From 4353eeb9288c95b98935e60928ec0b80de4e6145 Mon Sep 17 00:00:00 2001 From: Dave Parks Date: Tue, 24 May 2011 15:29:33 -0500 Subject: SH-469 GL_ARB_texture_multisample support --- indra/newview/llviewercontrol.cpp | 17 ++++++++++++++++- 1 file changed, 16 insertions(+), 1 deletion(-) (limited to 'indra/newview/llviewercontrol.cpp') diff --git a/indra/newview/llviewercontrol.cpp b/indra/newview/llviewercontrol.cpp index 379bbe614d..00a0b87d93 100644 --- a/indra/newview/llviewercontrol.cpp +++ b/indra/newview/llviewercontrol.cpp @@ -183,6 +183,21 @@ static bool handleReleaseGLBufferChanged(const LLSD& newvalue) return true; } +static bool handleFSAASamplesChanged(const LLSD& newvalue) +{ + if (gPipeline.isInit()) + { + gPipeline.releaseGLBuffers(); + gPipeline.createGLBuffers(); + + if (LLPipeline::sRenderDeferred) + { + LLViewerShaderMgr::instance()->setShaders(); + } + } + return true; +} + static bool handleAnisotropicChanged(const LLSD& newvalue) { LLImageGL::sGlobalUseAnisotropic = newvalue.asBoolean(); @@ -568,7 +583,7 @@ void settings_setup_listeners() 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("RenderFSAASamples")->getSignal()->connect(boost::bind(&handleFSAASamplesChanged, _2)); gSavedSettings.getControl("RenderAnisotropic")->getSignal()->connect(boost::bind(&handleAnisotropicChanged, _2)); gSavedSettings.getControl("RenderShadowResolutionScale")->getSignal()->connect(boost::bind(&handleReleaseGLBufferChanged, _2)); gSavedSettings.getControl("RenderGlow")->getSignal()->connect(boost::bind(&handleReleaseGLBufferChanged, _2)); -- cgit v1.3 From 53e890f4fdab1ab014b745c5d872b240b8c4ce76 Mon Sep 17 00:00:00 2001 From: Dave Parks Date: Thu, 26 May 2011 20:37:15 -0500 Subject: SH-1359 General octree tune-up -- increase maximum capacity of an octree node and rework requested octree node sizes for minimum triangles per frame and maximum triangles per drawing call. --- indra/llmath/lloctree.h | 11 +++-- indra/newview/app_settings/settings.xml | 75 +++++++++++++++++++++++++++++++-- indra/newview/lldrawable.cpp | 9 +++- indra/newview/lldrawable.h | 1 + indra/newview/llspatialpartition.cpp | 42 ++++++++---------- indra/newview/llviewercontrol.cpp | 16 +++++++ indra/newview/llviewerdisplay.cpp | 1 + indra/newview/llviewerobjectlist.cpp | 18 ++++++++ indra/newview/llviewerobjectlist.h | 1 + indra/newview/llvovolume.cpp | 48 +++++++++++++++------ indra/newview/pipeline.cpp | 28 +++++++++++- indra/newview/pipeline.h | 7 ++- 12 files changed, 208 insertions(+), 49 deletions(-) (limited to 'indra/newview/llviewercontrol.cpp') diff --git a/indra/llmath/lloctree.h b/indra/llmath/lloctree.h index fdfc24f8b7..e5ca47da69 100644 --- a/indra/llmath/lloctree.h +++ b/indra/llmath/lloctree.h @@ -35,12 +35,14 @@ #define OCT_ERRS LL_WARNS("OctreeErrors") -#define LL_OCTREE_PARANOIA_CHECK 0 + +extern U32 gOctreeMaxCapacity; +/*#define LL_OCTREE_PARANOIA_CHECK 0 #if LL_DARWIN #define LL_OCTREE_MAX_CAPACITY 32 #else #define LL_OCTREE_MAX_CAPACITY 128 -#endif +#endif*/ template class LLOctreeNode; @@ -74,6 +76,7 @@ template class LLOctreeNode : public LLTreeNode { public: + typedef LLOctreeTraveler oct_traveler; typedef LLTreeTraveler tree_traveler; typedef typename std::set > element_list; @@ -294,8 +297,8 @@ public: //is it here? if (isInside(data->getPositionGroup())) { - if ((getElementCount() < LL_OCTREE_MAX_CAPACITY && contains(data->getBinRadius()) || - (data->getBinRadius() > getSize()[0] && parent && parent->getElementCount() >= LL_OCTREE_MAX_CAPACITY))) + if ((getElementCount() < gOctreeMaxCapacity && contains(data->getBinRadius()) || + (data->getBinRadius() > getSize()[0] && parent && parent->getElementCount() >= gOctreeMaxCapacity))) { //it belongs here #if LL_OCTREE_PARANOIA_CHECK //if this is a redundant insertion, error out (should never happen) diff --git a/indra/newview/app_settings/settings.xml b/indra/newview/app_settings/settings.xml index 78db307d64..2c6d75e3c9 100644 --- a/indra/newview/app_settings/settings.xml +++ b/indra/newview/app_settings/settings.xml @@ -7091,7 +7091,76 @@ Value 1 - RenderAnisotropic + + OctreeMaxNodeCapacity + + Comment + Maximum number of elements to store in a single octree node + Persist + 1 + Type + U32 + Value + 128 + + + OctreeStaticObjectSizeFactor + + Comment + Multiplier on static object size for determining octree node size + Persist + 1 + Type + S32 + Value + 1 + + + OctreeAlphaDistanceFactor + + Comment + Multiplier on alpha object distance for determining octree node size + Persist + 1 + Type + Vector3 + Value + + 0.05 + 0.0 + 0.0 + + + + OctreeAttachmentSizeFactor + + Comment + Multiplier on attachment size for determining octree node size + Persist + 1 + Type + S32 + Value + 4 + + + OctreeDistanceFactor + + Comment + Multiplier on distance for determining octree node size + Persist + 1 + Type + Vector3 + Value + + 0.05 + 0.0 + 0.0 + + + + RenderAnisotropic Comment Render textures using anisotropic filtering @@ -7188,7 +7257,7 @@ Type F32 Value - 1.0 + 1.0 RenderAvatarVP @@ -8589,7 +8658,7 @@ Type S32 Value - 8192 + 65536 RenderMaxVBOSize diff --git a/indra/newview/lldrawable.cpp b/indra/newview/lldrawable.cpp index 7db13868d6..ad3710843c 100644 --- a/indra/newview/lldrawable.cpp +++ b/indra/newview/lldrawable.cpp @@ -94,7 +94,9 @@ void LLDrawable::init() mRenderType = 0; mCurrentScale = LLVector3(1,1,1); mDistanceWRTCamera = 0.0f; - + mPositionGroup.clear(); + mExtents[0].clear(); + mExtents[1].clear(); mQuietCount = 0; mState = 0; @@ -587,7 +589,10 @@ void LLDrawable::setRadius(F32 radius) void LLDrawable::moveUpdatePipeline(BOOL moved) { - makeActive(); + if (moved) + { + makeActive(); + } // Update the face centers. for (S32 i = 0; i < getNumFaces(); i++) diff --git a/indra/newview/lldrawable.h b/indra/newview/lldrawable.h index 9ebe1a45b4..e268640a21 100644 --- a/indra/newview/lldrawable.h +++ b/indra/newview/lldrawable.h @@ -276,6 +276,7 @@ public: REBUILD_SHADOW = 0x02000000, HAS_ALPHA = 0x04000000, RIGGED = 0x08000000, + PARTITION_MOVE = 0x10000000, } EDrawableFlags; private: //aligned members diff --git a/indra/newview/llspatialpartition.cpp b/indra/newview/llspatialpartition.cpp index a812ba4e03..8143d6a41f 100644 --- a/indra/newview/llspatialpartition.cpp +++ b/indra/newview/llspatialpartition.cpp @@ -69,6 +69,7 @@ U32 LLSpatialGroup::sNodeCount = 0; std::set LLSpatialGroup::sPendingQueries; +U32 gOctreeMaxCapacity; BOOL LLSpatialGroup::sNoDelete = FALSE; @@ -630,7 +631,7 @@ BOOL LLSpatialGroup::updateInGroup(LLDrawable *drawablep, BOOL immediate) if (mOctreeNode->isInside(drawablep->getPositionGroup()) && (mOctreeNode->contains(drawablep) || (drawablep->getBinRadius() > mOctreeNode->getSize()[0] && - parent && parent->getElementCount() >= LL_OCTREE_MAX_CAPACITY))) + parent && parent->getElementCount() >= gOctreeMaxCapacity))) { unbound(); setState(OBJECT_DIRTY); @@ -2515,7 +2516,7 @@ void renderOctree(LLSpatialGroup* group) //coded by buffer usage and activity gGL.setSceneBlendType(LLRender::BT_ADD_WITH_ALPHA); LLVector4 col; - if (group->mBuilt > 0.f) + /*if (group->mBuilt > 0.f) { group->mBuilt -= 2.f * gFrameIntervalSeconds; if (group->mBufferUsage == GL_STATIC_DRAW_ARB) @@ -2584,7 +2585,7 @@ void renderOctree(LLSpatialGroup* group) gGL.color4f(1,1,1,1); } } - else + else*/ { if (group->mBufferUsage == GL_STATIC_DRAW_ARB && !group->getData().empty() && group->mSpatialPartition->mRenderByGroup) @@ -2604,33 +2605,24 @@ void renderOctree(LLSpatialGroup* group) size.mul(1.01f); size.add(fudge); - { - LLGLDepthTest depth(GL_TRUE, GL_FALSE); - drawBox(group->mObjectBounds[0], fudge); - } + //{ + // LLGLDepthTest depth(GL_TRUE, GL_FALSE); + // drawBox(group->mObjectBounds[0], fudge); + //} gGL.setSceneBlendType(LLRender::BT_ALPHA); - if (group->mBuilt <= 0.f) + //if (group->mBuilt <= 0.f) { //draw opaque outline - gGL.color4f(col.mV[0], col.mV[1], col.mV[2], 1.f); - drawBoxOutline(group->mObjectBounds[0], group->mObjectBounds[1]); + //gGL.color4f(col.mV[0], col.mV[1], col.mV[2], 1.f); + //drawBoxOutline(group->mObjectBounds[0], group->mObjectBounds[1]); - if (group->mOctreeNode->isLeaf()) - { - gGL.color4f(1,1,1,1); - } - else - { - gGL.color4f(0,1,1,1); - } - + gGL.color4f(0,1,1,1); drawBoxOutline(group->mBounds[0],group->mBounds[1]); - - + //draw bounding box for draw info - if (group->mSpatialPartition->mRenderByGroup) + /*if (group->mSpatialPartition->mRenderByGroup) { gGL.color4f(1.0f, 0.75f, 0.25f, 0.6f); for (LLSpatialGroup::draw_map_t::iterator i = group->mDrawMap.begin(); i != group->mDrawMap.end(); ++i) @@ -2647,7 +2639,7 @@ void renderOctree(LLSpatialGroup* group) drawBoxOutline(center, size); } } - } + }*/ } // LLSpatialGroup::OctreeNode* node = group->mOctreeNode; @@ -2690,7 +2682,7 @@ void renderVisibility(LLSpatialGroup* group, LLCamera* camera) gGL.color4f(0.f, 0.75f, 0.f, 0.5f); pushBufferVerts(group, LLVertexBuffer::MAP_VERTEX); } - else if (camera && group->mOcclusionVerts.notNull()) + /*else if (camera && group->mOcclusionVerts.notNull()) { LLVertexBuffer::unbind(); group->mOcclusionVerts->setBuffer(LLVertexBuffer::MAP_VERTEX); @@ -2702,7 +2694,7 @@ void renderVisibility(LLSpatialGroup* group, LLCamera* camera) glColor4f(1.0f, 1.f, 1.f, 1.0f); group->mOcclusionVerts->drawRange(LLRender::TRIANGLE_FAN, 0, 7, 8, get_box_fan_indices(camera, group->mBounds[0])); glPolygonMode(GL_FRONT_AND_BACK, GL_FILL); - } + }*/ } } diff --git a/indra/newview/llviewercontrol.cpp b/indra/newview/llviewercontrol.cpp index 00a0b87d93..61200db2b9 100644 --- a/indra/newview/llviewercontrol.cpp +++ b/indra/newview/llviewercontrol.cpp @@ -57,6 +57,7 @@ #include "llworld.h" #include "pipeline.h" #include "llviewerjoystick.h" +#include "llviewerobjectlist.h" #include "llviewerparcelmgr.h" #include "llparcel.h" #include "llkeyboard.h" @@ -372,6 +373,16 @@ static bool handleResetVertexBuffersChanged(const LLSD&) return true; } +static bool handleRepartition(const LLSD&) +{ + if (gPipeline.isInit()) + { + gOctreeMaxCapacity = gSavedSettings.getU32("OctreeMaxNodeCapacity"); + gObjectList.repartitionObjects(); + } + return true; +} + static bool handleRenderDynamicLODChanged(const LLSD& newvalue) { LLPipeline::sDynamicLOD = newvalue.asBoolean(); @@ -575,6 +586,11 @@ 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("OctreeStaticObjectSizeFactor")->getSignal()->connect(boost::bind(&handleRepartition, _2)); + gSavedSettings.getControl("OctreeDistanceFactor")->getSignal()->connect(boost::bind(&handleRepartition, _2)); + gSavedSettings.getControl("OctreeMaxNodeCapacity")->getSignal()->connect(boost::bind(&handleRepartition, _2)); + gSavedSettings.getControl("OctreeAlphaDistanceFactor")->getSignal()->connect(boost::bind(&handleRepartition, _2)); + gSavedSettings.getControl("OctreeAttachmentSizeFactor")->getSignal()->connect(boost::bind(&handleRepartition, _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)); diff --git a/indra/newview/llviewerdisplay.cpp b/indra/newview/llviewerdisplay.cpp index 10e96525ce..5dadd2c33b 100644 --- a/indra/newview/llviewerdisplay.cpp +++ b/indra/newview/llviewerdisplay.cpp @@ -582,6 +582,7 @@ void display(BOOL rebuild, F32 zoom_factor, int subfield, BOOL for_snapshot) LLMemType mt_ug(LLMemType::MTYPE_DISPLAY_UPDATE_GEOM); const F32 max_geom_update_time = 0.005f*10.f*gFrameIntervalSeconds; // 50 ms/second update time gPipeline.createObjects(max_geom_update_time); + gPipeline.processPartitionQ(); gPipeline.updateGeom(max_geom_update_time); stop_glerror(); } diff --git a/indra/newview/llviewerobjectlist.cpp b/indra/newview/llviewerobjectlist.cpp index 78b06b0764..45c6777ae8 100644 --- a/indra/newview/llviewerobjectlist.cpp +++ b/indra/newview/llviewerobjectlist.cpp @@ -1497,6 +1497,24 @@ void LLViewerObjectList::shiftObjects(const LLVector3 &offset) LLWorld::getInstance()->shiftRegions(offset); } +void LLViewerObjectList::repartitionObjects() +{ + for (vobj_list_t::iterator iter = mObjects.begin(); iter != mObjects.end(); ++iter) + { + LLViewerObject* objectp = *iter; + if (!objectp->isDead()) + { + LLDrawable* drawable = objectp->mDrawable; + if (drawable && !drawable->isDead()) + { + drawable->updateBinRadius(); + drawable->updateSpatialExtents(); + drawable->movePartition(); + } + } + } +} + //debug code bool LLViewerObjectList::hasMapObjectInRegion(LLViewerRegion* regionp) { diff --git a/indra/newview/llviewerobjectlist.h b/indra/newview/llviewerobjectlist.h index 6f0d285f5d..9d1b5cb56f 100644 --- a/indra/newview/llviewerobjectlist.h +++ b/indra/newview/llviewerobjectlist.h @@ -105,6 +105,7 @@ public: void updateQuota( const LLUUID& objectId, const SelectionQuota& costs ); void shiftObjects(const LLVector3 &offset); + void repartitionObjects(); bool hasMapObjectInRegion(LLViewerRegion* regionp) ; void clearAllMapObjectsInRegion(LLViewerRegion* regionp) ; diff --git a/indra/newview/llvovolume.cpp b/indra/newview/llvovolume.cpp index 943de669f6..cc443d32fb 100644 --- a/indra/newview/llvovolume.cpp +++ b/indra/newview/llvovolume.cpp @@ -1245,7 +1245,19 @@ BOOL LLVOVolume::calcLOD() if (gPipeline.hasRenderDebugMask(LLPipeline::RENDER_DEBUG_LOD_INFO)) { - setDebugText(llformat("%.2f:%.2f, %d", debug_distance, radius, cur_detail)); + //setDebugText(llformat("%.2f:%.2f, %d", debug_distance, radius, cur_detail)); + + F32 bin_radius = getBinRadius(); + F32 node_size = 0.f; + + LLSpatialGroup* group = mDrawable->getSpatialGroup(); + if (group) + { + LLSpatialGroup::OctreeNode* node = group->mOctreeNode; + node_size = node->getSize()[0]; + } + + setDebugText(llformat("%.2f:%.2f", bin_radius, node_size)); } if (cur_detail != mLOD) @@ -1274,6 +1286,15 @@ BOOL LLVOVolume::updateLOD() gPipeline.markRebuild(mDrawable, LLDrawable::REBUILD_VOLUME, FALSE); mLODChanged = TRUE; } + else + { + F32 new_radius = getBinRadius(); + F32 old_radius = mDrawable->getBinRadius(); + if (new_radius < old_radius * 0.9f || new_radius > old_radius*1.1f) + { + gPipeline.markPartitionMove(mDrawable); + } + } lod_changed = lod_changed || LLViewerObject::updateLOD(); @@ -3187,6 +3208,10 @@ F32 LLVOVolume::getBinRadius() F32 scale = 1.f; + S32 size_factor = llmax(gSavedSettings.getS32("OctreeStaticObjectSizeFactor"), 1); + S32 attachment_size_factor = llmax(gSavedSettings.getS32("OctreeAttachmentSizeFactor"), 1); + LLVector3 distance_factor = gSavedSettings.getVector3("OctreeDistanceFactor"); + LLVector3 alpha_distance_factor = gSavedSettings.getVector3("OctreeAlphaDistanceFactor"); const LLVector4a* ext = mDrawable->getSpatialExtents(); BOOL shrink_wrap = mDrawable->isAnimating(); @@ -3216,6 +3241,8 @@ F32 LLVOVolume::getBinRadius() radius = llmin(bounds.mV[1], bounds.mV[2]); radius = llmin(radius, bounds.mV[0]); radius *= 0.5f; + radius *= 1.f+mDrawable->mDistanceWRTCamera*alpha_distance_factor[1]; + radius += mDrawable->mDistanceWRTCamera*alpha_distance_factor[0]; } else if (shrink_wrap) { @@ -3226,24 +3253,19 @@ F32 LLVOVolume::getBinRadius() } else if (mDrawable->isStatic()) { - /*if (mDrawable->getRadius() < 2.0f) - { - radius = 16.f; - } - else - { - radius = llmax(mDrawable->getRadius(), 32.f); - }*/ - - radius = (((S32) mDrawable->getRadius())/2+1)*8; + radius = llmax((S32) mDrawable->getRadius(), 1)*size_factor; + radius *= 1.f + mDrawable->mDistanceWRTCamera * distance_factor[1]; + radius += mDrawable->mDistanceWRTCamera * distance_factor[0]; } else if (mDrawable->getVObj()->isAttachment()) { - radius = (((S32) (mDrawable->getRadius()*4)+1))*2; + radius = llmax((S32) mDrawable->getRadius(),1)*attachment_size_factor; } else { - radius = 8.f; + radius = mDrawable->getRadius(); + radius *= 1.f + mDrawable->mDistanceWRTCamera * distance_factor[1]; + radius += mDrawable->mDistanceWRTCamera * distance_factor[0]; } return llclamp(radius*scale, 0.5f, 256.f); diff --git a/indra/newview/pipeline.cpp b/indra/newview/pipeline.cpp index 1214d1b545..fadae7c71e 100644 --- a/indra/newview/pipeline.cpp +++ b/indra/newview/pipeline.cpp @@ -392,6 +392,7 @@ void LLPipeline::init() { LLMemType mt(LLMemType::MTYPE_PIPELINE_INIT); + gOctreeMaxCapacity = gSavedSettings.getU32("OctreeMaxNodeCapacity"); sDynamicLOD = gSavedSettings.getBOOL("RenderDynamicLOD"); sRenderBump = gSavedSettings.getBOOL("RenderObjectBump"); sUseTriStrips = gSavedSettings.getBOOL("RenderUseTriStrips"); @@ -2519,6 +2520,32 @@ void LLPipeline::markGLRebuild(LLGLUpdate* glu) } } +void LLPipeline::markPartitionMove(LLDrawable* drawable) +{ + if (!drawable->isState(LLDrawable::PARTITION_MOVE) && + !drawable->getPositionGroup().equals3(LLVector4a::getZero())) + { + drawable->setState(LLDrawable::PARTITION_MOVE); + mPartitionQ.push_back(drawable); + } +} + +void LLPipeline::processPartitionQ() +{ + for (LLDrawable::drawable_list_t::iterator iter = mPartitionQ.begin(); iter != mPartitionQ.end(); ++iter) + { + LLDrawable* drawable = *iter; + if (!drawable->isDead()) + { + drawable->updateBinRadius(); + drawable->movePartition(); + } + drawable->clearState(LLDrawable::PARTITION_MOVE); + } + + mPartitionQ.clear(); +} + void LLPipeline::markRebuild(LLSpatialGroup* group, BOOL priority) { LLMemType mt(LLMemType::MTYPE_PIPELINE); @@ -5888,7 +5915,6 @@ LLSpatialPartition* LLPipeline::getSpatialPartition(LLViewerObject* vobj) return NULL; } - void LLPipeline::resetVertexBuffers(LLDrawable* drawable) { if (!drawable || drawable->isDead()) diff --git a/indra/newview/pipeline.h b/indra/newview/pipeline.h index ed4a803e7e..e967b8c31e 100644 --- a/indra/newview/pipeline.h +++ b/indra/newview/pipeline.h @@ -157,7 +157,8 @@ public: void markGLRebuild(LLGLUpdate* glu); void markRebuild(LLSpatialGroup* group, BOOL priority = FALSE); void markRebuild(LLDrawable *drawablep, LLDrawable::EDrawableFlags flag = LLDrawable::REBUILD_ALL, BOOL priority = FALSE); - + void markPartitionMove(LLDrawable* drawablep); + //get the object between start and end that's closest to start. LLViewerObject* lineSegmentIntersectInWorld(const LLVector3& start, const LLVector3& end, BOOL pick_transparent, @@ -211,6 +212,7 @@ public: void updateCull(LLCamera& camera, LLCullResult& result, S32 water_clip = 0, LLPlane* plane = NULL); //if water_clip is 0, ignore water plane, 1, cull to above plane, -1, cull to below plane void createObjects(F32 max_dtime); void createObject(LLViewerObject* vobj); + void processPartitionQ(); void updateGeom(F32 max_dtime); void updateGL(); void rebuildPriorityGroups(); @@ -636,6 +638,9 @@ protected: LLDrawable::drawable_list_t mBuildQ2; // non-priority LLSpatialGroup::sg_vector_t mGroupQ1; //priority LLSpatialGroup::sg_vector_t mGroupQ2; // non-priority + + LLDrawable::drawable_list_t mPartitionQ; //drawables that need to update their spatial partition radius + bool mGroupQ2Locked; bool mGroupQ1Locked; -- cgit v1.3 From 6992dbc1e32e1d8b803291aa1b87862fd6640c2a Mon Sep 17 00:00:00 2001 From: Dave Parks Date: Wed, 1 Jun 2011 23:46:04 -0500 Subject: SH-1682 Use GL_ARB_map_buffer_range to mitigate impact of mapping larger vertex buffer objects. Limit number of textures per batch to 6 (prevents frame stalls on NVIDIA). --- indra/llrender/llgl.cpp | 116 +++++++---- indra/llrender/llgl.h | 1 + indra/llrender/llglheaders.h | 12 ++ indra/llrender/llglslshader.cpp | 6 +- indra/llrender/llglslshader.h | 2 + indra/llrender/llimagegl.cpp | 8 +- indra/llrender/llrender.cpp | 10 +- indra/llrender/llvertexbuffer.cpp | 345 ++++++++++++++++++++++++-------- indra/llrender/llvertexbuffer.h | 54 +++-- indra/newview/app_settings/settings.xml | 17 +- indra/newview/lldrawpool.cpp | 2 +- indra/newview/lldrawpoolsimple.cpp | 27 +-- indra/newview/llface.cpp | 112 ++++++----- indra/newview/llspatialpartition.h | 4 +- indra/newview/llviewercontrol.cpp | 1 + indra/newview/llvovolume.cpp | 43 +++- 16 files changed, 505 insertions(+), 255 deletions(-) (limited to 'indra/newview/llviewercontrol.cpp') diff --git a/indra/llrender/llgl.cpp b/indra/llrender/llgl.cpp index a8d7df3bc8..a3aed4dd8a 100644 --- a/indra/llrender/llgl.cpp +++ b/indra/llrender/llgl.cpp @@ -127,6 +127,11 @@ PFNGLUNMAPBUFFERARBPROC glUnmapBufferARB = NULL; PFNGLGETBUFFERPARAMETERIVARBPROC glGetBufferParameterivARB = NULL; PFNGLGETBUFFERPOINTERVARBPROC glGetBufferPointervARB = NULL; +// GL_ARB_map_buffer_range +PFNGLMAPBUFFERRANGEPROC glMapBufferRange; +PFNGLFLUSHMAPPEDBUFFERRANGEPROC glFlushMappedBufferRange; + + // vertex object prototypes PFNGLNEWOBJECTBUFFERATIPROC glNewObjectBufferATI = NULL; PFNGLISOBJECTBUFFERATIPROC glIsObjectBufferATI = NULL; @@ -331,6 +336,7 @@ LLGLManager::LLGLManager() : mHasBlendFuncSeparate(FALSE), mHasVertexBufferObject(FALSE), + mHasMapBufferRange(FALSE), mHasPBuffer(FALSE), mHasShaderObjects(FALSE), mHasVertexShader(FALSE), @@ -761,6 +767,7 @@ void LLGLManager::initExtensions() mHasOcclusionQuery = ExtensionExists("GL_ARB_occlusion_query", gGLHExts.mSysExts); mHasOcclusionQuery2 = ExtensionExists("GL_ARB_occlusion_query2", gGLHExts.mSysExts); mHasVertexBufferObject = ExtensionExists("GL_ARB_vertex_buffer_object", gGLHExts.mSysExts); + mHasMapBufferRange = ExtensionExists("GL_ARB_map_buffer_range", gGLHExts.mSysExts); mHasDepthClamp = ExtensionExists("GL_ARB_depth_clamp", gGLHExts.mSysExts) || ExtensionExists("GL_NV_depth_clamp", gGLHExts.mSysExts); // mask out FBO support when packed_depth_stencil isn't there 'cause we need it for LLRenderTarget -Brad #ifdef GL_ARB_framebuffer_object @@ -955,6 +962,11 @@ void LLGLManager::initExtensions() mHasVertexBufferObject = FALSE; } } + if (mHasMapBufferRange) + { + glMapBufferRange = (PFNGLMAPBUFFERRANGEPROC) GLH_EXT_GET_PROC_ADDRESS("glMapBufferRange"); + glFlushMappedBufferRange = (PFNGLFLUSHMAPPEDBUFFERRANGEPROC) GLH_EXT_GET_PROC_ADDRESS("glFlushMappedBufferRange"); + } if (mHasFramebufferObject) { llinfos << "initExtensions() FramebufferObject-related procs..." << llendl; @@ -1411,10 +1423,6 @@ void LLGLState::checkTextureChannels(const std::string& msg) } } - GLint maxTextureUnits = 0; - glGetIntegerv(GL_MAX_TEXTURE_UNITS_ARB, &maxTextureUnits); - stop_glerror(); - static const char* label[] = { "GL_TEXTURE_2D", @@ -1449,69 +1457,91 @@ void LLGLState::checkTextureChannels(const std::string& msg) glh::matrix4f identity; identity.identity(); - for (GLint i = 1; i < maxTextureUnits; i++) + for (GLint i = 1; i < gGLManager.mNumTextureUnits; i++) { gGL.getTexUnit(i)->activate(); - glClientActiveTextureARB(GL_TEXTURE0_ARB+i); - stop_glerror(); - glGetIntegerv(GL_TEXTURE_STACK_DEPTH, &stackDepth); - stop_glerror(); - if (stackDepth != 1) + if (i < gGLManager.mNumTextureUnits) { - error = TRUE; - LL_WARNS("RenderState") << "Texture matrix stack corrupted." << LL_ENDL; + glClientActiveTextureARB(GL_TEXTURE0_ARB+i); + stop_glerror(); + glGetIntegerv(GL_TEXTURE_STACK_DEPTH, &stackDepth); + stop_glerror(); - if (gDebugSession) + if (stackDepth != 1) { - gFailLog << "Texture matrix stack corrupted." << std::endl; + error = TRUE; + LL_WARNS("RenderState") << "Texture matrix stack corrupted." << LL_ENDL; + + if (gDebugSession) + { + gFailLog << "Texture matrix stack corrupted." << std::endl; + } } - } - glGetFloatv(GL_TEXTURE_MATRIX, (GLfloat*) mat.m); - stop_glerror(); + glGetFloatv(GL_TEXTURE_MATRIX, (GLfloat*) mat.m); + stop_glerror(); - if (mat != identity) - { - error = TRUE; - LL_WARNS("RenderState") << "Texture matrix in channel " << i << " corrupt." << LL_ENDL; - if (gDebugSession) + if (mat != identity) { - gFailLog << "Texture matrix in channel " << i << " corrupt." << std::endl; + error = TRUE; + LL_WARNS("RenderState") << "Texture matrix in channel " << i << " corrupt." << LL_ENDL; + if (gDebugSession) + { + gFailLog << "Texture matrix in channel " << i << " corrupt." << std::endl; + } } - } - for (S32 j = (i == 0 ? 1 : 0); - j < 9; j++) - { - if (j == 8 && !gGLManager.mHasTextureRectangle || - j == 9 && !gGLManager.mHasTextureMultisample) + for (S32 j = (i == 0 ? 1 : 0); + j < 9; j++) { - continue; - } + if (j == 8 && !gGLManager.mHasTextureRectangle || + j == 9 && !gGLManager.mHasTextureMultisample) + { + continue; + } - if (glIsEnabled(value[j])) + if (glIsEnabled(value[j])) + { + error = TRUE; + LL_WARNS("RenderState") << "Texture channel " << i << " still has " << label[j] << " enabled." << LL_ENDL; + if (gDebugSession) + { + gFailLog << "Texture channel " << i << " still has " << label[j] << " enabled." << std::endl; + } + } + stop_glerror(); + } + + glGetFloatv(GL_TEXTURE_MATRIX, mat.m); + stop_glerror(); + + if (mat != identity) { error = TRUE; - LL_WARNS("RenderState") << "Texture channel " << i << " still has " << label[j] << " enabled." << LL_ENDL; + LL_WARNS("RenderState") << "Texture matrix " << i << " is not identity." << LL_ENDL; if (gDebugSession) { - gFailLog << "Texture channel " << i << " still has " << label[j] << " enabled." << std::endl; + gFailLog << "Texture matrix " << i << " is not identity." << std::endl; } } - stop_glerror(); } - glGetFloatv(GL_TEXTURE_MATRIX, mat.m); - stop_glerror(); - - if (mat != identity) { - error = TRUE; - LL_WARNS("RenderState") << "Texture matrix " << i << " is not identity." << LL_ENDL; - if (gDebugSession) + GLint tex = 0; + stop_glerror(); + glGetIntegerv(GL_TEXTURE_BINDING_2D, &tex); + stop_glerror(); + + if (tex != 0) { - gFailLog << "Texture matrix " << i << " is not identity." << std::endl; + error = TRUE; + LL_WARNS("RenderState") << "Texture channel " << i << " still has texture " << tex << " bound." << llendl; + + if (gDebugSession) + { + gFailLog << "Texture channel " << i << " still has texture " << tex << " bound." << std::endl; + } } } } diff --git a/indra/llrender/llgl.h b/indra/llrender/llgl.h index 420922cf06..d1bee00161 100644 --- a/indra/llrender/llgl.h +++ b/indra/llrender/llgl.h @@ -88,6 +88,7 @@ public: // ARB Extensions BOOL mHasVertexBufferObject; + BOOL mHasMapBufferRange; BOOL mHasPBuffer; BOOL mHasShaderObjects; BOOL mHasVertexShader; diff --git a/indra/llrender/llglheaders.h b/indra/llrender/llglheaders.h index 825d304d35..94f7a08c92 100644 --- a/indra/llrender/llglheaders.h +++ b/indra/llrender/llglheaders.h @@ -68,6 +68,10 @@ extern PFNGLUNMAPBUFFERARBPROC glUnmapBufferARB; extern PFNGLGETBUFFERPARAMETERIVARBPROC glGetBufferParameterivARB; extern PFNGLGETBUFFERPOINTERVARBPROC glGetBufferPointervARB; +// GL_ARB_map_buffer_range +extern PFNGLMAPBUFFERRANGEPROC glMapBufferRange; +extern PFNGLFLUSHMAPPEDBUFFERRANGEPROC glFlushMappedBufferRange; + // GL_ATI_vertex_array_object extern PFNGLNEWOBJECTBUFFERATIPROC glNewObjectBufferATI; extern PFNGLISOBJECTBUFFERATIPROC glIsObjectBufferATI; @@ -306,6 +310,10 @@ extern PFNGLUNMAPBUFFERARBPROC glUnmapBufferARB; extern PFNGLGETBUFFERPARAMETERIVARBPROC glGetBufferParameterivARB; extern PFNGLGETBUFFERPOINTERVARBPROC glGetBufferPointervARB; +// GL_ARB_map_buffer_range +extern PFNGLMAPBUFFERRANGEPROC glMapBufferRange; +extern PFNGLFLUSHMAPPEDBUFFERRANGEPROC glFlushMappedBufferRange; + // GL_ATI_vertex_array_object extern PFNGLNEWOBJECTBUFFERATIPROC glNewObjectBufferATI; extern PFNGLISOBJECTBUFFERATIPROC glIsObjectBufferATI; @@ -511,6 +519,10 @@ extern PFNGLUNMAPBUFFERARBPROC glUnmapBufferARB; extern PFNGLGETBUFFERPARAMETERIVARBPROC glGetBufferParameterivARB; extern PFNGLGETBUFFERPOINTERVARBPROC glGetBufferPointervARB; +// GL_ARB_map_buffer_range +extern PFNGLMAPBUFFERRANGEPROC glMapBufferRange; +extern PFNGLFLUSHMAPPEDBUFFERRANGEPROC glFlushMappedBufferRange; + // GL_ATI_vertex_array_object extern PFNGLNEWOBJECTBUFFERATIPROC glNewObjectBufferATI; extern PFNGLISOBJECTBUFFERATIPROC glIsObjectBufferATI; diff --git a/indra/llrender/llglslshader.cpp b/indra/llrender/llglslshader.cpp index 2dab757828..5473f23d86 100644 --- a/indra/llrender/llglslshader.cpp +++ b/indra/llrender/llglslshader.cpp @@ -48,6 +48,8 @@ using std::pair; using std::make_pair; using std::string; +U32 LLGLSLShader::sCurBoundShader = 0; + BOOL shouldChange(const LLVector4& v1, const LLVector4& v2) { return v1 != v2; @@ -367,7 +369,7 @@ void LLGLSLShader::bind() if (gGLManager.mHasShaderObjects) { glUseProgramObjectARB(mProgramObject); - + sCurBoundShader = mProgramObject; if (mUniformsDirty) { LLShaderMgr::instance()->updateShaderUniforms(this); @@ -390,6 +392,7 @@ void LLGLSLShader::unbind() } } glUseProgramObjectARB(0); + sCurBoundShader = 0; stop_glerror(); } } @@ -397,6 +400,7 @@ void LLGLSLShader::unbind() void LLGLSLShader::bindNoShader(void) { glUseProgramObjectARB(0); + sCurBoundShader = 0; } S32 LLGLSLShader::enableTexture(S32 uniform, LLTexUnit::eTextureType mode) diff --git a/indra/llrender/llglslshader.h b/indra/llrender/llglslshader.h index 51be56b295..4922eb6d67 100644 --- a/indra/llrender/llglslshader.h +++ b/indra/llrender/llglslshader.h @@ -66,6 +66,8 @@ public: LLGLSLShader(); + static GLhandleARB sCurBoundShader; + void unload(); BOOL createShader(std::vector * attributes, std::vector * uniforms); diff --git a/indra/llrender/llimagegl.cpp b/indra/llrender/llimagegl.cpp index 3a4139bace..60a5962234 100644 --- a/indra/llrender/llimagegl.cpp +++ b/indra/llrender/llimagegl.cpp @@ -1418,11 +1418,13 @@ void LLImageGL::deleteDeadTextures() { GLuint tex = sDeadTextureList.front(); sDeadTextureList.pop_front(); - for (int i = 0; i < gGLManager.mNumTextureUnits; i++) + for (int i = 0; i < gGLManager.mNumTextureImageUnits; i++) { - if (sCurrentBoundTextures[i] == tex) + LLTexUnit* tex_unit = gGL.getTexUnit(i); + + if (tex_unit->getCurrTexture() == tex) { - gGL.getTexUnit(i)->unbind(gGL.getTexUnit(i)->getCurrType()); + tex_unit->unbind(tex_unit->getCurrType()); stop_glerror(); } } diff --git a/indra/llrender/llrender.cpp b/indra/llrender/llrender.cpp index fdfcfe7fab..e91ceb873e 100644 --- a/indra/llrender/llrender.cpp +++ b/indra/llrender/llrender.cpp @@ -30,6 +30,7 @@ #include "llvertexbuffer.h" #include "llcubemap.h" +#include "llglslshader.h" #include "llimagegl.h" #include "llrendertarget.h" #include "lltexture.h" @@ -183,7 +184,8 @@ void LLTexUnit::enable(eTextureType type) mCurrTexType = type; gGL.flush(); - if (type != LLTexUnit::TT_MULTISAMPLE_TEXTURE && + if (LLGLSLShader::sCurBoundShader == 0 && + type != LLTexUnit::TT_MULTISAMPLE_TEXTURE && mIndex < gGLManager.mNumTextureUnits) { glEnable(sGLTextureType[type]); @@ -200,12 +202,10 @@ void LLTexUnit::disable(void) activate(); unbind(mCurrTexType); gGL.flush(); - if (mCurrTexType != LLTexUnit::TT_MULTISAMPLE_TEXTURE && + if (LLGLSLShader::sCurBoundShader == 0 && mCurrTexType != LLTexUnit::TT_MULTISAMPLE_TEXTURE && mIndex < gGLManager.mNumTextureUnits) { - stop_glerror(); glDisable(sGLTextureType[mCurrTexType]); - stop_glerror(); } mCurrTexType = TT_NONE; @@ -295,7 +295,7 @@ bool LLTexUnit::bind(LLImageGL* texture, bool for_rendering, bool forceBind) glBindTexture(sGLTextureType[texture->getTarget()], mCurrTexture); texture->updateBindStats(texture->mTextureMemory); mHasMipMaps = texture->mHasMipMaps; - if (texture->mTexOptionsDirty) + if (mIndex == 0 && texture->mTexOptionsDirty) { texture->mTexOptionsDirty = false; setTextureAddressMode(texture->mAddressMode); diff --git a/indra/llrender/llvertexbuffer.cpp b/indra/llrender/llvertexbuffer.cpp index f715a8e9ba..27cc88462a 100644 --- a/indra/llrender/llvertexbuffer.cpp +++ b/indra/llrender/llvertexbuffer.cpp @@ -934,8 +934,26 @@ void LLVertexBuffer::allocateClientIndexBuffer() } } +bool expand_region(LLVertexBuffer::MappedRegion& region, S32 index, S32 count) +{ + S32 end = index+count; + S32 region_end = region.mIndex+region.mCount; + + if (end < region.mIndex || + index > region_end) + { //gap exists, do not merge + return false; + } + + S32 new_end = llmax(end, region_end); + S32 new_index = llmin(index, region.mIndex); + region.mIndex = new_index; + region.mCount = new_end-new_index; + return true; +} + // Map for data access -U8* LLVertexBuffer::mapVertexBuffer(S32 type, S32 access) +U8* LLVertexBuffer::mapVertexBuffer(S32 type, S32 index, S32 count, bool map_range) { LLMemType mt2(LLMemType::MTYPE_VERTEX_MAP_BUFFER); if (mFinal) @@ -947,8 +965,44 @@ U8* LLVertexBuffer::mapVertexBuffer(S32 type, S32 access) llerrs << "LLVertexBuffer::mapVertexBuffer() called on unallocated buffer." << llendl; } - if (!mVertexLocked && useVBOs()) + if (useVBOs()) { + + if (!sDisableVBOMapping && gGLManager.mHasMapBufferRange) + { + if (count == -1) + { + count = mNumVerts; + } + + bool mapped = false; + //see if range is already mapped + for (U32 i = 0; i < mMappedVertexRegions.size(); ++i) + { + MappedRegion& region = mMappedVertexRegions[i]; + if (region.mType == type) + { + if (expand_region(region, index, count)) + { + mapped = true; + } + } + } + + if (!mapped) + { + //not already mapped, map new region + MappedRegion region(type, map_range ? -1 : index, count); + mMappedVertexRegions.push_back(region); + } + } + + if (mVertexLocked && map_range) + { + llerrs << "Attempted to map a specific range of a buffer that was already mapped." << llendl; + } + + if (!mVertexLocked) { LLMemType mt_v(LLMemType::MTYPE_VERTEX_MAP_BUFFER_VERTICES); setBuffer(0, type); @@ -957,61 +1011,91 @@ U8* LLVertexBuffer::mapVertexBuffer(S32 type, S32 access) if(sDisableVBOMapping) { + map_range = false; allocateClientVertexBuffer() ; } else { - U8* src = (U8*) glMapBufferARB(GL_ARRAY_BUFFER_ARB, GL_WRITE_ONLY_ARB); + U8* src = NULL; + if (gGLManager.mHasMapBufferRange) + { + if (map_range) + { + S32 offset = mOffsets[type] + sTypeSize[type]*index; + S32 length = (sTypeSize[type]*count+0xF) & ~0xF; + src = (U8*) glMapBufferRange(GL_ARRAY_BUFFER_ARB, offset, length, GL_MAP_WRITE_BIT | GL_MAP_FLUSH_EXPLICIT_BIT | GL_MAP_INVALIDATE_RANGE_BIT); + } + else + { + src = (U8*) glMapBufferRange(GL_ARRAY_BUFFER_ARB, 0, mSize, GL_MAP_WRITE_BIT | GL_MAP_FLUSH_EXPLICIT_BIT); + } + } + else + { + map_range = false; + src = (U8*) glMapBufferARB(GL_ARRAY_BUFFER_ARB, GL_WRITE_ONLY_ARB); + } + mMappedData = LL_NEXT_ALIGNED_ADDRESS(src); mAlignedOffset = mMappedData - src; stop_glerror(); } - } - - - if (!mMappedData) - { - log_glerror(); - - //check the availability of memory - U32 avail_phy_mem, avail_vir_mem; - LLMemoryInfo::getAvailableMemoryKB(avail_phy_mem, avail_vir_mem) ; - llinfos << "Available physical mwmory(KB): " << avail_phy_mem << llendl ; - llinfos << "Available virtual memory(KB): " << avail_vir_mem << llendl; - - if(!sDisableVBOMapping) - { - //-------------------- - //print out more debug info before crash - llinfos << "vertex buffer size: (num verts : num indices) = " << getNumVerts() << " : " << getNumIndices() << llendl ; - GLint size ; - glGetBufferParameterivARB(GL_ARRAY_BUFFER_ARB, GL_BUFFER_SIZE_ARB, &size) ; - llinfos << "GL_ARRAY_BUFFER_ARB size is " << size << llendl ; - //-------------------- + + if (!mMappedData) + { + log_glerror(); + + //check the availability of memory + U32 avail_phy_mem, avail_vir_mem; + LLMemoryInfo::getAvailableMemoryKB(avail_phy_mem, avail_vir_mem) ; + llinfos << "Available physical mwmory(KB): " << avail_phy_mem << llendl ; + llinfos << "Available virtual memory(KB): " << avail_vir_mem << llendl; + + if(!sDisableVBOMapping) + { + //-------------------- + //print out more debug info before crash + llinfos << "vertex buffer size: (num verts : num indices) = " << getNumVerts() << " : " << getNumIndices() << llendl ; + GLint size ; + glGetBufferParameterivARB(GL_ARRAY_BUFFER_ARB, GL_BUFFER_SIZE_ARB, &size) ; + llinfos << "GL_ARRAY_BUFFER_ARB size is " << size << llendl ; + //-------------------- + + GLint buff; + glGetIntegerv(GL_ARRAY_BUFFER_BINDING_ARB, &buff); + if ((GLuint)buff != mGLBuffer) + { + llerrs << "Invalid GL vertex buffer bound: " << buff << llendl; + } - GLint buff; - glGetIntegerv(GL_ARRAY_BUFFER_BINDING_ARB, &buff); - if ((GLuint)buff != mGLBuffer) + + llerrs << "glMapBuffer returned NULL (no vertex data)" << llendl; + } + else { - llerrs << "Invalid GL vertex buffer bound: " << buff << llendl; + llerrs << "memory allocation for vertex data failed." << llendl ; } - - - llerrs << "glMapBuffer returned NULL (no vertex data)" << llendl; - } - else - { - llerrs << "memory allocation for vertex data failed." << llendl ; } + sMappedCount++; } - sMappedCount++; + } + else + { + map_range = false; } - return mMappedData; + if (map_range) + { + return mMappedData; + } + else + { + return mMappedData+mOffsets[type]+sTypeSize[type]*index; + } } -U8* LLVertexBuffer::mapIndexBuffer(S32 access) +U8* LLVertexBuffer::mapIndexBuffer(S32 index, S32 count, bool map_range) { LLMemType mt2(LLMemType::MTYPE_VERTEX_MAP_BUFFER); if (mFinal) @@ -1023,8 +1107,40 @@ U8* LLVertexBuffer::mapIndexBuffer(S32 access) llerrs << "LLVertexBuffer::mapIndexBuffer() called on unallocated buffer." << llendl; } - if (!mIndexLocked && useVBOs()) + if (useVBOs()) { + if (!sDisableVBOMapping && gGLManager.mHasMapBufferRange) + { + if (count == -1) + { + count = mNumIndices; + } + + bool mapped = false; + //see if range is already mapped + for (U32 i = 0; i < mMappedIndexRegions.size(); ++i) + { + MappedRegion& region = mMappedIndexRegions[i]; + if (expand_region(region, index, count)) + { + mapped = true; + } + } + + if (!mapped) + { + //not already mapped, map new region + MappedRegion region(TYPE_INDEX, map_range ? -1 : index, count); + mMappedIndexRegions.push_back(region); + } + } + + if (mIndexLocked && map_range) + { + llerrs << "Attempted to map a specific range of a buffer that was already mapped." << llendl; + } + + if (!mIndexLocked) { LLMemType mt_v(LLMemType::MTYPE_VERTEX_MAP_BUFFER_INDICES); @@ -1034,12 +1150,32 @@ U8* LLVertexBuffer::mapIndexBuffer(S32 access) if(sDisableVBOMapping) { + map_range = false; allocateClientIndexBuffer() ; } else { - U8* src = (U8*) glMapBufferARB(GL_ELEMENT_ARRAY_BUFFER_ARB, GL_WRITE_ONLY_ARB); - mMappedIndexData = LL_NEXT_ALIGNED_ADDRESS(src); + U8* src = NULL; + if (gGLManager.mHasMapBufferRange) + { + if (map_range) + { + S32 offset = sizeof(U16)*index; + S32 length = sizeof(U16)*count; + src = (U8*) glMapBufferRange(GL_ELEMENT_ARRAY_BUFFER_ARB, offset, length, GL_MAP_WRITE_BIT | GL_MAP_FLUSH_EXPLICIT_BIT | GL_MAP_INVALIDATE_RANGE_BIT); + } + else + { + src = (U8*) glMapBufferRange(GL_ELEMENT_ARRAY_BUFFER_ARB, 0, sizeof(U16)*mNumIndices, GL_MAP_WRITE_BIT | GL_MAP_FLUSH_EXPLICIT_BIT); + } + } + else + { + map_range = false; + src = (U8*) glMapBufferARB(GL_ELEMENT_ARRAY_BUFFER_ARB, GL_WRITE_ONLY_ARB); + } + + mMappedIndexData = src; //LL_NEXT_ALIGNED_ADDRESS(src); mAlignedIndexOffset = mMappedIndexData - src; stop_glerror(); } @@ -1068,19 +1204,31 @@ U8* LLVertexBuffer::mapIndexBuffer(S32 access) sMappedCount++; } + else + { + map_range = false; + } - return mMappedIndexData ; + if (map_range) + { + return mMappedIndexData; + } + else + { + return mMappedIndexData + sizeof(U16)*index; + } } void LLVertexBuffer::unmapBuffer(S32 type) { LLMemType mt2(LLMemType::MTYPE_VERTEX_UNMAP_BUFFER); - if (!useVBOs()) + if (!useVBOs() || type == -2) { return ; //nothing to unmap } bool updated_all = false ; + if (mMappedData && mVertexLocked && type != TYPE_INDEX) { updated_all = (mIndexLocked && type < 0) ; //both vertex and index buffers done updating @@ -1093,6 +1241,23 @@ void LLVertexBuffer::unmapBuffer(S32 type) } else { + if (gGLManager.mHasMapBufferRange) + { + if (!mMappedVertexRegions.empty()) + { + stop_glerror(); + for (U32 i = 0; i < mMappedVertexRegions.size(); ++i) + { + const MappedRegion& region = mMappedVertexRegions[i]; + S32 offset = region.mIndex >= 0 ? mOffsets[region.mType]+sTypeSize[region.mType]*region.mIndex : 0; + S32 length = sTypeSize[region.mType]*region.mCount; + glFlushMappedBufferRange(GL_ARRAY_BUFFER_ARB, offset, length); + stop_glerror(); + } + + mMappedVertexRegions.clear(); + } + } stop_glerror(); glUnmapBufferARB(GL_ARRAY_BUFFER_ARB); stop_glerror(); @@ -1103,8 +1268,8 @@ void LLVertexBuffer::unmapBuffer(S32 type) mVertexLocked = FALSE ; sMappedCount--; } - - if(mMappedIndexData && mIndexLocked && (type < 0 || type == TYPE_INDEX)) + + if (mMappedIndexData && mIndexLocked && (type < 0 || type == TYPE_INDEX)) { if(sDisableVBOMapping) { @@ -1114,6 +1279,23 @@ void LLVertexBuffer::unmapBuffer(S32 type) } else { + + if (gGLManager.mHasMapBufferRange) + { + if (!mMappedIndexRegions.empty()) + { + for (U32 i = 0; i < mMappedIndexRegions.size(); ++i) + { + const MappedRegion& region = mMappedIndexRegions[i]; + S32 offset = region.mIndex >= 0 ? sizeof(U16)*region.mIndex : 0; + S32 length = sizeof(U16)*region.mCount; + glFlushMappedBufferRange(GL_ELEMENT_ARRAY_BUFFER_ARB, offset, length); + stop_glerror(); + } + + mMappedIndexRegions.clear(); + } + } stop_glerror(); glUnmapBufferARB(GL_ELEMENT_ARRAY_BUFFER_ARB); stop_glerror(); @@ -1152,19 +1334,21 @@ template struct VertexBufferStrider typedef LLStrider strider_t; static bool get(LLVertexBuffer& vbo, strider_t& strider, - S32 index) + S32 index, S32 count, bool map_range) { if (type == LLVertexBuffer::TYPE_INDEX) { S32 stride = sizeof(T); - if (vbo.mapIndexBuffer() == NULL) + U8* ptr = vbo.mapIndexBuffer(index, count, map_range); + + if (ptr == NULL) { llwarns << "mapIndexBuffer failed!" << llendl; return FALSE; } - strider = (T*)(vbo.getMappedIndices() + index*stride); + strider = (T*)ptr; strider.setStride(0); return TRUE; } @@ -1172,13 +1356,15 @@ template struct VertexBufferStrider { S32 stride = LLVertexBuffer::sTypeSize[type]; - if (vbo.mapVertexBuffer(type) == NULL) + U8* ptr = vbo.mapVertexBuffer(type, index, count, map_range); + + if (ptr == NULL) { llwarns << "mapVertexBuffer failed!" << llendl; return FALSE; } - strider = (T*)(vbo.getMappedData() + vbo.getOffset(type)+index*stride); + strider = (T*)ptr; strider.setStride(stride); return TRUE; } @@ -1190,55 +1376,48 @@ template struct VertexBufferStrider } }; -bool LLVertexBuffer::getVertexStrider(LLStrider& strider, S32 index) +bool LLVertexBuffer::getVertexStrider(LLStrider& strider, S32 index, S32 count, bool map_range) { - return VertexBufferStrider::get(*this, strider, index); + return VertexBufferStrider::get(*this, strider, index, count, map_range); } -bool LLVertexBuffer::getIndexStrider(LLStrider& strider, S32 index) +bool LLVertexBuffer::getIndexStrider(LLStrider& strider, S32 index, S32 count, bool map_range) { - return VertexBufferStrider::get(*this, strider, index); + return VertexBufferStrider::get(*this, strider, index, count, map_range); } -bool LLVertexBuffer::getTexCoord0Strider(LLStrider& strider, S32 index) +bool LLVertexBuffer::getTexCoord0Strider(LLStrider& strider, S32 index, S32 count, bool map_range) { - return VertexBufferStrider::get(*this, strider, index); + return VertexBufferStrider::get(*this, strider, index, count, map_range); } -bool LLVertexBuffer::getTexCoord1Strider(LLStrider& strider, S32 index) +bool LLVertexBuffer::getTexCoord1Strider(LLStrider& strider, S32 index, S32 count, bool map_range) { - return VertexBufferStrider::get(*this, strider, index); + return VertexBufferStrider::get(*this, strider, index, count, map_range); } -/*bool LLVertexBuffer::getTexCoord2Strider(LLStrider& strider, S32 index) -{ - return VertexBufferStrider::get(*this, strider, index); -} -bool LLVertexBuffer::getTexCoord3Strider(LLStrider& strider, S32 index) -{ - return VertexBufferStrider::get(*this, strider, index); -}*/ -bool LLVertexBuffer::getNormalStrider(LLStrider& strider, S32 index) + +bool LLVertexBuffer::getNormalStrider(LLStrider& strider, S32 index, S32 count, bool map_range) { - return VertexBufferStrider::get(*this, strider, index); + return VertexBufferStrider::get(*this, strider, index, count, map_range); } -bool LLVertexBuffer::getBinormalStrider(LLStrider& strider, S32 index) +bool LLVertexBuffer::getBinormalStrider(LLStrider& strider, S32 index, S32 count, bool map_range) { - return VertexBufferStrider::get(*this, strider, index); + return VertexBufferStrider::get(*this, strider, index, count, map_range); } -bool LLVertexBuffer::getColorStrider(LLStrider& strider, S32 index) +bool LLVertexBuffer::getColorStrider(LLStrider& strider, S32 index, S32 count, bool map_range) { - return VertexBufferStrider::get(*this, strider, index); + return VertexBufferStrider::get(*this, strider, index, count, map_range); } -bool LLVertexBuffer::getWeightStrider(LLStrider& strider, S32 index) +bool LLVertexBuffer::getWeightStrider(LLStrider& strider, S32 index, S32 count, bool map_range) { - return VertexBufferStrider::get(*this, strider, index); + return VertexBufferStrider::get(*this, strider, index, count, map_range); } -bool LLVertexBuffer::getWeight4Strider(LLStrider& strider, S32 index) +bool LLVertexBuffer::getWeight4Strider(LLStrider& strider, S32 index, S32 count, bool map_range) { - return VertexBufferStrider::get(*this, strider, index); + return VertexBufferStrider::get(*this, strider, index, count, map_range); } -bool LLVertexBuffer::getClothWeightStrider(LLStrider& strider, S32 index) +bool LLVertexBuffer::getClothWeightStrider(LLStrider& strider, S32 index, S32 count, bool map_range) { - return VertexBufferStrider::get(*this, strider, index); + return VertexBufferStrider::get(*this, strider, index, count, map_range); } //---------------------------------------------------------------------------- @@ -1510,11 +1689,3 @@ void LLVertexBuffer::setupVertexBuffer(U32 data_mask) const llglassertok(); } -void LLVertexBuffer::markDirty(U32 vert_index, U32 vert_count, U32 indices_index, U32 indices_count) -{ - // TODO: use GL_APPLE_flush_buffer_range here - /*if (useVBOs() && !mFilthy) - { - - }*/ -} diff --git a/indra/llrender/llvertexbuffer.h b/indra/llrender/llvertexbuffer.h index 0c4b241537..aa5df305a6 100644 --- a/indra/llrender/llvertexbuffer.h +++ b/indra/llrender/llvertexbuffer.h @@ -77,6 +77,18 @@ protected: class LLVertexBuffer : public LLRefCount { public: + class MappedRegion + { + public: + S32 mType; + S32 mIndex; + S32 mCount; + + MappedRegion(S32 type, S32 index, S32 count) + : mType(type), mIndex(index), mCount(count) + { } + }; + LLVertexBuffer(const LLVertexBuffer& rhs) { *this = rhs; @@ -177,8 +189,8 @@ public: LLVertexBuffer(U32 typemask, S32 usage); // map for data access - U8* mapVertexBuffer(S32 type = -1, S32 access = -1); - U8* mapIndexBuffer(S32 access = -1); + U8* mapVertexBuffer(S32 type, S32 index, S32 count, bool map_range); + U8* mapIndexBuffer(S32 index, S32 count, bool map_range); // set for rendering virtual void setBuffer(U32 data_mask, S32 type = -1); // calls setupVertexBuffer() if data_mask is not 0 @@ -193,16 +205,16 @@ public: // vb->getNormalStrider(norms); // setVertsNorms(verts, norms); // vb->unmapBuffer(); - bool getVertexStrider(LLStrider& strider, S32 index=0); - bool getIndexStrider(LLStrider& strider, S32 index=0); - bool getTexCoord0Strider(LLStrider& strider, S32 index=0); - bool getTexCoord1Strider(LLStrider& strider, S32 index=0); - bool getNormalStrider(LLStrider& strider, S32 index=0); - bool getBinormalStrider(LLStrider& strider, S32 index=0); - bool getColorStrider(LLStrider& strider, S32 index=0); - bool getWeightStrider(LLStrider& strider, S32 index=0); - bool getWeight4Strider(LLStrider& strider, S32 index=0); - bool getClothWeightStrider(LLStrider& strider, S32 index=0); + bool getVertexStrider(LLStrider& strider, S32 index=0, S32 count = -1, bool map_range = false); + bool getIndexStrider(LLStrider& strider, S32 index=0, S32 count = -1, bool map_range = false); + bool getTexCoord0Strider(LLStrider& strider, S32 index=0, S32 count = -1, bool map_range = false); + bool getTexCoord1Strider(LLStrider& strider, S32 index=0, S32 count = -1, bool map_range = false); + bool getNormalStrider(LLStrider& strider, S32 index=0, S32 count = -1, bool map_range = false); + bool getBinormalStrider(LLStrider& strider, S32 index=0, S32 count = -1, bool map_range = false); + bool getColorStrider(LLStrider& strider, S32 index=0, S32 count = -1, bool map_range = false); + bool getWeightStrider(LLStrider& strider, S32 index=0, S32 count = -1, bool map_range = false); + bool getWeight4Strider(LLStrider& strider, S32 index=0, S32 count = -1, bool map_range = false); + bool getClothWeightStrider(LLStrider& strider, S32 index=0, S32 count = -1, bool map_range = false); BOOL isEmpty() const { return mEmpty; } BOOL isLocked() const { return mVertexLocked || mIndexLocked; } @@ -222,8 +234,6 @@ public: S32 getOffset(S32 type) const { return mOffsets[type]; } S32 getUsage() const { return mUsage; } - void markDirty(U32 vert_index, U32 vert_count, U32 indices_index, U32 indices_count); - void draw(U32 mode, U32 count, U32 indices_offset) const; void drawArrays(U32 mode, U32 offset, U32 count) const; void drawRange(U32 mode, U32 start, U32 end, U32 count, U32 indices_offset) const; @@ -257,20 +267,8 @@ protected: BOOL mDynamicSize; // if TRUE, buffer has been resized at least once (and should be padded) S32 mOffsets[TYPE_MAX]; - class DirtyRegion - { - public: - U32 mIndex; - U32 mCount; - U32 mIndicesIndex; - U32 mIndicesCount; - - DirtyRegion(U32 vi, U32 vc, U32 ii, U32 ic) - : mIndex(vi), mCount(vc), mIndicesIndex(ii), mIndicesCount(ic) - { } - }; - - std::vector mDirtyRegions; //vector of dirty regions to rebuild + std::vector mMappedVertexRegions; + std::vector mMappedIndexRegions; public: static S32 sCount; diff --git a/indra/newview/app_settings/settings.xml b/indra/newview/app_settings/settings.xml index bed7e4705e..b7ba9e3607 100644 --- a/indra/newview/app_settings/settings.xml +++ b/indra/newview/app_settings/settings.xml @@ -7126,7 +7126,7 @@ Vector3 Value - 0.05 + 0.1 0.0 0.0 @@ -7154,7 +7154,7 @@ Vector3 Value - 0.25 + 0.01 0.0 0.0 @@ -7513,6 +7513,17 @@ Value 0 + RenderMaxTextureIndex + + Comment + Maximum texture index to use for indexed texture rendering. + Persist + 1 + Type + U32 + Value + 6 + RenderDebugTextureBind Comment @@ -9011,7 +9022,7 @@ Type Boolean Value - 1 + 0 RenderUseStreamVBO diff --git a/indra/newview/lldrawpool.cpp b/indra/newview/lldrawpool.cpp index d92d9e92a1..f5483d969d 100644 --- a/indra/newview/lldrawpool.cpp +++ b/indra/newview/lldrawpool.cpp @@ -191,7 +191,7 @@ void LLDrawPool::renderPostDeferred(S32 pass) //virtual void LLDrawPool::endRenderPass( S32 pass ) { - for (U32 i = 1; i < gGLManager.mNumTextureImageUnits; i++) + for (U32 i = 0; i < gGLManager.mNumTextureImageUnits; i++) { //dummy cleanup of any currently bound textures if (gGL.getTexUnit(i)->getCurrType() != LLTexUnit::TT_NONE) { diff --git a/indra/newview/lldrawpoolsimple.cpp b/indra/newview/lldrawpoolsimple.cpp index 6963fea26b..5dbb27cabb 100644 --- a/indra/newview/lldrawpoolsimple.cpp +++ b/indra/newview/lldrawpoolsimple.cpp @@ -71,14 +71,7 @@ void LLDrawPoolGlow::renderPostDeferred(S32 pass) void LLDrawPoolGlow::endPostDeferredPass(S32 pass) { gDeferredFullbrightProgram.unbind(); - for (U32 i = 0; i < 8; i++) - { - if (gGL.getTexUnit(i)->getCurrType() != LLTexUnit::TT_NONE) - { - gGL.getTexUnit(i)->unbind(gGL.getTexUnit(i)->getCurrType()); - gGL.getTexUnit(i)->disable(); - } - } + LLRenderPass::endRenderPass(pass); } void LLDrawPoolGlow::render(S32 pass) @@ -225,15 +218,6 @@ void LLDrawPoolSimple::endDeferredPass(S32 pass) LLRenderPass::endRenderPass(pass); gDeferredDiffuseProgram.unbind(); - - for (U32 i = 0; i < 8; i++) - { - if (gGL.getTexUnit(i)->getCurrType() != LLTexUnit::TT_NONE) - { - gGL.getTexUnit(i)->unbind(gGL.getTexUnit(i)->getCurrType()); - gGL.getTexUnit(i)->disable(); - } - } } void LLDrawPoolSimple::renderDeferred(S32 pass) @@ -368,14 +352,7 @@ void LLDrawPoolFullbright::renderPostDeferred(S32 pass) void LLDrawPoolFullbright::endPostDeferredPass(S32 pass) { gDeferredFullbrightProgram.unbind(); - for (U32 i = 0; i < 8; i++) - { - if (gGL.getTexUnit(i)->getCurrType() != LLTexUnit::TT_NONE) - { - gGL.getTexUnit(i)->unbind(gGL.getTexUnit(i)->getCurrType()); - gGL.getTexUnit(i)->disable(); - } - } + LLRenderPass::endRenderPass(pass); } void LLDrawPoolFullbright::beginRenderPass(S32 pass) diff --git a/indra/newview/llface.cpp b/indra/newview/llface.cpp index e30522d380..540ed054e9 100644 --- a/indra/newview/llface.cpp +++ b/indra/newview/llface.cpp @@ -429,11 +429,11 @@ U16 LLFace::getGeometryAvatar( if (mVertexBuffer.notNull()) { - mVertexBuffer->getVertexStrider (vertices, mGeomIndex); - mVertexBuffer->getNormalStrider (normals, mGeomIndex); - mVertexBuffer->getTexCoord0Strider (tex_coords, mGeomIndex); - mVertexBuffer->getWeightStrider(vertex_weights, mGeomIndex); - mVertexBuffer->getClothWeightStrider(clothing_weights, mGeomIndex); + mVertexBuffer->getVertexStrider (vertices, mGeomIndex, mGeomCount); + mVertexBuffer->getNormalStrider (normals, mGeomIndex, mGeomCount); + mVertexBuffer->getTexCoord0Strider (tex_coords, mGeomIndex, mGeomCount); + mVertexBuffer->getWeightStrider(vertex_weights, mGeomIndex, mGeomCount); + mVertexBuffer->getClothWeightStrider(clothing_weights, mGeomIndex, mGeomCount); } return mGeomIndex; @@ -446,17 +446,17 @@ U16 LLFace::getGeometry(LLStrider &vertices, LLStrider &no if (mVertexBuffer.notNull()) { - mVertexBuffer->getVertexStrider(vertices, mGeomIndex); + mVertexBuffer->getVertexStrider(vertices, mGeomIndex, mGeomCount); if (mVertexBuffer->hasDataType(LLVertexBuffer::TYPE_NORMAL)) { - mVertexBuffer->getNormalStrider(normals, mGeomIndex); + mVertexBuffer->getNormalStrider(normals, mGeomIndex, mGeomCount); } if (mVertexBuffer->hasDataType(LLVertexBuffer::TYPE_TEXCOORD0)) { - mVertexBuffer->getTexCoord0Strider(tex_coords, mGeomIndex); + mVertexBuffer->getTexCoord0Strider(tex_coords, mGeomIndex, mGeomCount); } - mVertexBuffer->getIndexStrider(indicesp, mIndicesIndex); + mVertexBuffer->getIndexStrider(indicesp, mIndicesIndex, mIndicesCount); } return mGeomIndex; @@ -1092,27 +1092,6 @@ BOOL LLFace::getGeometryVolume(const LLVolume& volume, const LLTextureEntry *tep = mVObjp->getTE(f); const U8 bump_code = tep ? tep->getBumpmap() : 0; - if (rebuild_pos) - { - mVertexBuffer->getVertexStrider(vert, mGeomIndex); - vertices = (LLVector4a*) vert.get(); - } - if (rebuild_normal) - { - mVertexBuffer->getNormalStrider(norm, mGeomIndex); - normals = (LLVector4a*) norm.get(); - } - if (rebuild_binormal) - { - mVertexBuffer->getBinormalStrider(binorm, mGeomIndex); - binormals = (LLVector4a*) binorm.get(); - } - if (rebuild_weights) - { - mVertexBuffer->getWeight4Strider(wght, mGeomIndex); - weights = (LLVector4a*) wght.get(); - } - F32 tcoord_xoffset = 0.f ; F32 tcoord_yoffset = 0.f ; F32 tcoord_xscale = 1.f ; @@ -1121,12 +1100,6 @@ BOOL LLFace::getGeometryVolume(const LLVolume& volume, if (rebuild_tcoord) { - mVertexBuffer->getTexCoord0Strider(tex_coords, mGeomIndex); - if (bump_code && mVertexBuffer->hasDataType(LLVertexBuffer::TYPE_TEXCOORD1)) - { - mVertexBuffer->getTexCoord1Strider(tex_coords2, mGeomIndex); - } - in_atlas = isAtlasInUse() ; if(in_atlas) { @@ -1139,11 +1112,7 @@ BOOL LLFace::getGeometryVolume(const LLVolume& volume, tcoord_yscale = tmp->mV[1] ; } } - if (rebuild_color) - { - mVertexBuffer->getColorStrider(colors, mGeomIndex); - } - + BOOL is_static = mDrawablep->isStatic(); BOOL is_global = is_static; @@ -1182,7 +1151,8 @@ BOOL LLFace::getGeometryVolume(const LLVolume& volume, // INDICES if (full_rebuild) { - mVertexBuffer->getIndexStrider(indicesp, mIndicesIndex); + mVertexBuffer->getIndexStrider(indicesp, mIndicesIndex, mIndicesCount, true); + __m128i* dst = (__m128i*) indicesp.get(); __m128i* src = (__m128i*) vf.mIndices; __m128i offset = _mm_set1_epi16(index_offset); @@ -1199,6 +1169,8 @@ BOOL LLFace::getGeometryVolume(const LLVolume& volume, { indicesp[i] = vf.mIndices[i]+index_offset; } + + mVertexBuffer->setBuffer(0); } LLMatrix4a mat_normal; @@ -1344,6 +1316,8 @@ BOOL LLFace::getGeometryVolume(const LLVolume& volume, if (!in_atlas && !do_bump) { //not in atlas or not bump mapped, might be able to do a cheap update + mVertexBuffer->getTexCoord0Strider(tex_coords, mGeomIndex, mGeomCount); + if (texgen != LLTextureEntry::TEX_GEN_PLANAR) { if (!do_tex_mat) @@ -1416,9 +1390,15 @@ BOOL LLFace::getGeometryVolume(const LLVolume& volume, } } } + + mVertexBuffer->setBuffer(0); } else { //either bump mapped or in atlas, just do the whole expensive loop + mVertexBuffer->getTexCoord0Strider(tex_coords, mGeomIndex, mGeomCount, true); + + std::vector bump_tc; + for (S32 i = 0; i < num_vertices; i++) { LLVector2 tc(vf.mTexCoords[i]); @@ -1549,8 +1529,20 @@ BOOL LLFace::getGeometryVolume(const LLVolume& volume, *tex_coords++ = tc; - - if (bump_code && mVertexBuffer->hasDataType(LLVertexBuffer::TYPE_TEXCOORD1)) + if (do_bump) + { + bump_tc.push_back(tc); + } + } + + mVertexBuffer->setBuffer(0); + + + if (do_bump) + { + mVertexBuffer->getTexCoord1Strider(tex_coords2, mGeomIndex, mGeomCount, true); + + for (S32 i = 0; i < num_vertices; i++) { LLVector4a tangent; tangent.setCross3(vf.mBinormals[i], vf.mNormals[i]); @@ -1572,16 +1564,22 @@ BOOL LLFace::getGeometryVolume(const LLVolume& volume, } binormal.normalize3fast(); + LLVector2 tc = bump_tc[i]; tc += LLVector2( bump_s_primary_light_ray.dot3(tangent).getF32(), bump_t_primary_light_ray.dot3(binormal).getF32() ); *tex_coords2++ = tc; - } + } + + mVertexBuffer->setBuffer(0); } } } if (rebuild_pos) { + mVertexBuffer->getVertexStrider(vert, mGeomIndex, mGeomCount, true); + vertices = (LLVector4a*) vert.get(); + LLMatrix4a mat_vert; mat_vert.loadu(mat_vert_in); @@ -1607,11 +1605,15 @@ BOOL LLFace::getGeometryVolume(const LLVolume& volume, index_dst += 4; } while (index_dst < index_end); - + + mVertexBuffer->setBuffer(0); } if (rebuild_normal) { + mVertexBuffer->getNormalStrider(norm, mGeomIndex, mGeomCount, true); + normals = (LLVector4a*) norm.get(); + for (S32 i = 0; i < num_vertices; i++) { LLVector4a normal; @@ -1619,10 +1621,15 @@ BOOL LLFace::getGeometryVolume(const LLVolume& volume, normal.normalize3fast(); normals[i] = normal; } + + mVertexBuffer->setBuffer(0); } if (rebuild_binormal) { + mVertexBuffer->getBinormalStrider(binorm, mGeomIndex, mGeomCount, true); + binormals = (LLVector4a*) binorm.get(); + for (S32 i = 0; i < num_vertices; i++) { LLVector4a binormal; @@ -1630,15 +1637,22 @@ BOOL LLFace::getGeometryVolume(const LLVolume& volume, binormal.normalize3fast(); binormals[i] = binormal; } + + mVertexBuffer->setBuffer(0); } if (rebuild_weights && vf.mWeights) { + mVertexBuffer->getWeight4Strider(wght, mGeomIndex, mGeomCount, true); + weights = (LLVector4a*) wght.get(); LLVector4a::memcpyNonAliased16((F32*) weights, (F32*) vf.mWeights, num_vertices*4*sizeof(F32)); + mVertexBuffer->setBuffer(0); } if (rebuild_color) { + mVertexBuffer->getColorStrider(colors, mGeomIndex, mGeomCount, true); + LLVector4a src; U32 vec[4]; @@ -1657,6 +1671,8 @@ BOOL LLFace::getGeometryVolume(const LLVolume& volume, { dst[i] = src; } + + mVertexBuffer->setBuffer(0); } if (rebuild_tcoord) @@ -2073,13 +2089,13 @@ S32 LLFace::getColors(LLStrider &colors) } // llassert(mGeomIndex >= 0); - mVertexBuffer->getColorStrider(colors, mGeomIndex); + mVertexBuffer->getColorStrider(colors, mGeomIndex, mGeomCount); return mGeomIndex; } S32 LLFace::getIndices(LLStrider &indicesp) { - mVertexBuffer->getIndexStrider(indicesp, mIndicesIndex); + mVertexBuffer->getIndexStrider(indicesp, mIndicesIndex, mIndicesCount); llassert(indicesp[0] != indicesp[1]); return mIndicesIndex; } diff --git a/indra/newview/llspatialpartition.h b/indra/newview/llspatialpartition.h index ae5d4fa463..db8a0c2992 100644 --- a/indra/newview/llspatialpartition.h +++ b/indra/newview/llspatialpartition.h @@ -209,7 +209,7 @@ public: typedef std::vector > drawmap_elem_t; typedef std::map draw_map_t; typedef std::vector > buffer_list_t; - typedef std::map, buffer_list_t> buffer_texture_map_t; + typedef std::map buffer_texture_map_t; typedef std::map buffer_map_t; typedef LLOctreeListener BaseType; @@ -401,7 +401,7 @@ protected: public: bridge_list_t mBridgeList; - buffer_map_t mBufferMap; //used by volume buffers to store unique buffers per texture + buffer_map_t mBufferMap; //used by volume buffers to attempt to reuse vertex buffers F32 mBuilt; OctreeNode* mOctreeNode; diff --git a/indra/newview/llviewercontrol.cpp b/indra/newview/llviewercontrol.cpp index 61200db2b9..87ca80260f 100644 --- a/indra/newview/llviewercontrol.cpp +++ b/indra/newview/llviewercontrol.cpp @@ -591,6 +591,7 @@ void settings_setup_listeners() gSavedSettings.getControl("OctreeMaxNodeCapacity")->getSignal()->connect(boost::bind(&handleRepartition, _2)); gSavedSettings.getControl("OctreeAlphaDistanceFactor")->getSignal()->connect(boost::bind(&handleRepartition, _2)); gSavedSettings.getControl("OctreeAttachmentSizeFactor")->getSignal()->connect(boost::bind(&handleRepartition, _2)); + gSavedSettings.getControl("RenderMaxTextureIndex")->getSignal()->connect(boost::bind(&handleResetVertexBuffersChanged, _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)); diff --git a/indra/newview/llvovolume.cpp b/indra/newview/llvovolume.cpp index 125283d474..58e38ea9b1 100644 --- a/indra/newview/llvovolume.cpp +++ b/indra/newview/llvovolume.cpp @@ -4426,6 +4426,33 @@ void LLVolumeGeometryManager::rebuildMesh(LLSpatialGroup* group) llassert(!group || !group->isState(LLSpatialGroup::NEW_DRAWINFO)); } +struct CompareBatchBreakerModified +{ + bool operator()(const LLFace* const& lhs, const LLFace* const& rhs) + { + const LLTextureEntry* lte = lhs->getTextureEntry(); + const LLTextureEntry* rte = rhs->getTextureEntry(); + + if (lte->getBumpmap() != rte->getBumpmap()) + { + return lte->getBumpmap() < rte->getBumpmap(); + } + else if (lte->getFullbright() != rte->getFullbright()) + { + return lte->getFullbright() < rte->getFullbright(); + } + else if (lte->getGlow() != rte->getGlow()) + { + return lte->getGlow() < rte->getGlow(); + } + else + { + return lhs->getTexture() < rhs->getTexture(); + } + + } +}; + void LLVolumeGeometryManager::genDrawInfo(LLSpatialGroup* group, U32 mask, std::vector& faces, BOOL distance_sort, BOOL batch_textures) { //calculate maximum number of vertices to store in a single buffer @@ -4435,7 +4462,7 @@ void LLVolumeGeometryManager::genDrawInfo(LLSpatialGroup* group, U32 mask, std:: if (!distance_sort) { //sort faces by things that break batches - std::sort(faces.begin(), faces.end(), LLFace::CompareBatchBreaker()); + std::sort(faces.begin(), faces.end(), CompareBatchBreakerModified()); } else { @@ -4462,6 +4489,8 @@ void LLVolumeGeometryManager::genDrawInfo(LLSpatialGroup* group, U32 mask, std:: texture_index_channels = gDeferredAlphaProgram.mFeatures.mIndexedTextureChannels; } + texture_index_channels = llmin(texture_index_channels, (S32) gSavedSettings.getU32("RenderMaxTextureIndex")); + while (face_iter != faces.end()) { @@ -4587,7 +4616,7 @@ void LLVolumeGeometryManager::genDrawInfo(LLSpatialGroup* group, U32 mask, std:: //create/delete/resize vertex buffer if needed LLVertexBuffer* buffer = NULL; - LLSpatialGroup::buffer_texture_map_t::iterator found_iter = group->mBufferMap[mask].find(tex); + LLSpatialGroup::buffer_texture_map_t::iterator found_iter = group->mBufferMap[mask].find(*face_iter); if (found_iter != group->mBufferMap[mask].end()) { @@ -4618,7 +4647,7 @@ void LLVolumeGeometryManager::genDrawInfo(LLSpatialGroup* group, U32 mask, std:: } } - buffer_map[mask][tex].push_back(buffer); + buffer_map[mask][*face_iter].push_back(buffer); //add face geometry @@ -4649,12 +4678,8 @@ void LLVolumeGeometryManager::genDrawInfo(LLSpatialGroup* group, U32 mask, std:: U32 te_idx = facep->getTEOffset(); - if (facep->getGeometryVolume(*volume, te_idx, - vobj->getRelativeXform(), vobj->getRelativeXformInvTrans(), index_offset)) - { - buffer->markDirty(facep->getGeomIndex(), facep->getGeomCount(), - facep->getIndicesStart(), facep->getIndicesCount()); - } + facep->getGeometryVolume(*volume, te_idx, + vobj->getRelativeXform(), vobj->getRelativeXformInvTrans(), index_offset); } } -- cgit v1.3