++ b/indra/newview/llviewershadermgr.cpp @@ -74,6 +74,7 @@ LLGLSLShader gObjectFullbrightProgram; LLGLSLShader gObjectFullbrightWaterProgram; LLGLSLShader gObjectFullbrightShinyProgram; +LLGLSLShader gObjectFullbrightShinyWaterProgram; LLGLSLShader gObjectShinyProgram; LLGLSLShader gObjectShinyWaterProgram; @@ -163,6 +164,7 @@ LLViewerShaderMgr::LLViewerShaderMgr() : mShaderList.push_back(&gObjectSimpleProgram); mShaderList.push_back(&gObjectFullbrightProgram); mShaderList.push_back(&gObjectFullbrightShinyProgram); + mShaderList.push_back(&gObjectFullbrightShinyWaterProgram); mShaderList.push_back(&gSkinnedObjectSimpleProgram); mShaderList.push_back(&gSkinnedObjectFullbrightProgram); mShaderList.push_back(&gSkinnedObjectFullbrightShinyProgram); @@ -576,6 +578,7 @@ void LLViewerShaderMgr::unloadShaders() gObjectShinyProgram.unload(); gObjectFullbrightShinyProgram.unload(); + gObjectFullbrightShinyWaterProgram.unload(); gObjectShinyWaterProgram.unload(); gSkinnedObjectSimpleProgram.unload(); @@ -685,7 +688,7 @@ BOOL LLViewerShaderMgr::loadBasicShaders() // (in order of shader function call depth for reference purposes, deepest level first) shaders.clear(); - shaders.reserve(12); + shaders.reserve(13); shaders.push_back( make_pair( "windlight/atmosphericsVarsF.glsl", mVertexShaderLevel[SHADER_WINDLIGHT] ) ); shaders.push_back( make_pair( "windlight/gammaF.glsl", mVertexShaderLevel[SHADER_WINDLIGHT]) ); shaders.push_back( make_pair( "windlight/atmosphericsF.glsl", mVertexShaderLevel[SHADER_WINDLIGHT] ) ); @@ -698,6 +701,7 @@ BOOL LLViewerShaderMgr::loadBasicShaders() shaders.push_back( make_pair( "lighting/lightShinyF.glsl", mVertexShaderLevel[SHADER_LIGHTING] ) ); shaders.push_back( make_pair( "lighting/lightFullbrightShinyF.glsl", mVertexShaderLevel[SHADER_LIGHTING] ) ); shaders.push_back( make_pair( "lighting/lightShinyWaterF.glsl", mVertexShaderLevel[SHADER_LIGHTING] ) ); + shaders.push_back( make_pair( "lighting/lightFullbrightShinyWaterF.glsl", mVertexShaderLevel[SHADER_LIGHTING] ) ); for (U32 i = 0; i < shaders.size(); i++) { @@ -1307,6 +1311,7 @@ BOOL LLViewerShaderMgr::loadShadersObject() { gObjectShinyProgram.unload(); gObjectFullbrightShinyProgram.unload(); + gObjectFullbrightShinyWaterProgram.unload(); gObjectShinyWaterProgram.unload(); gObjectSimpleProgram.unload(); gObjectSimpleWaterProgram.unload(); @@ -1430,6 +1435,23 @@ BOOL LLViewerShaderMgr::loadShadersObject() success = gObjectFullbrightShinyProgram.createShader(NULL, &mShinyUniforms); } + if (success) + { + gObjectFullbrightShinyWaterProgram.mName = "Fullbright Shiny Water Shader"; + gObjectFullbrightShinyWaterProgram.mFeatures.calculatesAtmospherics = true; + gObjectFullbrightShinyWaterProgram.mFeatures.isFullbright = true; + gObjectFullbrightShinyWaterProgram.mFeatures.isShiny = true; + gObjectFullbrightShinyWaterProgram.mFeatures.hasGamma = true; + gObjectFullbrightShinyWaterProgram.mFeatures.hasTransport = true; + gObjectFullbrightShinyWaterProgram.mFeatures.hasWaterFog = true; + gObjectFullbrightShinyWaterProgram.mShaderFiles.clear(); + gObjectFullbrightShinyWaterProgram.mShaderFiles.push_back(make_pair("objects/fullbrightShinyV.glsl", GL_VERTEX_SHADER_ARB)); + gObjectFullbrightShinyWaterProgram.mShaderFiles.push_back(make_pair("objects/fullbrightShinyWaterF.glsl", GL_FRAGMENT_SHADER_ARB)); + gObjectFullbrightShinyWaterProgram.mShaderLevel = mVertexShaderLevel[SHADER_OBJECT]; + gObjectFullbrightShinyWaterProgram.mShaderGroup = LLGLSLShader::SG_WATER; + success = gObjectFullbrightShinyWaterProgram.createShader(NULL, &mShinyUniforms); + } + if (success) { gSkinnedObjectSimpleProgram.mName = "Skinned Simple Shader"; @@ -1541,7 +1563,7 @@ BOOL LLViewerShaderMgr::loadShadersObject() gSkinnedObjectFullbrightShinyWaterProgram.mShaderGroup = LLGLSLShader::SG_WATER; gSkinnedObjectFullbrightShinyWaterProgram.mShaderFiles.clear(); gSkinnedObjectFullbrightShinyWaterProgram.mShaderFiles.push_back(make_pair("objects/fullbrightShinySkinnedV.glsl", GL_VERTEX_SHADER_ARB)); - gSkinnedObjectFullbrightShinyWaterProgram.mShaderFiles.push_back(make_pair("objects/fullbrightWaterF.glsl", GL_FRAGMENT_SHADER_ARB)); + gSkinnedObjectFullbrightShinyWaterProgram.mShaderFiles.push_back(make_pair("objects/fullbrightShinyWaterF.glsl", GL_FRAGMENT_SHADER_ARB)); gSkinnedObjectFullbrightShinyWaterProgram.mShaderLevel = mVertexShaderLevel[SHADER_OBJECT]; success = gSkinnedObjectFullbrightShinyWaterProgram.createShader(NULL, &mShinyUniforms); } diff --git a/indra/newview/llviewershadermgr.h b/indra/newview/llviewershadermgr.h index b6fa86081c..7962854cae 100644 --- a/indra/newview/llviewershadermgr.h +++ b/indra/newview/llviewershadermgr.h @@ -311,6 +311,7 @@ extern LLGLSLShader gObjectSimpleLODProgram; extern LLGLSLShader gObjectFullbrightLODProgram; extern LLGLSLShader gObjectFullbrightShinyProgram; +extern LLGLSLShader gObjectFullbrightShinyWaterProgram; extern LLGLSLShader gObjectShinyProgram; extern LLGLSLShader gObjectShinyWaterProgram; diff --git a/indra/newview/llvovolume.cpp b/indra/newview/llvovolume.cpp index 6fa97b6688..1397b64623 100644 --- a/indra/newview/llvovolume.cpp +++ b/indra/newview/llvovolume.cpp @@ -3546,7 +3546,7 @@ void LLVolumeGeometryManager::rebuildGeom(LLSpatialGroup* group) facep->mLastVertexBuffer = NULL; } - facep->setState(LLFace::RIGGED); + facep->setState(LLFace::RIGGED); //get drawpool of avatar with rigged face LLDrawPoolAvatar* pool = get_avatar_drawpool(vobj); @@ -3585,7 +3585,14 @@ void LLVolumeGeometryManager::rebuildGeom(LLSpatialGroup* group) } else { - pool->addRiggedFace(facep, LLDrawPoolAvatar::RIGGED_SHINY); + if (LLPipeline::sRenderDeferred) + { + pool->addRiggedFace(facep, LLDrawPoolAvatar::RIGGED_SIMPLE); + } + else + { + pool->addRiggedFace(facep, LLDrawPoolAvatar::RIGGED_SHINY); + } } } else -- cgit v1.3 From 2fea1d5d33ec1b41a3cfa4307a1bfa58d8014f88 Mon Sep 17 00:00:00 2001 From: Dave Parks Date: Thu, 19 Aug 2010 12:25:15 -0500 Subject: Integrate SIMD API from oreh/server-trunk-oreh --- indra/cmake/00-Common.cmake | 1 + indra/llcharacter/llkeyframemotion.cpp | 2 +- indra/llmath/CMakeLists.txt | 246 ++-- indra/llmath/llcamera.cpp | 2 +- indra/llmath/llmath.h | 1034 +++++++------- indra/llmath/lloctree.h | 19 +- indra/llmath/llquantize.h | 310 +++-- indra/llmath/llquaternion.cpp | 1921 +++++++++++++------------- indra/llmath/llquaternion.h | 1184 ++++++++-------- indra/llmath/llvolume.cpp | 123 +- indra/llmath/tests/v2math_test.cpp | 6 +- indra/llmath/tests/v3color_test.cpp | 6 +- indra/llmath/tests/v3dmath_test.cpp | 2 +- indra/llmath/tests/v3math_test.cpp | 4 +- indra/llmath/tests/v4color_test.cpp | 4 +- indra/llmath/tests/v4coloru_test.cpp | 2 +- indra/llmath/tests/v4math_test.cpp | 4 +- indra/llmath/v2math.cpp | 2 +- indra/llmath/v2math.h | 8 +- indra/llmath/v3color.h | 10 +- indra/llmath/v3dmath.h | 10 +- indra/llmath/v3math.h | 10 +- indra/llmath/v4color.h | 8 +- indra/llmath/v4coloru.h | 4 +- indra/llmath/v4math.h | 8 +- indra/llmessage/llsdmessagebuilder.cpp | 1 + indra/llmessage/lltemplatemessagebuilder.cpp | 1 + indra/llmessage/lltemplatemessagereader.cpp | 1 + indra/llrender/llrender.cpp | 4 +- indra/newview/llagent.cpp | 2 +- indra/newview/lldrawable.cpp | 10 +- indra/newview/lldrawpoolavatar.cpp | 190 ++- indra/newview/lldrawpoolavatar.h | 3 +- indra/newview/llface.cpp | 29 +- indra/newview/llface.h | 3 +- indra/newview/llflexibleobject.cpp | 2 +- indra/newview/llhudicon.cpp | 2 +- indra/newview/llpanelnearbymedia.cpp | 2 +- indra/newview/llpanelprimmediacontrols.cpp | 4 +- indra/newview/llselectmgr.cpp | 6 +- indra/newview/llspatialpartition.cpp | 44 +- indra/newview/llviewercamera.cpp | 2 +- indra/newview/llviewerjointmesh.cpp | 10 +- indra/newview/llviewerjoystick.cpp | 2 +- indra/newview/llviewermedia.cpp | 2 +- indra/newview/llviewerobject.cpp | 2 +- indra/newview/llviewerpartsim.cpp | 4 +- indra/newview/llviewertexture.cpp | 2 +- indra/newview/llvoavatar.cpp | 22 +- indra/newview/llvoavatar.h | 3 + indra/newview/llvograss.cpp | 2 +- indra/newview/llvosurfacepatch.cpp | 4 +- indra/newview/llvotextbubble.cpp | 4 +- indra/newview/llvotree.cpp | 6 +- indra/newview/llvovolume.cpp | 16 +- indra/newview/llworld.cpp | 4 +- indra/newview/pipeline.cpp | 14 +- 57 files changed, 2731 insertions(+), 2602 deletions(-) (limited to 'indra/newview/llvovolume.cpp') diff --git a/indra/cmake/00-Common.cmake b/indra/cmake/00-Common.cmake index 2c7bb77758..f10a61e1e7 100644 --- a/indra/cmake/00-Common.cmake +++ b/indra/cmake/00-Common.cmake @@ -68,6 +68,7 @@ if (WINDOWS) add_definitions( /Zc:wchar_t- + /arch:SSE2 ) endif (MSVC80 OR MSVC90) diff --git a/indra/llcharacter/llkeyframemotion.cpp b/indra/llcharacter/llkeyframemotion.cpp index 7bc9e97eb1..bce272082e 100644 --- a/indra/llcharacter/llkeyframemotion.cpp +++ b/indra/llcharacter/llkeyframemotion.cpp @@ -1151,7 +1151,7 @@ void LLKeyframeMotion::applyConstraint(JointConstraint* constraint, F32 time, U8 constraint->mPositions[joint_num] = new_pos; } constraint->mFixupDistanceRMS *= 1.f / (constraint->mTotalLength * (F32)(shared_data->mChainLength - 1)); - constraint->mFixupDistanceRMS = fsqrtf(constraint->mFixupDistanceRMS); + constraint->mFixupDistanceRMS = (F32) sqrt(constraint->mFixupDistanceRMS); //reset old joint rots for (joint_num = 0; joint_num <= shared_data->mChainLength; joint_num++) diff --git a/indra/llmath/CMakeLists.txt b/indra/llmath/CMakeLists.txt index dda07133d5..8d85765eb8 100644 --- a/indra/llmath/CMakeLists.txt +++ b/indra/llmath/CMakeLists.txt @@ -1,118 +1,128 @@ -# -*- cmake -*- - -project(llmath) - -include(00-Common) -include(LLCommon) - -include_directories( - ${LLCOMMON_INCLUDE_DIRS} - ) - -set(llmath_SOURCE_FILES - llbbox.cpp - llbboxlocal.cpp - llcamera.cpp - llcoordframe.cpp - llline.cpp - llmodularmath.cpp - llperlin.cpp - llquaternion.cpp - llrect.cpp - llsphere.cpp - llvolume.cpp - llvolumemgr.cpp - llvolumeoctree.cpp - llsdutil_math.cpp - m3math.cpp - m4math.cpp - raytrace.cpp - v2math.cpp - v3color.cpp - v3dmath.cpp - v3math.cpp - v4color.cpp - v4coloru.cpp - v4math.cpp - xform.cpp - ) - -set(llmath_HEADER_FILES - CMakeLists.txt - - camera.h - coordframe.h - llbbox.h - llbboxlocal.h - llcamera.h - llcoord.h - llcoordframe.h - llinterp.h - llline.h - llmath.h - llmodularmath.h - lloctree.h - llperlin.h - llplane.h - llquantize.h - llquaternion.h - llrect.h - llsphere.h - lltreenode.h - llv4math.h - llv4matrix3.h - llv4matrix4.h - llv4vector3.h - llvector4a.h - llmatrix4a.h - llvolume.h - llvolumemgr.h - llvolumeoctree.h - llsdutil_math.h - m3math.h - m4math.h - raytrace.h - v2math.h - v3color.h - v3dmath.h - v3math.h - v4color.h - v4coloru.h - v4math.h - xform.h - ) - -set_source_files_properties(${llmath_HEADER_FILES} - PROPERTIES HEADER_FILE_ONLY TRUE) - -list(APPEND llmath_SOURCE_FILES ${llmath_HEADER_FILES}) - -add_library (llmath ${llmath_SOURCE_FILES}) - -# Add tests -if (LL_TESTS) - include(LLAddBuildTest) - # UNIT TESTS - SET(llmath_TEST_SOURCE_FILES - llbboxlocal.cpp - llmodularmath.cpp - llrect.cpp - v2math.cpp - v3color.cpp - v4color.cpp - v4coloru.cpp - ) - LL_ADD_PROJECT_UNIT_TESTS(llmath "${llmath_TEST_SOURCE_FILES}") - - # INTEGRATION TESTS - set(test_libs llmath llcommon ${LLCOMMON_LIBRARIES} ${WINDOWS_LIBRARIES}) - # TODO: Some of these need refactoring to be proper Unit tests rather than Integration tests. - LL_ADD_INTEGRATION_TEST(llbbox llbbox.cpp "${test_libs}") - LL_ADD_INTEGRATION_TEST(llquaternion llquaternion.cpp "${test_libs}") - LL_ADD_INTEGRATION_TEST(mathmisc "" "${test_libs}") - LL_ADD_INTEGRATION_TEST(m3math "" "${test_libs}") - LL_ADD_INTEGRATION_TEST(v3dmath v3dmath.cpp "${test_libs}") - LL_ADD_INTEGRATION_TEST(v3math v3math.cpp "${test_libs}") - LL_ADD_INTEGRATION_TEST(v4math v4math.cpp "${test_libs}") - LL_ADD_INTEGRATION_TEST(xform xform.cpp "${test_libs}") -endif (LL_TESTS) +# -*- cmake -*- + +project(llmath) + +include(00-Common) +include(LLCommon) + +include_directories( + ${LLCOMMON_INCLUDE_DIRS} + ) + +set(llmath_SOURCE_FILES + llbbox.cpp + llbboxlocal.cpp + llcamera.cpp + llcoordframe.cpp + llline.cpp + llmatrix3a.cpp + llmodularmath.cpp + llperlin.cpp + llquaternion.cpp + llrect.cpp + llsphere.cpp + llvector4a.cpp + llvolume.cpp + llvolumemgr.cpp + llvolumeoctree.cpp + llsdutil_math.cpp + m3math.cpp + m4math.cpp + raytrace.cpp + v2math.cpp + v3color.cpp + v3dmath.cpp + v3math.cpp + v4color.cpp + v4coloru.cpp + v4math.cpp + xform.cpp + ) + +set(llmath_HEADER_FILES + CMakeLists.txt + + camera.h + coordframe.h + llbbox.h + llbboxlocal.h + llcamera.h + llcoord.h + llcoordframe.h + llinterp.h + llline.h + llmath.h + llmatrix3a.h + llmatrix3a.inl + llmodularmath.h + lloctree.h + llperlin.h + llplane.h + llquantize.h + llquaternion.h + llquaternion2.h + llquaternion2.inl + llrect.h + llsimdmath.h + llsimdtypes.h + llsimdtypes.inl + llsphere.h + lltreenode.h + llvector4a.h + llvector4a.inl + llvector4logical.h + llv4math.h + llv4matrix3.h + llv4matrix4.h + llv4vector3.h + llvolume.h + llvolumemgr.h + llvolumeoctree.h + llsdutil_math.h + m3math.h + m4math.h + raytrace.h + v2math.h + v3color.h + v3dmath.h + v3math.h + v4color.h + v4coloru.h + v4math.h + xform.h + ) + +set_source_files_properties(${llmath_HEADER_FILES} + PROPERTIES HEADER_FILE_ONLY TRUE) + +list(APPEND llmath_SOURCE_FILES ${llmath_HEADER_FILES}) + +add_library (llmath ${llmath_SOURCE_FILES}) + +# Add tests +if (LL_TESTS) + include(LLAddBuildTest) + # UNIT TESTS + SET(llmath_TEST_SOURCE_FILES + llbboxlocal.cpp + llmodularmath.cpp + llrect.cpp + v2math.cpp + v3color.cpp + v4color.cpp + v4coloru.cpp + ) + LL_ADD_PROJECT_UNIT_TESTS(llmath "${llmath_TEST_SOURCE_FILES}") + + # INTEGRATION TESTS + set(test_libs llmath llcommon ${LLCOMMON_LIBRARIES} ${WINDOWS_LIBRARIES}) + # TODO: Some of these need refactoring to be proper Unit tests rather than Integration tests. + LL_ADD_INTEGRATION_TEST(llbbox llbbox.cpp "${test_libs}") + LL_ADD_INTEGRATION_TEST(llquaternion llquaternion.cpp "${test_libs}") + LL_ADD_INTEGRATION_TEST(mathmisc "" "${test_libs}") + LL_ADD_INTEGRATION_TEST(m3math "" "${test_libs}") + LL_ADD_INTEGRATION_TEST(v3dmath v3dmath.cpp "${test_libs}") + LL_ADD_INTEGRATION_TEST(v3math v3math.cpp "${test_libs}") + LL_ADD_INTEGRATION_TEST(v4math v4math.cpp "${test_libs}") + LL_ADD_INTEGRATION_TEST(xform xform.cpp "${test_libs}") +endif (LL_TESTS) diff --git a/indra/llmath/llcamera.cpp b/indra/llmath/llcamera.cpp index 6b56e4870e..beb5c48624 100644 --- a/indra/llmath/llcamera.cpp +++ b/indra/llmath/llcamera.cpp @@ -77,7 +77,7 @@ const LLCamera& LLCamera::operator=(const LLCamera& rhs) { memcpy(this, &rhs, sizeof(LLCamera)); alignPlanes(); - LLVector4a::memcpyNonAliased16((F32*) mAgentPlanes, (F32*) rhs.mAgentPlanes, 4*7); + LLVector4a::memcpyNonAliased16((F32*) mAgentPlanes, (F32*) rhs.mAgentPlanes, 4*7*sizeof(F32)); return *this; } diff --git a/indra/llmath/llmath.h b/indra/llmath/llmath.h index c3c15e1374..742bbc4751 100644 --- a/indra/llmath/llmath.h +++ b/indra/llmath/llmath.h @@ -1,525 +1,509 @@ -/** - * @file llmath.h - * @brief Useful math constants and macros. - * - * $LicenseInfo:firstyear=2000&license=viewergpl$ - * - * Copyright (c) 2000-2009, Linden Research, Inc. - * - * Second Life Viewer Source Code - * The source code in this file ("Source Code") is provided by Linden Lab - * to you under the terms of the GNU General Public License, version 2.0 - * ("GPL"), unless you have obtained a separate licensing agreement - * ("Other License"), formally executed by you and Linden Lab. Terms of - * the GPL can be found in doc/GPL-license.txt in this distribution, or - * online at http://secondlifegrid.net/programs/open_source/licensing/gplv2 - * - * There are special exceptions to the terms and conditions of the GPL as - * it is applied to this Source Code. View the full text of the exception - * in the file doc/FLOSS-exception.txt in this software distribution, or - * online at - * http://secondlifegrid.net/programs/open_source/licensing/flossexception - * - * By copying, modifying or distributing this software, you acknowledge - * that you have read and understood your obligations described above, - * and agree to abide by those obligations. - * - * ALL LINDEN LAB SOURCE CODE IS PROVIDED "AS IS." LINDEN LAB MAKES NO - * WARRANTIES, EXPRESS, IMPLIED OR OTHERWISE, REGARDING ITS ACCURACY, - * COMPLETENESS OR PERFORMANCE. - * $/LicenseInfo$ - */ - -#ifndef LLMATH_H -#define LLMATH_H - -#include -#include -#include -#include "lldefs.h" -//#include "llstl.h" // *TODO: Remove when LLString is gone -//#include "llstring.h" // *TODO: Remove when LLString is gone -// lltut.h uses is_approx_equal_fraction(). This was moved to its own header -// file in llcommon so we can use lltut.h for llcommon tests without making -// llcommon depend on llmath. -#include "is_approx_equal_fraction.h" - -// work around for Windows & older gcc non-standard function names. -#if LL_WINDOWS -#include -#define llisnan(val) _isnan(val) -#define llfinite(val) _finite(val) -#elif (LL_LINUX && __GNUC__ <= 2) -#define llisnan(val) isnan(val) -#define llfinite(val) isfinite(val) -#elif LL_SOLARIS -#define llisnan(val) isnan(val) -#define llfinite(val) (val <= std::numeric_limits::max()) -#else -#define llisnan(val) std::isnan(val) -#define llfinite(val) std::isfinite(val) -#endif - -// Single Precision Floating Point Routines -#ifndef sqrtf -#define sqrtf(x) ((F32)sqrt((F64)(x))) -#endif -#ifndef fsqrtf -#define fsqrtf(x) sqrtf(x) -#endif - -#ifndef cosf -#define cosf(x) ((F32)cos((F64)(x))) -#endif -#ifndef sinf -#define sinf(x) ((F32)sin((F64)(x))) -#endif -#ifndef tanf -#define tanf(x) ((F32)tan((F64)(x))) -#endif -#ifndef acosf -#define acosf(x) ((F32)acos((F64)(x))) -#endif - -#ifndef powf -#define powf(x,y) ((F32)pow((F64)(x),(F64)(y))) -#endif -#ifndef expf -#define expf(x) ((F32)exp((F64)(x))) -#endif - -const F32 GRAVITY = -9.8f; - -// mathematical constants -const F32 F_PI = 3.1415926535897932384626433832795f; -const F32 F_TWO_PI = 6.283185307179586476925286766559f; -const F32 F_PI_BY_TWO = 1.5707963267948966192313216916398f; -const F32 F_SQRT_TWO_PI = 2.506628274631000502415765284811f; -const F32 F_E = 2.71828182845904523536f; -const F32 F_SQRT2 = 1.4142135623730950488016887242097f; -const F32 F_SQRT3 = 1.73205080756888288657986402541f; -const F32 OO_SQRT2 = 0.7071067811865475244008443621049f; -const F32 DEG_TO_RAD = 0.017453292519943295769236907684886f; -const F32 RAD_TO_DEG = 57.295779513082320876798154814105f; -const F32 F_APPROXIMATELY_ZERO = 0.00001f; -const F32 F_LN2 = 0.69314718056f; -const F32 OO_LN2 = 1.4426950408889634073599246810019f; - -const F32 F_ALMOST_ZERO = 0.0001f; -const F32 F_ALMOST_ONE = 1.0f - F_ALMOST_ZERO; - -// BUG: Eliminate in favor of F_APPROXIMATELY_ZERO above? -const F32 FP_MAG_THRESHOLD = 0.0000001f; - -// TODO: Replace with logic like is_approx_equal -inline BOOL is_approx_zero( F32 f ) { return (-F_APPROXIMATELY_ZERO < f) && (f < F_APPROXIMATELY_ZERO); } - -// These functions work by interpreting sign+exp+mantissa as an unsigned -// integer. -// For example: -// x = 1 00000010 00000000000000000000000 -// y = 1 00000001 11111111111111111111111 -// -// interpreted as ints = -// x = 10000001000000000000000000000000 -// y = 10000000111111111111111111111111 -// which is clearly a different of 1 in the least significant bit -// Values with the same exponent can be trivially shown to work. -// -// WARNING: Denormals of opposite sign do not work -// x = 1 00000000 00000000000000000000001 -// y = 0 00000000 00000000000000000000001 -// Although these values differ by 2 in the LSB, the sign bit makes -// the int comparison fail. -// -// WARNING: NaNs can compare equal -// There is no special treatment of exceptional values like NaNs -// -// WARNING: Infinity is comparable with F32_MAX and negative -// infinity is comparable with F32_MIN - -inline BOOL is_approx_equal(F32 x, F32 y) -{ - const S32 COMPARE_MANTISSA_UP_TO_BIT = 0x02; - return (std::abs((S32) ((U32&)x - (U32&)y) ) < COMPARE_MANTISSA_UP_TO_BIT); -} - -inline BOOL is_approx_equal(F64 x, F64 y) -{ - const S64 COMPARE_MANTISSA_UP_TO_BIT = 0x02; - return (std::abs((S32) ((U64&)x - (U64&)y) ) < COMPARE_MANTISSA_UP_TO_BIT); -} - -inline S32 llabs(const S32 a) -{ - return S32(std::labs(a)); -} - -inline F32 llabs(const F32 a) -{ - return F32(std::fabs(a)); -} - -inline F64 llabs(const F64 a) -{ - return F64(std::fabs(a)); -} - -inline S32 lltrunc( F32 f ) -{ -#if LL_WINDOWS && !defined( __INTEL_COMPILER ) - // Avoids changing the floating point control word. - // Add or subtract 0.5 - epsilon and then round - const static U32 zpfp[] = { 0xBEFFFFFF, 0x3EFFFFFF }; - S32 result; - __asm { - fld f - mov eax, f - shr eax, 29 - and eax, 4 - fadd dword ptr [zpfp + eax] - fistp result - } - return result; -#else - return (S32)f; -#endif -} - -inline S32 lltrunc( F64 f ) -{ - return (S32)f; -} - -inline S32 llfloor( F32 f ) -{ -#if LL_WINDOWS && !defined( __INTEL_COMPILER ) - // Avoids changing the floating point control word. - // Accurate (unlike Stereopsis version) for all values between S32_MIN and S32_MAX and slightly faster than Stereopsis version. - // Add -(0.5 - epsilon) and then round - const U32 zpfp = 0xBEFFFFFF; - S32 result; - __asm { - fld f - fadd dword ptr [zpfp] - fistp result - } - return result; -#else - return (S32)floorf(f); -#endif -} - - -inline S32 llceil( F32 f ) -{ - // This could probably be optimized, but this works. - return (S32)ceil(f); -} - - -#ifndef BOGUS_ROUND -// Use this round. Does an arithmetic round (0.5 always rounds up) -inline S32 llround(const F32 val) -{ - return llfloor(val + 0.5f); -} - -#else // BOGUS_ROUND -// Old llround implementation - does banker's round (toward nearest even in the case of a 0.5. -// Not using this because we don't have a consistent implementation on both platforms, use -// llfloor(val + 0.5f), which is consistent on all platforms. -inline S32 llround(const F32 val) -{ - #if LL_WINDOWS - // Note: assumes that the floating point control word is set to rounding mode (the default) - S32 ret_val; - _asm fld val - _asm fistp ret_val; - return ret_val; - #elif LL_LINUX - // Note: assumes that the floating point control word is set - // to rounding mode (the default) - S32 ret_val; - __asm__ __volatile__( "flds %1 \n\t" - "fistpl %0 \n\t" - : "=m" (ret_val) - : "m" (val) ); - return ret_val; - #else - return llfloor(val + 0.5f); - #endif -} - -// A fast arithmentic round on intel, from Laurent de Soras http://ldesoras.free.fr -inline int round_int(double x) -{ - const float round_to_nearest = 0.5f; - int i; - __asm - { - fld x - fadd st, st (0) - fadd round_to_nearest - fistp i - sar i, 1 - } - return (i); -} -#endif // BOGUS_ROUND - -inline F32 llround( F32 val, F32 nearest ) -{ - return F32(floor(val * (1.0f / nearest) + 0.5f)) * nearest; -} - -inline F64 llround( F64 val, F64 nearest ) -{ - return F64(floor(val * (1.0 / nearest) + 0.5)) * nearest; -} - -// these provide minimum peak error -// -// avg error = -0.013049 -// peak error = -31.4 dB -// RMS error = -28.1 dB - -const F32 FAST_MAG_ALPHA = 0.960433870103f; -const F32 FAST_MAG_BETA = 0.397824734759f; - -// these provide minimum RMS error -// -// avg error = 0.000003 -// peak error = -32.6 dB -// RMS error = -25.7 dB -// -//const F32 FAST_MAG_ALPHA = 0.948059448969f; -//const F32 FAST_MAG_BETA = 0.392699081699f; - -inline F32 fastMagnitude(F32 a, F32 b) -{ - a = (a > 0) ? a : -a; - b = (b > 0) ? b : -b; - return(FAST_MAG_ALPHA * llmax(a,b) + FAST_MAG_BETA * llmin(a,b)); -} - - - -//////////////////// -// -// Fast F32/S32 conversions -// -// Culled from www.stereopsis.com/FPU.html - -const F64 LL_DOUBLE_TO_FIX_MAGIC = 68719476736.0*1.5; //2^36 * 1.5, (52-_shiftamt=36) uses limited precisicion to floor -const S32 LL_SHIFT_AMOUNT = 16; //16.16 fixed point representation, - -// Endian dependent code -#ifdef LL_LITTLE_ENDIAN - #define LL_EXP_INDEX 1 - #define LL_MAN_INDEX 0 -#else - #define LL_EXP_INDEX 0 - #define LL_MAN_INDEX 1 -#endif - -/* Deprecated: use llround(), lltrunc(), or llfloor() instead -// ================================================================================================ -// Real2Int -// ================================================================================================ -inline S32 F64toS32(F64 val) -{ - val = val + LL_DOUBLE_TO_FIX_MAGIC; - return ((S32*)&val)[LL_MAN_INDEX] >> LL_SHIFT_AMOUNT; -} - -// ================================================================================================ -// Real2Int -// ================================================================================================ -inline S32 F32toS32(F32 val) -{ - return F64toS32 ((F64)val); -} -*/ - -//////////////////////////////////////////////// -// -// Fast exp and log -// - -// Implementation of fast exp() approximation (from a paper by Nicol N. Schraudolph -// http://www.inf.ethz.ch/~schraudo/pubs/exp.pdf -static union -{ - double d; - struct - { -#ifdef LL_LITTLE_ENDIAN - S32 j, i; -#else - S32 i, j; -#endif - } n; -} LLECO; // not sure what the name means - -#define LL_EXP_A (1048576 * OO_LN2) // use 1512775 for integer -#define LL_EXP_C (60801) // this value of C good for -4 < y < 4 - -#define LL_FAST_EXP(y) (LLECO.n.i = llround(F32(LL_EXP_A*(y))) + (1072693248 - LL_EXP_C), LLECO.d) - - - -inline F32 llfastpow(const F32 x, const F32 y) -{ - return (F32)(LL_FAST_EXP(y * log(x))); -} - - -inline F32 snap_to_sig_figs(F32 foo, S32 sig_figs) -{ - // compute the power of ten - F32 bar = 1.f; - for (S32 i = 0; i < sig_figs; i++) - { - bar *= 10.f; - } - - foo = (F32)llround(foo * bar); - - // shift back - foo /= bar; - return foo; -} - -inline F32 lerp(F32 a, F32 b, F32 u) -{ - return a + ((b - a) * u); -} - -inline F32 lerp2d(F32 x00, F32 x01, F32 x10, F32 x11, F32 u, F32 v) -{ - F32 a = x00 + (x01-x00)*u; - F32 b = x10 + (x11-x10)*u; - F32 r = a + (b-a)*v; - return r; -} - -inline F32 ramp(F32 x, F32 a, F32 b) -{ - return (a == b) ? 0.0f : ((a - x) / (a - b)); -} - -inline F32 rescale(F32 x, F32 x1, F32 x2, F32 y1, F32 y2) -{ - return lerp(y1, y2, ramp(x, x1, x2)); -} - -inline F32 clamp_rescale(F32 x, F32 x1, F32 x2, F32 y1, F32 y2) -{ - if (y1 < y2) - { - return llclamp(rescale(x,x1,x2,y1,y2),y1,y2); - } - else - { - return llclamp(rescale(x,x1,x2,y1,y2),y2,y1); - } -} - - -inline F32 cubic_step( F32 x, F32 x0, F32 x1, F32 s0, F32 s1 ) -{ - if (x <= x0) - return s0; - - if (x >= x1) - return s1; - - F32 f = (x - x0) / (x1 - x0); - - return s0 + (s1 - s0) * (f * f) * (3.0f - 2.0f * f); -} - -inline F32 cubic_step( F32 x ) -{ - x = llclampf(x); - - return (x * x) * (3.0f - 2.0f * x); -} - -inline F32 quadratic_step( F32 x, F32 x0, F32 x1, F32 s0, F32 s1 ) -{ - if (x <= x0) - return s0; - - if (x >= x1) - return s1; - - F32 f = (x - x0) / (x1 - x0); - F32 f_squared = f * f; - - return (s0 * (1.f - f_squared)) + ((s1 - s0) * f_squared); -} - -inline F32 llsimple_angle(F32 angle) -{ - while(angle <= -F_PI) - angle += F_TWO_PI; - while(angle > F_PI) - angle -= F_TWO_PI; - return angle; -} - -//SDK - Renamed this to get_lower_power_two, since this is what this actually does. -inline U32 get_lower_power_two(U32 val, U32 max_power_two) -{ - if(!max_power_two) - { - max_power_two = 1 << 31 ; - } - if(max_power_two & (max_power_two - 1)) - { - return 0 ; - } - - for(; val < max_power_two ; max_power_two >>= 1) ; - - return max_power_two ; -} - -// calculate next highest power of two, limited by max_power_two -// This is taken from a brilliant little code snipped on http://acius2.blogspot.com/2007/11/calculating-next-power-of-2.html -// Basically we convert the binary to a solid string of 1's with the same -// number of digits, then add one. We subtract 1 initially to handle -// the case where the number passed in is actually a power of two. -// WARNING: this only works with 32 bit ints. -inline U32 get_next_power_two(U32 val, U32 max_power_two) -{ - if(!max_power_two) - { - max_power_two = 1 << 31 ; - } - - if(val >= max_power_two) - { - return max_power_two; - } - - val--; - val = (val >> 1) | val; - val = (val >> 2) | val; - val = (val >> 4) | val; - val = (val >> 8) | val; - val = (val >> 16) | val; - val++; - - return val; -} - -//get the gaussian value given the linear distance from axis x and guassian value o -inline F32 llgaussian(F32 x, F32 o) -{ - return 1.f/(F_SQRT_TWO_PI*o)*powf(F_E, -(x*x)/(2*o*o)); -} - -#endif +/** + * @file llmath.h + * @brief Useful math constants and macros. + * + * $LicenseInfo:firstyear=2000&license=viewergpl$ + * + * Copyright (c) 2000-2009, Linden Research, Inc. + * + * Second Life Viewer Source Code + * The source code in this file ("Source Code") is provided by Linden Lab + * to you under the terms of the GNU General Public License, version 2.0 + * ("GPL"), unless you have obtained a separate licensing agreement + * ("Other License"), formally executed by you and Linden Lab. Terms of + * the GPL can be found in doc/GPL-license.txt in this distribution, or + * online at http://secondlifegrid.net/programs/open_source/licensing/gplv2 + * + * There are special exceptions to the terms and conditions of the GPL as + * it is applied to this Source Code. View the full text of the exception + * in the file doc/FLOSS-exception.txt in this software distribution, or + * online at + * http://secondlifegrid.net/programs/open_source/licensing/flossexception + * + * By copying, modifying or distributing this software, you acknowledge + * that you have read and understood your obligations described above, + * and agree to abide by those obligations. + * + * ALL LINDEN LAB SOURCE CODE IS PROVIDED "AS IS." LINDEN LAB MAKES NO + * WARRANTIES, EXPRESS, IMPLIED OR OTHERWISE, REGARDING ITS ACCURACY, + * COMPLETENESS OR PERFORMANCE. + * $/LicenseInfo$ + */ + +#ifndef LLMATH_H +#define LLMATH_H + +#include +#include +#include "lldefs.h" +//#include "llstl.h" // *TODO: Remove when LLString is gone +//#include "llstring.h" // *TODO: Remove when LLString is gone +// lltut.h uses is_approx_equal_fraction(). This was moved to its own header +// file in llcommon so we can use lltut.h for llcommon tests without making +// llcommon depend on llmath. +#include "is_approx_equal_fraction.h" + +// work around for Windows & older gcc non-standard function names. +#if LL_WINDOWS +#include +#define llisnan(val) _isnan(val) +#define llfinite(val) _finite(val) +#elif (LL_LINUX && __GNUC__ <= 2) +#define llisnan(val) isnan(val) +#define llfinite(val) isfinite(val) +#elif LL_SOLARIS +#define llisnan(val) isnan(val) +#define llfinite(val) (val <= std::numeric_limits::max()) +#else +#define llisnan(val) std::isnan(val) +#define llfinite(val) std::isfinite(val) +#endif + +// Single Precision Floating Point Routines +// (There used to be more defined here, but they appeared to be redundant and +// were breaking some other includes. Removed by Falcon, reviewed by Andrew, 11/25/09) +/*#ifndef tanf +#define tanf(x) ((F32)tan((F64)(x))) +#endif*/ + +const F32 GRAVITY = -9.8f; + +// mathematical constants +const F32 F_PI = 3.1415926535897932384626433832795f; +const F32 F_TWO_PI = 6.283185307179586476925286766559f; +const F32 F_PI_BY_TWO = 1.5707963267948966192313216916398f; +const F32 F_SQRT_TWO_PI = 2.506628274631000502415765284811f; +const F32 F_E = 2.71828182845904523536f; +const F32 F_SQRT2 = 1.4142135623730950488016887242097f; +const F32 F_SQRT3 = 1.73205080756888288657986402541f; +const F32 OO_SQRT2 = 0.7071067811865475244008443621049f; +const F32 DEG_TO_RAD = 0.017453292519943295769236907684886f; +const F32 RAD_TO_DEG = 57.295779513082320876798154814105f; +const F32 F_APPROXIMATELY_ZERO = 0.00001f; +const F32 F_LN2 = 0.69314718056f; +const F32 OO_LN2 = 1.4426950408889634073599246810019f; + +const F32 F_ALMOST_ZERO = 0.0001f; +const F32 F_ALMOST_ONE = 1.0f - F_ALMOST_ZERO; + +// BUG: Eliminate in favor of F_APPROXIMATELY_ZERO above? +const F32 FP_MAG_THRESHOLD = 0.0000001f; + +// TODO: Replace with logic like is_approx_equal +inline BOOL is_approx_zero( F32 f ) { return (-F_APPROXIMATELY_ZERO < f) && (f < F_APPROXIMATELY_ZERO); } + +// These functions work by interpreting sign+exp+mantissa as an unsigned +// integer. +// For example: +// x = 1 00000010 00000000000000000000000 +// y = 1 00000001 11111111111111111111111 +// +// interpreted as ints = +// x = 10000001000000000000000000000000 +// y = 10000000111111111111111111111111 +// which is clearly a different of 1 in the least significant bit +// Values with the same exponent can be trivially shown to work. +// +// WARNING: Denormals of opposite sign do not work +// x = 1 00000000 00000000000000000000001 +// y = 0 00000000 00000000000000000000001 +// Although these values differ by 2 in the LSB, the sign bit makes +// the int comparison fail. +// +// WARNING: NaNs can compare equal +// There is no special treatment of exceptional values like NaNs +// +// WARNING: Infinity is comparable with F32_MAX and negative +// infinity is comparable with F32_MIN + +inline BOOL is_approx_equal(F32 x, F32 y) +{ + const S32 COMPARE_MANTISSA_UP_TO_BIT = 0x02; + return (std::abs((S32) ((U32&)x - (U32&)y) ) < COMPARE_MANTISSA_UP_TO_BIT); +} + +inline BOOL is_approx_equal(F64 x, F64 y) +{ + const S64 COMPARE_MANTISSA_UP_TO_BIT = 0x02; + return (std::abs((S32) ((U64&)x - (U64&)y) ) < COMPARE_MANTISSA_UP_TO_BIT); +} + +inline S32 llabs(const S32 a) +{ + return S32(std::labs(a)); +} + +inline F32 llabs(const F32 a) +{ + return F32(std::fabs(a)); +} + +inline F64 llabs(const F64 a) +{ + return F64(std::fabs(a)); +} + +inline S32 lltrunc( F32 f ) +{ +#if LL_WINDOWS && !defined( __INTEL_COMPILER ) + // Avoids changing the floating point control word. + // Add or subtract 0.5 - epsilon and then round + const static U32 zpfp[] = { 0xBEFFFFFF, 0x3EFFFFFF }; + S32 result; + __asm { + fld f + mov eax, f + shr eax, 29 + and eax, 4 + fadd dword ptr [zpfp + eax] + fistp result + } + return result; +#else + return (S32)f; +#endif +} + +inline S32 lltrunc( F64 f ) +{ + return (S32)f; +} + +inline S32 llfloor( F32 f ) +{ +#if LL_WINDOWS && !defined( __INTEL_COMPILER ) + // Avoids changing the floating point control word. + // Accurate (unlike Stereopsis version) for all values between S32_MIN and S32_MAX and slightly faster than Stereopsis version. + // Add -(0.5 - epsilon) and then round + const U32 zpfp = 0xBEFFFFFF; + S32 result; + __asm { + fld f + fadd dword ptr [zpfp] + fistp result + } + return result; +#else + return (S32)floor(f); +#endif +} + + +inline S32 llceil( F32 f ) +{ + // This could probably be optimized, but this works. + return (S32)ceil(f); +} + + +#ifndef BOGUS_ROUND +// Use this round. Does an arithmetic round (0.5 always rounds up) +inline S32 llround(const F32 val) +{ + return llfloor(val + 0.5f); +} + +#else // BOGUS_ROUND +// Old llround implementation - does banker's round (toward nearest even in the case of a 0.5. +// Not using this because we don't have a consistent implementation on both platforms, use +// llfloor(val + 0.5f), which is consistent on all platforms. +inline S32 llround(const F32 val) +{ + #if LL_WINDOWS + // Note: assumes that the floating point control word is set to rounding mode (the default) + S32 ret_val; + _asm fld val + _asm fistp ret_val; + return ret_val; + #elif LL_LINUX + // Note: assumes that the floating point control word is set + // to rounding mode (the default) + S32 ret_val; + __asm__ __volatile__( "flds %1 \n\t" + "fistpl %0 \n\t" + : "=m" (ret_val) + : "m" (val) ); + return ret_val; + #else + return llfloor(val + 0.5f); + #endif +} + +// A fast arithmentic round on intel, from Laurent de Soras http://ldesoras.free.fr +inline int round_int(double x) +{ + const float round_to_nearest = 0.5f; + int i; + __asm + { + fld x + fadd st, st (0) + fadd round_to_nearest + fistp i + sar i, 1 + } + return (i); +} +#endif // BOGUS_ROUND + +inline F32 llround( F32 val, F32 nearest ) +{ + return F32(floor(val * (1.0f / nearest) + 0.5f)) * nearest; +} + +inline F64 llround( F64 val, F64 nearest ) +{ + return F64(floor(val * (1.0 / nearest) + 0.5)) * nearest; +} + +// these provide minimum peak error +// +// avg error = -0.013049 +// peak error = -31.4 dB +// RMS error = -28.1 dB + +const F32 FAST_MAG_ALPHA = 0.960433870103f; +const F32 FAST_MAG_BETA = 0.397824734759f; + +// these provide minimum RMS error +// +// avg error = 0.000003 +// peak error = -32.6 dB +// RMS error = -25.7 dB +// +//const F32 FAST_MAG_ALPHA = 0.948059448969f; +//const F32 FAST_MAG_BETA = 0.392699081699f; + +inline F32 fastMagnitude(F32 a, F32 b) +{ + a = (a > 0) ? a : -a; + b = (b > 0) ? b : -b; + return(FAST_MAG_ALPHA * llmax(a,b) + FAST_MAG_BETA * llmin(a,b)); +} + + + +//////////////////// +// +// Fast F32/S32 conversions +// +// Culled from www.stereopsis.com/FPU.html + +const F64 LL_DOUBLE_TO_FIX_MAGIC = 68719476736.0*1.5; //2^36 * 1.5, (52-_shiftamt=36) uses limited precisicion to floor +const S32 LL_SHIFT_AMOUNT = 16; //16.16 fixed point representation, + +// Endian dependent code +#ifdef LL_LITTLE_ENDIAN + #define LL_EXP_INDEX 1 + #define LL_MAN_INDEX 0 +#else + #define LL_EXP_INDEX 0 + #define LL_MAN_INDEX 1 +#endif + +/* Deprecated: use llround(), lltrunc(), or llfloor() instead +// ================================================================================================ +// Real2Int +// ================================================================================================ +inline S32 F64toS32(F64 val) +{ + val = val + LL_DOUBLE_TO_FIX_MAGIC; + return ((S32*)&val)[LL_MAN_INDEX] >> LL_SHIFT_AMOUNT; +} + +// ================================================================================================ +// Real2Int +// ================================================================================================ +inline S32 F32toS32(F32 val) +{ + return F64toS32 ((F64)val); +} +*/ + +//////////////////////////////////////////////// +// +// Fast exp and log +// + +// Implementation of fast exp() approximation (from a paper by Nicol N. Schraudolph +// http://www.inf.ethz.ch/~schraudo/pubs/exp.pdf +static union +{ + double d; + struct + { +#ifdef LL_LITTLE_ENDIAN + S32 j, i; +#else + S32 i, j; +#endif + } n; +} LLECO; // not sure what the name means + +#define LL_EXP_A (1048576 * OO_LN2) // use 1512775 for integer +#define LL_EXP_C (60801) // this value of C good for -4 < y < 4 + +#define LL_FAST_EXP(y) (LLECO.n.i = llround(F32(LL_EXP_A*(y))) + (1072693248 - LL_EXP_C), LLECO.d) + + + +inline F32 llfastpow(const F32 x, const F32 y) +{ + return (F32)(LL_FAST_EXP(y * log(x))); +} + + +inline F32 snap_to_sig_figs(F32 foo, S32 sig_figs) +{ + // compute the power of ten + F32 bar = 1.f; + for (S32 i = 0; i < sig_figs; i++) + { + bar *= 10.f; + } + + //F32 new_foo = (F32)llround(foo * bar); + // the llround() implementation sucks. Don't us it. + + F32 sign = (foo > 0.f) ? 1.f : -1.f; + F32 new_foo = F32( S64(foo * bar + sign * 0.5f)); + new_foo /= bar; + + return new_foo; +} + +inline F32 lerp(F32 a, F32 b, F32 u) +{ + return a + ((b - a) * u); +} + +inline F32 lerp2d(F32 x00, F32 x01, F32 x10, F32 x11, F32 u, F32 v) +{ + F32 a = x00 + (x01-x00)*u; + F32 b = x10 + (x11-x10)*u; + F32 r = a + (b-a)*v; + return r; +} + +inline F32 ramp(F32 x, F32 a, F32 b) +{ + return (a == b) ? 0.0f : ((a - x) / (a - b)); +} + +inline F32 rescale(F32 x, F32 x1, F32 x2, F32 y1, F32 y2) +{ + return lerp(y1, y2, ramp(x, x1, x2)); +} + +inline F32 clamp_rescale(F32 x, F32 x1, F32 x2, F32 y1, F32 y2) +{ + if (y1 < y2) + { + return llclamp(rescale(x,x1,x2,y1,y2),y1,y2); + } + else + { + return llclamp(rescale(x,x1,x2,y1,y2),y2,y1); + } +} + + +inline F32 cubic_step( F32 x, F32 x0, F32 x1, F32 s0, F32 s1 ) +{ + if (x <= x0) + return s0; + + if (x >= x1) + return s1; + + F32 f = (x - x0) / (x1 - x0); + + return s0 + (s1 - s0) * (f * f) * (3.0f - 2.0f * f); +} + +inline F32 cubic_step( F32 x ) +{ + x = llclampf(x); + + return (x * x) * (3.0f - 2.0f * x); +} + +inline F32 quadratic_step( F32 x, F32 x0, F32 x1, F32 s0, F32 s1 ) +{ + if (x <= x0) + return s0; + + if (x >= x1) + return s1; + + F32 f = (x - x0) / (x1 - x0); + F32 f_squared = f * f; + + return (s0 * (1.f - f_squared)) + ((s1 - s0) * f_squared); +} + +inline F32 llsimple_angle(F32 angle) +{ + while(angle <= -F_PI) + angle += F_TWO_PI; + while(angle > F_PI) + angle -= F_TWO_PI; + return angle; +} + +//SDK - Renamed this to get_lower_power_two, since this is what this actually does. +inline U32 get_lower_power_two(U32 val, U32 max_power_two) +{ + if(!max_power_two) + { + max_power_two = 1 << 31 ; + } + if(max_power_two & (max_power_two - 1)) + { + return 0 ; + } + + for(; val < max_power_two ; max_power_two >>= 1) ; + + return max_power_two ; +} + +// calculate next highest power of two, limited by max_power_two +// This is taken from a brilliant little code snipped on http://acius2.blogspot.com/2007/11/calculating-next-power-of-2.html +// Basically we convert the binary to a solid string of 1's with the same +// number of digits, then add one. We subtract 1 initially to handle +// the case where the number passed in is actually a power of two. +// WARNING: this only works with 32 bit ints. +inline U32 get_next_power_two(U32 val, U32 max_power_two) +{ + if(!max_power_two) + { + max_power_two = 1 << 31 ; + } + + if(val >= max_power_two) + { + return max_power_two; + } + + val--; + val = (val >> 1) | val; + val = (val >> 2) | val; + val = (val >> 4) | val; + val = (val >> 8) | val; + val = (val >> 16) | val; + val++; + + return val; +} + +//get the gaussian value given the linear distance from axis x and guassian value o +inline F32 llgaussian(F32 x, F32 o) +{ + return 1.f/(F_SQRT_TWO_PI*o)*powf(F_E, -(x*x)/(2*o*o)); +} + +// Include simd math header +#include "llsimdmath.h" + +#endif diff --git a/indra/llmath/lloctree.h b/indra/llmath/lloctree.h index 59828ae565..432e9fbcd8 100644 --- a/indra/llmath/lloctree.h +++ b/indra/llmath/lloctree.h @@ -142,7 +142,7 @@ public: S32 getOctant(const LLVector4a& pos) const //get the octant pos is in { - return pos.greaterThan4(mD[CENTER]).getComparisonMask() & 0x7; + return pos.greaterThan(mD[CENTER]).getGatheredBits() & 0x7; } inline bool isInside(const LLVector4a& pos, const F32& rad) const @@ -157,13 +157,13 @@ public: bool isInside(const LLVector4a& pos) const { - S32 gt = pos.greaterThan4(mD[MAX]).getComparisonMask() & 0x7; + S32 gt = pos.greaterThan(mD[MAX]).getGatheredBits() & 0x7; if (gt) { return false; } - S32 lt = pos.lessEqual4(mD[MIN]).getComparisonMask() & 0x7; + S32 lt = pos.lessEqual(mD[MIN]).getGatheredBits() & 0x7; if (lt) { return false; @@ -206,13 +206,13 @@ public: { const LLVector4a& pos = data->getPositionGroup(); - LLVector4a gt = pos.greaterThan4(center); + LLVector4a gt = pos.greaterThan(center); LLVector4a up; - up.mQ = _mm_and_ps(size.mQ, gt.mQ); + up = _mm_and_ps(size, gt); LLVector4a down; - down.mQ = _mm_andnot_ps(gt.mQ, size.mQ); + down = _mm_andnot_ps(gt, size); center.add(up); center.sub(down); @@ -326,9 +326,8 @@ public: LLVector4a val; val.setSub(center, getCenter()); val.setAbs(val); - LLVector4a app_zero; - app_zero.mQ = F_APPROXIMATELY_ZERO_4A; - S32 lt = val.lessThan4(app_zero).getComparisonMask() & 0x7; + + S32 lt = val.lessThan(LLVector4a::getEpsilon()).getGatheredBits() & 0x7; if( lt == 0x7 ) { @@ -642,7 +641,7 @@ public: LLVector4a val; val.setSub(v, BaseType::mD[BaseType::CENTER]); val.setAbs(val); - S32 lt = val.lessThan4(MAX_MAG).getComparisonMask() & 0x7; + S32 lt = val.lessThan(MAX_MAG).getGatheredBits() & 0x7; if (lt != 0x7) { diff --git a/indra/llmath/llquantize.h b/indra/llmath/llquantize.h index 2192427f07..000d8a060f 100644 --- a/indra/llmath/llquantize.h +++ b/indra/llmath/llquantize.h @@ -1,152 +1,158 @@ -/** - * @file llquantize.h - * @brief useful routines for quantizing floats to various length ints - * and back out again - * - * $LicenseInfo:firstyear=2001&license=viewergpl$ - * - * Copyright (c) 2001-2009, Linden Research, Inc. - * - * Second Life Viewer Source Code - * The source code in this file ("Source Code") is provided by Linden Lab - * to you under the terms of the GNU General Public License, version 2.0 - * ("GPL"), unless you have obtained a separate licensing agreement - * ("Other License"), formally executed by you and Linden Lab. Terms of - * the GPL can be found in doc/GPL-license.txt in this distribution, or - * online at http://secondlifegrid.net/programs/open_source/licensing/gplv2 - * - * There are special exceptions to the terms and conditions of the GPL as - * it is applied to this Source Code. View the full text of the exception - * in the file doc/FLOSS-exception.txt in this software distribution, or - * online at - * http://secondlifegrid.net/programs/open_source/licensing/flossexception - * - * By copying, modifying or distributing this software, you acknowledge - * that you have read and understood your obligations described above, - * and agree to abide by those obligations. - * - * ALL LINDEN LAB SOURCE CODE IS PROVIDED "AS IS." LINDEN LAB MAKES NO - * WARRANTIES, EXPRESS, IMPLIED OR OTHERWISE, REGARDING ITS ACCURACY, - * COMPLETENESS OR PERFORMANCE. - * $/LicenseInfo$ - */ - -#ifndef LL_LLQUANTIZE_H -#define LL_LLQUANTIZE_H - -const U16 U16MAX = 65535; -const F32 OOU16MAX = 1.f/(F32)(U16MAX); - -const U8 U8MAX = 255; -const F32 OOU8MAX = 1.f/(F32)(U8MAX); - -const U8 FIRSTVALIDCHAR = 54; -const U8 MAXSTRINGVAL = U8MAX - FIRSTVALIDCHAR; //we don't allow newline or null - - -inline U16 F32_to_U16_ROUND(F32 val, F32 lower, F32 upper) -{ - val = llclamp(val, lower, upper); - // make sure that the value is positive and normalized to <0, 1> - val -= lower; - val /= (upper - lower); - - // round the value. Sreturn the U16 - return (U16)(llround(val*U16MAX)); -} - - -inline U16 F32_to_U16(F32 val, F32 lower, F32 upper) -{ - val = llclamp(val, lower, upper); - // make sure that the value is positive and normalized to <0, 1> - val -= lower; - val /= (upper - lower); - - // return the U16 - return (U16)(llfloor(val*U16MAX)); -} - -inline F32 U16_to_F32(U16 ival, F32 lower, F32 upper) -{ - F32 val = ival*OOU16MAX; - F32 delta = (upper - lower); - val *= delta; - val += lower; - - F32 max_error = delta*OOU16MAX; - - // make sure that zero's come through as zero - if (fabsf(val) < max_error) - val = 0.f; - - return val; -} - - -inline U8 F32_to_U8_ROUND(F32 val, F32 lower, F32 upper) -{ - val = llclamp(val, lower, upper); - // make sure that the value is positive and normalized to <0, 1> - val -= lower; - val /= (upper - lower); - - // return the rounded U8 - return (U8)(llround(val*U8MAX)); -} - - -inline U8 F32_to_U8(F32 val, F32 lower, F32 upper) -{ - val = llclamp(val, lower, upper); - // make sure that the value is positive and normalized to <0, 1> - val -= lower; - val /= (upper - lower); - - // return the U8 - return (U8)(llfloor(val*U8MAX)); -} - -inline F32 U8_to_F32(U8 ival, F32 lower, F32 upper) -{ - F32 val = ival*OOU8MAX; - F32 delta = (upper - lower); - val *= delta; - val += lower; - - F32 max_error = delta*OOU8MAX; - - // make sure that zero's come through as zero - if (fabsf(val) < max_error) - val = 0.f; - - return val; -} - -inline U8 F32_TO_STRING(F32 val, F32 lower, F32 upper) -{ - val = llclamp(val, lower, upper); //[lower, upper] - // make sure that the value is positive and normalized to <0, 1> - val -= lower; //[0, upper-lower] - val /= (upper - lower); //[0,1] - val = val * MAXSTRINGVAL; //[0, MAXSTRINGVAL] - val = floor(val + 0.5f); //[0, MAXSTRINGVAL] - - U8 stringVal = (U8)(val) + FIRSTVALIDCHAR; //[FIRSTVALIDCHAR, MAXSTRINGVAL + FIRSTVALIDCHAR] - return stringVal; -} - -inline F32 STRING_TO_F32(U8 ival, F32 lower, F32 upper) -{ - // remove empty space left for NULL, newline, etc. - ival -= FIRSTVALIDCHAR; //[0, MAXSTRINGVAL] - - F32 val = (F32)ival * (1.f / (F32)MAXSTRINGVAL); //[0, 1] - F32 delta = (upper - lower); - val *= delta; //[0, upper - lower] - val += lower; //[lower, upper] - - return val; -} - -#endif +/** + * @file llquantize.h + * @brief useful routines for quantizing floats to various length ints + * and back out again + * + * $LicenseInfo:firstyear=2001&license=viewergpl$ + * + * Copyright (c) 2001-2009, Linden Research, Inc. + * + * Second Life Viewer Source Code + * The source code in this file ("Source Code") is provided by Linden Lab + * to you under the terms of the GNU General Public License, version 2.0 + * ("GPL"), unless you have obtained a separate licensing agreement + * ("Other License"), formally executed by you and Linden Lab. Terms of + * the GPL can be found in doc/GPL-license.txt in this distribution, or + * online at http://secondlifegrid.net/programs/open_source/licensing/gplv2 + * + * There are special exceptions to the terms and conditions of the GPL as + * it is applied to this Source Code. View the full text of the exception + * in the file doc/FLOSS-exception.txt in this software distribution, or + * online at + * http://secondlifegrid.net/programs/open_source/licensing/flossexception + * + * By copying, modifying or distributing this software, you acknowledge + * that you have read and understood your obligations described above, + * and agree to abide by those obligations. + * + * ALL LINDEN LAB SOURCE CODE IS PROVIDED "AS IS." LINDEN LAB MAKES NO + * WARRANTIES, EXPRESS, IMPLIED OR OTHERWISE, REGARDING ITS ACCURACY, + * COMPLETENESS OR PERFORMANCE. + * $/LicenseInfo$ + */ + +#ifndef LL_LLQUANTIZE_H +#define LL_LLQUANTIZE_H + +const U16 U16MAX = 65535; +LL_ALIGN_16( const F32 F_U16MAX_4A[4] ) = { 65535.f, 65535.f, 65535.f, 65535.f }; + +const F32 OOU16MAX = 1.f/(F32)(U16MAX); +LL_ALIGN_16( const F32 F_OOU16MAX_4A[4] ) = { OOU16MAX, OOU16MAX, OOU16MAX, OOU16MAX }; + +const U8 U8MAX = 255; +LL_ALIGN_16( const F32 F_U8MAX_4A[4] ) = { 255.f, 255.f, 255.f, 255.f }; + +const F32 OOU8MAX = 1.f/(F32)(U8MAX); +LL_ALIGN_16( const F32 F_OOU8MAX_4A[4] ) = { OOU8MAX, OOU8MAX, OOU8MAX, OOU8MAX }; + +const U8 FIRSTVALIDCHAR = 54; +const U8 MAXSTRINGVAL = U8MAX - FIRSTVALIDCHAR; //we don't allow newline or null + + +inline U16 F32_to_U16_ROUND(F32 val, F32 lower, F32 upper) +{ + val = llclamp(val, lower, upper); + // make sure that the value is positive and normalized to <0, 1> + val -= lower; + val /= (upper - lower); + + // round the value. Sreturn the U16 + return (U16)(llround(val*U16MAX)); +} + + +inline U16 F32_to_U16(F32 val, F32 lower, F32 upper) +{ + val = llclamp(val, lower, upper); + // make sure that the value is positive and normalized to <0, 1> + val -= lower; + val /= (upper - lower); + + // return the U16 + return (U16)(llfloor(val*U16MAX)); +} + +inline F32 U16_to_F32(U16 ival, F32 lower, F32 upper) +{ + F32 val = ival*OOU16MAX; + F32 delta = (upper - lower); + val *= delta; + val += lower; + + F32 max_error = delta*OOU16MAX; + + // make sure that zero's come through as zero + if (fabsf(val) < max_error) + val = 0.f; + + return val; +} + + +inline U8 F32_to_U8_ROUND(F32 val, F32 lower, F32 upper) +{ + val = llclamp(val, lower, upper); + // make sure that the value is positive and normalized to <0, 1> + val -= lower; + val /= (upper - lower); + + // return the rounded U8 + return (U8)(llround(val*U8MAX)); +} + + +inline U8 F32_to_U8(F32 val, F32 lower, F32 upper) +{ + val = llclamp(val, lower, upper); + // make sure that the value is positive and normalized to <0, 1> + val -= lower; + val /= (upper - lower); + + // return the U8 + return (U8)(llfloor(val*U8MAX)); +} + +inline F32 U8_to_F32(U8 ival, F32 lower, F32 upper) +{ + F32 val = ival*OOU8MAX; + F32 delta = (upper - lower); + val *= delta; + val += lower; + + F32 max_error = delta*OOU8MAX; + + // make sure that zero's come through as zero + if (fabsf(val) < max_error) + val = 0.f; + + return val; +} + +inline U8 F32_TO_STRING(F32 val, F32 lower, F32 upper) +{ + val = llclamp(val, lower, upper); //[lower, upper] + // make sure that the value is positive and normalized to <0, 1> + val -= lower; //[0, upper-lower] + val /= (upper - lower); //[0,1] + val = val * MAXSTRINGVAL; //[0, MAXSTRINGVAL] + val = floor(val + 0.5f); //[0, MAXSTRINGVAL] + + U8 stringVal = (U8)(val) + FIRSTVALIDCHAR; //[FIRSTVALIDCHAR, MAXSTRINGVAL + FIRSTVALIDCHAR] + return stringVal; +} + +inline F32 STRING_TO_F32(U8 ival, F32 lower, F32 upper) +{ + // remove empty space left for NULL, newline, etc. + ival -= FIRSTVALIDCHAR; //[0, MAXSTRINGVAL] + + F32 val = (F32)ival * (1.f / (F32)MAXSTRINGVAL); //[0, 1] + F32 delta = (upper - lower); + val *= delta; //[0, upper - lower] + val += lower; //[lower, upper] + + return val; +} + +#endif diff --git a/indra/llmath/llquaternion.cpp b/indra/llmath/llquaternion.cpp index fdcc19d657..efdc10e2c6 100644 --- a/indra/llmath/llquaternion.cpp +++ b/indra/llmath/llquaternion.cpp @@ -1,960 +1,961 @@ -/** - * @file llquaternion.cpp - * @brief LLQuaternion class implementation. - * - * $LicenseInfo:firstyear=2000&license=viewergpl$ - * - * Copyright (c) 2000-2009, Linden Research, Inc. - * - * Second Life Viewer Source Code - * The source code in this file ("Source Code") is provided by Linden Lab - * to you under the terms of the GNU General Public License, version 2.0 - * ("GPL"), unless you have obtained a separate licensing agreement - * ("Other License"), formally executed by you and Linden Lab. Terms of - * the GPL can be found in doc/GPL-license.txt in this distribution, or - * online at http://secondlifegrid.net/programs/open_source/licensing/gplv2 - * - * There are special exceptions to the terms and conditions of the GPL as - * it is applied to this Source Code. View the full text of the exception - * in the file doc/FLOSS-exception.txt in this software distribution, or - * online at - * http://secondlifegrid.net/programs/open_source/licensing/flossexception - * - * By copying, modifying or distributing this software, you acknowledge - * that you have read and understood your obligations described above, - * and agree to abide by those obligations. - * - * ALL LINDEN LAB SOURCE CODE IS PROVIDED "AS IS." LINDEN LAB MAKES NO - * WARRANTIES, EXPRESS, IMPLIED OR OTHERWISE, REGARDING ITS ACCURACY, - * COMPLETENESS OR PERFORMANCE. - * $/LicenseInfo$ - */ - -#include "linden_common.h" - -#include "llquaternion.h" - -#include "llmath.h" // for F_PI -//#include "vmath.h" -#include "v3math.h" -#include "v3dmath.h" -#include "v4math.h" -#include "m4math.h" -#include "m3math.h" -#include "llquantize.h" - -// WARNING: Don't use this for global const definitions! using this -// at the top of a *.cpp file might not give you what you think. -const LLQuaternion LLQuaternion::DEFAULT; - -// Constructors - -LLQuaternion::LLQuaternion(const LLMatrix4 &mat) -{ - *this = mat.quaternion(); - normalize(); -} - -LLQuaternion::LLQuaternion(const LLMatrix3 &mat) -{ - *this = mat.quaternion(); - normalize(); -} - -LLQuaternion::LLQuaternion(F32 angle, const LLVector4 &vec) -{ - LLVector3 v(vec.mV[VX], vec.mV[VY], vec.mV[VZ]); - v.normalize(); - - F32 c, s; - c = cosf(angle*0.5f); - s = sinf(angle*0.5f); - - mQ[VX] = v.mV[VX] * s; - mQ[VY] = v.mV[VY] * s; - mQ[VZ] = v.mV[VZ] * s; - mQ[VW] = c; - normalize(); -} - -LLQuaternion::LLQuaternion(F32 angle, const LLVector3 &vec) -{ - LLVector3 v(vec); - v.normalize(); - - F32 c, s; - c = cosf(angle*0.5f); - s = sinf(angle*0.5f); - - mQ[VX] = v.mV[VX] * s; - mQ[VY] = v.mV[VY] * s; - mQ[VZ] = v.mV[VZ] * s; - mQ[VW] = c; - normalize(); -} - -LLQuaternion::LLQuaternion(const LLVector3 &x_axis, - const LLVector3 &y_axis, - const LLVector3 &z_axis) -{ - LLMatrix3 mat; - mat.setRows(x_axis, y_axis, z_axis); - *this = mat.quaternion(); - normalize(); -} - -// Quatizations -void LLQuaternion::quantize16(F32 lower, F32 upper) -{ - F32 x = mQ[VX]; - F32 y = mQ[VY]; - F32 z = mQ[VZ]; - F32 s = mQ[VS]; - - x = U16_to_F32(F32_to_U16_ROUND(x, lower, upper), lower, upper); - y = U16_to_F32(F32_to_U16_ROUND(y, lower, upper), lower, upper); - z = U16_to_F32(F32_to_U16_ROUND(z, lower, upper), lower, upper); - s = U16_to_F32(F32_to_U16_ROUND(s, lower, upper), lower, upper); - - mQ[VX] = x; - mQ[VY] = y; - mQ[VZ] = z; - mQ[VS] = s; - - normalize(); -} - -void LLQuaternion::quantize8(F32 lower, F32 upper) -{ - mQ[VX] = U8_to_F32(F32_to_U8_ROUND(mQ[VX], lower, upper), lower, upper); - mQ[VY] = U8_to_F32(F32_to_U8_ROUND(mQ[VY], lower, upper), lower, upper); - mQ[VZ] = U8_to_F32(F32_to_U8_ROUND(mQ[VZ], lower, upper), lower, upper); - mQ[VS] = U8_to_F32(F32_to_U8_ROUND(mQ[VS], lower, upper), lower, upper); - - normalize(); -} - -// LLVector3 Magnitude and Normalization Functions - - -// Set LLQuaternion routines - -const LLQuaternion& LLQuaternion::setAngleAxis(F32 angle, F32 x, F32 y, F32 z) -{ - LLVector3 vec(x, y, z); - vec.normalize(); - - angle *= 0.5f; - F32 c, s; - c = cosf(angle); - s = sinf(angle); - - mQ[VX] = vec.mV[VX]*s; - mQ[VY] = vec.mV[VY]*s; - mQ[VZ] = vec.mV[VZ]*s; - mQ[VW] = c; - - normalize(); - return (*this); -} - -const LLQuaternion& LLQuaternion::setAngleAxis(F32 angle, const LLVector3 &vec) -{ - LLVector3 v(vec); - v.normalize(); - - angle *= 0.5f; - F32 c, s; - c = cosf(angle); - s = sinf(angle); - - mQ[VX] = v.mV[VX]*s; - mQ[VY] = v.mV[VY]*s; - mQ[VZ] = v.mV[VZ]*s; - mQ[VW] = c; - - normalize(); - return (*this); -} - -const LLQuaternion& LLQuaternion::setAngleAxis(F32 angle, const LLVector4 &vec) -{ - LLVector3 v(vec.mV[VX], vec.mV[VY], vec.mV[VZ]); - v.normalize(); - - F32 c, s; - c = cosf(angle*0.5f); - s = sinf(angle*0.5f); - - mQ[VX] = v.mV[VX]*s; - mQ[VY] = v.mV[VY]*s; - mQ[VZ] = v.mV[VZ]*s; - mQ[VW] = c; - - normalize(); - return (*this); -} - -const LLQuaternion& LLQuaternion::setEulerAngles(F32 roll, F32 pitch, F32 yaw) -{ - LLMatrix3 rot_mat(roll, pitch, yaw); - rot_mat.orthogonalize(); - *this = rot_mat.quaternion(); - - normalize(); - return (*this); -} - -// deprecated -const LLQuaternion& LLQuaternion::set(const LLMatrix3 &mat) -{ - *this = mat.quaternion(); - normalize(); - return (*this); -} - -// deprecated -const LLQuaternion& LLQuaternion::set(const LLMatrix4 &mat) -{ - *this = mat.quaternion(); - normalize(); - return (*this); -} - -// deprecated -const LLQuaternion& LLQuaternion::setQuat(F32 angle, F32 x, F32 y, F32 z) -{ - LLVector3 vec(x, y, z); - vec.normalize(); - - angle *= 0.5f; - F32 c, s; - c = cosf(angle); - s = sinf(angle); - - mQ[VX] = vec.mV[VX]*s; - mQ[VY] = vec.mV[VY]*s; - mQ[VZ] = vec.mV[VZ]*s; - mQ[VW] = c; - - normalize(); - return (*this); -} - -// deprecated -const LLQuaternion& LLQuaternion::setQuat(F32 angle, const LLVector3 &vec) -{ - LLVector3 v(vec); - v.normalize(); - - angle *= 0.5f; - F32 c, s; - c = cosf(angle); - s = sinf(angle); - - mQ[VX] = v.mV[VX]*s; - mQ[VY] = v.mV[VY]*s; - mQ[VZ] = v.mV[VZ]*s; - mQ[VW] = c; - - normalize(); - return (*this); -} - -const LLQuaternion& LLQuaternion::setQuat(F32 angle, const LLVector4 &vec) -{ - LLVector3 v(vec.mV[VX], vec.mV[VY], vec.mV[VZ]); - v.normalize(); - - F32 c, s; - c = cosf(angle*0.5f); - s = sinf(angle*0.5f); - - mQ[VX] = v.mV[VX]*s; - mQ[VY] = v.mV[VY]*s; - mQ[VZ] = v.mV[VZ]*s; - mQ[VW] = c; - - normalize(); - return (*this); -} - -const LLQuaternion& LLQuaternion::setQuat(F32 roll, F32 pitch, F32 yaw) -{ - LLMatrix3 rot_mat(roll, pitch, yaw); - rot_mat.orthogonalize(); - *this = rot_mat.quaternion(); - - normalize(); - return (*this); -} - -const LLQuaternion& LLQuaternion::setQuat(const LLMatrix3 &mat) -{ - *this = mat.quaternion(); - normalize(); - return (*this); -} - -const LLQuaternion& LLQuaternion::setQuat(const LLMatrix4 &mat) -{ - *this = mat.quaternion(); - normalize(); - return (*this); -//#if 1 -// // NOTE: LLQuaternion's are actually inverted with respect to -// // the matrices, so this code also assumes inverted quaternions -// // (-x, -y, -z, w). The result is that roll,pitch,yaw are applied -// // in reverse order (yaw,pitch,roll). -// F64 cosX = cos(roll); -// F64 cosY = cos(pitch); -// F64 cosZ = cos(yaw); -// -// F64 sinX = sin(roll); -// F64 sinY = sin(pitch); -// F64 sinZ = sin(yaw); -// -// mQ[VW] = (F32)sqrt(cosY*cosZ - sinX*sinY*sinZ + cosX*cosZ + cosX*cosY + 1.0)*.5; -// if (fabs(mQ[VW]) < F_APPROXIMATELY_ZERO) -// { -// // null rotation, any axis will do -// mQ[VX] = 0.0f; -// mQ[VY] = 1.0f; -// mQ[VZ] = 0.0f; -// } -// else -// { -// F32 inv_s = 1.0f / (4.0f * mQ[VW]); -// mQ[VX] = (F32)-(-sinX*cosY - cosX*sinY*sinZ - sinX*cosZ) * inv_s; -// mQ[VY] = (F32)-(-cosX*sinY*cosZ + sinX*sinZ - sinY) * inv_s; -// mQ[VZ] = (F32)-(-cosY*sinZ - sinX*sinY*cosZ - cosX*sinZ) * inv_s; -// } -// -//#else // This only works on a certain subset of roll/pitch/yaw -// -// F64 cosX = cosf(roll/2.0); -// F64 cosY = cosf(pitch/2.0); -// F64 cosZ = cosf(yaw/2.0); -// -// F64 sinX = sinf(roll/2.0); -// F64 sinY = sinf(pitch/2.0); -// F64 sinZ = sinf(yaw/2.0); -// -// mQ[VW] = (F32)(cosX*cosY*cosZ + sinX*sinY*sinZ); -// mQ[VX] = (F32)(sinX*cosY*cosZ - cosX*sinY*sinZ); -// mQ[VY] = (F32)(cosX*sinY*cosZ + sinX*cosY*sinZ); -// mQ[VZ] = (F32)(cosX*cosY*sinZ - sinX*sinY*cosZ); -//#endif -// -// normalize(); -// return (*this); -} - -// SJB: This code is correct for a logicly stored (non-transposed) matrix; -// Our matrices are stored transposed, OpenGL style, so this generates the -// INVERSE matrix, or the CORRECT matrix form an INVERSE quaternion. -// Because we use similar logic in LLMatrix3::quaternion(), -// we are internally consistant so everything works OK :) -LLMatrix3 LLQuaternion::getMatrix3(void) const -{ - LLMatrix3 mat; - F32 xx, xy, xz, xw, yy, yz, yw, zz, zw; - - xx = mQ[VX] * mQ[VX]; - xy = mQ[VX] * mQ[VY]; - xz = mQ[VX] * mQ[VZ]; - xw = mQ[VX] * mQ[VW]; - - yy = mQ[VY] * mQ[VY]; - yz = mQ[VY] * mQ[VZ]; - yw = mQ[VY] * mQ[VW]; - - zz = mQ[VZ] * mQ[VZ]; - zw = mQ[VZ] * mQ[VW]; - - mat.mMatrix[0][0] = 1.f - 2.f * ( yy + zz ); - mat.mMatrix[0][1] = 2.f * ( xy + zw ); - mat.mMatrix[0][2] = 2.f * ( xz - yw ); - - mat.mMatrix[1][0] = 2.f * ( xy - zw ); - mat.mMatrix[1][1] = 1.f - 2.f * ( xx + zz ); - mat.mMatrix[1][2] = 2.f * ( yz + xw ); - - mat.mMatrix[2][0] = 2.f * ( xz + yw ); - mat.mMatrix[2][1] = 2.f * ( yz - xw ); - mat.mMatrix[2][2] = 1.f - 2.f * ( xx + yy ); - - return mat; -} - -LLMatrix4 LLQuaternion::getMatrix4(void) const -{ - LLMatrix4 mat; - F32 xx, xy, xz, xw, yy, yz, yw, zz, zw; - - xx = mQ[VX] * mQ[VX]; - xy = mQ[VX] * mQ[VY]; - xz = mQ[VX] * mQ[VZ]; - xw = mQ[VX] * mQ[VW]; - - yy = mQ[VY] * mQ[VY]; - yz = mQ[VY] * mQ[VZ]; - yw = mQ[VY] * mQ[VW]; - - zz = mQ[VZ] * mQ[VZ]; - zw = mQ[VZ] * mQ[VW]; - - mat.mMatrix[0][0] = 1.f - 2.f * ( yy + zz ); - mat.mMatrix[0][1] = 2.f * ( xy + zw ); - mat.mMatrix[0][2] = 2.f * ( xz - yw ); - - mat.mMatrix[1][0] = 2.f * ( xy - zw ); - mat.mMatrix[1][1] = 1.f - 2.f * ( xx + zz ); - mat.mMatrix[1][2] = 2.f * ( yz + xw ); - - mat.mMatrix[2][0] = 2.f * ( xz + yw ); - mat.mMatrix[2][1] = 2.f * ( yz - xw ); - mat.mMatrix[2][2] = 1.f - 2.f * ( xx + yy ); - - // TODO -- should we set the translation portion to zero? - - return mat; -} - - - - -// Other useful methods - - -// calculate the shortest rotation from a to b -void LLQuaternion::shortestArc(const LLVector3 &a, const LLVector3 &b) -{ - // Make a local copy of both vectors. - LLVector3 vec_a = a; - LLVector3 vec_b = b; - - // Make sure neither vector is zero length. Also normalize - // the vectors while we are at it. - F32 vec_a_mag = vec_a.normalize(); - F32 vec_b_mag = vec_b.normalize(); - if (vec_a_mag < F_APPROXIMATELY_ZERO || - vec_b_mag < F_APPROXIMATELY_ZERO) - { - // Can't calculate a rotation from this. - // Just return ZERO_ROTATION instead. - loadIdentity(); - return; - } - - // Create an axis to rotate around, and the cos of the angle to rotate. - LLVector3 axis = vec_a % vec_b; - F32 cos_theta = vec_a * vec_b; - - // Check the angle between the vectors to see if they are parallel or anti-parallel. - if (cos_theta > 1.0 - F_APPROXIMATELY_ZERO) - { - // a and b are parallel. No rotation is necessary. - loadIdentity(); - } - else if (cos_theta < -1.0 + F_APPROXIMATELY_ZERO) - { - // a and b are anti-parallel. - // Rotate 180 degrees around some orthogonal axis. - // Find the projection of the x-axis onto a, and try - // using the vector between the projection and the x-axis - // as the orthogonal axis. - LLVector3 proj = vec_a.mV[VX] / (vec_a * vec_a) * vec_a; - LLVector3 ortho_axis(1.f, 0.f, 0.f); - ortho_axis -= proj; - - // Turn this into an orthonormal axis. - F32 ortho_length = ortho_axis.normalize(); - // If the axis' length is 0, then our guess at an orthogonal axis - // was wrong (a is parallel to the x-axis). - if (ortho_length < F_APPROXIMATELY_ZERO) - { - // Use the z-axis instead. - ortho_axis.setVec(0.f, 0.f, 1.f); - } - - // Construct a quaternion from this orthonormal axis. - mQ[VX] = ortho_axis.mV[VX]; - mQ[VY] = ortho_axis.mV[VY]; - mQ[VZ] = ortho_axis.mV[VZ]; - mQ[VW] = 0.f; - } - else - { - // a and b are NOT parallel or anti-parallel. - // Return the rotation between these vectors. - F32 theta = (F32)acos(cos_theta); - - setAngleAxis(theta, axis); - } -} - -// constrains rotation to a cone angle specified in radians -const LLQuaternion &LLQuaternion::constrain(F32 radians) -{ - const F32 cos_angle_lim = cosf( radians/2 ); // mQ[VW] limit - const F32 sin_angle_lim = sinf( radians/2 ); // rotation axis length limit - - if (mQ[VW] < 0.f) - { - mQ[VX] *= -1.f; - mQ[VY] *= -1.f; - mQ[VZ] *= -1.f; - mQ[VW] *= -1.f; - } - - // if rotation angle is greater than limit (cos is less than limit) - if( mQ[VW] < cos_angle_lim ) - { - mQ[VW] = cos_angle_lim; - F32 axis_len = sqrtf( mQ[VX]*mQ[VX] + mQ[VY]*mQ[VY] + mQ[VZ]*mQ[VZ] ); // sin(theta/2) - F32 axis_mult_fact = sin_angle_lim / axis_len; - mQ[VX] *= axis_mult_fact; - mQ[VY] *= axis_mult_fact; - mQ[VZ] *= axis_mult_fact; - } - - return *this; -} - -// Operators - -std::ostream& operator<<(std::ostream &s, const LLQuaternion &a) -{ - s << "{ " - << a.mQ[VX] << ", " << a.mQ[VY] << ", " << a.mQ[VZ] << ", " << a.mQ[VW] - << " }"; - return s; -} - - -// Does NOT renormalize the result -LLQuaternion operator*(const LLQuaternion &a, const LLQuaternion &b) -{ -// LLQuaternion::mMultCount++; - - LLQuaternion q( - b.mQ[3] * a.mQ[0] + b.mQ[0] * a.mQ[3] + b.mQ[1] * a.mQ[2] - b.mQ[2] * a.mQ[1], - b.mQ[3] * a.mQ[1] + b.mQ[1] * a.mQ[3] + b.mQ[2] * a.mQ[0] - b.mQ[0] * a.mQ[2], - b.mQ[3] * a.mQ[2] + b.mQ[2] * a.mQ[3] + b.mQ[0] * a.mQ[1] - b.mQ[1] * a.mQ[0], - b.mQ[3] * a.mQ[3] - b.mQ[0] * a.mQ[0] - b.mQ[1] * a.mQ[1] - b.mQ[2] * a.mQ[2] - ); - return q; -} - -/* -LLMatrix4 operator*(const LLMatrix4 &m, const LLQuaternion &q) -{ - LLMatrix4 qmat(q); - return (m*qmat); -} -*/ - - - -LLVector4 operator*(const LLVector4 &a, const LLQuaternion &rot) -{ - F32 rw = - rot.mQ[VX] * a.mV[VX] - rot.mQ[VY] * a.mV[VY] - rot.mQ[VZ] * a.mV[VZ]; - F32 rx = rot.mQ[VW] * a.mV[VX] + rot.mQ[VY] * a.mV[VZ] - rot.mQ[VZ] * a.mV[VY]; - F32 ry = rot.mQ[VW] * a.mV[VY] + rot.mQ[VZ] * a.mV[VX] - rot.mQ[VX] * a.mV[VZ]; - F32 rz = rot.mQ[VW] * a.mV[VZ] + rot.mQ[VX] * a.mV[VY] - rot.mQ[VY] * a.mV[VX]; - - F32 nx = - rw * rot.mQ[VX] + rx * rot.mQ[VW] - ry * rot.mQ[VZ] + rz * rot.mQ[VY]; - F32 ny = - rw * rot.mQ[VY] + ry * rot.mQ[VW] - rz * rot.mQ[VX] + rx * rot.mQ[VZ]; - F32 nz = - rw * rot.mQ[VZ] + rz * rot.mQ[VW] - rx * rot.mQ[VY] + ry * rot.mQ[VX]; - - return LLVector4(nx, ny, nz, a.mV[VW]); -} - -LLVector3 operator*(const LLVector3 &a, const LLQuaternion &rot) -{ - F32 rw = - rot.mQ[VX] * a.mV[VX] - rot.mQ[VY] * a.mV[VY] - rot.mQ[VZ] * a.mV[VZ]; - F32 rx = rot.mQ[VW] * a.mV[VX] + rot.mQ[VY] * a.mV[VZ] - rot.mQ[VZ] * a.mV[VY]; - F32 ry = rot.mQ[VW] * a.mV[VY] + rot.mQ[VZ] * a.mV[VX] - rot.mQ[VX] * a.mV[VZ]; - F32 rz = rot.mQ[VW] * a.mV[VZ] + rot.mQ[VX] * a.mV[VY] - rot.mQ[VY] * a.mV[VX]; - - F32 nx = - rw * rot.mQ[VX] + rx * rot.mQ[VW] - ry * rot.mQ[VZ] + rz * rot.mQ[VY]; - F32 ny = - rw * rot.mQ[VY] + ry * rot.mQ[VW] - rz * rot.mQ[VX] + rx * rot.mQ[VZ]; - F32 nz = - rw * rot.mQ[VZ] + rz * rot.mQ[VW] - rx * rot.mQ[VY] + ry * rot.mQ[VX]; - - return LLVector3(nx, ny, nz); -} - -LLVector3d operator*(const LLVector3d &a, const LLQuaternion &rot) -{ - F64 rw = - rot.mQ[VX] * a.mdV[VX] - rot.mQ[VY] * a.mdV[VY] - rot.mQ[VZ] * a.mdV[VZ]; - F64 rx = rot.mQ[VW] * a.mdV[VX] + rot.mQ[VY] * a.mdV[VZ] - rot.mQ[VZ] * a.mdV[VY]; - F64 ry = rot.mQ[VW] * a.mdV[VY] + rot.mQ[VZ] * a.mdV[VX] - rot.mQ[VX] * a.mdV[VZ]; - F64 rz = rot.mQ[VW] * a.mdV[VZ] + rot.mQ[VX] * a.mdV[VY] - rot.mQ[VY] * a.mdV[VX]; - - F64 nx = - rw * rot.mQ[VX] + rx * rot.mQ[VW] - ry * rot.mQ[VZ] + rz * rot.mQ[VY]; - F64 ny = - rw * rot.mQ[VY] + ry * rot.mQ[VW] - rz * rot.mQ[VX] + rx * rot.mQ[VZ]; - F64 nz = - rw * rot.mQ[VZ] + rz * rot.mQ[VW] - rx * rot.mQ[VY] + ry * rot.mQ[VX]; - - return LLVector3d(nx, ny, nz); -} - -F32 dot(const LLQuaternion &a, const LLQuaternion &b) -{ - return a.mQ[VX] * b.mQ[VX] + - a.mQ[VY] * b.mQ[VY] + - a.mQ[VZ] * b.mQ[VZ] + - a.mQ[VW] * b.mQ[VW]; -} - -// DEMO HACK: This lerp is probably inocrrect now due intermediate normalization -// it should look more like the lerp below -#if 0 -// linear interpolation -LLQuaternion lerp(F32 t, const LLQuaternion &p, const LLQuaternion &q) -{ - LLQuaternion r; - r = t * (q - p) + p; - r.normalize(); - return r; -} -#endif - -// lerp from identity to q -LLQuaternion lerp(F32 t, const LLQuaternion &q) -{ - LLQuaternion r; - r.mQ[VX] = t * q.mQ[VX]; - r.mQ[VY] = t * q.mQ[VY]; - r.mQ[VZ] = t * q.mQ[VZ]; - r.mQ[VW] = t * (q.mQ[VZ] - 1.f) + 1.f; - r.normalize(); - return r; -} - -LLQuaternion lerp(F32 t, const LLQuaternion &p, const LLQuaternion &q) -{ - LLQuaternion r; - F32 inv_t; - - inv_t = 1.f - t; - - r.mQ[VX] = t * q.mQ[VX] + (inv_t * p.mQ[VX]); - r.mQ[VY] = t * q.mQ[VY] + (inv_t * p.mQ[VY]); - r.mQ[VZ] = t * q.mQ[VZ] + (inv_t * p.mQ[VZ]); - r.mQ[VW] = t * q.mQ[VW] + (inv_t * p.mQ[VW]); - r.normalize(); - return r; -} - - -// spherical linear interpolation -LLQuaternion slerp( F32 u, const LLQuaternion &a, const LLQuaternion &b ) -{ - // cosine theta = dot product of a and b - F32 cos_t = a.mQ[0]*b.mQ[0] + a.mQ[1]*b.mQ[1] + a.mQ[2]*b.mQ[2] + a.mQ[3]*b.mQ[3]; - - // if b is on opposite hemisphere from a, use -a instead - int bflip; - if (cos_t < 0.0f) - { - cos_t = -cos_t; - bflip = TRUE; - } - else - bflip = FALSE; - - // if B is (within precision limits) the same as A, - // just linear interpolate between A and B. - F32 alpha; // interpolant - F32 beta; // 1 - interpolant - if (1.0f - cos_t < 0.00001f) - { - beta = 1.0f - u; - alpha = u; - } - else - { - F32 theta = acosf(cos_t); - F32 sin_t = sinf(theta); - beta = sinf(theta - u*theta) / sin_t; - alpha = sinf(u*theta) / sin_t; - } - - if (bflip) - beta = -beta; - - // interpolate - LLQuaternion ret; - ret.mQ[0] = beta*a.mQ[0] + alpha*b.mQ[0]; - ret.mQ[1] = beta*a.mQ[1] + alpha*b.mQ[1]; - ret.mQ[2] = beta*a.mQ[2] + alpha*b.mQ[2]; - ret.mQ[3] = beta*a.mQ[3] + alpha*b.mQ[3]; - - return ret; -} - -// lerp whenever possible -LLQuaternion nlerp(F32 t, const LLQuaternion &a, const LLQuaternion &b) -{ - if (dot(a, b) < 0.f) - { - return slerp(t, a, b); - } - else - { - return lerp(t, a, b); - } -} - -LLQuaternion nlerp(F32 t, const LLQuaternion &q) -{ - if (q.mQ[VW] < 0.f) - { - return slerp(t, q); - } - else - { - return lerp(t, q); - } -} - -// slerp from identity quaternion to another quaternion -LLQuaternion slerp(F32 t, const LLQuaternion &q) -{ - F32 c = q.mQ[VW]; - if (1.0f == t || 1.0f == c) - { - // the trivial cases - return q; - } - - LLQuaternion r; - F32 s, angle, stq, stp; - - s = (F32) sqrt(1.f - c*c); - - if (c < 0.0f) - { - // when c < 0.0 then theta > PI/2 - // since quat and -quat are the same rotation we invert one of - // p or q to reduce unecessary spins - // A equivalent way to do it is to convert acos(c) as if it had - // been negative, and to negate stp - angle = (F32) acos(-c); - stp = -(F32) sin(angle * (1.f - t)); - stq = (F32) sin(angle * t); - } - else - { - angle = (F32) acos(c); - stp = (F32) sin(angle * (1.f - t)); - stq = (F32) sin(angle * t); - } - - r.mQ[VX] = (q.mQ[VX] * stq) / s; - r.mQ[VY] = (q.mQ[VY] * stq) / s; - r.mQ[VZ] = (q.mQ[VZ] * stq) / s; - r.mQ[VW] = (stp + q.mQ[VW] * stq) / s; - - return r; -} - -LLQuaternion mayaQ(F32 xRot, F32 yRot, F32 zRot, LLQuaternion::Order order) -{ - LLQuaternion xQ( xRot*DEG_TO_RAD, LLVector3(1.0f, 0.0f, 0.0f) ); - LLQuaternion yQ( yRot*DEG_TO_RAD, LLVector3(0.0f, 1.0f, 0.0f) ); - LLQuaternion zQ( zRot*DEG_TO_RAD, LLVector3(0.0f, 0.0f, 1.0f) ); - LLQuaternion ret; - switch( order ) - { - case LLQuaternion::XYZ: - ret = xQ * yQ * zQ; - break; - case LLQuaternion::YZX: - ret = yQ * zQ * xQ; - break; - case LLQuaternion::ZXY: - ret = zQ * xQ * yQ; - break; - case LLQuaternion::XZY: - ret = xQ * zQ * yQ; - break; - case LLQuaternion::YXZ: - ret = yQ * xQ * zQ; - break; - case LLQuaternion::ZYX: - ret = zQ * yQ * xQ; - break; - } - return ret; -} - -const char *OrderToString( const LLQuaternion::Order order ) -{ - const char *p = NULL; - switch( order ) - { - default: - case LLQuaternion::XYZ: - p = "XYZ"; - break; - case LLQuaternion::YZX: - p = "YZX"; - break; - case LLQuaternion::ZXY: - p = "ZXY"; - break; - case LLQuaternion::XZY: - p = "XZY"; - break; - case LLQuaternion::YXZ: - p = "YXZ"; - break; - case LLQuaternion::ZYX: - p = "ZYX"; - break; - } - return p; -} - -LLQuaternion::Order StringToOrder( const char *str ) -{ - if (strncmp(str, "XYZ", 3)==0 || strncmp(str, "xyz", 3)==0) - return LLQuaternion::XYZ; - - if (strncmp(str, "YZX", 3)==0 || strncmp(str, "yzx", 3)==0) - return LLQuaternion::YZX; - - if (strncmp(str, "ZXY", 3)==0 || strncmp(str, "zxy", 3)==0) - return LLQuaternion::ZXY; - - if (strncmp(str, "XZY", 3)==0 || strncmp(str, "xzy", 3)==0) - return LLQuaternion::XZY; - - if (strncmp(str, "YXZ", 3)==0 || strncmp(str, "yxz", 3)==0) - return LLQuaternion::YXZ; - - if (strncmp(str, "ZYX", 3)==0 || strncmp(str, "zyx", 3)==0) - return LLQuaternion::ZYX; - - return LLQuaternion::XYZ; -} - -void LLQuaternion::getAngleAxis(F32* angle, LLVector3 &vec) const -{ - F32 cos_a = mQ[VW]; - if (cos_a > 1.0f) cos_a = 1.0f; - if (cos_a < -1.0f) cos_a = -1.0f; - - F32 sin_a = (F32) sqrt( 1.0f - cos_a * cos_a ); - - if ( fabs( sin_a ) < 0.0005f ) - sin_a = 1.0f; - else - sin_a = 1.f/sin_a; - - F32 temp_angle = 2.0f * (F32) acos( cos_a ); - if (temp_angle > F_PI) - { - // The (angle,axis) pair should never have angles outside [PI, -PI] - // since we want the _shortest_ (angle,axis) solution. - // Since acos is defined for [0, PI], and we multiply by 2.0, we - // can push the angle outside the acceptible range. - // When this happens we set the angle to the other portion of a - // full 2PI rotation, and negate the axis, which reverses the - // direction of the rotation (by the right-hand rule). - *angle = 2.f * F_PI - temp_angle; - vec.mV[VX] = - mQ[VX] * sin_a; - vec.mV[VY] = - mQ[VY] * sin_a; - vec.mV[VZ] = - mQ[VZ] * sin_a; - } - else - { - *angle = temp_angle; - vec.mV[VX] = mQ[VX] * sin_a; - vec.mV[VY] = mQ[VY] * sin_a; - vec.mV[VZ] = mQ[VZ] * sin_a; - } -} - - -// quaternion does not need to be normalized -void LLQuaternion::getEulerAngles(F32 *roll, F32 *pitch, F32 *yaw) const -{ - LLMatrix3 rot_mat(*this); - rot_mat.orthogonalize(); - rot_mat.getEulerAngles(roll, pitch, yaw); - -// // NOTE: LLQuaternion's are actually inverted with respect to -// // the matrices, so this code also assumes inverted quaternions -// // (-x, -y, -z, w). The result is that roll,pitch,yaw are applied -// // in reverse order (yaw,pitch,roll). -// F32 x = -mQ[VX], y = -mQ[VY], z = -mQ[VZ], w = mQ[VW]; -// F64 m20 = 2.0*(x*z-y*w); -// if (1.0f - fabsf(m20) < F_APPROXIMATELY_ZERO) -// { -// *roll = 0.0f; -// *pitch = (F32)asin(m20); -// *yaw = (F32)atan2(2.0*(x*y-z*w), 1.0 - 2.0*(x*x+z*z)); -// } -// else -// { -// *roll = (F32)atan2(-2.0*(y*z+x*w), 1.0-2.0*(x*x+y*y)); -// *pitch = (F32)asin(m20); -// *yaw = (F32)atan2(-2.0*(x*y+z*w), 1.0-2.0*(y*y+z*z)); -// } -} - -// Saves space by using the fact that our quaternions are normalized -LLVector3 LLQuaternion::packToVector3() const -{ - if( mQ[VW] >= 0 ) - { - return LLVector3( mQ[VX], mQ[VY], mQ[VZ] ); - } - else - { - return LLVector3( -mQ[VX], -mQ[VY], -mQ[VZ] ); - } -} - -// Saves space by using the fact that our quaternions are normalized -void LLQuaternion::unpackFromVector3( const LLVector3& vec ) -{ - mQ[VX] = vec.mV[VX]; - mQ[VY] = vec.mV[VY]; - mQ[VZ] = vec.mV[VZ]; - F32 t = 1.f - vec.magVecSquared(); - if( t > 0 ) - { - mQ[VW] = sqrt( t ); - } - else - { - // Need this to avoid trying to find the square root of a negative number due - // to floating point error. - mQ[VW] = 0; - } -} - -BOOL LLQuaternion::parseQuat(const std::string& buf, LLQuaternion* value) -{ - if( buf.empty() || value == NULL) - { - return FALSE; - } - - LLQuaternion quat; - S32 count = sscanf( buf.c_str(), "%f %f %f %f", quat.mQ + 0, quat.mQ + 1, quat.mQ + 2, quat.mQ + 3 ); - if( 4 == count ) - { - value->set( quat ); - return TRUE; - } - - return FALSE; -} - - -// End +/** + * @file llquaternion.cpp + * @brief LLQuaternion class implementation. + * + * $LicenseInfo:firstyear=2000&license=viewergpl$ + * + * Copyright (c) 2000-2009, Linden Research, Inc. + * + * Second Life Viewer Source Code + * The source code in this file ("Source Code") is provided by Linden Lab + * to you under the terms of the GNU General Public License, version 2.0 + * ("GPL"), unless you have obtained a separate licensing agreement + * ("Other License"), formally executed by you and Linden Lab. Terms of + * the GPL can be found in doc/GPL-license.txt in this distribution, or + * online at http://secondlifegrid.net/programs/open_source/licensing/gplv2 + * + * There are special exceptions to the terms and conditions of the GPL as + * it is applied to this Source Code. View the full text of the exception + * in the file doc/FLOSS-exception.txt in this software distribution, or + * online at + * http://secondlifegrid.net/programs/open_source/licensing/flossexception + * + * By copying, modifying or distributing this software, you acknowledge + * that you have read and understood your obligations described above, + * and agree to abide by those obligations. + * + * ALL LINDEN LAB SOURCE CODE IS PROVIDED "AS IS." LINDEN LAB MAKES NO + * WARRANTIES, EXPRESS, IMPLIED OR OTHERWISE, REGARDING ITS ACCURACY, + * COMPLETENESS OR PERFORMANCE. + * $/LicenseInfo$ + */ + +#include "linden_common.h" + +#include "llmath.h" // for F_PI + +#include "llquaternion.h" + +//#include "vmath.h" +#include "v3math.h" +#include "v3dmath.h" +#include "v4math.h" +#include "m4math.h" +#include "m3math.h" +#include "llquantize.h" + +// WARNING: Don't use this for global const definitions! using this +// at the top of a *.cpp file might not give you what you think. +const LLQuaternion LLQuaternion::DEFAULT; + +// Constructors + +LLQuaternion::LLQuaternion(const LLMatrix4 &mat) +{ + *this = mat.quaternion(); + normalize(); +} + +LLQuaternion::LLQuaternion(const LLMatrix3 &mat) +{ + *this = mat.quaternion(); + normalize(); +} + +LLQuaternion::LLQuaternion(F32 angle, const LLVector4 &vec) +{ + LLVector3 v(vec.mV[VX], vec.mV[VY], vec.mV[VZ]); + v.normalize(); + + F32 c, s; + c = cosf(angle*0.5f); + s = sinf(angle*0.5f); + + mQ[VX] = v.mV[VX] * s; + mQ[VY] = v.mV[VY] * s; + mQ[VZ] = v.mV[VZ] * s; + mQ[VW] = c; + normalize(); +} + +LLQuaternion::LLQuaternion(F32 angle, const LLVector3 &vec) +{ + LLVector3 v(vec); + v.normalize(); + + F32 c, s; + c = cosf(angle*0.5f); + s = sinf(angle*0.5f); + + mQ[VX] = v.mV[VX] * s; + mQ[VY] = v.mV[VY] * s; + mQ[VZ] = v.mV[VZ] * s; + mQ[VW] = c; + normalize(); +} + +LLQuaternion::LLQuaternion(const LLVector3 &x_axis, + const LLVector3 &y_axis, + const LLVector3 &z_axis) +{ + LLMatrix3 mat; + mat.setRows(x_axis, y_axis, z_axis); + *this = mat.quaternion(); + normalize(); +} + +// Quatizations +void LLQuaternion::quantize16(F32 lower, F32 upper) +{ + F32 x = mQ[VX]; + F32 y = mQ[VY]; + F32 z = mQ[VZ]; + F32 s = mQ[VS]; + + x = U16_to_F32(F32_to_U16_ROUND(x, lower, upper), lower, upper); + y = U16_to_F32(F32_to_U16_ROUND(y, lower, upper), lower, upper); + z = U16_to_F32(F32_to_U16_ROUND(z, lower, upper), lower, upper); + s = U16_to_F32(F32_to_U16_ROUND(s, lower, upper), lower, upper); + + mQ[VX] = x; + mQ[VY] = y; + mQ[VZ] = z; + mQ[VS] = s; + + normalize(); +} + +void LLQuaternion::quantize8(F32 lower, F32 upper) +{ + mQ[VX] = U8_to_F32(F32_to_U8_ROUND(mQ[VX], lower, upper), lower, upper); + mQ[VY] = U8_to_F32(F32_to_U8_ROUND(mQ[VY], lower, upper), lower, upper); + mQ[VZ] = U8_to_F32(F32_to_U8_ROUND(mQ[VZ], lower, upper), lower, upper); + mQ[VS] = U8_to_F32(F32_to_U8_ROUND(mQ[VS], lower, upper), lower, upper); + + normalize(); +} + +// LLVector3 Magnitude and Normalization Functions + + +// Set LLQuaternion routines + +const LLQuaternion& LLQuaternion::setAngleAxis(F32 angle, F32 x, F32 y, F32 z) +{ + LLVector3 vec(x, y, z); + vec.normalize(); + + angle *= 0.5f; + F32 c, s; + c = cosf(angle); + s = sinf(angle); + + mQ[VX] = vec.mV[VX]*s; + mQ[VY] = vec.mV[VY]*s; + mQ[VZ] = vec.mV[VZ]*s; + mQ[VW] = c; + + normalize(); + return (*this); +} + +const LLQuaternion& LLQuaternion::setAngleAxis(F32 angle, const LLVector3 &vec) +{ + LLVector3 v(vec); + v.normalize(); + + angle *= 0.5f; + F32 c, s; + c = cosf(angle); + s = sinf(angle); + + mQ[VX] = v.mV[VX]*s; + mQ[VY] = v.mV[VY]*s; + mQ[VZ] = v.mV[VZ]*s; + mQ[VW] = c; + + normalize(); + return (*this); +} + +const LLQuaternion& LLQuaternion::setAngleAxis(F32 angle, const LLVector4 &vec) +{ + LLVector3 v(vec.mV[VX], vec.mV[VY], vec.mV[VZ]); + v.normalize(); + + F32 c, s; + c = cosf(angle*0.5f); + s = sinf(angle*0.5f); + + mQ[VX] = v.mV[VX]*s; + mQ[VY] = v.mV[VY]*s; + mQ[VZ] = v.mV[VZ]*s; + mQ[VW] = c; + + normalize(); + return (*this); +} + +const LLQuaternion& LLQuaternion::setEulerAngles(F32 roll, F32 pitch, F32 yaw) +{ + LLMatrix3 rot_mat(roll, pitch, yaw); + rot_mat.orthogonalize(); + *this = rot_mat.quaternion(); + + normalize(); + return (*this); +} + +// deprecated +const LLQuaternion& LLQuaternion::set(const LLMatrix3 &mat) +{ + *this = mat.quaternion(); + normalize(); + return (*this); +} + +// deprecated +const LLQuaternion& LLQuaternion::set(const LLMatrix4 &mat) +{ + *this = mat.quaternion(); + normalize(); + return (*this); +} + +// deprecated +const LLQuaternion& LLQuaternion::setQuat(F32 angle, F32 x, F32 y, F32 z) +{ + LLVector3 vec(x, y, z); + vec.normalize(); + + angle *= 0.5f; + F32 c, s; + c = cosf(angle); + s = sinf(angle); + + mQ[VX] = vec.mV[VX]*s; + mQ[VY] = vec.mV[VY]*s; + mQ[VZ] = vec.mV[VZ]*s; + mQ[VW] = c; + + normalize(); + return (*this); +} + +// deprecated +const LLQuaternion& LLQuaternion::setQuat(F32 angle, const LLVector3 &vec) +{ + LLVector3 v(vec); + v.normalize(); + + angle *= 0.5f; + F32 c, s; + c = cosf(angle); + s = sinf(angle); + + mQ[VX] = v.mV[VX]*s; + mQ[VY] = v.mV[VY]*s; + mQ[VZ] = v.mV[VZ]*s; + mQ[VW] = c; + + normalize(); + return (*this); +} + +const LLQuaternion& LLQuaternion::setQuat(F32 angle, const LLVector4 &vec) +{ + LLVector3 v(vec.mV[VX], vec.mV[VY], vec.mV[VZ]); + v.normalize(); + + F32 c, s; + c = cosf(angle*0.5f); + s = sinf(angle*0.5f); + + mQ[VX] = v.mV[VX]*s; + mQ[VY] = v.mV[VY]*s; + mQ[VZ] = v.mV[VZ]*s; + mQ[VW] = c; + + normalize(); + return (*this); +} + +const LLQuaternion& LLQuaternion::setQuat(F32 roll, F32 pitch, F32 yaw) +{ + LLMatrix3 rot_mat(roll, pitch, yaw); + rot_mat.orthogonalize(); + *this = rot_mat.quaternion(); + + normalize(); + return (*this); +} + +const LLQuaternion& LLQuaternion::setQuat(const LLMatrix3 &mat) +{ + *this = mat.quaternion(); + normalize(); + return (*this); +} + +const LLQuaternion& LLQuaternion::setQuat(const LLMatrix4 &mat) +{ + *this = mat.quaternion(); + normalize(); + return (*this); +//#if 1 +// // NOTE: LLQuaternion's are actually inverted with respect to +// // the matrices, so this code also assumes inverted quaternions +// // (-x, -y, -z, w). The result is that roll,pitch,yaw are applied +// // in reverse order (yaw,pitch,roll). +// F64 cosX = cos(roll); +// F64 cosY = cos(pitch); +// F64 cosZ = cos(yaw); +// +// F64 sinX = sin(roll); +// F64 sinY = sin(pitch); +// F64 sinZ = sin(yaw); +// +// mQ[VW] = (F32)sqrt(cosY*cosZ - sinX*sinY*sinZ + cosX*cosZ + cosX*cosY + 1.0)*.5; +// if (fabs(mQ[VW]) < F_APPROXIMATELY_ZERO) +// { +// // null rotation, any axis will do +// mQ[VX] = 0.0f; +// mQ[VY] = 1.0f; +// mQ[VZ] = 0.0f; +// } +// else +// { +// F32 inv_s = 1.0f / (4.0f * mQ[VW]); +// mQ[VX] = (F32)-(-sinX*cosY - cosX*sinY*sinZ - sinX*cosZ) * inv_s; +// mQ[VY] = (F32)-(-cosX*sinY*cosZ + sinX*sinZ - sinY) * inv_s; +// mQ[VZ] = (F32)-(-cosY*sinZ - sinX*sinY*cosZ - cosX*sinZ) * inv_s; +// } +// +//#else // This only works on a certain subset of roll/pitch/yaw +// +// F64 cosX = cosf(roll/2.0); +// F64 cosY = cosf(pitch/2.0); +// F64 cosZ = cosf(yaw/2.0); +// +// F64 sinX = sinf(roll/2.0); +// F64 sinY = sinf(pitch/2.0); +// F64 sinZ = sinf(yaw/2.0); +// +// mQ[VW] = (F32)(cosX*cosY*cosZ + sinX*sinY*sinZ); +// mQ[VX] = (F32)(sinX*cosY*cosZ - cosX*sinY*sinZ); +// mQ[VY] = (F32)(cosX*sinY*cosZ + sinX*cosY*sinZ); +// mQ[VZ] = (F32)(cosX*cosY*sinZ - sinX*sinY*cosZ); +//#endif +// +// normalize(); +// return (*this); +} + +// SJB: This code is correct for a logicly stored (non-transposed) matrix; +// Our matrices are stored transposed, OpenGL style, so this generates the +// INVERSE matrix, or the CORRECT matrix form an INVERSE quaternion. +// Because we use similar logic in LLMatrix3::quaternion(), +// we are internally consistant so everything works OK :) +LLMatrix3 LLQuaternion::getMatrix3(void) const +{ + LLMatrix3 mat; + F32 xx, xy, xz, xw, yy, yz, yw, zz, zw; + + xx = mQ[VX] * mQ[VX]; + xy = mQ[VX] * mQ[VY]; + xz = mQ[VX] * mQ[VZ]; + xw = mQ[VX] * mQ[VW]; + + yy = mQ[VY] * mQ[VY]; + yz = mQ[VY] * mQ[VZ]; + yw = mQ[VY] * mQ[VW]; + + zz = mQ[VZ] * mQ[VZ]; + zw = mQ[VZ] * mQ[VW]; + + mat.mMatrix[0][0] = 1.f - 2.f * ( yy + zz ); + mat.mMatrix[0][1] = 2.f * ( xy + zw ); + mat.mMatrix[0][2] = 2.f * ( xz - yw ); + + mat.mMatrix[1][0] = 2.f * ( xy - zw ); + mat.mMatrix[1][1] = 1.f - 2.f * ( xx + zz ); + mat.mMatrix[1][2] = 2.f * ( yz + xw ); + + mat.mMatrix[2][0] = 2.f * ( xz + yw ); + mat.mMatrix[2][1] = 2.f * ( yz - xw ); + mat.mMatrix[2][2] = 1.f - 2.f * ( xx + yy ); + + return mat; +} + +LLMatrix4 LLQuaternion::getMatrix4(void) const +{ + LLMatrix4 mat; + F32 xx, xy, xz, xw, yy, yz, yw, zz, zw; + + xx = mQ[VX] * mQ[VX]; + xy = mQ[VX] * mQ[VY]; + xz = mQ[VX] * mQ[VZ]; + xw = mQ[VX] * mQ[VW]; + + yy = mQ[VY] * mQ[VY]; + yz = mQ[VY] * mQ[VZ]; + yw = mQ[VY] * mQ[VW]; + + zz = mQ[VZ] * mQ[VZ]; + zw = mQ[VZ] * mQ[VW]; + + mat.mMatrix[0][0] = 1.f - 2.f * ( yy + zz ); + mat.mMatrix[0][1] = 2.f * ( xy + zw ); + mat.mMatrix[0][2] = 2.f * ( xz - yw ); + + mat.mMatrix[1][0] = 2.f * ( xy - zw ); + mat.mMatrix[1][1] = 1.f - 2.f * ( xx + zz ); + mat.mMatrix[1][2] = 2.f * ( yz + xw ); + + mat.mMatrix[2][0] = 2.f * ( xz + yw ); + mat.mMatrix[2][1] = 2.f * ( yz - xw ); + mat.mMatrix[2][2] = 1.f - 2.f * ( xx + yy ); + + // TODO -- should we set the translation portion to zero? + + return mat; +} + + + + +// Other useful methods + + +// calculate the shortest rotation from a to b +void LLQuaternion::shortestArc(const LLVector3 &a, const LLVector3 &b) +{ + // Make a local copy of both vectors. + LLVector3 vec_a = a; + LLVector3 vec_b = b; + + // Make sure neither vector is zero length. Also normalize + // the vectors while we are at it. + F32 vec_a_mag = vec_a.normalize(); + F32 vec_b_mag = vec_b.normalize(); + if (vec_a_mag < F_APPROXIMATELY_ZERO || + vec_b_mag < F_APPROXIMATELY_ZERO) + { + // Can't calculate a rotation from this. + // Just return ZERO_ROTATION instead. + loadIdentity(); + return; + } + + // Create an axis to rotate around, and the cos of the angle to rotate. + LLVector3 axis = vec_a % vec_b; + F32 cos_theta = vec_a * vec_b; + + // Check the angle between the vectors to see if they are parallel or anti-parallel. + if (cos_theta > 1.0 - F_APPROXIMATELY_ZERO) + { + // a and b are parallel. No rotation is necessary. + loadIdentity(); + } + else if (cos_theta < -1.0 + F_APPROXIMATELY_ZERO) + { + // a and b are anti-parallel. + // Rotate 180 degrees around some orthogonal axis. + // Find the projection of the x-axis onto a, and try + // using the vector between the projection and the x-axis + // as the orthogonal axis. + LLVector3 proj = vec_a.mV[VX] / (vec_a * vec_a) * vec_a; + LLVector3 ortho_axis(1.f, 0.f, 0.f); + ortho_axis -= proj; + + // Turn this into an orthonormal axis. + F32 ortho_length = ortho_axis.normalize(); + // If the axis' length is 0, then our guess at an orthogonal axis + // was wrong (a is parallel to the x-axis). + if (ortho_length < F_APPROXIMATELY_ZERO) + { + // Use the z-axis instead. + ortho_axis.setVec(0.f, 0.f, 1.f); + } + + // Construct a quaternion from this orthonormal axis. + mQ[VX] = ortho_axis.mV[VX]; + mQ[VY] = ortho_axis.mV[VY]; + mQ[VZ] = ortho_axis.mV[VZ]; + mQ[VW] = 0.f; + } + else + { + // a and b are NOT parallel or anti-parallel. + // Return the rotation between these vectors. + F32 theta = (F32)acos(cos_theta); + + setAngleAxis(theta, axis); + } +} + +// constrains rotation to a cone angle specified in radians +const LLQuaternion &LLQuaternion::constrain(F32 radians) +{ + const F32 cos_angle_lim = cosf( radians/2 ); // mQ[VW] limit + const F32 sin_angle_lim = sinf( radians/2 ); // rotation axis length limit + + if (mQ[VW] < 0.f) + { + mQ[VX] *= -1.f; + mQ[VY] *= -1.f; + mQ[VZ] *= -1.f; + mQ[VW] *= -1.f; + } + + // if rotation angle is greater than limit (cos is less than limit) + if( mQ[VW] < cos_angle_lim ) + { + mQ[VW] = cos_angle_lim; + F32 axis_len = sqrtf( mQ[VX]*mQ[VX] + mQ[VY]*mQ[VY] + mQ[VZ]*mQ[VZ] ); // sin(theta/2) + F32 axis_mult_fact = sin_angle_lim / axis_len; + mQ[VX] *= axis_mult_fact; + mQ[VY] *= axis_mult_fact; + mQ[VZ] *= axis_mult_fact; + } + + return *this; +} + +// Operators + +std::ostream& operator<<(std::ostream &s, const LLQuaternion &a) +{ + s << "{ " + << a.mQ[VX] << ", " << a.mQ[VY] << ", " << a.mQ[VZ] << ", " << a.mQ[VW] + << " }"; + return s; +} + + +// Does NOT renormalize the result +LLQuaternion operator*(const LLQuaternion &a, const LLQuaternion &b) +{ +// LLQuaternion::mMultCount++; + + LLQuaternion q( + b.mQ[3] * a.mQ[0] + b.mQ[0] * a.mQ[3] + b.mQ[1] * a.mQ[2] - b.mQ[2] * a.mQ[1], + b.mQ[3] * a.mQ[1] + b.mQ[1] * a.mQ[3] + b.mQ[2] * a.mQ[0] - b.mQ[0] * a.mQ[2], + b.mQ[3] * a.mQ[2] + b.mQ[2] * a.mQ[3] + b.mQ[0] * a.mQ[1] - b.mQ[1] * a.mQ[0], + b.mQ[3] * a.mQ[3] - b.mQ[0] * a.mQ[0] - b.mQ[1] * a.mQ[1] - b.mQ[2] * a.mQ[2] + ); + return q; +} + +/* +LLMatrix4 operator*(const LLMatrix4 &m, const LLQuaternion &q) +{ + LLMatrix4 qmat(q); + return (m*qmat); +} +*/ + + + +LLVector4 operator*(const LLVector4 &a, const LLQuaternion &rot) +{ + F32 rw = - rot.mQ[VX] * a.mV[VX] - rot.mQ[VY] * a.mV[VY] - rot.mQ[VZ] * a.mV[VZ]; + F32 rx = rot.mQ[VW] * a.mV[VX] + rot.mQ[VY] * a.mV[VZ] - rot.mQ[VZ] * a.mV[VY]; + F32 ry = rot.mQ[VW] * a.mV[VY] + rot.mQ[VZ] * a.mV[VX] - rot.mQ[VX] * a.mV[VZ]; + F32 rz = rot.mQ[VW] * a.mV[VZ] + rot.mQ[VX] * a.mV[VY] - rot.mQ[VY] * a.mV[VX]; + + F32 nx = - rw * rot.mQ[VX] + rx * rot.mQ[VW] - ry * rot.mQ[VZ] + rz * rot.mQ[VY]; + F32 ny = - rw * rot.mQ[VY] + ry * rot.mQ[VW] - rz * rot.mQ[VX] + rx * rot.mQ[VZ]; + F32 nz = - rw * rot.mQ[VZ] + rz * rot.mQ[VW] - rx * rot.mQ[VY] + ry * rot.mQ[VX]; + + return LLVector4(nx, ny, nz, a.mV[VW]); +} + +LLVector3 operator*(const LLVector3 &a, const LLQuaternion &rot) +{ + F32 rw = - rot.mQ[VX] * a.mV[VX] - rot.mQ[VY] * a.mV[VY] - rot.mQ[VZ] * a.mV[VZ]; + F32 rx = rot.mQ[VW] * a.mV[VX] + rot.mQ[VY] * a.mV[VZ] - rot.mQ[VZ] * a.mV[VY]; + F32 ry = rot.mQ[VW] * a.mV[VY] + rot.mQ[VZ] * a.mV[VX] - rot.mQ[VX] * a.mV[VZ]; + F32 rz = rot.mQ[VW] * a.mV[VZ] + rot.mQ[VX] * a.mV[VY] - rot.mQ[VY] * a.mV[VX]; + + F32 nx = - rw * rot.mQ[VX] + rx * rot.mQ[VW] - ry * rot.mQ[VZ] + rz * rot.mQ[VY]; + F32 ny = - rw * rot.mQ[VY] + ry * rot.mQ[VW] - rz * rot.mQ[VX] + rx * rot.mQ[VZ]; + F32 nz = - rw * rot.mQ[VZ] + rz * rot.mQ[VW] - rx * rot.mQ[VY] + ry * rot.mQ[VX]; + + return LLVector3(nx, ny, nz); +} + +LLVector3d operator*(const LLVector3d &a, const LLQuaternion &rot) +{ + F64 rw = - rot.mQ[VX] * a.mdV[VX] - rot.mQ[VY] * a.mdV[VY] - rot.mQ[VZ] * a.mdV[VZ]; + F64 rx = rot.mQ[VW] * a.mdV[VX] + rot.mQ[VY] * a.mdV[VZ] - rot.mQ[VZ] * a.mdV[VY]; + F64 ry = rot.mQ[VW] * a.mdV[VY] + rot.mQ[VZ] * a.mdV[VX] - rot.mQ[VX] * a.mdV[VZ]; + F64 rz = rot.mQ[VW] * a.mdV[VZ] + rot.mQ[VX] * a.mdV[VY] - rot.mQ[VY] * a.mdV[VX]; + + F64 nx = - rw * rot.mQ[VX] + rx * rot.mQ[VW] - ry * rot.mQ[VZ] + rz * rot.mQ[VY]; + F64 ny = - rw * rot.mQ[VY] + ry * rot.mQ[VW] - rz * rot.mQ[VX] + rx * rot.mQ[VZ]; + F64 nz = - rw * rot.mQ[VZ] + rz * rot.mQ[VW] - rx * rot.mQ[VY] + ry * rot.mQ[VX]; + + return LLVector3d(nx, ny, nz); +} + +F32 dot(const LLQuaternion &a, const LLQuaternion &b) +{ + return a.mQ[VX] * b.mQ[VX] + + a.mQ[VY] * b.mQ[VY] + + a.mQ[VZ] * b.mQ[VZ] + + a.mQ[VW] * b.mQ[VW]; +} + +// DEMO HACK: This lerp is probably inocrrect now due intermediate normalization +// it should look more like the lerp below +#if 0 +// linear interpolation +LLQuaternion lerp(F32 t, const LLQuaternion &p, const LLQuaternion &q) +{ + LLQuaternion r; + r = t * (q - p) + p; + r.normalize(); + return r; +} +#endif + +// lerp from identity to q +LLQuaternion lerp(F32 t, const LLQuaternion &q) +{ + LLQuaternion r; + r.mQ[VX] = t * q.mQ[VX]; + r.mQ[VY] = t * q.mQ[VY]; + r.mQ[VZ] = t * q.mQ[VZ]; + r.mQ[VW] = t * (q.mQ[VZ] - 1.f) + 1.f; + r.normalize(); + return r; +} + +LLQuaternion lerp(F32 t, const LLQuaternion &p, const LLQuaternion &q) +{ + LLQuaternion r; + F32 inv_t; + + inv_t = 1.f - t; + + r.mQ[VX] = t * q.mQ[VX] + (inv_t * p.mQ[VX]); + r.mQ[VY] = t * q.mQ[VY] + (inv_t * p.mQ[VY]); + r.mQ[VZ] = t * q.mQ[VZ] + (inv_t * p.mQ[VZ]); + r.mQ[VW] = t * q.mQ[VW] + (inv_t * p.mQ[VW]); + r.normalize(); + return r; +} + + +// spherical linear interpolation +LLQuaternion slerp( F32 u, const LLQuaternion &a, const LLQuaternion &b ) +{ + // cosine theta = dot product of a and b + F32 cos_t = a.mQ[0]*b.mQ[0] + a.mQ[1]*b.mQ[1] + a.mQ[2]*b.mQ[2] + a.mQ[3]*b.mQ[3]; + + // if b is on opposite hemisphere from a, use -a instead + int bflip; + if (cos_t < 0.0f) + { + cos_t = -cos_t; + bflip = TRUE; + } + else + bflip = FALSE; + + // if B is (within precision limits) the same as A, + // just linear interpolate between A and B. + F32 alpha; // interpolant + F32 beta; // 1 - interpolant + if (1.0f - cos_t < 0.00001f) + { + beta = 1.0f - u; + alpha = u; + } + else + { + F32 theta = acosf(cos_t); + F32 sin_t = sinf(theta); + beta = sinf(theta - u*theta) / sin_t; + alpha = sinf(u*theta) / sin_t; + } + + if (bflip) + beta = -beta; + + // interpolate + LLQuaternion ret; + ret.mQ[0] = beta*a.mQ[0] + alpha*b.mQ[0]; + ret.mQ[1] = beta*a.mQ[1] + alpha*b.mQ[1]; + ret.mQ[2] = beta*a.mQ[2] + alpha*b.mQ[2]; + ret.mQ[3] = beta*a.mQ[3] + alpha*b.mQ[3]; + + return ret; +} + +// lerp whenever possible +LLQuaternion nlerp(F32 t, const LLQuaternion &a, const LLQuaternion &b) +{ + if (dot(a, b) < 0.f) + { + return slerp(t, a, b); + } + else + { + return lerp(t, a, b); + } +} + +LLQuaternion nlerp(F32 t, const LLQuaternion &q) +{ + if (q.mQ[VW] < 0.f) + { + return slerp(t, q); + } + else + { + return lerp(t, q); + } +} + +// slerp from identity quaternion to another quaternion +LLQuaternion slerp(F32 t, const LLQuaternion &q) +{ + F32 c = q.mQ[VW]; + if (1.0f == t || 1.0f == c) + { + // the trivial cases + return q; + } + + LLQuaternion r; + F32 s, angle, stq, stp; + + s = (F32) sqrt(1.f - c*c); + + if (c < 0.0f) + { + // when c < 0.0 then theta > PI/2 + // since quat and -quat are the same rotation we invert one of + // p or q to reduce unecessary spins + // A equivalent way to do it is to convert acos(c) as if it had + // been negative, and to negate stp + angle = (F32) acos(-c); + stp = -(F32) sin(angle * (1.f - t)); + stq = (F32) sin(angle * t); + } + else + { + angle = (F32) acos(c); + stp = (F32) sin(angle * (1.f - t)); + stq = (F32) sin(angle * t); + } + + r.mQ[VX] = (q.mQ[VX] * stq) / s; + r.mQ[VY] = (q.mQ[VY] * stq) / s; + r.mQ[VZ] = (q.mQ[VZ] * stq) / s; + r.mQ[VW] = (stp + q.mQ[VW] * stq) / s; + + return r; +} + +LLQuaternion mayaQ(F32 xRot, F32 yRot, F32 zRot, LLQuaternion::Order order) +{ + LLQuaternion xQ( xRot*DEG_TO_RAD, LLVector3(1.0f, 0.0f, 0.0f) ); + LLQuaternion yQ( yRot*DEG_TO_RAD, LLVector3(0.0f, 1.0f, 0.0f) ); + LLQuaternion zQ( zRot*DEG_TO_RAD, LLVector3(0.0f, 0.0f, 1.0f) ); + LLQuaternion ret; + switch( order ) + { + case LLQuaternion::XYZ: + ret = xQ * yQ * zQ; + break; + case LLQuaternion::YZX: + ret = yQ * zQ * xQ; + break; + case LLQuaternion::ZXY: + ret = zQ * xQ * yQ; + break; + case LLQuaternion::XZY: + ret = xQ * zQ * yQ; + break; + case LLQuaternion::YXZ: + ret = yQ * xQ * zQ; + break; + case LLQuaternion::ZYX: + ret = zQ * yQ * xQ; + break; + } + return ret; +} + +const char *OrderToString( const LLQuaternion::Order order ) +{ + const char *p = NULL; + switch( order ) + { + default: + case LLQuaternion::XYZ: + p = "XYZ"; + break; + case LLQuaternion::YZX: + p = "YZX"; + break; + case LLQuaternion::ZXY: + p = "ZXY"; + break; + case LLQuaternion::XZY: + p = "XZY"; + break; + case LLQuaternion::YXZ: + p = "YXZ"; + break; + case LLQuaternion::ZYX: + p = "ZYX"; + break; + } + return p; +} + +LLQuaternion::Order StringToOrder( const char *str ) +{ + if (strncmp(str, "XYZ", 3)==0 || strncmp(str, "xyz", 3)==0) + return LLQuaternion::XYZ; + + if (strncmp(str, "YZX", 3)==0 || strncmp(str, "yzx", 3)==0) + return LLQuaternion::YZX; + + if (strncmp(str, "ZXY", 3)==0 || strncmp(str, "zxy", 3)==0) + return LLQuaternion::ZXY; + + if (strncmp(str, "XZY", 3)==0 || strncmp(str, "xzy", 3)==0) + return LLQuaternion::XZY; + + if (strncmp(str, "YXZ", 3)==0 || strncmp(str, "yxz", 3)==0) + return LLQuaternion::YXZ; + + if (strncmp(str, "ZYX", 3)==0 || strncmp(str, "zyx", 3)==0) + return LLQuaternion::ZYX; + + return LLQuaternion::XYZ; +} + +void LLQuaternion::getAngleAxis(F32* angle, LLVector3 &vec) const +{ + F32 cos_a = mQ[VW]; + if (cos_a > 1.0f) cos_a = 1.0f; + if (cos_a < -1.0f) cos_a = -1.0f; + + F32 sin_a = (F32) sqrt( 1.0f - cos_a * cos_a ); + + if ( fabs( sin_a ) < 0.0005f ) + sin_a = 1.0f; + else + sin_a = 1.f/sin_a; + + F32 temp_angle = 2.0f * (F32) acos( cos_a ); + if (temp_angle > F_PI) + { + // The (angle,axis) pair should never have angles outside [PI, -PI] + // since we want the _shortest_ (angle,axis) solution. + // Since acos is defined for [0, PI], and we multiply by 2.0, we + // can push the angle outside the acceptible range. + // When this happens we set the angle to the other portion of a + // full 2PI rotation, and negate the axis, which reverses the + // direction of the rotation (by the right-hand rule). + *angle = 2.f * F_PI - temp_angle; + vec.mV[VX] = - mQ[VX] * sin_a; + vec.mV[VY] = - mQ[VY] * sin_a; + vec.mV[VZ] = - mQ[VZ] * sin_a; + } + else + { + *angle = temp_angle; + vec.mV[VX] = mQ[VX] * sin_a; + vec.mV[VY] = mQ[VY] * sin_a; + vec.mV[VZ] = mQ[VZ] * sin_a; + } +} + + +// quaternion does not need to be normalized +void LLQuaternion::getEulerAngles(F32 *roll, F32 *pitch, F32 *yaw) const +{ + LLMatrix3 rot_mat(*this); + rot_mat.orthogonalize(); + rot_mat.getEulerAngles(roll, pitch, yaw); + +// // NOTE: LLQuaternion's are actually inverted with respect to +// // the matrices, so this code also assumes inverted quaternions +// // (-x, -y, -z, w). The result is that roll,pitch,yaw are applied +// // in reverse order (yaw,pitch,roll). +// F32 x = -mQ[VX], y = -mQ[VY], z = -mQ[VZ], w = mQ[VW]; +// F64 m20 = 2.0*(x*z-y*w); +// if (1.0f - fabsf(m20) < F_APPROXIMATELY_ZERO) +// { +// *roll = 0.0f; +// *pitch = (F32)asin(m20); +// *yaw = (F32)atan2(2.0*(x*y-z*w), 1.0 - 2.0*(x*x+z*z)); +// } +// else +// { +// *roll = (F32)atan2(-2.0*(y*z+x*w), 1.0-2.0*(x*x+y*y)); +// *pitch = (F32)asin(m20); +// *yaw = (F32)atan2(-2.0*(x*y+z*w), 1.0-2.0*(y*y+z*z)); +// } +} + +// Saves space by using the fact that our quaternions are normalized +LLVector3 LLQuaternion::packToVector3() const +{ + if( mQ[VW] >= 0 ) + { + return LLVector3( mQ[VX], mQ[VY], mQ[VZ] ); + } + else + { + return LLVector3( -mQ[VX], -mQ[VY], -mQ[VZ] ); + } +} + +// Saves space by using the fact that our quaternions are normalized +void LLQuaternion::unpackFromVector3( const LLVector3& vec ) +{ + mQ[VX] = vec.mV[VX]; + mQ[VY] = vec.mV[VY]; + mQ[VZ] = vec.mV[VZ]; + F32 t = 1.f - vec.magVecSquared(); + if( t > 0 ) + { + mQ[VW] = sqrt( t ); + } + else + { + // Need this to avoid trying to find the square root of a negative number due + // to floating point error. + mQ[VW] = 0; + } +} + +BOOL LLQuaternion::parseQuat(const std::string& buf, LLQuaternion* value) +{ + if( buf.empty() || value == NULL) + { + return FALSE; + } + + LLQuaternion quat; + S32 count = sscanf( buf.c_str(), "%f %f %f %f", quat.mQ + 0, quat.mQ + 1, quat.mQ + 2, quat.mQ + 3 ); + if( 4 == count ) + { + value->set( quat ); + return TRUE; + } + + return FALSE; +} + + +// End diff --git a/indra/llmath/llquaternion.h b/indra/llmath/llquaternion.h index 0769f29f23..bbd4326483 100644 --- a/indra/llmath/llquaternion.h +++ b/indra/llmath/llquaternion.h @@ -1,590 +1,594 @@ -/** - * @file llquaternion.h - * @brief LLQuaternion class header file. - * - * $LicenseInfo:firstyear=2000&license=viewergpl$ - * - * Copyright (c) 2000-2009, Linden Research, Inc. - * - * Second Life Viewer Source Code - * The source code in this file ("Source Code") is provided by Linden Lab - * to you under the terms of the GNU General Public License, version 2.0 - * ("GPL"), unless you have obtained a separate licensing agreement - * ("Other License"), formally executed by you and Linden Lab. Terms of - * the GPL can be found in doc/GPL-license.txt in this distribution, or - * online at http://secondlifegrid.net/programs/open_source/licensing/gplv2 - * - * There are special exceptions to the terms and conditions of the GPL as - * it is applied to this Source Code. View the full text of the exception - * in the file doc/FLOSS-exception.txt in this software distribution, or - * online at - * http://secondlifegrid.net/programs/open_source/licensing/flossexception - * - * By copying, modifying or distributing this software, you acknowledge - * that you have read and understood your obligations described above, - * and agree to abide by those obligations. - * - * ALL LINDEN LAB SOURCE CODE IS PROVIDED "AS IS." LINDEN LAB MAKES NO - * WARRANTIES, EXPRESS, IMPLIED OR OTHERWISE, REGARDING ITS ACCURACY, - * COMPLETENESS OR PERFORMANCE. - * $/LicenseInfo$ - */ - -#ifndef LLQUATERNION_H -#define LLQUATERNION_H - -#include "llmath.h" - -class LLVector4; -class LLVector3; -class LLVector3d; -class LLMatrix4; -class LLMatrix3; - -// NOTA BENE: Quaternion code is written assuming Unit Quaternions!!!! -// Moreover, it is written assuming that all vectors and matricies -// passed as arguments are normalized and unitary respectively. -// VERY VERY VERY VERY BAD THINGS will happen if these assumptions fail. - -static const U32 LENGTHOFQUAT = 4; - -class LLQuaternion -{ -public: - F32 mQ[LENGTHOFQUAT]; - - static const LLQuaternion DEFAULT; - - LLQuaternion(); // Initializes Quaternion to (0,0,0,1) - explicit LLQuaternion(const LLMatrix4 &mat); // Initializes Quaternion from Matrix4 - explicit LLQuaternion(const LLMatrix3 &mat); // Initializes Quaternion from Matrix3 - LLQuaternion(F32 x, F32 y, F32 z, F32 w); // Initializes Quaternion to normalize(x, y, z, w) - LLQuaternion(F32 angle, const LLVector4 &vec); // Initializes Quaternion to axis_angle2quat(angle, vec) - LLQuaternion(F32 angle, const LLVector3 &vec); // Initializes Quaternion to axis_angle2quat(angle, vec) - LLQuaternion(const F32 *q); // Initializes Quaternion to normalize(x, y, z, w) - LLQuaternion(const LLVector3 &x_axis, - const LLVector3 &y_axis, - const LLVector3 &z_axis); // Initializes Quaternion from Matrix3 = [x_axis ; y_axis ; z_axis] - - BOOL isIdentity() const; - BOOL isNotIdentity() const; - BOOL isFinite() const; // checks to see if all values of LLQuaternion are finite - void quantize16(F32 lower, F32 upper); // changes the vector to reflect quatization - void quantize8(F32 lower, F32 upper); // changes the vector to reflect quatization - void loadIdentity(); // Loads the quaternion that represents the identity rotation - - const LLQuaternion& set(F32 x, F32 y, F32 z, F32 w); // Sets Quaternion to normalize(x, y, z, w) - const LLQuaternion& set(const LLQuaternion &quat); // Copies Quaternion - const LLQuaternion& set(const F32 *q); // Sets Quaternion to normalize(quat[VX], quat[VY], quat[VZ], quat[VW]) - const LLQuaternion& set(const LLMatrix3 &mat); // Sets Quaternion to mat2quat(mat) - const LLQuaternion& set(const LLMatrix4 &mat); // Sets Quaternion to mat2quat(mat) - - const LLQuaternion& setAngleAxis(F32 angle, F32 x, F32 y, F32 z); // Sets Quaternion to axis_angle2quat(angle, x, y, z) - const LLQuaternion& setAngleAxis(F32 angle, const LLVector3 &vec); // Sets Quaternion to axis_angle2quat(angle, vec) - const LLQuaternion& setAngleAxis(F32 angle, const LLVector4 &vec); // Sets Quaternion to axis_angle2quat(angle, vec) - const LLQuaternion& setEulerAngles(F32 roll, F32 pitch, F32 yaw); // Sets Quaternion to euler2quat(pitch, yaw, roll) - - const LLQuaternion& setQuatInit(F32 x, F32 y, F32 z, F32 w); // deprecated - const LLQuaternion& setQuat(const LLQuaternion &quat); // deprecated - const LLQuaternion& setQuat(const F32 *q); // deprecated - const LLQuaternion& setQuat(const LLMatrix3 &mat); // deprecated - const LLQuaternion& setQuat(const LLMatrix4 &mat); // deprecated - const LLQuaternion& setQuat(F32 angle, F32 x, F32 y, F32 z); // deprecated - const LLQuaternion& setQuat(F32 angle, const LLVector3 &vec); // deprecated - const LLQuaternion& setQuat(F32 angle, const LLVector4 &vec); // deprecated - const LLQuaternion& setQuat(F32 roll, F32 pitch, F32 yaw); // deprecated - - LLMatrix4 getMatrix4(void) const; // Returns the Matrix4 equivalent of Quaternion - LLMatrix3 getMatrix3(void) const; // Returns the Matrix3 equivalent of Quaternion - void getAngleAxis(F32* angle, F32* x, F32* y, F32* z) const; // returns rotation in radians about axis x,y,z - void getAngleAxis(F32* angle, LLVector3 &vec) const; - void getEulerAngles(F32 *roll, F32* pitch, F32 *yaw) const; - - F32 normalize(); // Normalizes Quaternion and returns magnitude - F32 normQuat(); // deprecated - - const LLQuaternion& conjugate(void); // Conjugates Quaternion and returns result - const LLQuaternion& conjQuat(void); // deprecated - - // Other useful methods - const LLQuaternion& transpose(); // transpose (same as conjugate) - const LLQuaternion& transQuat(); // deprecated - - void shortestArc(const LLVector3 &a, const LLVector3 &b); // shortest rotation from a to b - const LLQuaternion& constrain(F32 radians); // constrains rotation to a cone angle specified in radians - - // Standard operators - friend std::ostream& operator<<(std::ostream &s, const LLQuaternion &a); // Prints a - friend LLQuaternion operator+(const LLQuaternion &a, const LLQuaternion &b); // Addition - friend LLQuaternion operator-(const LLQuaternion &a, const LLQuaternion &b); // Subtraction - friend LLQuaternion operator-(const LLQuaternion &a); // Negation - friend LLQuaternion operator*(F32 a, const LLQuaternion &q); // Scale - friend LLQuaternion operator*(const LLQuaternion &q, F32 b); // Scale - friend LLQuaternion operator*(const LLQuaternion &a, const LLQuaternion &b); // Returns a * b - friend LLQuaternion operator~(const LLQuaternion &a); // Returns a* (Conjugate of a) - bool operator==(const LLQuaternion &b) const; // Returns a == b - bool operator!=(const LLQuaternion &b) const; // Returns a != b - - friend const LLQuaternion& operator*=(LLQuaternion &a, const LLQuaternion &b); // Returns a * b - - friend LLVector4 operator*(const LLVector4 &a, const LLQuaternion &rot); // Rotates a by rot - friend LLVector3 operator*(const LLVector3 &a, const LLQuaternion &rot); // Rotates a by rot - friend LLVector3d operator*(const LLVector3d &a, const LLQuaternion &rot); // Rotates a by rot - - // Non-standard operators - friend F32 dot(const LLQuaternion &a, const LLQuaternion &b); - friend LLQuaternion lerp(F32 t, const LLQuaternion &p, const LLQuaternion &q); // linear interpolation (t = 0 to 1) from p to q - friend LLQuaternion lerp(F32 t, const LLQuaternion &q); // linear interpolation (t = 0 to 1) from identity to q - friend LLQuaternion slerp(F32 t, const LLQuaternion &p, const LLQuaternion &q); // spherical linear interpolation from p to q - friend LLQuaternion slerp(F32 t, const LLQuaternion &q); // spherical linear interpolation from identity to q - friend LLQuaternion nlerp(F32 t, const LLQuaternion &p, const LLQuaternion &q); // normalized linear interpolation from p to q - friend LLQuaternion nlerp(F32 t, const LLQuaternion &q); // normalized linear interpolation from p to q - - LLVector3 packToVector3() const; // Saves space by using the fact that our quaternions are normalized - void unpackFromVector3(const LLVector3& vec); // Saves space by using the fact that our quaternions are normalized - - enum Order { - XYZ = 0, - YZX = 1, - ZXY = 2, - XZY = 3, - YXZ = 4, - ZYX = 5 - }; - // Creates a quaternions from maya's rotation representation, - // which is 3 rotations (in DEGREES) in the specified order - friend LLQuaternion mayaQ(F32 x, F32 y, F32 z, Order order); - - // Conversions between Order and strings like "xyz" or "ZYX" - friend const char *OrderToString( const Order order ); - friend Order StringToOrder( const char *str ); - - static BOOL parseQuat(const std::string& buf, LLQuaternion* value); - - // For debugging, only - //static U32 mMultCount; -}; - -// checker -inline BOOL LLQuaternion::isFinite() const -{ - return (llfinite(mQ[VX]) && llfinite(mQ[VY]) && llfinite(mQ[VZ]) && llfinite(mQ[VS])); -} - -inline BOOL LLQuaternion::isIdentity() const -{ - return - ( mQ[VX] == 0.f ) && - ( mQ[VY] == 0.f ) && - ( mQ[VZ] == 0.f ) && - ( mQ[VS] == 1.f ); -} - -inline BOOL LLQuaternion::isNotIdentity() const -{ - return - ( mQ[VX] != 0.f ) || - ( mQ[VY] != 0.f ) || - ( mQ[VZ] != 0.f ) || - ( mQ[VS] != 1.f ); -} - - - -inline LLQuaternion::LLQuaternion(void) -{ - mQ[VX] = 0.f; - mQ[VY] = 0.f; - mQ[VZ] = 0.f; - mQ[VS] = 1.f; -} - -inline LLQuaternion::LLQuaternion(F32 x, F32 y, F32 z, F32 w) -{ - mQ[VX] = x; - mQ[VY] = y; - mQ[VZ] = z; - mQ[VS] = w; - - //RN: don't normalize this case as its used mainly for temporaries during calculations - //normalize(); - /* - F32 mag = sqrtf(mQ[VX]*mQ[VX] + mQ[VY]*mQ[VY] + mQ[VZ]*mQ[VZ] + mQ[VS]*mQ[VS]); - mag -= 1.f; - mag = fabs(mag); - llassert(mag < 10.f*FP_MAG_THRESHOLD); - */ -} - -inline LLQuaternion::LLQuaternion(const F32 *q) -{ - mQ[VX] = q[VX]; - mQ[VY] = q[VY]; - mQ[VZ] = q[VZ]; - mQ[VS] = q[VW]; - - normalize(); - /* - F32 mag = sqrtf(mQ[VX]*mQ[VX] + mQ[VY]*mQ[VY] + mQ[VZ]*mQ[VZ] + mQ[VS]*mQ[VS]); - mag -= 1.f; - mag = fabs(mag); - llassert(mag < FP_MAG_THRESHOLD); - */ -} - - -inline void LLQuaternion::loadIdentity() -{ - mQ[VX] = 0.0f; - mQ[VY] = 0.0f; - mQ[VZ] = 0.0f; - mQ[VW] = 1.0f; -} - - -inline const LLQuaternion& LLQuaternion::set(F32 x, F32 y, F32 z, F32 w) -{ - mQ[VX] = x; - mQ[VY] = y; - mQ[VZ] = z; - mQ[VS] = w; - normalize(); - return (*this); -} - -inline const LLQuaternion& LLQuaternion::set(const LLQuaternion &quat) -{ - mQ[VX] = quat.mQ[VX]; - mQ[VY] = quat.mQ[VY]; - mQ[VZ] = quat.mQ[VZ]; - mQ[VW] = quat.mQ[VW]; - normalize(); - return (*this); -} - -inline const LLQuaternion& LLQuaternion::set(const F32 *q) -{ - mQ[VX] = q[VX]; - mQ[VY] = q[VY]; - mQ[VZ] = q[VZ]; - mQ[VS] = q[VW]; - normalize(); - return (*this); -} - - -// deprecated -inline const LLQuaternion& LLQuaternion::setQuatInit(F32 x, F32 y, F32 z, F32 w) -{ - mQ[VX] = x; - mQ[VY] = y; - mQ[VZ] = z; - mQ[VS] = w; - normalize(); - return (*this); -} - -// deprecated -inline const LLQuaternion& LLQuaternion::setQuat(const LLQuaternion &quat) -{ - mQ[VX] = quat.mQ[VX]; - mQ[VY] = quat.mQ[VY]; - mQ[VZ] = quat.mQ[VZ]; - mQ[VW] = quat.mQ[VW]; - normalize(); - return (*this); -} - -// deprecated -inline const LLQuaternion& LLQuaternion::setQuat(const F32 *q) -{ - mQ[VX] = q[VX]; - mQ[VY] = q[VY]; - mQ[VZ] = q[VZ]; - mQ[VS] = q[VW]; - normalize(); - return (*this); -} - -// There may be a cheaper way that avoids the sqrt. -// Does sin_a = VX*VX + VY*VY + VZ*VZ? -// Copied from Matrix and Quaternion FAQ 1.12 -inline void LLQuaternion::getAngleAxis(F32* angle, F32* x, F32* y, F32* z) const -{ - F32 cos_a = mQ[VW]; - if (cos_a > 1.0f) cos_a = 1.0f; - if (cos_a < -1.0f) cos_a = -1.0f; - - F32 sin_a = (F32) sqrt( 1.0f - cos_a * cos_a ); - - if ( fabs( sin_a ) < 0.0005f ) - sin_a = 1.0f; - else - sin_a = 1.f/sin_a; - - F32 temp_angle = 2.0f * (F32) acos( cos_a ); - if (temp_angle > F_PI) - { - // The (angle,axis) pair should never have angles outside [PI, -PI] - // since we want the _shortest_ (angle,axis) solution. - // Since acos is defined for [0, PI], and we multiply by 2.0, we - // can push the angle outside the acceptible range. - // When this happens we set the angle to the other portion of a - // full 2PI rotation, and negate the axis, which reverses the - // direction of the rotation (by the right-hand rule). - *angle = 2.f * F_PI - temp_angle; - *x = - mQ[VX] * sin_a; - *y = - mQ[VY] * sin_a; - *z = - mQ[VZ] * sin_a; - } - else - { - *angle = temp_angle; - *x = mQ[VX] * sin_a; - *y = mQ[VY] * sin_a; - *z = mQ[VZ] * sin_a; - } -} - -inline const LLQuaternion& LLQuaternion::conjugate() -{ - mQ[VX] *= -1.f; - mQ[VY] *= -1.f; - mQ[VZ] *= -1.f; - return (*this); -} - -inline const LLQuaternion& LLQuaternion::conjQuat() -{ - mQ[VX] *= -1.f; - mQ[VY] *= -1.f; - mQ[VZ] *= -1.f; - return (*this); -} - -// Transpose -inline const LLQuaternion& LLQuaternion::transpose() -{ - mQ[VX] *= -1.f; - mQ[VY] *= -1.f; - mQ[VZ] *= -1.f; - return (*this); -} - -// deprecated -inline const LLQuaternion& LLQuaternion::transQuat() -{ - mQ[VX] *= -1.f; - mQ[VY] *= -1.f; - mQ[VZ] *= -1.f; - return (*this); -} - - -inline LLQuaternion operator+(const LLQuaternion &a, const LLQuaternion &b) -{ - return LLQuaternion( - a.mQ[VX] + b.mQ[VX], - a.mQ[VY] + b.mQ[VY], - a.mQ[VZ] + b.mQ[VZ], - a.mQ[VW] + b.mQ[VW] ); -} - - -inline LLQuaternion operator-(const LLQuaternion &a, const LLQuaternion &b) -{ - return LLQuaternion( - a.mQ[VX] - b.mQ[VX], - a.mQ[VY] - b.mQ[VY], - a.mQ[VZ] - b.mQ[VZ], - a.mQ[VW] - b.mQ[VW] ); -} - - -inline LLQuaternion operator-(const LLQuaternion &a) -{ - return LLQuaternion( - -a.mQ[VX], - -a.mQ[VY], - -a.mQ[VZ], - -a.mQ[VW] ); -} - - -inline LLQuaternion operator*(F32 a, const LLQuaternion &q) -{ - return LLQuaternion( - a * q.mQ[VX], - a * q.mQ[VY], - a * q.mQ[VZ], - a * q.mQ[VW] ); -} - - -inline LLQuaternion operator*(const LLQuaternion &q, F32 a) -{ - return LLQuaternion( - a * q.mQ[VX], - a * q.mQ[VY], - a * q.mQ[VZ], - a * q.mQ[VW] ); -} - -inline LLQuaternion operator~(const LLQuaternion &a) -{ - LLQuaternion q(a); - q.conjQuat(); - return q; -} - -inline bool LLQuaternion::operator==(const LLQuaternion &b) const -{ - return ( (mQ[VX] == b.mQ[VX]) - &&(mQ[VY] == b.mQ[VY]) - &&(mQ[VZ] == b.mQ[VZ]) - &&(mQ[VS] == b.mQ[VS])); -} - -inline bool LLQuaternion::operator!=(const LLQuaternion &b) const -{ - return ( (mQ[VX] != b.mQ[VX]) - ||(mQ[VY] != b.mQ[VY]) - ||(mQ[VZ] != b.mQ[VZ]) - ||(mQ[VS] != b.mQ[VS])); -} - -inline const LLQuaternion& operator*=(LLQuaternion &a, const LLQuaternion &b) -{ -#if 1 - LLQuaternion q( - b.mQ[3] * a.mQ[0] + b.mQ[0] * a.mQ[3] + b.mQ[1] * a.mQ[2] - b.mQ[2] * a.mQ[1], - b.mQ[3] * a.mQ[1] + b.mQ[1] * a.mQ[3] + b.mQ[2] * a.mQ[0] - b.mQ[0] * a.mQ[2], - b.mQ[3] * a.mQ[2] + b.mQ[2] * a.mQ[3] + b.mQ[0] * a.mQ[1] - b.mQ[1] * a.mQ[0], - b.mQ[3] * a.mQ[3] - b.mQ[0] * a.mQ[0] - b.mQ[1] * a.mQ[1] - b.mQ[2] * a.mQ[2] - ); - a = q; -#else - a = a * b; -#endif - return a; -} - -const F32 ONE_PART_IN_A_MILLION = 0.000001f; - -inline F32 LLQuaternion::normalize() -{ - F32 mag = sqrtf(mQ[VX]*mQ[VX] + mQ[VY]*mQ[VY] + mQ[VZ]*mQ[VZ] + mQ[VS]*mQ[VS]); - - if (mag > FP_MAG_THRESHOLD) - { - // Floating point error can prevent some quaternions from achieving - // exact unity length. When trying to renormalize such quaternions we - // can oscillate between multiple quantized states. To prevent such - // drifts we only renomalize if the length is far enough from unity. - if (fabs(1.f - mag) > ONE_PART_IN_A_MILLION) - { - F32 oomag = 1.f/mag; - mQ[VX] *= oomag; - mQ[VY] *= oomag; - mQ[VZ] *= oomag; - mQ[VS] *= oomag; - } - } - else - { - // we were given a very bad quaternion so we set it to identity - mQ[VX] = 0.f; - mQ[VY] = 0.f; - mQ[VZ] = 0.f; - mQ[VS] = 1.f; - } - - return mag; -} - -// deprecated -inline F32 LLQuaternion::normQuat() -{ - F32 mag = sqrtf(mQ[VX]*mQ[VX] + mQ[VY]*mQ[VY] + mQ[VZ]*mQ[VZ] + mQ[VS]*mQ[VS]); - - if (mag > FP_MAG_THRESHOLD) - { - if (fabs(1.f - mag) > ONE_PART_IN_A_MILLION) - { - // only renormalize if length not close enough to 1.0 already - F32 oomag = 1.f/mag; - mQ[VX] *= oomag; - mQ[VY] *= oomag; - mQ[VZ] *= oomag; - mQ[VS] *= oomag; - } - } - else - { - mQ[VX] = 0.f; - mQ[VY] = 0.f; - mQ[VZ] = 0.f; - mQ[VS] = 1.f; - } - - return mag; -} - -LLQuaternion::Order StringToOrder( const char *str ); - -// Some notes about Quaternions - -// What is a Quaternion? -// --------------------- -// A quaternion is a point in 4-dimensional complex space. -// Q = { Qx, Qy, Qz, Qw } -// -// -// Why Quaternions? -// ---------------- -// The set of quaternions that make up the the 4-D unit sphere -// can be mapped to the set of all rotations in 3-D space. Sometimes -// it is easier to describe/manipulate rotations in quaternion space -// than rotation-matrix space. -// -// -// How Quaternions? -// ---------------- -// In order to take advantage of quaternions we need to know how to -// go from rotation-matricies to quaternions and back. We also have -// to agree what variety of rotations we're generating. -// -// Consider the equation... v' = v * R -// -// There are two ways to think about rotations of vectors. -// 1) v' is the same vector in a different reference frame -// 2) v' is a new vector in the same reference frame -// -// bookmark -- which way are we using? -// -// -// Quaternion from Angle-Axis: -// --------------------------- -// Suppose we wanted to represent a rotation of some angle (theta) -// about some axis ({Ax, Ay, Az})... -// -// axis of rotation = {Ax, Ay, Az} -// angle_of_rotation = theta -// -// s = sin(0.5 * theta) -// c = cos(0.5 * theta) -// Q = { s * Ax, s * Ay, s * Az, c } -// -// -// 3x3 Matrix from Quaternion -// -------------------------- -// -// | | -// | 1 - 2 * (y^2 + z^2) 2 * (x * y + z * w) 2 * (y * w - x * z) | -// | | -// M = | 2 * (x * y - z * w) 1 - 2 * (x^2 + z^2) 2 * (y * z + x * w) | -// | | -// | 2 * (x * z + y * w) 2 * (y * z - x * w) 1 - 2 * (x^2 + y^2) | -// | | - -#endif +/** + * @file llquaternion.h + * @brief LLQuaternion class header file. + * + * $LicenseInfo:firstyear=2000&license=viewergpl$ + * + * Copyright (c) 2000-2009, Linden Research, Inc. + * + * Second Life Viewer Source Code + * The source code in this file ("Source Code") is provided by Linden Lab + * to you under the terms of the GNU General Public License, version 2.0 + * ("GPL"), unless you have obtained a separate licensing agreement + * ("Other License"), formally executed by you and Linden Lab. Terms of + * the GPL can be found in doc/GPL-license.txt in this distribution, or + * online at http://secondlifegrid.net/programs/open_source/licensing/gplv2 + * + * There are special exceptions to the terms and conditions of the GPL as + * it is applied to this Source Code. View the full text of the exception + * in the file doc/FLOSS-exception.txt in this software distribution, or + * online at + * http://secondlifegrid.net/programs/open_source/licensing/flossexception + * + * By copying, modifying or distributing this software, you acknowledge + * that you have read and understood your obligations described above, + * and agree to abide by those obligations. + * + * ALL LINDEN LAB SOURCE CODE IS PROVIDED "AS IS." LINDEN LAB MAKES NO + * WARRANTIES, EXPRESS, IMPLIED OR OTHERWISE, REGARDING ITS ACCURACY, + * COMPLETENESS OR PERFORMANCE. + * $/LicenseInfo$ + */ + +#ifndef LLQUATERNION_H +#define LLQUATERNION_H + +#include + +#ifndef LLMATH_H //enforce specific include order to avoid tangling inline dependencies +#error "Please include llmath.h first." +#endif + +class LLVector4; +class LLVector3; +class LLVector3d; +class LLMatrix4; +class LLMatrix3; + +// NOTA BENE: Quaternion code is written assuming Unit Quaternions!!!! +// Moreover, it is written assuming that all vectors and matricies +// passed as arguments are normalized and unitary respectively. +// VERY VERY VERY VERY BAD THINGS will happen if these assumptions fail. + +static const U32 LENGTHOFQUAT = 4; + +class LLQuaternion +{ +public: + F32 mQ[LENGTHOFQUAT]; + + static const LLQuaternion DEFAULT; + + LLQuaternion(); // Initializes Quaternion to (0,0,0,1) + explicit LLQuaternion(const LLMatrix4 &mat); // Initializes Quaternion from Matrix4 + explicit LLQuaternion(const LLMatrix3 &mat); // Initializes Quaternion from Matrix3 + LLQuaternion(F32 x, F32 y, F32 z, F32 w); // Initializes Quaternion to normalize(x, y, z, w) + LLQuaternion(F32 angle, const LLVector4 &vec); // Initializes Quaternion to axis_angle2quat(angle, vec) + LLQuaternion(F32 angle, const LLVector3 &vec); // Initializes Quaternion to axis_angle2quat(angle, vec) + LLQuaternion(const F32 *q); // Initializes Quaternion to normalize(x, y, z, w) + LLQuaternion(const LLVector3 &x_axis, + const LLVector3 &y_axis, + const LLVector3 &z_axis); // Initializes Quaternion from Matrix3 = [x_axis ; y_axis ; z_axis] + + BOOL isIdentity() const; + BOOL isNotIdentity() const; + BOOL isFinite() const; // checks to see if all values of LLQuaternion are finite + void quantize16(F32 lower, F32 upper); // changes the vector to reflect quatization + void quantize8(F32 lower, F32 upper); // changes the vector to reflect quatization + void loadIdentity(); // Loads the quaternion that represents the identity rotation + + const LLQuaternion& set(F32 x, F32 y, F32 z, F32 w); // Sets Quaternion to normalize(x, y, z, w) + const LLQuaternion& set(const LLQuaternion &quat); // Copies Quaternion + const LLQuaternion& set(const F32 *q); // Sets Quaternion to normalize(quat[VX], quat[VY], quat[VZ], quat[VW]) + const LLQuaternion& set(const LLMatrix3 &mat); // Sets Quaternion to mat2quat(mat) + const LLQuaternion& set(const LLMatrix4 &mat); // Sets Quaternion to mat2quat(mat) + + const LLQuaternion& setAngleAxis(F32 angle, F32 x, F32 y, F32 z); // Sets Quaternion to axis_angle2quat(angle, x, y, z) + const LLQuaternion& setAngleAxis(F32 angle, const LLVector3 &vec); // Sets Quaternion to axis_angle2quat(angle, vec) + const LLQuaternion& setAngleAxis(F32 angle, const LLVector4 &vec); // Sets Quaternion to axis_angle2quat(angle, vec) + const LLQuaternion& setEulerAngles(F32 roll, F32 pitch, F32 yaw); // Sets Quaternion to euler2quat(pitch, yaw, roll) + + const LLQuaternion& setQuatInit(F32 x, F32 y, F32 z, F32 w); // deprecated + const LLQuaternion& setQuat(const LLQuaternion &quat); // deprecated + const LLQuaternion& setQuat(const F32 *q); // deprecated + const LLQuaternion& setQuat(const LLMatrix3 &mat); // deprecated + const LLQuaternion& setQuat(const LLMatrix4 &mat); // deprecated + const LLQuaternion& setQuat(F32 angle, F32 x, F32 y, F32 z); // deprecated + const LLQuaternion& setQuat(F32 angle, const LLVector3 &vec); // deprecated + const LLQuaternion& setQuat(F32 angle, const LLVector4 &vec); // deprecated + const LLQuaternion& setQuat(F32 roll, F32 pitch, F32 yaw); // deprecated + + LLMatrix4 getMatrix4(void) const; // Returns the Matrix4 equivalent of Quaternion + LLMatrix3 getMatrix3(void) const; // Returns the Matrix3 equivalent of Quaternion + void getAngleAxis(F32* angle, F32* x, F32* y, F32* z) const; // returns rotation in radians about axis x,y,z + void getAngleAxis(F32* angle, LLVector3 &vec) const; + void getEulerAngles(F32 *roll, F32* pitch, F32 *yaw) const; + + F32 normalize(); // Normalizes Quaternion and returns magnitude + F32 normQuat(); // deprecated + + const LLQuaternion& conjugate(void); // Conjugates Quaternion and returns result + const LLQuaternion& conjQuat(void); // deprecated + + // Other useful methods + const LLQuaternion& transpose(); // transpose (same as conjugate) + const LLQuaternion& transQuat(); // deprecated + + void shortestArc(const LLVector3 &a, const LLVector3 &b); // shortest rotation from a to b + const LLQuaternion& constrain(F32 radians); // constrains rotation to a cone angle specified in radians + + // Standard operators + friend std::ostream& operator<<(std::ostream &s, const LLQuaternion &a); // Prints a + friend LLQuaternion operator+(const LLQuaternion &a, const LLQuaternion &b); // Addition + friend LLQuaternion operator-(const LLQuaternion &a, const LLQuaternion &b); // Subtraction + friend LLQuaternion operator-(const LLQuaternion &a); // Negation + friend LLQuaternion operator*(F32 a, const LLQuaternion &q); // Scale + friend LLQuaternion operator*(const LLQuaternion &q, F32 b); // Scale + friend LLQuaternion operator*(const LLQuaternion &a, const LLQuaternion &b); // Returns a * b + friend LLQuaternion operator~(const LLQuaternion &a); // Returns a* (Conjugate of a) + bool operator==(const LLQuaternion &b) const; // Returns a == b + bool operator!=(const LLQuaternion &b) const; // Returns a != b + + friend const LLQuaternion& operator*=(LLQuaternion &a, const LLQuaternion &b); // Returns a * b + + friend LLVector4 operator*(const LLVector4 &a, const LLQuaternion &rot); // Rotates a by rot + friend LLVector3 operator*(const LLVector3 &a, const LLQuaternion &rot); // Rotates a by rot + friend LLVector3d operator*(const LLVector3d &a, const LLQuaternion &rot); // Rotates a by rot + + // Non-standard operators + friend F32 dot(const LLQuaternion &a, const LLQuaternion &b); + friend LLQuaternion lerp(F32 t, const LLQuaternion &p, const LLQuaternion &q); // linear interpolation (t = 0 to 1) from p to q + friend LLQuaternion lerp(F32 t, const LLQuaternion &q); // linear interpolation (t = 0 to 1) from identity to q + friend LLQuaternion slerp(F32 t, const LLQuaternion &p, const LLQuaternion &q); // spherical linear interpolation from p to q + friend LLQuaternion slerp(F32 t, const LLQuaternion &q); // spherical linear interpolation from identity to q + friend LLQuaternion nlerp(F32 t, const LLQuaternion &p, const LLQuaternion &q); // normalized linear interpolation from p to q + friend LLQuaternion nlerp(F32 t, const LLQuaternion &q); // normalized linear interpolation from p to q + + LLVector3 packToVector3() const; // Saves space by using the fact that our quaternions are normalized + void unpackFromVector3(const LLVector3& vec); // Saves space by using the fact that our quaternions are normalized + + enum Order { + XYZ = 0, + YZX = 1, + ZXY = 2, + XZY = 3, + YXZ = 4, + ZYX = 5 + }; + // Creates a quaternions from maya's rotation representation, + // which is 3 rotations (in DEGREES) in the specified order + friend LLQuaternion mayaQ(F32 x, F32 y, F32 z, Order order); + + // Conversions between Order and strings like "xyz" or "ZYX" + friend const char *OrderToString( const Order order ); + friend Order StringToOrder( const char *str ); + + static BOOL parseQuat(const std::string& buf, LLQuaternion* value); + + // For debugging, only + //static U32 mMultCount; +}; + +// checker +inline BOOL LLQuaternion::isFinite() const +{ + return (llfinite(mQ[VX]) && llfinite(mQ[VY]) && llfinite(mQ[VZ]) && llfinite(mQ[VS])); +} + +inline BOOL LLQuaternion::isIdentity() const +{ + return + ( mQ[VX] == 0.f ) && + ( mQ[VY] == 0.f ) && + ( mQ[VZ] == 0.f ) && + ( mQ[VS] == 1.f ); +} + +inline BOOL LLQuaternion::isNotIdentity() const +{ + return + ( mQ[VX] != 0.f ) || + ( mQ[VY] != 0.f ) || + ( mQ[VZ] != 0.f ) || + ( mQ[VS] != 1.f ); +} + + + +inline LLQuaternion::LLQuaternion(void) +{ + mQ[VX] = 0.f; + mQ[VY] = 0.f; + mQ[VZ] = 0.f; + mQ[VS] = 1.f; +} + +inline LLQuaternion::LLQuaternion(F32 x, F32 y, F32 z, F32 w) +{ + mQ[VX] = x; + mQ[VY] = y; + mQ[VZ] = z; + mQ[VS] = w; + + //RN: don't normalize this case as its used mainly for temporaries during calculations + //normalize(); + /* + F32 mag = sqrtf(mQ[VX]*mQ[VX] + mQ[VY]*mQ[VY] + mQ[VZ]*mQ[VZ] + mQ[VS]*mQ[VS]); + mag -= 1.f; + mag = fabs(mag); + llassert(mag < 10.f*FP_MAG_THRESHOLD); + */ +} + +inline LLQuaternion::LLQuaternion(const F32 *q) +{ + mQ[VX] = q[VX]; + mQ[VY] = q[VY]; + mQ[VZ] = q[VZ]; + mQ[VS] = q[VW]; + + normalize(); + /* + F32 mag = sqrtf(mQ[VX]*mQ[VX] + mQ[VY]*mQ[VY] + mQ[VZ]*mQ[VZ] + mQ[VS]*mQ[VS]); + mag -= 1.f; + mag = fabs(mag); + llassert(mag < FP_MAG_THRESHOLD); + */ +} + + +inline void LLQuaternion::loadIdentity() +{ + mQ[VX] = 0.0f; + mQ[VY] = 0.0f; + mQ[VZ] = 0.0f; + mQ[VW] = 1.0f; +} + + +inline const LLQuaternion& LLQuaternion::set(F32 x, F32 y, F32 z, F32 w) +{ + mQ[VX] = x; + mQ[VY] = y; + mQ[VZ] = z; + mQ[VS] = w; + normalize(); + return (*this); +} + +inline const LLQuaternion& LLQuaternion::set(const LLQuaternion &quat) +{ + mQ[VX] = quat.mQ[VX]; + mQ[VY] = quat.mQ[VY]; + mQ[VZ] = quat.mQ[VZ]; + mQ[VW] = quat.mQ[VW]; + normalize(); + return (*this); +} + +inline const LLQuaternion& LLQuaternion::set(const F32 *q) +{ + mQ[VX] = q[VX]; + mQ[VY] = q[VY]; + mQ[VZ] = q[VZ]; + mQ[VS] = q[VW]; + normalize(); + return (*this); +} + + +// deprecated +inline const LLQuaternion& LLQuaternion::setQuatInit(F32 x, F32 y, F32 z, F32 w) +{ + mQ[VX] = x; + mQ[VY] = y; + mQ[VZ] = z; + mQ[VS] = w; + normalize(); + return (*this); +} + +// deprecated +inline const LLQuaternion& LLQuaternion::setQuat(const LLQuaternion &quat) +{ + mQ[VX] = quat.mQ[VX]; + mQ[VY] = quat.mQ[VY]; + mQ[VZ] = quat.mQ[VZ]; + mQ[VW] = quat.mQ[VW]; + normalize(); + return (*this); +} + +// deprecated +inline const LLQuaternion& LLQuaternion::setQuat(const F32 *q) +{ + mQ[VX] = q[VX]; + mQ[VY] = q[VY]; + mQ[VZ] = q[VZ]; + mQ[VS] = q[VW]; + normalize(); + return (*this); +} + +// There may be a cheaper way that avoids the sqrt. +// Does sin_a = VX*VX + VY*VY + VZ*VZ? +// Copied from Matrix and Quaternion FAQ 1.12 +inline void LLQuaternion::getAngleAxis(F32* angle, F32* x, F32* y, F32* z) const +{ + F32 cos_a = mQ[VW]; + if (cos_a > 1.0f) cos_a = 1.0f; + if (cos_a < -1.0f) cos_a = -1.0f; + + F32 sin_a = (F32) sqrt( 1.0f - cos_a * cos_a ); + + if ( fabs( sin_a ) < 0.0005f ) + sin_a = 1.0f; + else + sin_a = 1.f/sin_a; + + F32 temp_angle = 2.0f * (F32) acos( cos_a ); + if (temp_angle > F_PI) + { + // The (angle,axis) pair should never have angles outside [PI, -PI] + // since we want the _shortest_ (angle,axis) solution. + // Since acos is defined for [0, PI], and we multiply by 2.0, we + // can push the angle outside the acceptible range. + // When this happens we set the angle to the other portion of a + // full 2PI rotation, and negate the axis, which reverses the + // direction of the rotation (by the right-hand rule). + *angle = 2.f * F_PI - temp_angle; + *x = - mQ[VX] * sin_a; + *y = - mQ[VY] * sin_a; + *z = - mQ[VZ] * sin_a; + } + else + { + *angle = temp_angle; + *x = mQ[VX] * sin_a; + *y = mQ[VY] * sin_a; + *z = mQ[VZ] * sin_a; + } +} + +inline const LLQuaternion& LLQuaternion::conjugate() +{ + mQ[VX] *= -1.f; + mQ[VY] *= -1.f; + mQ[VZ] *= -1.f; + return (*this); +} + +inline const LLQuaternion& LLQuaternion::conjQuat() +{ + mQ[VX] *= -1.f; + mQ[VY] *= -1.f; + mQ[VZ] *= -1.f; + return (*this); +} + +// Transpose +inline const LLQuaternion& LLQuaternion::transpose() +{ + mQ[VX] *= -1.f; + mQ[VY] *= -1.f; + mQ[VZ] *= -1.f; + return (*this); +} + +// deprecated +inline const LLQuaternion& LLQuaternion::transQuat() +{ + mQ[VX] *= -1.f; + mQ[VY] *= -1.f; + mQ[VZ] *= -1.f; + return (*this); +} + + +inline LLQuaternion operator+(const LLQuaternion &a, const LLQuaternion &b) +{ + return LLQuaternion( + a.mQ[VX] + b.mQ[VX], + a.mQ[VY] + b.mQ[VY], + a.mQ[VZ] + b.mQ[VZ], + a.mQ[VW] + b.mQ[VW] ); +} + + +inline LLQuaternion operator-(const LLQuaternion &a, const LLQuaternion &b) +{ + return LLQuaternion( + a.mQ[VX] - b.mQ[VX], + a.mQ[VY] - b.mQ[VY], + a.mQ[VZ] - b.mQ[VZ], + a.mQ[VW] - b.mQ[VW] ); +} + + +inline LLQuaternion operator-(const LLQuaternion &a) +{ + return LLQuaternion( + -a.mQ[VX], + -a.mQ[VY], + -a.mQ[VZ], + -a.mQ[VW] ); +} + + +inline LLQuaternion operator*(F32 a, const LLQuaternion &q) +{ + return LLQuaternion( + a * q.mQ[VX], + a * q.mQ[VY], + a * q.mQ[VZ], + a * q.mQ[VW] ); +} + + +inline LLQuaternion operator*(const LLQuaternion &q, F32 a) +{ + return LLQuaternion( + a * q.mQ[VX], + a * q.mQ[VY], + a * q.mQ[VZ], + a * q.mQ[VW] ); +} + +inline LLQuaternion operator~(const LLQuaternion &a) +{ + LLQuaternion q(a); + q.conjQuat(); + return q; +} + +inline bool LLQuaternion::operator==(const LLQuaternion &b) const +{ + return ( (mQ[VX] == b.mQ[VX]) + &&(mQ[VY] == b.mQ[VY]) + &&(mQ[VZ] == b.mQ[VZ]) + &&(mQ[VS] == b.mQ[VS])); +} + +inline bool LLQuaternion::operator!=(const LLQuaternion &b) const +{ + return ( (mQ[VX] != b.mQ[VX]) + ||(mQ[VY] != b.mQ[VY]) + ||(mQ[VZ] != b.mQ[VZ]) + ||(mQ[VS] != b.mQ[VS])); +} + +inline const LLQuaternion& operator*=(LLQuaternion &a, const LLQuaternion &b) +{ +#if 1 + LLQuaternion q( + b.mQ[3] * a.mQ[0] + b.mQ[0] * a.mQ[3] + b.mQ[1] * a.mQ[2] - b.mQ[2] * a.mQ[1], + b.mQ[3] * a.mQ[1] + b.mQ[1] * a.mQ[3] + b.mQ[2] * a.mQ[0] - b.mQ[0] * a.mQ[2], + b.mQ[3] * a.mQ[2] + b.mQ[2] * a.mQ[3] + b.mQ[0] * a.mQ[1] - b.mQ[1] * a.mQ[0], + b.mQ[3] * a.mQ[3] - b.mQ[0] * a.mQ[0] - b.mQ[1] * a.mQ[1] - b.mQ[2] * a.mQ[2] + ); + a = q; +#else + a = a * b; +#endif + return a; +} + +const F32 ONE_PART_IN_A_MILLION = 0.000001f; + +inline F32 LLQuaternion::normalize() +{ + F32 mag = sqrtf(mQ[VX]*mQ[VX] + mQ[VY]*mQ[VY] + mQ[VZ]*mQ[VZ] + mQ[VS]*mQ[VS]); + + if (mag > FP_MAG_THRESHOLD) + { + // Floating point error can prevent some quaternions from achieving + // exact unity length. When trying to renormalize such quaternions we + // can oscillate between multiple quantized states. To prevent such + // drifts we only renomalize if the length is far enough from unity. + if (fabs(1.f - mag) > ONE_PART_IN_A_MILLION) + { + F32 oomag = 1.f/mag; + mQ[VX] *= oomag; + mQ[VY] *= oomag; + mQ[VZ] *= oomag; + mQ[VS] *= oomag; + } + } + else + { + // we were given a very bad quaternion so we set it to identity + mQ[VX] = 0.f; + mQ[VY] = 0.f; + mQ[VZ] = 0.f; + mQ[VS] = 1.f; + } + + return mag; +} + +// deprecated +inline F32 LLQuaternion::normQuat() +{ + F32 mag = sqrtf(mQ[VX]*mQ[VX] + mQ[VY]*mQ[VY] + mQ[VZ]*mQ[VZ] + mQ[VS]*mQ[VS]); + + if (mag > FP_MAG_THRESHOLD) + { + if (fabs(1.f - mag) > ONE_PART_IN_A_MILLION) + { + // only renormalize if length not close enough to 1.0 already + F32 oomag = 1.f/mag; + mQ[VX] *= oomag; + mQ[VY] *= oomag; + mQ[VZ] *= oomag; + mQ[VS] *= oomag; + } + } + else + { + mQ[VX] = 0.f; + mQ[VY] = 0.f; + mQ[VZ] = 0.f; + mQ[VS] = 1.f; + } + + return mag; +} + +LLQuaternion::Order StringToOrder( const char *str ); + +// Some notes about Quaternions + +// What is a Quaternion? +// --------------------- +// A quaternion is a point in 4-dimensional complex space. +// Q = { Qx, Qy, Qz, Qw } +// +// +// Why Quaternions? +// ---------------- +// The set of quaternions that make up the the 4-D unit sphere +// can be mapped to the set of all rotations in 3-D space. Sometimes +// it is easier to describe/manipulate rotations in quaternion space +// than rotation-matrix space. +// +// +// How Quaternions? +// ---------------- +// In order to take advantage of quaternions we need to know how to +// go from rotation-matricies to quaternions and back. We also have +// to agree what variety of rotations we're generating. +// +// Consider the equation... v' = v * R +// +// There are two ways to think about rotations of vectors. +// 1) v' is the same vector in a different reference frame +// 2) v' is a new vector in the same reference frame +// +// bookmark -- which way are we using? +// +// +// Quaternion from Angle-Axis: +// --------------------------- +// Suppose we wanted to represent a rotation of some angle (theta) +// about some axis ({Ax, Ay, Az})... +// +// axis of rotation = {Ax, Ay, Az} +// angle_of_rotation = theta +// +// s = sin(0.5 * theta) +// c = cos(0.5 * theta) +// Q = { s * Ax, s * Ay, s * Az, c } +// +// +// 3x3 Matrix from Quaternion +// -------------------------- +// +// | | +// | 1 - 2 * (y^2 + z^2) 2 * (x * y + z * w) 2 * (y * w - x * z) | +// | | +// M = | 2 * (x * y - z * w) 1 - 2 * (x^2 + z^2) 2 * (y * z + x * w) | +// | | +// | 2 * (x * z + y * w) 2 * (y * z - x * w) 1 - 2 * (x^2 + y^2) | +// | | + +#endif diff --git a/indra/llmath/llvolume.cpp b/indra/llmath/llvolume.cpp index bba0a6d089..ab9f8c4c24 100644 --- a/indra/llmath/llvolume.cpp +++ b/indra/llmath/llvolume.cpp @@ -45,7 +45,7 @@ #include "v4math.h" #include "m4math.h" #include "m3math.h" -#include "llmatrix4a.h" +#include "llmatrix3a.h" #include "lloctree.h" #include "lldarray.h" #include "llvolume.h" @@ -53,6 +53,7 @@ #include "llstl.h" #include "llsdserialize.h" #include "llvector4a.h" +#include "llmatrix4a.h" #define DEBUG_SILHOUETTE_BINORMALS 0 #define DEBUG_SILHOUETTE_NORMALS 0 // TomY: Use this to display normals using the silhouette @@ -161,7 +162,7 @@ BOOL LLTriangleRayIntersect(const LLVector4a& vert0, const LLVector4a& vert1, co LLVector4a det; det.setAllDot3(edge1, pvec); - if (det.greaterEqual4(LLVector4a::getApproximatelyZero()).getComparisonMask() & 0x7) + if (det.greaterEqual(LLVector4a::getEpsilon()).getGatheredBits() & 0x7) { /* calculate distance from vert0 to ray origin */ LLVector4a tvec; @@ -171,8 +172,8 @@ BOOL LLTriangleRayIntersect(const LLVector4a& vert0, const LLVector4a& vert1, co LLVector4a u; u.setAllDot3(tvec,pvec); - if ((u.greaterEqual4(LLVector4a::getZero()).getComparisonMask() & 0x7) && - (u.lessEqual4(det).getComparisonMask() & 0x7)) + if ((u.greaterEqual(LLVector4a::getZero()).getGatheredBits() & 0x7) && + (u.lessEqual(det).getGatheredBits() & 0x7)) { /* prepare to test V parameter */ LLVector4a qvec; @@ -188,8 +189,8 @@ BOOL LLTriangleRayIntersect(const LLVector4a& vert0, const LLVector4a& vert1, co LLVector4a sum_uv; sum_uv.setAdd(u, v); - S32 v_gequal = v.greaterEqual4(LLVector4a::getZero()).getComparisonMask() & 0x7; - S32 sum_lequal = sum_uv.lessEqual4(det).getComparisonMask() & 0x7; + S32 v_gequal = v.greaterEqual(LLVector4a::getZero()).getGatheredBits() & 0x7; + S32 sum_lequal = sum_uv.lessEqual(det).getGatheredBits() & 0x7; if (v_gequal && sum_lequal) { @@ -230,7 +231,7 @@ BOOL LLTriangleRayIntersectTwoSided(const LLVector4a& vert0, const LLVector4a& v pvec.setCross3(dir, edge2); /* if determinant is near zero, ray lies in plane of triangle */ - F32 det = edge1.dot3(pvec); + F32 det = edge1.dot3(pvec).getF32(); if (det > -F_APPROXIMATELY_ZERO && det < F_APPROXIMATELY_ZERO) @@ -245,7 +246,7 @@ BOOL LLTriangleRayIntersectTwoSided(const LLVector4a& vert0, const LLVector4a& v tvec.setSub(orig, vert0); /* calculate U parameter and test bounds */ - u = (tvec.dot3(pvec)) * inv_det; + u = (tvec.dot3(pvec).getF32()) * inv_det; if (u < 0.f || u > 1.f) { return FALSE; @@ -255,7 +256,7 @@ BOOL LLTriangleRayIntersectTwoSided(const LLVector4a& vert0, const LLVector4a& v tvec.sub(edge1); /* calculate V parameter and test bounds */ - v = (dir.dot3(tvec)) * inv_det; + v = (dir.dot3(tvec).getF32()) * inv_det; if (v < 0.f || u + v > 1.f) { @@ -263,7 +264,7 @@ BOOL LLTriangleRayIntersectTwoSided(const LLVector4a& vert0, const LLVector4a& v } /* calculate t, ray intersects triangle */ - t = (edge2.dot3(tvec)) * inv_det; + t = (edge2.dot3(tvec).getF32()) * inv_det; intersection_a = u; intersection_b = v; @@ -326,20 +327,20 @@ public: //stretch by triangles in node tri = *iter; - min.setMin(*tri->mV[0]); - min.setMin(*tri->mV[1]); - min.setMin(*tri->mV[2]); + min.setMin(min, *tri->mV[0]); + min.setMin(min, *tri->mV[1]); + min.setMin(min, *tri->mV[2]); - max.setMax(*tri->mV[0]); - max.setMax(*tri->mV[1]); - max.setMax(*tri->mV[2]); + max.setMax(max, *tri->mV[0]); + max.setMax(max, *tri->mV[1]); + max.setMax(max, *tri->mV[2]); } for (S32 i = 0; i < branch->getChildCount(); ++i) { //stretch by child extents LLVolumeOctreeListener* child = (LLVolumeOctreeListener*) branch->getChild(i)->getListener(0); - min.setMin(child->mExtents[0]); - max.setMax(child->mExtents[1]); + min.setMin(min, child->mExtents[0]); + max.setMax(min, child->mExtents[1]); } } else if (branch->getChildCount() != 0) @@ -352,8 +353,8 @@ public: for (S32 i = 1; i < branch->getChildCount(); ++i) { //stretch by child extents child = (LLVolumeOctreeListener*) branch->getChild(i)->getListener(0); - min.setMin(child->mExtents[0]); - max.setMax(child->mExtents[1]); + min.setMin(min, child->mExtents[0]); + max.setMax(max, child->mExtents[1]); } } else @@ -2011,7 +2012,7 @@ const LLVolumeFace::VertexData& LLVolumeFace::VertexData::operator=(const LLVolu if (this != &rhs) { init(); - LLVector4a::memcpyNonAliased16((F32*) mData, (F32*) rhs.mData, 8); + LLVector4a::memcpyNonAliased16((F32*) mData, (F32*) rhs.mData, 8*sizeof(F32)); mTexCoord = rhs.mTexCoord; } return *this; @@ -2055,8 +2056,8 @@ void LLVolumeFace::VertexData::setNormal(const LLVector4a& norm) bool LLVolumeFace::VertexData::operator<(const LLVolumeFace::VertexData& rhs)const { - const F32* lp = this->getPosition().getF32(); - const F32* rp = rhs.getPosition().getF32(); + const F32* lp = this->getPosition().getF32ptr(); + const F32* rp = rhs.getPosition().getF32ptr(); if (lp[0] != rp[0]) { @@ -2073,8 +2074,8 @@ bool LLVolumeFace::VertexData::operator<(const LLVolumeFace::VertexData& rhs)con return lp[2] < rp[2]; } - lp = getNormal().getF32(); - rp = rhs.getNormal().getF32(); + lp = getNormal().getF32ptr(); + rp = rhs.getNormal().getF32ptr(); if (lp[0] != rp[0]) { @@ -2101,23 +2102,23 @@ bool LLVolumeFace::VertexData::operator<(const LLVolumeFace::VertexData& rhs)con bool LLVolumeFace::VertexData::operator==(const LLVolumeFace::VertexData& rhs)const { - return mData[POSITION].equal3(rhs.getPosition()) && - mData[NORMAL].equal3(rhs.getNormal()) && + return mData[POSITION].equals3(rhs.getPosition()) && + mData[NORMAL].equals3(rhs.getNormal()) && mTexCoord == rhs.mTexCoord; } bool LLVolumeFace::VertexData::compareNormal(const LLVolumeFace::VertexData& rhs, F32 angle_cutoff) const { bool retval = false; - if (rhs.mData[POSITION].equal3(mData[POSITION]) && rhs.mTexCoord == mTexCoord) + if (rhs.mData[POSITION].equals3(mData[POSITION]) && rhs.mTexCoord == mTexCoord) { if (angle_cutoff > 1.f) { - retval = (mData[NORMAL].equal3(rhs.mData[NORMAL])); + retval = (mData[NORMAL].equals3(rhs.mData[NORMAL])); } else { - F32 cur_angle = rhs.mData[NORMAL].dot3(mData[NORMAL]); + F32 cur_angle = rhs.mData[NORMAL].dot3(mData[NORMAL]).getF32(); retval = cur_angle > angle_cutoff; } } @@ -2331,8 +2332,8 @@ bool LLVolume::unpackVolumeFaces(std::istream& is, S32 size) } else { - min.setMin(*pos_out); - max.setMax(*pos_out); + min.setMin(min, *pos_out); + max.setMax(max, *pos_out); } pos_out++; @@ -2944,7 +2945,7 @@ void sculpt_calc_mesh_resolution(U16 width, U16 height, U8 type, F32 detail, S32 ratio = (F32) width / (F32) height; - s = (S32)fsqrtf(((F32)vertices / ratio)); + s = (S32)(F32) sqrt(((F32)vertices / ratio)); s = llmax(s, 4); // no degenerate sizes, please t = vertices / s; @@ -5280,16 +5281,15 @@ LLVolumeFace& LLVolumeFace::operator=(const LLVolumeFace& src) freeData(); - LLVector4a::memcpyNonAliased16((F32*) mExtents, (F32*) src.mExtents, 12); + LLVector4a::memcpyNonAliased16((F32*) mExtents, (F32*) src.mExtents, 12*sizeof(F32)); resizeVertices(src.mNumVertices); resizeIndices(src.mNumIndices); if (mNumVertices) { - S32 vert_size = mNumVertices*4; + S32 vert_size = mNumVertices*4*sizeof(F32); S32 tc_size = (mNumVertices*8+0xF) & ~0xF; - tc_size /= 4; LLVector4a::memcpyNonAliased16((F32*) mPositions, (F32*) src.mPositions, vert_size); LLVector4a::memcpyNonAliased16((F32*) mNormals, (F32*) src.mNormals, vert_size); @@ -5322,8 +5322,7 @@ LLVolumeFace& LLVolumeFace::operator=(const LLVolumeFace& src) if (mNumIndices) { S32 idx_size = (mNumIndices*2+0xF) & ~0xF; - idx_size /= 4; - + LLVector4a::memcpyNonAliased16((F32*) mIndices, (F32*) src.mIndices, idx_size); } @@ -5388,9 +5387,9 @@ void LLVolumeFace::getVertexData(U16 index, LLVolumeFace::VertexData& cv) bool LLVolumeFace::VertexMapData::operator==(const LLVolumeFace::VertexData& rhs) const { - return getPosition().equal3(rhs.getPosition()) && + return getPosition().equals3(rhs.getPosition()) && mTexCoord == rhs.mTexCoord && - getNormal().equal3(rhs.getNormal()); + getNormal().equals3(rhs.getNormal()); } bool LLVolumeFace::VertexMapData::ComparePosition::operator()(const LLVector3& a, const LLVector3& b) const @@ -5423,7 +5422,7 @@ void LLVolumeFace::optimize(F32 angle_cutoff) getVertexData(index, cv); BOOL found = FALSE; - VertexMapData::PointMap::iterator point_iter = point_map.find(LLVector3(cv.getPosition().getF32())); + VertexMapData::PointMap::iterator point_iter = point_map.find(LLVector3(cv.getPosition().getF32ptr())); if (point_iter != point_map.end()) { //duplicate point might exist for (U32 j = 0; j < point_iter->second.size(); ++j) @@ -5455,7 +5454,7 @@ void LLVolumeFace::optimize(F32 angle_cutoff) } else { - point_map[LLVector3(d.getPosition().getF32())].push_back(d); + point_map[LLVector3(d.getPosition().getF32ptr())].push_back(d); } } } @@ -5491,12 +5490,12 @@ void LLVolumeFace::createOctree() tri->mIndex[2] = mIndices[i+2]; LLVector4a min = v0; - min.setMin(v1); - min.setMin(v2); + min.setMin(min, v1); + min.setMin(min, v2); LLVector4a max = v0; - max.setMax(v1); - max.setMax(v2); + max.setMax(max, v1); + max.setMax(max, v2); LLVector4a center; center.setAdd(min, max); @@ -5507,7 +5506,7 @@ void LLVolumeFace::createOctree() LLVector4a size; size.setSub(max,min); - tri->mRadius = size.length3() * 0.5f; + tri->mRadius = size.getLength3().getF32() * 0.5f; mOctree->insert(tri); } @@ -5655,12 +5654,13 @@ BOOL LLVolumeFace::createUnCutCubeCap(LLVolume* volume, BOOL partial_build) if (gx == 0 && gy == 0) { - min = max = newVert.getPosition(); + min = newVert.getPosition(); + max = min; } else { - min.setMin(newVert.getPosition()); - max.setMax(newVert.getPosition()); + min.setMin(min, newVert.getPosition()); + max.setMax(max, newVert.getPosition()); } } } @@ -5795,7 +5795,8 @@ BOOL LLVolumeFace::createCap(LLVolume* volume, BOOL partial_build) if (i == 0) { - min = max = pos[i]; + max = pos[i]; + min = max; min_uv = max_uv = tc[i]; } else @@ -5848,8 +5849,8 @@ BOOL LLVolumeFace::createCap(LLVolume* volume, BOOL partial_build) for (S32 i = 0; i < num_vertices; i++) { - binorm[i].load4a((F32*) &binormal.mQ); - norm[i].load4a((F32*) &normal.mQ); + binorm[i].load4a(binormal.getF32ptr()); + norm[i].load4a(normal.getF32ptr()); } if (partial_build) @@ -6186,7 +6187,7 @@ void LLVolumeFace::pushVertex(const LLVector4a& pos, const LLVector4a& norm, con LLVector4a* dst = (LLVector4a*) ll_aligned_malloc_16(new_size); if (mPositions) { - LLVector4a::memcpyNonAliased16((F32*) dst, (F32*) mPositions, old_size/4); + LLVector4a::memcpyNonAliased16((F32*) dst, (F32*) mPositions, old_size); ll_aligned_free_16(mPositions); } mPositions = dst; @@ -6195,7 +6196,7 @@ void LLVolumeFace::pushVertex(const LLVector4a& pos, const LLVector4a& norm, con dst = (LLVector4a*) ll_aligned_malloc_16(new_size); if (mNormals) { - LLVector4a::memcpyNonAliased16((F32*) dst, (F32*) mNormals, old_size/4); + LLVector4a::memcpyNonAliased16((F32*) dst, (F32*) mNormals, old_size); ll_aligned_free_16(mNormals); } mNormals = dst; @@ -6209,7 +6210,7 @@ void LLVolumeFace::pushVertex(const LLVector4a& pos, const LLVector4a& norm, con LLVector2* dst = (LLVector2*) ll_aligned_malloc_16(new_size); if (mTexCoords) { - LLVector4a::memcpyNonAliased16((F32*) dst, (F32*) mTexCoords, old_size/4); + LLVector4a::memcpyNonAliased16((F32*) dst, (F32*) mTexCoords, old_size); ll_aligned_free_16(mTexCoords); } } @@ -6268,7 +6269,7 @@ void LLVolumeFace::pushIndex(const U16& idx) U16* dst = (U16*) ll_aligned_malloc_16(new_size); if (mIndices) { - LLVector4a::memcpyNonAliased16((F32*) dst, (F32*) mIndices, old_size/4); + LLVector4a::memcpyNonAliased16((F32*) dst, (F32*) mIndices, old_size); ll_aligned_free_16(mIndices); } mIndices = dst; @@ -6319,9 +6320,9 @@ void LLVolumeFace::appendFace(const LLVolumeFace& face, LLMatrix4& mat_in, LLMat if (mNumVertices > 0) { //copy old buffers - LLVector4a::memcpyNonAliased16((F32*) new_pos, (F32*) mPositions, mNumVertices*4); - LLVector4a::memcpyNonAliased16((F32*) new_norm, (F32*) mNormals, mNumVertices*4); - LLVector4a::memcpyNonAliased16((F32*) new_tc, (F32*) mTexCoords, mNumVertices*2); + LLVector4a::memcpyNonAliased16((F32*) new_pos, (F32*) mPositions, mNumVertices*4*sizeof(F32)); + LLVector4a::memcpyNonAliased16((F32*) new_norm, (F32*) mNormals, mNumVertices*4*sizeof(F32)); + LLVector4a::memcpyNonAliased16((F32*) new_tc, (F32*) mTexCoords, mNumVertices*2*sizeof(F32)); } //free old buffer space @@ -6382,7 +6383,7 @@ void LLVolumeFace::appendFace(const LLVolumeFace& face, LLMatrix4& mat_in, LLMat if (mNumIndices > 0) { //copy old index buffer S32 old_size = (mNumIndices*2+0xF) & ~0xF; - LLVector4a::memcpyNonAliased16((F32*) new_indices, (F32*) mIndices, old_size/4); + LLVector4a::memcpyNonAliased16((F32*) new_indices, (F32*) mIndices, old_size); } //free old index buffer diff --git a/indra/llmath/tests/v2math_test.cpp b/indra/llmath/tests/v2math_test.cpp index 4660fcb955..c745b9989e 100644 --- a/indra/llmath/tests/v2math_test.cpp +++ b/indra/llmath/tests/v2math_test.cpp @@ -91,7 +91,7 @@ namespace tut F32 x = 2.2345f, y = 3.5678f ; LLVector2 vec2(x,y); ensure("magVecSquared:Fail ", is_approx_equal(vec2.magVecSquared(), (x*x + y*y))); - ensure("magVec:Fail ", is_approx_equal(vec2.magVec(), fsqrtf(x*x + y*y))); + ensure("magVec:Fail ", is_approx_equal(vec2.magVec(), (F32) sqrt(x*x + y*y))); } template<> template<> @@ -413,7 +413,7 @@ namespace tut ensure_equals("dist_vec_squared values are not equal",val2, val1); val1 = dist_vec(vec2, vec3); - val2 = fsqrtf((x1 - x2)*(x1 - x2) + (y1 - y2)* (y1 - y2)); + val2 = (F32) sqrt((x1 - x2)*(x1 - x2) + (y1 - y2)* (y1 - y2)); ensure_equals("dist_vec values are not equal",val2, val1); } @@ -437,7 +437,7 @@ namespace tut LLVector2 vec2(x1, y1); F32 vecMag = vec2.normVec(); - F32 mag = fsqrtf(x1*x1 + y1*y1); + F32 mag = (F32) sqrt(x1*x1 + y1*y1); F32 oomag = 1.f / mag; val1 = x1 * oomag; diff --git a/indra/llmath/tests/v3color_test.cpp b/indra/llmath/tests/v3color_test.cpp index 316b6e392f..0efba8e9f3 100644 --- a/indra/llmath/tests/v3color_test.cpp +++ b/indra/llmath/tests/v3color_test.cpp @@ -99,7 +99,7 @@ namespace tut F32 r = 2.3436212f, g = 1231.f, b = 4.7849321232f; LLColor3 llcolor3(r,g,b); ensure("magVecSquared:Fail ", is_approx_equal(llcolor3.magVecSquared(), (r*r + g*g + b*b))); - ensure("magVec:Fail ", is_approx_equal(llcolor3.magVec(), fsqrtf(r*r + g*g + b*b))); + ensure("magVec:Fail ", is_approx_equal(llcolor3.magVec(), (F32) sqrt(r*r + g*g + b*b))); } template<> template<> @@ -109,7 +109,7 @@ namespace tut F32 val1, val2,val3; LLColor3 llcolor3(r,g,b); F32 vecMag = llcolor3.normVec(); - F32 mag = fsqrtf(r*r + g*g + b*b); + F32 mag = (F32) sqrt(r*r + g*g + b*b); F32 oomag = 1.f / mag; val1 = r * oomag; val2 = g * oomag; @@ -292,7 +292,7 @@ namespace tut F32 r1 =1.f, g1 = 2.f,b1 = 1.2f, r2 = -2.3f, g2 = 1.11f, b2 = 1234.234f; LLColor3 llcolor3(r1,g1,b1),llcolor3a(r2,g2,b2); F32 val = distVec(llcolor3,llcolor3a); - ensure("distVec failed ", is_approx_equal(fsqrtf((r1-r2)*(r1-r2) + (g1-g2)*(g1-g2) + (b1-b2)*(b1-b2)) ,val)); + ensure("distVec failed ", is_approx_equal((F32) sqrt((r1-r2)*(r1-r2) + (g1-g2)*(g1-g2) + (b1-b2)*(b1-b2)) ,val)); F32 val1 = distVec_squared(llcolor3,llcolor3a); ensure("distVec_squared failed ", is_approx_equal(((r1-r2)*(r1-r2) + (g1-g2)*(g1-g2) + (b1-b2)*(b1-b2)) ,val1)); diff --git a/indra/llmath/tests/v3dmath_test.cpp b/indra/llmath/tests/v3dmath_test.cpp index e7c949186c..894b6200f5 100644 --- a/indra/llmath/tests/v3dmath_test.cpp +++ b/indra/llmath/tests/v3dmath_test.cpp @@ -409,7 +409,7 @@ namespace tut LLVector3d vec3D(x,y,z); F64 res = (x*x + y*y + z*z) - vec3D.magVecSquared(); ensure("1:magVecSquared:Fail ", ((-F_APPROXIMATELY_ZERO <= res)&& (res <=F_APPROXIMATELY_ZERO))); - res = fsqrtf(x*x + y*y + z*z) - vec3D.magVec(); + res = (F32) sqrt(x*x + y*y + z*z) - vec3D.magVec(); ensure("2:magVec: Fail ", ((-F_APPROXIMATELY_ZERO <= res)&& (res <=F_APPROXIMATELY_ZERO))); } diff --git a/indra/llmath/tests/v3math_test.cpp b/indra/llmath/tests/v3math_test.cpp index 7faf076243..d5c8dd2f9c 100644 --- a/indra/llmath/tests/v3math_test.cpp +++ b/indra/llmath/tests/v3math_test.cpp @@ -155,7 +155,7 @@ namespace tut F32 x = 2.32f, y = 1.212f, z = -.12f; LLVector3 vec3(x,y,z); ensure("1:magVecSquared:Fail ", is_approx_equal(vec3.magVecSquared(), (x*x + y*y + z*z))); - ensure("2:magVec:Fail ", is_approx_equal(vec3.magVec(), fsqrtf(x*x + y*y + z*z))); + ensure("2:magVec:Fail ", is_approx_equal(vec3.magVec(), (F32) sqrt(x*x + y*y + z*z))); } template<> template<> @@ -515,7 +515,7 @@ namespace tut F32 val1,val2; LLVector3 vec3(x1,y1,z1),vec3a(x2,y2,z2); val1 = dist_vec(vec3,vec3a); - val2 = fsqrtf((x1 - x2)*(x1 - x2) + (y1 - y2)* (y1 - y2) + (z1 - z2)* (z1 -z2)); + val2 = (F32) sqrt((x1 - x2)*(x1 - x2) + (y1 - y2)* (y1 - y2) + (z1 - z2)* (z1 -z2)); ensure_equals("1:dist_vec: Fail ",val2, val1); val1 = dist_vec_squared(vec3,vec3a); val2 =((x1 - x2)*(x1 - x2) + (y1 - y2)* (y1 - y2) + (z1 - z2)* (z1 -z2)); diff --git a/indra/llmath/tests/v4color_test.cpp b/indra/llmath/tests/v4color_test.cpp index 33921e0f0f..636446027a 100644 --- a/indra/llmath/tests/v4color_test.cpp +++ b/indra/llmath/tests/v4color_test.cpp @@ -161,7 +161,7 @@ namespace tut F32 r = 0x20, g = 0xFFFF, b = 0xFF; LLColor4 llcolor4(r,g,b); ensure("magVecSquared:Fail ", is_approx_equal(llcolor4.magVecSquared(), (r*r + g*g + b*b))); - ensure("magVec:Fail ", is_approx_equal(llcolor4.magVec(), fsqrtf(r*r + g*g + b*b))); + ensure("magVec:Fail ", is_approx_equal(llcolor4.magVec(), (F32) sqrt(r*r + g*g + b*b))); } template<> template<> @@ -170,7 +170,7 @@ namespace tut F32 r = 0x20, g = 0xFFFF, b = 0xFF; LLColor4 llcolor4(r,g,b); F32 vecMag = llcolor4.normVec(); - F32 mag = fsqrtf(r*r + g*g + b*b); + F32 mag = (F32) sqrt(r*r + g*g + b*b); F32 oomag = 1.f / mag; F32 val1 = r * oomag, val2 = g * oomag, val3 = b * oomag; ensure("1:normVec failed ", (is_approx_equal(val1, llcolor4.mV[0]) && is_approx_equal(val2, llcolor4.mV[1]) && is_approx_equal(val3, llcolor4.mV[2]) && is_approx_equal(vecMag, mag))); diff --git a/indra/llmath/tests/v4coloru_test.cpp b/indra/llmath/tests/v4coloru_test.cpp index 9f71cfc8cc..b3dbfece34 100644 --- a/indra/llmath/tests/v4coloru_test.cpp +++ b/indra/llmath/tests/v4coloru_test.cpp @@ -141,7 +141,7 @@ namespace tut U8 r = 0x12, g = 0xFF, b = 0xAF; LLColor4U llcolor4u(r,g,b); ensure("magVecSquared:Fail ", is_approx_equal(llcolor4u.magVecSquared(), (F32)(r*r + g*g + b*b))); - ensure("magVec:Fail ", is_approx_equal(llcolor4u.magVec(), fsqrtf(r*r + g*g + b*b))); + ensure("magVec:Fail ", is_approx_equal(llcolor4u.magVec(), (F32) sqrt((F32) (r*r + g*g + b*b)))); } template<> template<> diff --git a/indra/llmath/tests/v4math_test.cpp b/indra/llmath/tests/v4math_test.cpp index fe051c27e9..e919c90efa 100644 --- a/indra/llmath/tests/v4math_test.cpp +++ b/indra/llmath/tests/v4math_test.cpp @@ -102,7 +102,7 @@ namespace tut { F32 x = 10.f, y = -2.3f, z = -.023f; LLVector4 vec4(x,y,z); - ensure("magVec:Fail ", is_approx_equal(vec4.magVec(), fsqrtf(x*x + y*y + z*z))); + ensure("magVec:Fail ", is_approx_equal(vec4.magVec(), (F32) sqrt(x*x + y*y + z*z))); ensure("magVecSquared:Fail ", is_approx_equal(vec4.magVecSquared(), (x*x + y*y + z*z))); } @@ -343,7 +343,7 @@ namespace tut F32 val1,val2; LLVector4 vec4(x1,y1,z1),vec4a(x2,y2,z2); val1 = dist_vec(vec4,vec4a); - val2 = fsqrtf((x1 - x2)*(x1 - x2) + (y1 - y2)* (y1 - y2) + (z1 - z2)* (z1 -z2)); + val2 = (F32) sqrt((x1 - x2)*(x1 - x2) + (y1 - y2)* (y1 - y2) + (z1 - z2)* (z1 -z2)); ensure_equals("dist_vec: Fail ",val2, val1); val1 = dist_vec_squared(vec4,vec4a); val2 =((x1 - x2)*(x1 - x2) + (y1 - y2)* (y1 - y2) + (z1 - z2)* (z1 -z2)); diff --git a/indra/llmath/v2math.cpp b/indra/llmath/v2math.cpp index 220336e0c2..2603127f75 100644 --- a/indra/llmath/v2math.cpp +++ b/indra/llmath/v2math.cpp @@ -92,7 +92,7 @@ F32 dist_vec(const LLVector2 &a, const LLVector2 &b) { F32 x = a.mV[0] - b.mV[0]; F32 y = a.mV[1] - b.mV[1]; - return fsqrtf( x*x + y*y ); + return (F32) sqrt( x*x + y*y ); } F32 dist_vec_squared(const LLVector2 &a, const LLVector2 &b) diff --git a/indra/llmath/v2math.h b/indra/llmath/v2math.h index ae26c85ce4..35fd1b6048 100644 --- a/indra/llmath/v2math.h +++ b/indra/llmath/v2math.h @@ -225,7 +225,7 @@ inline void LLVector2::setVec(const F32 *vec) inline F32 LLVector2::length(void) const { - return fsqrtf(mV[0]*mV[0] + mV[1]*mV[1]); + return (F32) sqrt(mV[0]*mV[0] + mV[1]*mV[1]); } inline F32 LLVector2::lengthSquared(void) const @@ -235,7 +235,7 @@ inline F32 LLVector2::lengthSquared(void) const inline F32 LLVector2::normalize(void) { - F32 mag = fsqrtf(mV[0]*mV[0] + mV[1]*mV[1]); + F32 mag = (F32) sqrt(mV[0]*mV[0] + mV[1]*mV[1]); F32 oomag; if (mag > FP_MAG_THRESHOLD) @@ -262,7 +262,7 @@ inline bool LLVector2::isFinite() const // deprecated inline F32 LLVector2::magVec(void) const { - return fsqrtf(mV[0]*mV[0] + mV[1]*mV[1]); + return (F32) sqrt(mV[0]*mV[0] + mV[1]*mV[1]); } // deprecated @@ -274,7 +274,7 @@ inline F32 LLVector2::magVecSquared(void) const // deprecated inline F32 LLVector2::normVec(void) { - F32 mag = fsqrtf(mV[0]*mV[0] + mV[1]*mV[1]); + F32 mag = (F32) sqrt(mV[0]*mV[0] + mV[1]*mV[1]); F32 oomag; if (mag > FP_MAG_THRESHOLD) diff --git a/indra/llmath/v3color.h b/indra/llmath/v3color.h index 1915d80502..95a3de8b62 100644 --- a/indra/llmath/v3color.h +++ b/indra/llmath/v3color.h @@ -284,7 +284,7 @@ inline F32 LLColor3::brightness(void) const inline F32 LLColor3::length(void) const { - return fsqrtf(mV[0]*mV[0] + mV[1]*mV[1] + mV[2]*mV[2]); + return (F32) sqrt(mV[0]*mV[0] + mV[1]*mV[1] + mV[2]*mV[2]); } inline F32 LLColor3::lengthSquared(void) const @@ -294,7 +294,7 @@ inline F32 LLColor3::lengthSquared(void) const inline F32 LLColor3::normalize(void) { - F32 mag = fsqrtf(mV[0]*mV[0] + mV[1]*mV[1] + mV[2]*mV[2]); + F32 mag = (F32) sqrt(mV[0]*mV[0] + mV[1]*mV[1] + mV[2]*mV[2]); F32 oomag; if (mag) @@ -310,7 +310,7 @@ inline F32 LLColor3::normalize(void) // deprecated inline F32 LLColor3::magVec(void) const { - return fsqrtf(mV[0]*mV[0] + mV[1]*mV[1] + mV[2]*mV[2]); + return (F32) sqrt(mV[0]*mV[0] + mV[1]*mV[1] + mV[2]*mV[2]); } // deprecated @@ -322,7 +322,7 @@ inline F32 LLColor3::magVecSquared(void) const // deprecated inline F32 LLColor3::normVec(void) { - F32 mag = fsqrtf(mV[0]*mV[0] + mV[1]*mV[1] + mV[2]*mV[2]); + F32 mag = (F32) sqrt(mV[0]*mV[0] + mV[1]*mV[1] + mV[2]*mV[2]); F32 oomag; if (mag) @@ -444,7 +444,7 @@ inline F32 distVec(const LLColor3 &a, const LLColor3 &b) F32 x = a.mV[0] - b.mV[0]; F32 y = a.mV[1] - b.mV[1]; F32 z = a.mV[2] - b.mV[2]; - return fsqrtf( x*x + y*y + z*z ); + return (F32) sqrt( x*x + y*y + z*z ); } inline F32 distVec_squared(const LLColor3 &a, const LLColor3 &b) diff --git a/indra/llmath/v3dmath.h b/indra/llmath/v3dmath.h index 6ab31e8a41..ab253de064 100644 --- a/indra/llmath/v3dmath.h +++ b/indra/llmath/v3dmath.h @@ -240,7 +240,7 @@ inline const LLVector3d& LLVector3d::setVec(const F64 *vec) inline F64 LLVector3d::normVec(void) { - F64 mag = fsqrtf(mdV[0]*mdV[0] + mdV[1]*mdV[1] + mdV[2]*mdV[2]); + F64 mag = (F32) sqrt(mdV[0]*mdV[0] + mdV[1]*mdV[1] + mdV[2]*mdV[2]); F64 oomag; if (mag > FP_MAG_THRESHOLD) @@ -262,7 +262,7 @@ inline F64 LLVector3d::normVec(void) inline F64 LLVector3d::normalize(void) { - F64 mag = fsqrtf(mdV[0]*mdV[0] + mdV[1]*mdV[1] + mdV[2]*mdV[2]); + F64 mag = (F32) sqrt(mdV[0]*mdV[0] + mdV[1]*mdV[1] + mdV[2]*mdV[2]); F64 oomag; if (mag > FP_MAG_THRESHOLD) @@ -286,7 +286,7 @@ inline F64 LLVector3d::normalize(void) inline F64 LLVector3d::magVec(void) const { - return fsqrtf(mdV[0]*mdV[0] + mdV[1]*mdV[1] + mdV[2]*mdV[2]); + return (F32) sqrt(mdV[0]*mdV[0] + mdV[1]*mdV[1] + mdV[2]*mdV[2]); } inline F64 LLVector3d::magVecSquared(void) const @@ -296,7 +296,7 @@ inline F64 LLVector3d::magVecSquared(void) const inline F64 LLVector3d::length(void) const { - return fsqrtf(mdV[0]*mdV[0] + mdV[1]*mdV[1] + mdV[2]*mdV[2]); + return (F32) sqrt(mdV[0]*mdV[0] + mdV[1]*mdV[1] + mdV[2]*mdV[2]); } inline F64 LLVector3d::lengthSquared(void) const @@ -406,7 +406,7 @@ inline F64 dist_vec(const LLVector3d &a, const LLVector3d &b) F64 x = a.mdV[0] - b.mdV[0]; F64 y = a.mdV[1] - b.mdV[1]; F64 z = a.mdV[2] - b.mdV[2]; - return fsqrtf( x*x + y*y + z*z ); + return (F32) sqrt( x*x + y*y + z*z ); } inline F64 dist_vec_squared(const LLVector3d &a, const LLVector3d &b) diff --git a/indra/llmath/v3math.h b/indra/llmath/v3math.h index 75c860a91e..5d483a8753 100644 --- a/indra/llmath/v3math.h +++ b/indra/llmath/v3math.h @@ -282,7 +282,7 @@ inline void LLVector3::setVec(const F32 *vec) inline F32 LLVector3::normalize(void) { - F32 mag = fsqrtf(mV[0]*mV[0] + mV[1]*mV[1] + mV[2]*mV[2]); + F32 mag = (F32) sqrt(mV[0]*mV[0] + mV[1]*mV[1] + mV[2]*mV[2]); F32 oomag; if (mag > FP_MAG_THRESHOLD) @@ -305,7 +305,7 @@ inline F32 LLVector3::normalize(void) // deprecated inline F32 LLVector3::normVec(void) { - F32 mag = fsqrtf(mV[0]*mV[0] + mV[1]*mV[1] + mV[2]*mV[2]); + F32 mag = (F32) sqrt(mV[0]*mV[0] + mV[1]*mV[1] + mV[2]*mV[2]); F32 oomag; if (mag > FP_MAG_THRESHOLD) @@ -329,7 +329,7 @@ inline F32 LLVector3::normVec(void) inline F32 LLVector3::length(void) const { - return fsqrtf(mV[0]*mV[0] + mV[1]*mV[1] + mV[2]*mV[2]); + return (F32) sqrt(mV[0]*mV[0] + mV[1]*mV[1] + mV[2]*mV[2]); } inline F32 LLVector3::lengthSquared(void) const @@ -339,7 +339,7 @@ inline F32 LLVector3::lengthSquared(void) const inline F32 LLVector3::magVec(void) const { - return fsqrtf(mV[0]*mV[0] + mV[1]*mV[1] + mV[2]*mV[2]); + return (F32) sqrt(mV[0]*mV[0] + mV[1]*mV[1] + mV[2]*mV[2]); } inline F32 LLVector3::magVecSquared(void) const @@ -473,7 +473,7 @@ inline F32 dist_vec(const LLVector3 &a, const LLVector3 &b) F32 x = a.mV[0] - b.mV[0]; F32 y = a.mV[1] - b.mV[1]; F32 z = a.mV[2] - b.mV[2]; - return fsqrtf( x*x + y*y + z*z ); + return (F32) sqrt( x*x + y*y + z*z ); } inline F32 dist_vec_squared(const LLVector3 &a, const LLVector3 &b) diff --git a/indra/llmath/v4color.h b/indra/llmath/v4color.h index 6b63b976b0..dd92e1cc63 100644 --- a/indra/llmath/v4color.h +++ b/indra/llmath/v4color.h @@ -392,7 +392,7 @@ inline const LLColor4& LLColor4::setAlpha(F32 a) inline F32 LLColor4::length(void) const { - return fsqrtf(mV[VX]*mV[VX] + mV[VY]*mV[VY] + mV[VZ]*mV[VZ]); + return (F32) sqrt(mV[VX]*mV[VX] + mV[VY]*mV[VY] + mV[VZ]*mV[VZ]); } inline F32 LLColor4::lengthSquared(void) const @@ -402,7 +402,7 @@ inline F32 LLColor4::lengthSquared(void) const inline F32 LLColor4::normalize(void) { - F32 mag = fsqrtf(mV[VX]*mV[VX] + mV[VY]*mV[VY] + mV[VZ]*mV[VZ]); + F32 mag = (F32) sqrt(mV[VX]*mV[VX] + mV[VY]*mV[VY] + mV[VZ]*mV[VZ]); F32 oomag; if (mag) @@ -418,7 +418,7 @@ inline F32 LLColor4::normalize(void) // deprecated inline F32 LLColor4::magVec(void) const { - return fsqrtf(mV[VX]*mV[VX] + mV[VY]*mV[VY] + mV[VZ]*mV[VZ]); + return (F32) sqrt(mV[VX]*mV[VX] + mV[VY]*mV[VY] + mV[VZ]*mV[VZ]); } // deprecated @@ -430,7 +430,7 @@ inline F32 LLColor4::magVecSquared(void) const // deprecated inline F32 LLColor4::normVec(void) { - F32 mag = fsqrtf(mV[VX]*mV[VX] + mV[VY]*mV[VY] + mV[VZ]*mV[VZ]); + F32 mag = (F32) sqrt(mV[VX]*mV[VX] + mV[VY]*mV[VY] + mV[VZ]*mV[VZ]); F32 oomag; if (mag) diff --git a/indra/llmath/v4coloru.h b/indra/llmath/v4coloru.h index 4ec5a345eb..08245403a1 100644 --- a/indra/llmath/v4coloru.h +++ b/indra/llmath/v4coloru.h @@ -300,7 +300,7 @@ inline const LLColor4U& LLColor4U::setAlpha(U8 a) inline F32 LLColor4U::length(void) const { - return fsqrtf( ((F32)mV[VX]) * mV[VX] + ((F32)mV[VY]) * mV[VY] + ((F32)mV[VZ]) * mV[VZ] ); + return (F32) sqrt( ((F32)mV[VX]) * mV[VX] + ((F32)mV[VY]) * mV[VY] + ((F32)mV[VZ]) * mV[VZ] ); } inline F32 LLColor4U::lengthSquared(void) const @@ -311,7 +311,7 @@ inline F32 LLColor4U::lengthSquared(void) const // deprecated inline F32 LLColor4U::magVec(void) const { - return fsqrtf( ((F32)mV[VX]) * mV[VX] + ((F32)mV[VY]) * mV[VY] + ((F32)mV[VZ]) * mV[VZ] ); + return (F32) sqrt( ((F32)mV[VX]) * mV[VX] + ((F32)mV[VY]) * mV[VY] + ((F32)mV[VZ]) * mV[VZ] ); } // deprecated diff --git a/indra/llmath/v4math.h b/indra/llmath/v4math.h index 4c82e6b629..72a477ed20 100644 --- a/indra/llmath/v4math.h +++ b/indra/llmath/v4math.h @@ -321,7 +321,7 @@ inline void LLVector4::setVec(const F32 *vec) inline F32 LLVector4::length(void) const { - return fsqrtf(mV[VX]*mV[VX] + mV[VY]*mV[VY] + mV[VZ]*mV[VZ]); + return (F32) sqrt(mV[VX]*mV[VX] + mV[VY]*mV[VY] + mV[VZ]*mV[VZ]); } inline F32 LLVector4::lengthSquared(void) const @@ -331,7 +331,7 @@ inline F32 LLVector4::lengthSquared(void) const inline F32 LLVector4::magVec(void) const { - return fsqrtf(mV[VX]*mV[VX] + mV[VY]*mV[VY] + mV[VZ]*mV[VZ]); + return (F32) sqrt(mV[VX]*mV[VX] + mV[VY]*mV[VY] + mV[VZ]*mV[VZ]); } inline F32 LLVector4::magVecSquared(void) const @@ -463,7 +463,7 @@ inline LLVector4 lerp(const LLVector4 &a, const LLVector4 &b, F32 u) inline F32 LLVector4::normalize(void) { - F32 mag = fsqrtf(mV[VX]*mV[VX] + mV[VY]*mV[VY] + mV[VZ]*mV[VZ]); + F32 mag = (F32) sqrt(mV[VX]*mV[VX] + mV[VY]*mV[VY] + mV[VZ]*mV[VZ]); F32 oomag; if (mag > FP_MAG_THRESHOLD) @@ -486,7 +486,7 @@ inline F32 LLVector4::normalize(void) // deprecated inline F32 LLVector4::normVec(void) { - F32 mag = fsqrtf(mV[VX]*mV[VX] + mV[VY]*mV[VY] + mV[VZ]*mV[VZ]); + F32 mag = (F32) sqrt(mV[VX]*mV[VX] + mV[VY]*mV[VY] + mV[VZ]*mV[VZ]); F32 oomag; if (mag > FP_MAG_THRESHOLD) diff --git a/indra/llmessage/llsdmessagebuilder.cpp b/indra/llmessage/llsdmessagebuilder.cpp index 6e41b03895..2680dd0b77 100755 --- a/indra/llmessage/llsdmessagebuilder.cpp +++ b/indra/llmessage/llsdmessagebuilder.cpp @@ -35,6 +35,7 @@ #include "llsdmessagebuilder.h" #include "llmessagetemplate.h" +#include "llmath.h" #include "llquaternion.h" #include "llsdutil.h" #include "llsdutil_math.h" diff --git a/indra/llmessage/lltemplatemessagebuilder.cpp b/indra/llmessage/lltemplatemessagebuilder.cpp index fa02456d90..6f9707ed57 100644 --- a/indra/llmessage/lltemplatemessagebuilder.cpp +++ b/indra/llmessage/lltemplatemessagebuilder.cpp @@ -35,6 +35,7 @@ #include "lltemplatemessagebuilder.h" #include "llmessagetemplate.h" +#include "llmath.h" #include "llquaternion.h" #include "u64.h" #include "v3dmath.h" diff --git a/indra/llmessage/lltemplatemessagereader.cpp b/indra/llmessage/lltemplatemessagereader.cpp index 8f56cf2521..2b67e9f6fa 100644 --- a/indra/llmessage/lltemplatemessagereader.cpp +++ b/indra/llmessage/lltemplatemessagereader.cpp @@ -36,6 +36,7 @@ #include "llfasttimer.h" #include "llmessagebuilder.h" #include "llmessagetemplate.h" +#include "llmath.h" #include "llquaternion.h" #include "message.h" #include "u64.h" diff --git a/indra/llrender/llrender.cpp b/indra/llrender/llrender.cpp index 70601663e6..2b1a9bbf94 100644 --- a/indra/llrender/llrender.cpp +++ b/indra/llrender/llrender.cpp @@ -924,7 +924,7 @@ LLVector3 LLRender::getUITranslation() { llerrs << "UI offset stack empty." << llendl; } - return LLVector3(mUIOffset[mUIStackDepth-1].getF32()); + return LLVector3(mUIOffset[mUIStackDepth-1].getF32ptr()); } LLVector3 LLRender::getUIScale() @@ -933,7 +933,7 @@ LLVector3 LLRender::getUIScale() { llerrs << "UI scale stack empty." << llendl; } - return LLVector3(mUIScale[mUIStackDepth-1].getF32()); + return LLVector3(mUIScale[mUIStackDepth-1].getF32ptr()); } diff --git a/indra/newview/llagent.cpp b/indra/newview/llagent.cpp index 72d51540ef..5fae5b893f 100644 --- a/indra/newview/llagent.cpp +++ b/indra/newview/llagent.cpp @@ -1210,7 +1210,7 @@ void LLAgent::startAutoPilotGlobal(const LLVector3d &target_global, const std::s else { // Guess at a reasonable stop distance. - mAutoPilotStopDistance = fsqrtf( distance ); + mAutoPilotStopDistance = (F32) sqrt( distance ); if (mAutoPilotStopDistance < 0.5f) { mAutoPilotStopDistance = 0.5f; diff --git a/indra/newview/lldrawable.cpp b/indra/newview/lldrawable.cpp index 8d0a06378f..68f52e04bc 100644 --- a/indra/newview/lldrawable.cpp +++ b/indra/newview/lldrawable.cpp @@ -743,7 +743,7 @@ void LLDrawable::updateDistance(LLCamera& camera, bool force_update) } else { - pos = LLVector3(getPositionGroup().getF32()); + pos = LLVector3(getPositionGroup().getF32ptr()); } pos -= camera.getOrigin(); @@ -824,7 +824,7 @@ void LLDrawable::shiftPos(const LLVector4a &shift_vector) for (S32 i = 0; i < getNumFaces(); i++) { LLFace *facep = getFace(i); - facep->mCenterAgent += LLVector3(shift_vector.getF32()); + facep->mCenterAgent += LLVector3(shift_vector.getF32ptr()); facep->mExtents[0].add(shift_vector); facep->mExtents[1].add(shift_vector); @@ -1153,13 +1153,13 @@ void LLSpatialBridge::updateSpatialExtents() LLVector4a max; max.setAdd(center, delta); - newMin.setMin(min); - newMax.setMax(max); + newMin.setMin(newMin, min); + newMax.setMax(newMax, max); } LLVector4a diagonal; diagonal.setSub(newMax, newMin); - mRadius = diagonal.length3() * 0.5f; + mRadius = diagonal.getLength3().getF32() * 0.5f; mPositionGroup->setAdd(newMin,newMax); mPositionGroup->mul(0.5f); diff --git a/indra/newview/lldrawpoolavatar.cpp b/indra/newview/lldrawpoolavatar.cpp index af082eea3e..d698624c15 100644 --- a/indra/newview/lldrawpoolavatar.cpp +++ b/indra/newview/lldrawpoolavatar.cpp @@ -37,6 +37,7 @@ #include "llvoavatar.h" #include "m3math.h" +#include "llmatrix4a.h" #include "llagent.h" //for gAgent.needsRenderAvatar() #include "lldrawable.h" @@ -495,7 +496,7 @@ S32 LLDrawPoolAvatar::getNumPasses() } else { - return 3; + return 4; } #else if (LLPipeline::sImpostorRender) @@ -814,26 +815,46 @@ void LLDrawPoolAvatar::endSkinned() #if LL_MESH_ENABLED void LLDrawPoolAvatar::beginRiggedSimple() { - if (LLPipeline::sUnderWaterRender) + if (sShaderLevel > 0) { - sVertexProgram = &gSkinnedObjectSimpleWaterProgram; + if (LLPipeline::sUnderWaterRender) + { + sVertexProgram = &gSkinnedObjectSimpleWaterProgram; + } + else + { + sVertexProgram = &gSkinnedObjectSimpleProgram; + } } else { - sVertexProgram = &gSkinnedObjectSimpleProgram; + if (LLPipeline::sUnderWaterRender) + { + sVertexProgram = &gObjectSimpleWaterProgram; + } + else + { + sVertexProgram = &gObjectSimpleProgram; + } } - sDiffuseChannel = 0; - sVertexProgram->bind(); - LLVertexBuffer::sWeight4Loc = sVertexProgram->getAttribLocation(LLViewerShaderMgr::OBJECT_WEIGHT); + if (sShaderLevel > 0 || gPipeline.canUseVertexShaders()) + { + sDiffuseChannel = 0; + sVertexProgram->bind(); + LLVertexBuffer::sWeight4Loc = sVertexProgram->getAttribLocation(LLViewerShaderMgr::OBJECT_WEIGHT); + } } void LLDrawPoolAvatar::endRiggedSimple() { LLVertexBuffer::unbind(); - sVertexProgram->unbind(); - sVertexProgram = NULL; - LLVertexBuffer::sWeight4Loc = -1; + if (sShaderLevel > 0 || gPipeline.canUseVertexShaders()) + { + sVertexProgram->unbind(); + sVertexProgram = NULL; + LLVertexBuffer::sWeight4Loc = -1; + } } void LLDrawPoolAvatar::beginRiggedAlpha() @@ -1281,8 +1302,10 @@ void LLDrawPoolAvatar::renderAvatars(LLVOAvatar* single_avatar, S32 pass) } #if LL_MESH_ENABLED -void LLDrawPoolAvatar::updateRiggedFaceVertexBuffer(LLFace* face, const LLMeshSkinInfo* skin, LLVolume* volume, const LLVolumeFace& vol_face) +void LLDrawPoolAvatar::updateRiggedFaceVertexBuffer(LLVOAvatar* avatar, LLFace* face, const LLMeshSkinInfo* skin, LLVolume* volume, const LLVolumeFace& vol_face) { + LLVertexBuffer* buffer = face->mVertexBuffer; + U32 data_mask = 0; for (U32 i = 0; i < face->mRiggedIndex.size(); ++i) { @@ -1292,17 +1315,23 @@ void LLDrawPoolAvatar::updateRiggedFaceVertexBuffer(LLFace* face, const LLMeshSk } } - LLVertexBuffer* buff = face->mVertexBuffer; - - if (!buff || - buff->getTypeMask() != data_mask || - buff->getRequestedVerts() != vol_face.mNumVertices) + if (!buffer || + buffer->getTypeMask() != data_mask || + buffer->getRequestedVerts() != vol_face.mNumVertices) { face->setGeomIndex(0); face->setIndicesIndex(0); face->setSize(vol_face.mNumVertices, vol_face.mNumIndices, true); - face->mVertexBuffer = new LLVertexBuffer(data_mask, 0); + if (sShaderLevel > 0) + { + face->mVertexBuffer = new LLVertexBuffer(data_mask, GL_DYNAMIC_DRAW_ARB); + } + else + { + face->mVertexBuffer = new LLVertexBuffer(data_mask, GL_STREAM_DRAW_ARB); + } + face->mVertexBuffer->allocateBuffer(face->getGeomCount(), face->getIndicesCount(), true); U16 offset = 0; @@ -1319,7 +1348,83 @@ void LLDrawPoolAvatar::updateRiggedFaceVertexBuffer(LLFace* face, const LLMeshSk LLMatrix3 mat_normal(mat3); face->getGeometryVolume(*volume, face->getTEOffset(), mat_vert, mat_normal, offset, true); - buff = face->mVertexBuffer; + buffer = face->mVertexBuffer; + } + + if (sShaderLevel <= 0 && face->mLastSkinTime < avatar->getLastSkinTime()) + { //perform software vertex skinning for this face + LLStrider position; + LLStrider normal; + + buffer->getVertexStrider(position); + buffer->getNormalStrider(normal); + + LLVector4a* pos = (LLVector4a*) position.get(); + LLVector4a* norm = (LLVector4a*) normal.get(); + + //build matrix palette + LLMatrix4a mp[64]; + LLMatrix4* mat = (LLMatrix4*) mp; + + for (U32 j = 0; j < skin->mJointNames.size(); ++j) + { + LLJoint* joint = avatar->getJoint(skin->mJointNames[j]); + if (joint) + { + mat[j] = skin->mInvBindMatrix[j]; + mat[j] *= joint->getWorldMatrix(); + } + } + + LLVector4a* weight = vol_face.mWeights; + + LLMatrix4a bind_shape_matrix; + bind_shape_matrix.loadu(skin->mBindShapeMatrix); + + for (U32 j = 0; j < buffer->getRequestedVerts(); ++j) + { + LLMatrix4a final_mat; + final_mat.clear(); + + S32 idx[4]; + + LLVector4 wght; + + F32 scale = 0.f; + for (U32 k = 0; k < 4; k++) + { + F32 w = weight[j][k]; + + idx[k] = (S32) floorf(w); + wght[k] = w - floorf(w); + scale += wght[k]; + } + + wght *= 1.f/scale; + + for (U32 k = 0; k < 4; k++) + { + F32 w = wght[k]; + + LLMatrix4a src; + src.setMul(mp[idx[k]], w); + + final_mat.add(src); + } + + + LLVector4a& v = vol_face.mPositions[j]; + LLVector4a t; + LLVector4a dst; + bind_shape_matrix.affineTransform(v, t); + final_mat.affineTransform(t, dst); + pos[j] = dst; + + LLVector4a& n = vol_face.mNormals[j]; + bind_shape_matrix.rotate(n, t); + final_mat.rotate(t, dst); + norm[j] = dst; + } } } @@ -1371,7 +1476,7 @@ void LLDrawPoolAvatar::renderRigged(LLVOAvatar* avatar, U32 type, bool glow) stop_glerror(); const LLVolumeFace& vol_face = volume->getVolumeFace(te); - updateRiggedFaceVertexBuffer(face, skin, volume, vol_face); + updateRiggedFaceVertexBuffer(avatar, face, skin, volume, vol_face); stop_glerror(); @@ -1381,30 +1486,37 @@ void LLDrawPoolAvatar::renderRigged(LLVOAvatar* avatar, U32 type, bool glow) if (buff) { - LLMatrix4 mat[64]; + if (sShaderLevel > 0) + { //upload matrix palette to shader + LLMatrix4 mat[64]; - for (U32 i = 0; i < skin->mJointNames.size(); ++i) - { - LLJoint* joint = avatar->getJoint(skin->mJointNames[i]); - if (joint) + for (U32 i = 0; i < skin->mJointNames.size(); ++i) { - mat[i] = skin->mInvBindMatrix[i]; - mat[i] *= joint->getWorldMatrix(); + LLJoint* joint = avatar->getJoint(skin->mJointNames[i]); + if (joint) + { + mat[i] = skin->mInvBindMatrix[i]; + mat[i] *= joint->getWorldMatrix(); + } } + + stop_glerror(); + + LLDrawPoolAvatar::sVertexProgram->uniformMatrix4fv("matrixPalette", + skin->mJointNames.size(), + FALSE, + (GLfloat*) mat[0].mMatrix); + LLDrawPoolAvatar::sVertexProgram->uniformMatrix4fv("matrixPalette[0]", + skin->mJointNames.size(), + FALSE, + (GLfloat*) mat[0].mMatrix); + + stop_glerror(); + } + else + { + data_mask &= ~LLVertexBuffer::MAP_WEIGHT4; } - - stop_glerror(); - - LLDrawPoolAvatar::sVertexProgram->uniformMatrix4fv("matrixPalette", - skin->mJointNames.size(), - FALSE, - (GLfloat*) mat[0].mMatrix); - LLDrawPoolAvatar::sVertexProgram->uniformMatrix4fv("matrixPalette[0]", - skin->mJointNames.size(), - FALSE, - (GLfloat*) mat[0].mMatrix); - - stop_glerror(); buff->setBuffer(data_mask); diff --git a/indra/newview/lldrawpoolavatar.h b/indra/newview/lldrawpoolavatar.h index b01394534b..d08ae04516 100644 --- a/indra/newview/lldrawpoolavatar.h +++ b/indra/newview/lldrawpoolavatar.h @@ -138,7 +138,8 @@ public: void endDeferredRiggedSimple(); void endDeferredRiggedBump(); - void updateRiggedFaceVertexBuffer(LLFace* facep, + void updateRiggedFaceVertexBuffer(LLVOAvatar* avatar, + LLFace* facep, const LLMeshSkinInfo* skin, LLVolume* volume, const LLVolumeFace& vol_face); diff --git a/indra/newview/llface.cpp b/indra/newview/llface.cpp index 8533f9710c..5d7d3387a4 100644 --- a/indra/newview/llface.cpp +++ b/indra/newview/llface.cpp @@ -106,8 +106,8 @@ void planarProjection(LLVector2 &tc, const LLVector4a& normal, LLVector4a tangent; tangent.setCross3(binormal,normal); - tc.mV[1] = -((tangent.dot3(vec))*2 - 0.5f); - tc.mV[0] = 1.0f+((binormal.dot3(vec))*2 - 0.5f); + tc.mV[1] = -((tangent.dot3(vec).getF32())*2 - 0.5f); + tc.mV[0] = 1.0f+((binormal.dot3(vec).getF32())*2 - 0.5f); } void sphericalProjection(LLVector2 &tc, const LLVector4a& normal, @@ -156,6 +156,7 @@ void LLFace::init(LLDrawable* drawablep, LLViewerObject* objp) mLastUpdateTime = gFrameTimeSeconds; mLastMoveTime = 0.f; + mLastSkinTime = gFrameTimeSeconds; mVSize = 0.f; mPixelArea = 16.f; mState = GLOBAL; @@ -828,8 +829,8 @@ BOOL LLFace::genVolumeBBoxes(const LLVolume &volume, S32 f, LLVector4a max; max.setAdd(center, delta); - newMin.setMin(min); - newMax.setMax(max); + newMin.setMin(newMin,min); + newMax.setMax(newMax,max); } if (!mDrawablep->isActive()) @@ -844,11 +845,11 @@ BOOL LLFace::genVolumeBBoxes(const LLVolume &volume, S32 f, t.mul(0.5f); //VECTORIZE THIS - mCenterLocal.set(t.getF32()); + mCenterLocal.set(t.getF32ptr()); t.setSub(newMax,newMin); t.mul(0.5f); - mBoundingSphereRadius = t.length3(); + mBoundingSphereRadius = t.getLength3().getF32(); updateCenterAgent(); } @@ -1313,7 +1314,7 @@ BOOL LLFace::getGeometryVolume(const LLVolume& volume, { if (!do_xform) { - LLVector4a::memcpyNonAliased16((F32*) tex_coords.get(), (F32*) vf.mTexCoords, num_vertices*2); + LLVector4a::memcpyNonAliased16((F32*) tex_coords.get(), (F32*) vf.mTexCoords, num_vertices*2*sizeof(F32)); } else { @@ -1529,13 +1530,13 @@ BOOL LLFace::getGeometryVolume(const LLVolume& volume, if (mDrawablep->isActive()) { LLVector3 t; - t.set(binormal.getF32()); + t.set(binormal.getF32ptr()); t *= bump_quat; binormal.load3(t.mV); } binormal.normalize3fast(); - tc += LLVector2( bump_s_primary_light_ray.dot3(tangent), bump_t_primary_light_ray.dot3(binormal) ); + tc += LLVector2( bump_s_primary_light_ray.dot3(tangent).getF32(), bump_t_primary_light_ray.dot3(binormal).getF32() ); *tex_coords2++ = tc; } @@ -1583,7 +1584,7 @@ BOOL LLFace::getGeometryVolume(const LLVolume& volume, if (rebuild_weights && vf.mWeights) { - LLVector4a::memcpyNonAliased16((F32*) weights, (F32*) vf.mWeights, num_vertices*4); + LLVector4a::memcpyNonAliased16((F32*) weights, (F32*) vf.mWeights, num_vertices*4*sizeof(F32)); } if (rebuild_color) @@ -1705,21 +1706,21 @@ BOOL LLFace::calcPixelArea(F32& cos_angle_to_view_dir, F32& radius) LLViewerCamera* camera = LLViewerCamera::getInstance(); - F32 size_squared = size.dot3(size); + F32 size_squared = size.dot3(size).getF32(); LLVector4a lookAt; LLVector4a t; t.load3(camera->getOrigin().mV); lookAt.setSub(center, t); - F32 dist = lookAt.length3(); + F32 dist = lookAt.getLength3().getF32(); lookAt.normalize3fast() ; //get area of circle around node - F32 app_angle = atanf(fsqrtf(size_squared) / dist); + F32 app_angle = atanf((F32) sqrt(size_squared) / dist); radius = app_angle*LLDrawable::sCurPixelAngle; mPixelArea = radius*radius * 3.14159f; LLVector4a x_axis; x_axis.load3(camera->getXAxis().mV); - cos_angle_to_view_dir = lookAt.dot3(x_axis); + cos_angle_to_view_dir = lookAt.dot3(x_axis).getF32(); //if has media, check if the face is out of the view frustum. if(hasMedia()) diff --git a/indra/newview/llface.h b/indra/newview/llface.h index b6a67c7fc4..a7d7889350 100644 --- a/indra/newview/llface.h +++ b/indra/newview/llface.h @@ -240,6 +240,7 @@ public: LLPointer mVertexBuffer; LLPointer mLastVertexBuffer; F32 mLastUpdateTime; + F32 mLastSkinTime; F32 mLastMoveTime; LLMatrix4* mTextureMatrix; LLDrawInfo* mDrawInfo; @@ -274,7 +275,7 @@ private: S32 mReferenceIndex; std::vector mRiggedIndex; - + F32 mVSize; F32 mPixelArea; diff --git a/indra/newview/llflexibleobject.cpp b/indra/newview/llflexibleobject.cpp index 8be4e34748..62336b03bf 100644 --- a/indra/newview/llflexibleobject.cpp +++ b/indra/newview/llflexibleobject.cpp @@ -100,7 +100,7 @@ void LLVolumeImplFlexible::onParameterChanged(U16 param_type, LLNetworkData *dat void LLVolumeImplFlexible::onShift(const LLVector4a &shift_vector) { //VECTORIZE THIS - LLVector3 shift(shift_vector.getF32()); + LLVector3 shift(shift_vector.getF32ptr()); for (int section = 0; section < (1<getInterest()); // proximity distance is actually distance squared -- display it as straight distance. - debug_str += llformat("%g/", fsqrtf(impl->getProximityDistance())); + debug_str += llformat("%g/", (F32) sqrt(impl->getProximityDistance())); // s += llformat("%g/", (float)impl->getCPUUsage()); // s += llformat("%g/", (float)impl->getApproximateTextureInterest()); diff --git a/indra/newview/llpanelprimmediacontrols.cpp b/indra/newview/llpanelprimmediacontrols.cpp index 98fbebbc5d..d84ac2e4c8 100644 --- a/indra/newview/llpanelprimmediacontrols.cpp +++ b/indra/newview/llpanelprimmediacontrols.cpp @@ -576,8 +576,8 @@ void LLPanelPrimMediaControls::updateShape() const LLVolumeFace& vf = volume->getVolumeFace(mTargetObjectFace); LLVector3 ext[2]; - ext[0].set(vf.mExtents[0].getF32()); - ext[1].set(vf.mExtents[1].getF32()); + ext[0].set(vf.mExtents[0].getF32ptr()); + ext[1].set(vf.mExtents[1].getF32ptr()); LLVector3 center = (ext[0]+ext[1])*0.5f; LLVector3 size = (ext[1]-ext[0])*0.5f; diff --git a/indra/newview/llselectmgr.cpp b/indra/newview/llselectmgr.cpp index 42f09f7396..92903a6aa9 100644 --- a/indra/newview/llselectmgr.cpp +++ b/indra/newview/llselectmgr.cpp @@ -1124,13 +1124,13 @@ void LLSelectMgr::getGrid(LLVector3& origin, LLQuaternion &rotation, LLVector3 & size.setSub(max_extents, min_extents); size.mul(0.5f); - mGridOrigin.set(center.getF32()); + mGridOrigin.set(center.getF32ptr()); LLDrawable* drawable = first_grid_object->mDrawable; if (drawable && drawable->isActive()) { mGridOrigin = mGridOrigin * first_grid_object->getRenderMatrix(); } - mGridScale.set(size.getF32()); + mGridScale.set(size.getF32ptr()); } } else // GRID_MODE_WORLD or just plain default @@ -3543,7 +3543,7 @@ void LLSelectMgr::deselectAllIfTooFar() { if (mDebugSelectMgr) { - llinfos << "Selection manager: auto-deselecting, select_dist = " << fsqrtf(select_dist_sq) << llendl; + llinfos << "Selection manager: auto-deselecting, select_dist = " << (F32) sqrt(select_dist_sq) << llendl; llinfos << "agent pos global = " << gAgent.getPositionGlobal() << llendl; llinfos << "selection pos global = " << selectionCenter << llendl; } diff --git a/indra/newview/llspatialpartition.cpp b/indra/newview/llspatialpartition.cpp index 2335c7bb8e..e55ec03356 100644 --- a/indra/newview/llspatialpartition.cpp +++ b/indra/newview/llspatialpartition.cpp @@ -231,7 +231,7 @@ U8* get_box_fan_indices(LLCamera* camera, const LLVector4a& center) LLVector4a origin; origin.load3(camera->getOrigin().mV); - S32 cypher = center.greaterThan4(origin).getComparisonMask() & 0x7; + S32 cypher = center.greaterThan(origin).getGatheredBits() & 0x7; return sOcclusionIndices+cypher*8; } @@ -253,7 +253,7 @@ void LLSpatialGroup::buildOcclusion() r2.splat(0.25f); r2.add(mBounds[1]); - r.setMin(r2); + r.setMin(r, r2); LLVector4a* v = mOcclusionVerts; const LLVector4a& c = mBounds[0]; @@ -775,8 +775,8 @@ BOOL LLSpatialGroup::boundObjects(BOOL empty, LLVector4a& minOut, LLVector4a& ma } else { - minOut.setMin(newMin); - maxOut.setMax(newMax); + minOut.setMin(minOut, newMin); + maxOut.setMax(maxOut, newMax); } return TRUE; @@ -1220,8 +1220,8 @@ void LLSpatialGroup::updateDistance(LLCamera &camera) #endif if (!getData().empty()) { - mRadius = mSpatialPartition->mRenderByGroup ? mObjectBounds[1].length3() : - (F32) mOctreeNode->getSize().length3(); + mRadius = mSpatialPartition->mRenderByGroup ? mObjectBounds[1].getLength3().getF32() : + (F32) mOctreeNode->getSize().getLength3().getF32(); mDistance = mSpatialPartition->calcDistance(this, camera); mPixelArea = mSpatialPartition->calcPixelArea(this, camera); } @@ -1241,7 +1241,7 @@ F32 LLSpatialPartition::calcDistance(LLSpatialGroup* group, LLCamera& camera) { LLVector4a v = eye; - dist = eye.length3(); + dist = eye.getLength3().getF32(); eye.normalize3fast(); if (!group->isState(LLSpatialGroup::ALPHA_DIRTY)) @@ -1253,7 +1253,7 @@ F32 LLSpatialPartition::calcDistance(LLSpatialGroup* group, LLCamera& camera) LLVector4a diff; diff.setSub(view_angle, *group->mLastUpdateViewAngle); - if (diff.length3() > 0.64f) + if (diff.getLength3().getF32() > 0.64f) { *group->mViewAngle = view_angle; *group->mLastUpdateViewAngle = view_angle; @@ -1279,11 +1279,11 @@ F32 LLSpatialPartition::calcDistance(LLSpatialGroup* group, LLCamera& camera) t.mul(group->mObjectBounds[1]); v.sub(t); - group->mDepth = v.dot3(ata); + group->mDepth = v.dot3(ata).getF32(); } else { - dist = eye.length3(); + dist = eye.getLength3().getF32(); } if (dist < 16.f) @@ -1497,8 +1497,8 @@ BOOL LLSpatialGroup::rebound() const LLVector4a& max = group->mExtents[1]; const LLVector4a& min = group->mExtents[0]; - newMax.setMax(max); - newMin.setMin(min); + newMax.setMax(newMax, max); + newMin.setMin(newMin, min); } boundObjects(FALSE, newMin, newMax); @@ -2196,8 +2196,8 @@ BOOL LLSpatialPartition::getVisibleExtents(LLCamera& camera, LLVector3& visMin, LLOctreeCullVisExtents vis(&camera, visMina, visMaxa); vis.traverse(mOctree); - visMin.set(visMina.getF32()); - visMax.set(visMaxa.getF32()); + visMin.set(visMina.getF32ptr()); + visMax.set(visMaxa.getF32ptr()); return vis.mEmpty; } @@ -2280,13 +2280,13 @@ BOOL earlyFail(LLCamera* camera, LLSpatialGroup* group) LLVector4a max; max.setAdd(c,r); - S32 lt = e.lessThan4(min).getComparisonMask() & 0x7; + S32 lt = e.lessThan(min).getGatheredBits() & 0x7; if (lt) { return FALSE; } - S32 gt = e.greaterThan4(max).getComparisonMask() & 0x7; + S32 gt = e.greaterThan(max).getGatheredBits() & 0x7; if (gt) { return FALSE; @@ -2745,8 +2745,8 @@ void renderNormals(LLDrawable* drawablep) p.setAdd(face.mPositions[j], n); gGL.color4f(1,1,1,1); - gGL.vertex3fv(face.mPositions[j].getF32()); - gGL.vertex3fv(p.getF32()); + gGL.vertex3fv(face.mPositions[j].getF32ptr()); + gGL.vertex3fv(p.getF32ptr()); if (face.mBinormals) { @@ -2754,8 +2754,8 @@ void renderNormals(LLDrawable* drawablep) p.setAdd(face.mPositions[j], n); gGL.color4f(0,1,1,1); - gGL.vertex3fv(face.mPositions[j].getF32()); - gGL.vertex3fv(p.getF32()); + gGL.vertex3fv(face.mPositions[j].getF32ptr()); + gGL.vertex3fv(p.getF32ptr()); } } @@ -3024,8 +3024,8 @@ public: LLVolumeOctreeListener* vl = (LLVolumeOctreeListener*) branch->getListener(0); LLVector3 center, size; - center.set(vl->mBounds[0].getF32()); - size.set(vl->mBounds[1].getF32()); + center.set(vl->mBounds[0].getF32ptr()); + size.set(vl->mBounds[1].getF32ptr()); drawBoxOutline(center, size); } diff --git a/indra/newview/llviewercamera.cpp b/indra/newview/llviewercamera.cpp index cef7c4abbb..917185ed04 100644 --- a/indra/newview/llviewercamera.cpp +++ b/indra/newview/llviewercamera.cpp @@ -810,7 +810,7 @@ BOOL LLViewerCamera::areVertsVisible(LLViewerObject* volumep, BOOL all_verts) render_mata.affineTransform(t, vec); } - BOOL in_frustum = pointInFrustum(LLVector3(vec.getF32())) > 0; + BOOL in_frustum = pointInFrustum(LLVector3(vec.getF32ptr())) > 0; if (( !in_frustum && all_verts) || (in_frustum && !all_verts)) diff --git a/indra/newview/llviewerjointmesh.cpp b/indra/newview/llviewerjointmesh.cpp index 9e35b5cc51..0a181e09bb 100644 --- a/indra/newview/llviewerjointmesh.cpp +++ b/indra/newview/llviewerjointmesh.cpp @@ -726,8 +726,8 @@ void LLViewerJointMesh::updateFaceData(LLFace *face, F32 pixel_area, BOOL damp_w U32 words = num_verts*4; - LLVector4a::memcpyNonAliased16(v, (F32*) mMesh->getCoords(), words); - LLVector4a::memcpyNonAliased16(n, (F32*) mMesh->getNormals(), words); + LLVector4a::memcpyNonAliased16(v, (F32*) mMesh->getCoords(), words*sizeof(F32)); + LLVector4a::memcpyNonAliased16(n, (F32*) mMesh->getNormals(), words*sizeof(F32)); if (!terse_update) @@ -740,9 +740,9 @@ void LLViewerJointMesh::updateFaceData(LLFace *face, F32 pixel_area, BOOL damp_w F32* vw = (F32*) vertex_weightsp.get(); F32* cw = (F32*) clothing_weightsp.get(); - LLVector4a::memcpyNonAliased16(tc, (F32*) mMesh->getTexCoords(), num_verts*2); - LLVector4a::memcpyNonAliased16(vw, (F32*) mMesh->getWeights(), num_verts); - LLVector4a::memcpyNonAliased16(cw, (F32*) mMesh->getClothingWeights(), num_verts*4); + LLVector4a::memcpyNonAliased16(tc, (F32*) mMesh->getTexCoords(), num_verts*2*sizeof(F32)); + LLVector4a::memcpyNonAliased16(vw, (F32*) mMesh->getWeights(), num_verts*sizeof(F32)); + LLVector4a::memcpyNonAliased16(cw, (F32*) mMesh->getClothingWeights(), num_verts*4*sizeof(F32)); } const U32 idx_count = mMesh->getNumFaces()*3; diff --git a/indra/newview/llviewerjoystick.cpp b/indra/newview/llviewerjoystick.cpp index 79d8fc7df9..16a6022b86 100644 --- a/indra/newview/llviewerjoystick.cpp +++ b/indra/newview/llviewerjoystick.cpp @@ -764,7 +764,7 @@ void LLViewerJoystick::moveAvatar(bool reset) sDelta[RX_I] += (cur_delta[RX_I] - sDelta[RX_I]) * time * feather; sDelta[RY_I] += (cur_delta[RY_I] - sDelta[RY_I]) * time * feather; - handleRun(fsqrtf(sDelta[Z_I]*sDelta[Z_I] + sDelta[X_I]*sDelta[X_I])); + handleRun((F32) sqrt(sDelta[Z_I]*sDelta[Z_I] + sDelta[X_I]*sDelta[X_I])); // Allow forward/backward movement some priority if (dom_axis == Z_I) diff --git a/indra/newview/llviewermedia.cpp b/indra/newview/llviewermedia.cpp index 178d928f57..303f339f7d 100644 --- a/indra/newview/llviewermedia.cpp +++ b/indra/newview/llviewermedia.cpp @@ -865,7 +865,7 @@ void LLViewerMedia::updateMedia(void *dummy_arg) // Set the low priority size for downsampling to approximately the size the texture is displayed at. { - F32 approximate_interest_dimension = fsqrtf(pimpl->getInterest()); + F32 approximate_interest_dimension = (F32) sqrt(pimpl->getInterest()); pimpl->setLowPrioritySizeLimit(llround(approximate_interest_dimension)); } diff --git a/indra/newview/llviewerobject.cpp b/indra/newview/llviewerobject.cpp index 2b89deaa53..9587fbafb1 100644 --- a/indra/newview/llviewerobject.cpp +++ b/indra/newview/llviewerobject.cpp @@ -2973,7 +2973,7 @@ F32 LLViewerObject::getBinRadius() const LLVector4a* ext = mDrawable->getSpatialExtents(); LLVector4a diff; diff.setSub(ext[1], ext[0]); - return diff.length3(); + return diff.getLength3().getF32(); } return getScale().magVec(); diff --git a/indra/newview/llviewerpartsim.cpp b/indra/newview/llviewerpartsim.cpp index 41848e8b7a..4759454ee7 100644 --- a/indra/newview/llviewerpartsim.cpp +++ b/indra/newview/llviewerpartsim.cpp @@ -161,8 +161,8 @@ LLViewerPartGroup::LLViewerPartGroup(const LLVector3 ¢er_agent, const F32 bo if (group != NULL) { - LLVector3 center(group->mOctreeNode->getCenter().getF32()); - LLVector3 size(group->mOctreeNode->getSize().getF32()); + LLVector3 center(group->mOctreeNode->getCenter().getF32ptr()); + LLVector3 size(group->mOctreeNode->getSize().getF32ptr()); size += LLVector3(0.01f, 0.01f, 0.01f); mMinObjPos = center - size; mMaxObjPos = center + size; diff --git a/indra/newview/llviewertexture.cpp b/indra/newview/llviewertexture.cpp index 719c5b0da5..d50efe89dd 100644 --- a/indra/newview/llviewertexture.cpp +++ b/indra/newview/llviewertexture.cpp @@ -1580,7 +1580,7 @@ F32 LLViewerFetchedTexture::calcDecodePriority() S32 cur_discard = getCurrentDiscardLevelForFetching(); bool have_all_data = (cur_discard >= 0 && (cur_discard <= mDesiredDiscardLevel)); - F32 pixel_priority = fsqrtf(mMaxVirtualSize); + F32 pixel_priority = (F32) sqrt(mMaxVirtualSize); F32 priority = 0.f; diff --git a/indra/newview/llvoavatar.cpp b/indra/newview/llvoavatar.cpp index abf22b5e5a..e50d0fce49 100644 --- a/indra/newview/llvoavatar.cpp +++ b/indra/newview/llvoavatar.cpp @@ -675,6 +675,7 @@ LLVOAvatar::LLVOAvatar(const LLUUID& id, mTexHairColor( NULL ), mTexEyeColor( NULL ), mNeedsSkin(FALSE), + mLastSkinTime(0.f), mUpdatePeriod(1), mFullyLoaded(FALSE), mPreviousFullyLoaded(FALSE), @@ -1356,7 +1357,7 @@ void LLVOAvatar::updateSpatialExtents(LLVector4a& newMin, LLVector4a &newMax) if (isImpostor() && !needsImpostorUpdate()) { LLVector3 delta = getRenderPosition() - - ((LLVector3(mDrawable->getPositionGroup().getF32())-mImpostorOffset)); + ((LLVector3(mDrawable->getPositionGroup().getF32ptr())-mImpostorOffset)); newMin.load3( (mLastAnimExtents[0] + delta).mV); newMax.load3( (mLastAnimExtents[1] + delta).mV); @@ -1364,12 +1365,12 @@ void LLVOAvatar::updateSpatialExtents(LLVector4a& newMin, LLVector4a &newMax) else { getSpatialExtents(newMin,newMax); - mLastAnimExtents[0].set(newMin.getF32()); - mLastAnimExtents[1].set(newMax.getF32()); + mLastAnimExtents[0].set(newMin.getF32ptr()); + mLastAnimExtents[1].set(newMax.getF32ptr()); LLVector4a pos_group; pos_group.setAdd(newMin,newMax); pos_group.mul(0.5f); - mImpostorOffset = LLVector3(pos_group.getF32())-getRenderPosition(); + mImpostorOffset = LLVector3(pos_group.getF32ptr())-getRenderPosition(); mDrawable->setPositionGroup(pos_group); } } @@ -1435,7 +1436,7 @@ void LLVOAvatar::getSpatialExtents(LLVector4a& newMin, LLVector4a& newMax) distance.setSub(ext[1], ext[0]); LLVector4a max_span(max_attachment_span); - S32 lt = distance.lessThan4(max_span).getComparisonMask() & 0x7; + S32 lt = distance.lessThan(max_span).getGatheredBits() & 0x7; // Only add the prim to spatial extents calculations if it isn't a megaprim. // max_attachment_span calculated at the start of the function @@ -2533,14 +2534,14 @@ void LLVOAvatar::idleUpdateMisc(bool detailed_update) getSpatialExtents(ext[0], ext[1]); LLVector4a diff; diff.setSub(ext[1], mImpostorExtents[1]); - if (diff.length3() > 0.05f) + if (diff.getLength3().getF32() > 0.05f) { mNeedsImpostorUpdate = TRUE; } else { diff.setSub(ext[0], mImpostorExtents[0]); - if (diff.length3() > 0.05f) + if (diff.getLength3().getF32() > 0.05f) { mNeedsImpostorUpdate = TRUE; } @@ -3887,7 +3888,8 @@ U32 LLVOAvatar::renderSkinned(EAvatarRenderPass pass) mMeshLOD[MESH_ID_HAIR]->updateJointGeometry(); } mNeedsSkin = FALSE; - + mLastSkinTime = gFrameTimeSeconds; + LLVertexBuffer* vb = mDrawable->getFace(0)->mVertexBuffer; if (vb) { @@ -4231,7 +4233,7 @@ void LLVOAvatar::updateTextures() if (gPipeline.hasRenderDebugMask(LLPipeline::RENDER_DEBUG_TEXTURE_AREA)) { - setDebugText(llformat("%4.0f:%4.0f", fsqrtf(mMinPixelArea),fsqrtf(mMaxPixelArea))); + setDebugText(llformat("%4.0f:%4.0f", (F32) sqrt(mMinPixelArea),(F32) sqrt(mMaxPixelArea))); } } @@ -5443,7 +5445,7 @@ void LLVOAvatar::setPixelAreaAndAngle(LLAgent &agent) } else { - F32 radius = size.length3(); + F32 radius = size.getLength3().getF32(); mAppAngle = (F32) atan2( radius, range) * RAD_TO_DEG; } diff --git a/indra/newview/llvoavatar.h b/indra/newview/llvoavatar.h index 95b0665f7d..94b564fc8f 100644 --- a/indra/newview/llvoavatar.h +++ b/indra/newview/llvoavatar.h @@ -345,6 +345,7 @@ public: U32 renderImpostor(LLColor4U color = LLColor4U(255,255,255,255), S32 diffuse_channel = 0); U32 renderRigid(); U32 renderSkinned(EAvatarRenderPass pass); + F32 getLastSkinTime() { return mLastSkinTime; } U32 renderSkinnedAttachments(); U32 renderTransparent(BOOL first_pass); void renderCollisionVolumes(); @@ -357,6 +358,8 @@ private: bool shouldAlphaMask(); BOOL mNeedsSkin; // avatar has been animated and verts have not been updated + F32 mLastSkinTime; //value of gFrameTimeSeconds at last skin update + S32 mUpdatePeriod; S32 mNumInitFaces; //number of faces generated when creating the avatar drawable, does not inculde splitted faces due to long vertex buffer. diff --git a/indra/newview/llvograss.cpp b/indra/newview/llvograss.cpp index fe1e36cbe8..65829b213e 100644 --- a/indra/newview/llvograss.cpp +++ b/indra/newview/llvograss.cpp @@ -328,7 +328,7 @@ void LLVOGrass::updateTextures() { if (gPipeline.hasRenderDebugMask(LLPipeline::RENDER_DEBUG_TEXTURE_AREA)) { - setDebugText(llformat("%4.0f", fsqrtf(mPixelArea))); + setDebugText(llformat("%4.0f", (F32) sqrt(mPixelArea))); } getTEImage(0)->addTextureStats(mPixelArea); } diff --git a/indra/newview/llvosurfacepatch.cpp b/indra/newview/llvosurfacepatch.cpp index 63f662c5a1..c047758a59 100644 --- a/indra/newview/llvosurfacepatch.cpp +++ b/indra/newview/llvosurfacepatch.cpp @@ -999,8 +999,8 @@ BOOL LLVOSurfacePatch::lineSegmentIntersect(const LLVector3& start, const LLVect const LLVector4a* exta = mDrawable->getSpatialExtents(); LLVector3 ext[2]; - ext[0].set(exta[0].getF32()); - ext[1].set(exta[1].getF32()); + ext[0].set(exta[0].getF32ptr()); + ext[1].set(exta[1].getF32ptr()); F32 rad = (delta*tdelta).magVecSquared(); diff --git a/indra/newview/llvotextbubble.cpp b/indra/newview/llvotextbubble.cpp index 339da3c0bf..e790373d02 100644 --- a/indra/newview/llvotextbubble.cpp +++ b/indra/newview/llvotextbubble.cpp @@ -254,8 +254,8 @@ void LLVOTextBubble::getGeometry(S32 idx, LLVector2* dst_tc = (LLVector2*) texcoordsp.get(); LLVector2* src_tc = (LLVector2*) face.mTexCoords; - LLVector4a::memcpyNonAliased16((F32*) dst_norm, (F32*) src_norm, face.mNumVertices*4); - LLVector4a::memcpyNonAliased16((F32*) dst_tc, (F32*) src_tc, face.mNumVertices*2); + LLVector4a::memcpyNonAliased16((F32*) dst_norm, (F32*) src_norm, face.mNumVertices*4*sizeof(F32)); + LLVector4a::memcpyNonAliased16((F32*) dst_tc, (F32*) src_tc, face.mNumVertices*2*sizeof(F32)); for (U32 i = 0; i < face.mNumVertices; i++) diff --git a/indra/newview/llvotree.cpp b/indra/newview/llvotree.cpp index eb790b04cc..e3b4efb9dd 100644 --- a/indra/newview/llvotree.cpp +++ b/indra/newview/llvotree.cpp @@ -472,7 +472,7 @@ void LLVOTree::updateTextures() { if (gPipeline.hasRenderDebugMask(LLPipeline::RENDER_DEBUG_TEXTURE_AREA)) { - setDebugText(llformat("%4.0f", fsqrtf(mPixelArea))); + setDebugText(llformat("%4.0f", (F32) sqrt(mPixelArea))); } mTreeImagep->addTextureStats(mPixelArea); } @@ -1278,8 +1278,8 @@ BOOL LLVOTree::lineSegmentIntersect(const LLVector3& start, const LLVector3& end //VECTORIZE THIS LLVector3 ext[2]; - ext[0].set(exta[0].getF32()); - ext[1].set(exta[1].getF32()); + ext[0].set(exta[0].getF32ptr()); + ext[1].set(exta[1].getF32ptr()); LLVector3 center = (ext[1]+ext[0])*0.5f; LLVector3 size = (ext[1]-ext[0]); diff --git a/indra/newview/llvovolume.cpp b/indra/newview/llvovolume.cpp index 1397b64623..128fd15142 100644 --- a/indra/newview/llvovolume.cpp +++ b/indra/newview/llvovolume.cpp @@ -696,7 +696,7 @@ void LLVOVolume::updateTextureVirtualSize() const LLTextureEntry *te = face->getTextureEntry(); LLViewerTexture *imagep = face->getTexture(); if (!imagep || !te || - face->mExtents[0].equal3(face->mExtents[1])) + face->mExtents[0].equals3(face->mExtents[1])) { continue; } @@ -820,15 +820,15 @@ void LLVOVolume::updateTextureVirtualSize() if (gPipeline.hasRenderDebugMask(LLPipeline::RENDER_DEBUG_TEXTURE_AREA)) { - setDebugText(llformat("%.0f:%.0f", fsqrtf(min_vsize),fsqrtf(max_vsize))); + setDebugText(llformat("%.0f:%.0f", (F32) sqrt(min_vsize),(F32) sqrt(max_vsize))); } else if (gPipeline.hasRenderDebugMask(LLPipeline::RENDER_DEBUG_TEXTURE_PRIORITY)) { - setDebugText(llformat("%.0f:%.0f", fsqrtf(min_vsize),fsqrtf(max_vsize))); + setDebugText(llformat("%.0f:%.0f", (F32) sqrt(min_vsize),(F32) sqrt(max_vsize))); } else if (gPipeline.hasRenderDebugMask(LLPipeline::RENDER_DEBUG_FACE_AREA)) { - setDebugText(llformat("%.0f:%.0f", fsqrtf(min_vsize),fsqrtf(max_vsize))); + setDebugText(llformat("%.0f:%.0f", (F32) sqrt(min_vsize),(F32) sqrt(max_vsize))); } if (mPixelArea == 0) @@ -1355,8 +1355,8 @@ BOOL LLVOVolume::genBBoxes(BOOL force_global) } else { - min.setMin(face->mExtents[0]); - max.setMax(face->mExtents[1]); + min.setMin(min, face->mExtents[0]); + max.setMax(max, face->mExtents[1]); } } } @@ -1864,7 +1864,7 @@ LLVector3 LLVOVolume::getApproximateFaceNormal(U8 face_id) result.add(face.mNormals[i]); } - LLVector3 ret((F32*) &result.mQ); + LLVector3 ret(result.getF32ptr()); ret = volumeDirectionToAgent(ret); ret.normVec(); } @@ -3075,7 +3075,7 @@ F32 LLVOVolume::getBinRadius() LLVector4a rad; rad.setSub(ext[1], ext[0]); - radius = rad.length3()*0.5f; + radius = rad.getLength3().getF32()*0.5f; } else if (mDrawable->isStatic()) { diff --git a/indra/newview/llworld.cpp b/indra/newview/llworld.cpp index a1e4df8a66..53eca0d08e 100644 --- a/indra/newview/llworld.cpp +++ b/indra/newview/llworld.cpp @@ -591,7 +591,7 @@ void LLWorld::updateVisibilities() region_list_t::iterator curiter = iter++; LLViewerRegion* regionp = *curiter; F32 height = regionp->getLand().getMaxZ() - regionp->getLand().getMinZ(); - F32 radius = 0.5f*fsqrtf(height * height + diagonal_squared); + F32 radius = 0.5f*(F32) sqrt(height * height + diagonal_squared); if (!regionp->getLand().hasZData() || LLViewerCamera::getInstance()->sphereInFrustum(regionp->getCenterAgent(), radius)) { @@ -612,7 +612,7 @@ void LLWorld::updateVisibilities() } F32 height = regionp->getLand().getMaxZ() - regionp->getLand().getMinZ(); - F32 radius = 0.5f*fsqrtf(height * height + diagonal_squared); + F32 radius = 0.5f*(F32) sqrt(height * height + diagonal_squared); if (LLViewerCamera::getInstance()->sphereInFrustum(regionp->getCenterAgent(), radius)) { regionp->calculateCameraDistance(); diff --git a/indra/newview/pipeline.cpp b/indra/newview/pipeline.cpp index fec7da1dd0..b0a7b1ce83 100644 --- a/indra/newview/pipeline.cpp +++ b/indra/newview/pipeline.cpp @@ -1522,7 +1522,7 @@ F32 LLPipeline::calcPixelArea(const LLVector4a& center, const LLVector4a& size, LLVector4a lookAt; lookAt.setSub(center, origin); - F32 dist = lookAt.length3(); + F32 dist = lookAt.getLength3().getF32(); //ramp down distance for nearby objects //shrink dist by dist/16. @@ -1534,7 +1534,7 @@ F32 LLPipeline::calcPixelArea(const LLVector4a& center, const LLVector4a& size, } //get area of circle around node - F32 app_angle = atanf(size.length3()/dist); + F32 app_angle = atanf(size.getLength3().getF32()/dist); F32 radius = app_angle*LLDrawable::sCurPixelAngle; return radius*radius * F_PI; } @@ -4671,7 +4671,7 @@ static F32 calc_light_dist(LLVOVolume* light, const LLVector3& cam_pos, F32 max_ { return max_dist; } - F32 dist = fsqrtf(dist2); + F32 dist = (F32) sqrt(dist2); dist *= 1.f / inten; dist -= radius; if (selected) @@ -6980,7 +6980,7 @@ void LLPipeline::renderDeferredLighting() LLVector4a center; center.load3(drawablep->getPositionAgent().mV); - const F32* c = center.getF32(); + const F32* c = center.getF32ptr(); F32 s = volume->getLightRadius()*1.5f; LLColor3 col = volume->getLightColor(); @@ -7078,7 +7078,7 @@ void LLPipeline::renderDeferredLighting() LLVector4a center; center.load3(drawablep->getPositionAgent().mV); - const F32* c = center.getF32(); + const F32* c = center.getF32ptr(); F32 s = volume->getLightRadius()*1.5f; sVisibleLightCount++; @@ -9184,8 +9184,8 @@ void LLPipeline::generateImpostor(LLVOAvatar* avatar) up.mul(up); up.normalize3fast(); - tdim.mV[0] = fabsf(half_height.dot3(left)); - tdim.mV[1] = fabsf(half_height.dot3(up)); + tdim.mV[0] = fabsf(half_height.dot3(left).getF32()); + tdim.mV[1] = fabsf(half_height.dot3(up).getF32()); glMatrixMode(GL_PROJECTION); glPushMatrix(); -- cgit v1.3