From 51bb3c15c8ac6c85ed1a7e8526ba6b60794ac29e Mon Sep 17 00:00:00 2001 From: Dave Parks Date: Thu, 19 May 2011 20:38:39 -0500 Subject: SH-469 WIP -- get rid of LLMultiSampleBuffer and use GL_ARB_texture_multisample instead. --- indra/llrender/llimagegl.cpp | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'indra/llrender/llimagegl.cpp') diff --git a/indra/llrender/llimagegl.cpp b/indra/llrender/llimagegl.cpp index d408077c68..30b66ae542 100644 --- a/indra/llrender/llimagegl.cpp +++ b/indra/llrender/llimagegl.cpp @@ -1417,7 +1417,7 @@ void LLImageGL::deleteDeadTextures() { if (sCurrentBoundTextures[i] == tex) { - gGL.getTexUnit(i)->unbind(LLTexUnit::TT_TEXTURE); + gGL.getTexUnit(i)->unbind(gGL.getTexUnit(i)->getCurrType()); stop_glerror(); } } -- cgit v1.3 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/llrender/llimagegl.cpp | 7 +- indra/llrender/llimagegl.h | 2 +- indra/llrender/llrendertarget.cpp | 4 +- indra/llrender/llshadermgr.cpp | 6 + indra/llrender/llshadermgr.h | 3 + .../shaders/class1/deferred/blurLightF.glsl | 4 +- .../shaders/class1/deferred/blurLightMSF.glsl | 4 +- .../shaders/class1/deferred/multiPointLightF.glsl | 2 +- .../class1/deferred/multiPointLightMSF.glsl | 137 ++++++++++++ .../shaders/class1/deferred/multiSpotLightMSF.glsl | 233 ++++++++++++++++++++ .../shaders/class1/deferred/pointLightF.glsl | 2 +- .../shaders/class1/deferred/pointLightMSF.glsl | 96 ++++---- .../shaders/class1/deferred/pointLightV.glsl | 7 +- .../shaders/class1/deferred/postDeferredF.glsl | 2 +- .../shaders/class1/deferred/postDeferredMSF.glsl | 4 +- .../class1/deferred/postDeferredNoDoFMSF.glsl | 13 +- .../shaders/class1/deferred/softenLightF.glsl | 2 +- .../shaders/class1/deferred/softenLightMSF.glsl | 14 +- .../shaders/class1/deferred/spotLightMSF.glsl | 234 ++++++++++++++++++++ .../shaders/class1/deferred/starsF.glsl | 19 ++ .../shaders/class1/deferred/starsV.glsl | 17 ++ .../shaders/class1/deferred/sunLightSSAOF.glsl | 2 +- .../shaders/class1/deferred/sunLightSSAOMSF.glsl | 2 - .../shaders/class1/effects/glowExtractMSF.glsl | 38 ++++ .../shaders/class2/deferred/edgeF.glsl | 2 +- .../shaders/class2/deferred/edgeMSF.glsl | 16 +- .../shaders/class2/deferred/multiSpotLightF.glsl | 2 +- .../shaders/class2/deferred/multiSpotLightMSF.glsl | 244 ++++++++++++++++++++ .../shaders/class2/deferred/softenLightF.glsl | 52 +---- .../shaders/class2/deferred/softenLightMSF.glsl | 1 - .../shaders/class2/deferred/spotLightMSF.glsl | 245 +++++++++++++++++++++ .../shaders/class2/deferred/sunLightF.glsl | 2 +- .../shaders/class2/deferred/sunLightMSF.glsl | 202 +++++++++++++++++ .../shaders/class2/deferred/sunLightSSAOF.glsl | 2 +- .../shaders/class2/deferred/sunLightSSAOMSF.glsl | 241 ++++++++++++++++++++ indra/newview/lldrawpoolwlsky.cpp | 7 +- indra/newview/llviewercontrol.cpp | 17 +- indra/newview/llviewershadermgr.cpp | 83 ++++++- indra/newview/llviewershadermgr.h | 2 +- indra/newview/pipeline.cpp | 14 +- 40 files changed, 1827 insertions(+), 159 deletions(-) create mode 100644 indra/newview/app_settings/shaders/class1/deferred/multiPointLightMSF.glsl create mode 100644 indra/newview/app_settings/shaders/class1/deferred/multiSpotLightMSF.glsl create mode 100644 indra/newview/app_settings/shaders/class1/deferred/spotLightMSF.glsl create mode 100644 indra/newview/app_settings/shaders/class1/deferred/starsF.glsl create mode 100644 indra/newview/app_settings/shaders/class1/deferred/starsV.glsl create mode 100644 indra/newview/app_settings/shaders/class1/effects/glowExtractMSF.glsl create mode 100644 indra/newview/app_settings/shaders/class2/deferred/multiSpotLightMSF.glsl create mode 100644 indra/newview/app_settings/shaders/class2/deferred/spotLightMSF.glsl create mode 100644 indra/newview/app_settings/shaders/class2/deferred/sunLightMSF.glsl create mode 100644 indra/newview/app_settings/shaders/class2/deferred/sunLightSSAOMSF.glsl (limited to 'indra/llrender/llimagegl.cpp') diff --git a/indra/llrender/llimagegl.cpp b/indra/llrender/llimagegl.cpp index 30b66ae542..3a4139bace 100644 --- a/indra/llrender/llimagegl.cpp +++ b/indra/llrender/llimagegl.cpp @@ -1083,12 +1083,17 @@ void LLImageGL::generateTextures(S32 numTextures, U32 *textures) } // static -void LLImageGL::deleteTextures(S32 numTextures, U32 *textures) +void LLImageGL::deleteTextures(S32 numTextures, U32 *textures, bool immediate) { for (S32 i = 0; i < numTextures; i++) { sDeadTextureList.push_back(textures[i]); } + + if (immediate) + { + LLImageGL::deleteDeadTextures(); + } } // static diff --git a/indra/llrender/llimagegl.h b/indra/llrender/llimagegl.h index 6c980984c0..2cfb15b0d9 100644 --- a/indra/llrender/llimagegl.h +++ b/indra/llrender/llimagegl.h @@ -98,7 +98,7 @@ public: // These 3 functions currently wrap glGenTextures(), glDeleteTextures(), and glTexImage2D() // for tracking purposes and will be deprecated in the future static void generateTextures(S32 numTextures, U32 *textures); - static void deleteTextures(S32 numTextures, U32 *textures); + static void deleteTextures(S32 numTextures, U32 *textures, bool immediate = false); static void setManualImage(U32 target, S32 miplevel, S32 intformat, S32 width, S32 height, U32 pixformat, U32 pixtype, const void *pixels); BOOL createGLTexture() ; diff --git a/indra/llrender/llrendertarget.cpp b/indra/llrender/llrendertarget.cpp index 64caf77b87..ddd6072fff 100644 --- a/indra/llrender/llrendertarget.cpp +++ b/indra/llrender/llrendertarget.cpp @@ -289,7 +289,7 @@ void LLRenderTarget::release() } else { - LLImageGL::deleteTextures(1, &mDepth); + LLImageGL::deleteTextures(1, &mDepth, true); stop_glerror(); } mDepth = 0; @@ -318,7 +318,7 @@ void LLRenderTarget::release() if (mTex.size() > 0) { - LLImageGL::deleteTextures(mTex.size(), &mTex[0]); + LLImageGL::deleteTextures(mTex.size(), &mTex[0], true); mTex.clear(); } diff --git a/indra/llrender/llshadermgr.cpp b/indra/llrender/llshadermgr.cpp index 98a0a93084..b8f9c60ca9 100644 --- a/indra/llrender/llshadermgr.cpp +++ b/indra/llrender/llshadermgr.cpp @@ -374,6 +374,12 @@ GLhandleARB LLShaderMgr::loadShaderFile(const std::string& filename, S32 & shade GLcharARB* text[1024]; GLuint count = 0; + //copy preprocessor definitions into buffer + for (std::map::iterator iter = mDefinitions.begin(); iter != mDefinitions.end(); ++iter) + { + std::string define = "#define " + iter->first + " " + iter->second + "\n"; + text[count++] = (GLcharARB *) strdup(define.c_str()); + } //copy file into memory while( fgets((char *)buff, 1024, file) != NULL && count < LL_ARRAY_SIZE(buff) ) diff --git a/indra/llrender/llshadermgr.h b/indra/llrender/llshadermgr.h index c54c4608d7..9ee11b98c8 100644 --- a/indra/llrender/llshadermgr.h +++ b/indra/llrender/llshadermgr.h @@ -60,6 +60,9 @@ public: std::vector mReservedUniforms; + //preprocessor definitions (name/value) + std::map mDefinitions; + protected: // our parameter manager singleton instance diff --git a/indra/newview/app_settings/shaders/class1/deferred/blurLightF.glsl b/indra/newview/app_settings/shaders/class1/deferred/blurLightF.glsl index d9f021b114..98d034c2b5 100644 --- a/indra/newview/app_settings/shaders/class1/deferred/blurLightF.glsl +++ b/indra/newview/app_settings/shaders/class1/deferred/blurLightF.glsl @@ -26,7 +26,7 @@ uniform vec2 screen_res; vec4 getPosition(vec2 pos_screen) { - float depth = texture2DRect(depthMap, pos_screen.xy).a; + float depth = texture2DRect(depthMap, pos_screen.xy).r; vec2 sc = pos_screen.xy*2.0; sc /= screen_res; sc -= vec2(1.0,1.0); @@ -39,7 +39,7 @@ vec4 getPosition(vec2 pos_screen) void main() { - vec2 tc = vary_fragcoord.xy; + vec2 tc = vary_fragcoord.xy; vec3 norm = texture2DRect(normalMap, tc).xyz; norm = vec3((norm.xy-0.5)*2.0,norm.z); // unpack norm vec3 pos = getPosition(tc).xyz; diff --git a/indra/newview/app_settings/shaders/class1/deferred/blurLightMSF.glsl b/indra/newview/app_settings/shaders/class1/deferred/blurLightMSF.glsl index 113197c871..4366376d99 100644 --- a/indra/newview/app_settings/shaders/class1/deferred/blurLightMSF.glsl +++ b/indra/newview/app_settings/shaders/class1/deferred/blurLightMSF.glsl @@ -28,12 +28,12 @@ uniform vec2 screen_res; vec4 texture2DMS(sampler2DMS tex, ivec2 tc) { vec4 ret = vec4(0,0,0,0); - for (int i = 0; i < 4; i++) + for (int i = 0; i < samples; i++) { ret += texelFetch(tex, tc, i); } - return ret * 0.25; + return ret/samples; } vec4 getPosition(ivec2 pos_screen) diff --git a/indra/newview/app_settings/shaders/class1/deferred/multiPointLightF.glsl b/indra/newview/app_settings/shaders/class1/deferred/multiPointLightF.glsl index 609fc4f14f..e10fb105e8 100644 --- a/indra/newview/app_settings/shaders/class1/deferred/multiPointLightF.glsl +++ b/indra/newview/app_settings/shaders/class1/deferred/multiPointLightF.glsl @@ -36,7 +36,7 @@ uniform mat4 inv_proj; vec4 getPosition(vec2 pos_screen) { - float depth = texture2DRect(depthMap, pos_screen.xy).a; + float depth = texture2DRect(depthMap, pos_screen.xy).r; vec2 sc = pos_screen.xy*2.0; sc /= screen_res; sc -= vec2(1.0,1.0); diff --git a/indra/newview/app_settings/shaders/class1/deferred/multiPointLightMSF.glsl b/indra/newview/app_settings/shaders/class1/deferred/multiPointLightMSF.glsl new file mode 100644 index 0000000000..16e97109f9 --- /dev/null +++ b/indra/newview/app_settings/shaders/class1/deferred/multiPointLightMSF.glsl @@ -0,0 +1,137 @@ +/** + * @file multiPointLightF.glsl + * + * $LicenseInfo:firstyear=2007&license=viewerlgpl$ + * $/LicenseInfo$ + */ + +#version 120 + +#extension GL_ARB_texture_rectangle : enable +#extension GL_ARB_texture_multisample : enable + +uniform sampler2DMS depthMap; +uniform sampler2DMS diffuseRect; +uniform sampler2DMS specularRect; +uniform sampler2DMS normalMap; +uniform sampler2D noiseMap; +uniform sampler2D lightFunc; + + +uniform vec3 env_mat[3]; +uniform float sun_wash; + +uniform int light_count; + +#define MAX_LIGHT_COUNT 16 +uniform vec4 light[MAX_LIGHT_COUNT]; +uniform vec4 light_col[MAX_LIGHT_COUNT]; + +varying vec4 vary_fragcoord; +uniform vec2 screen_res; + +uniform float far_z; + +uniform mat4 inv_proj; + +vec4 getPosition(ivec2 pos_screen, int sample) +{ + float depth = texelFetch(depthMap, pos_screen, sample).r; + vec2 sc = vec2(pos_screen.xy)*2.0; + sc /= screen_res; + sc -= vec2(1.0,1.0); + vec4 ndc = vec4(sc.x, sc.y, 2.0*depth-1.0, 1.0); + vec4 pos = inv_proj * ndc; + pos /= pos.w; + pos.w = 1.0; + return pos; +} + +void main() +{ + vec2 frag = (vary_fragcoord.xy*0.5+0.5)*screen_res; + ivec2 itc = ivec2(frag); + + int wght = 0; + vec3 fcol = vec3(0,0,0); + + for (int s = 0; s < samples; ++s) + { + vec3 pos = getPosition(itc, s).xyz; + if (pos.z >= far_z) + { + vec3 norm = texelFetch(normalMap, itc, s).xyz; + norm = vec3((norm.xy-0.5)*2.0,norm.z); // unpack norm + norm = normalize(norm); + vec4 spec = texelFetch(specularRect, itc, s); + vec3 diff = texelFetch(diffuseRect, itc, s).rgb; + float noise = texture2D(noiseMap, frag.xy/128.0).b; + vec3 out_col = vec3(0,0,0); + vec3 npos = normalize(-pos); + + // As of OSX 10.6.7 ATI Apple's crash when using a variable size loop + for (int i = 0; i < MAX_LIGHT_COUNT; ++i) + { + bool light_contrib = (i < light_count); + + vec3 lv = light[i].xyz-pos; + float dist2 = dot(lv,lv); + dist2 /= light[i].w; + if (dist2 > 1.0) + { + light_contrib = false; + } + + float da = dot(norm, lv); + if (da < 0.0) + { + light_contrib = false; + } + + if (light_contrib) + { + lv = normalize(lv); + da = dot(norm, lv); + + float fa = light_col[i].a+1.0; + float dist_atten = clamp(1.0-(dist2-1.0*(1.0-fa))/fa, 0.0, 1.0); + dist_atten *= noise; + + float lit = da * dist_atten; + + vec3 col = light_col[i].rgb*lit*diff; + //vec3 col = vec3(dist2, light_col[i].a, lit); + + if (spec.a > 0.0) + { + //vec3 ref = dot(pos+lv, norm); + + float sa = dot(normalize(lv+npos),norm); + + if (sa > 0.0) + { + sa = texture2D(lightFunc,vec2(sa, spec.a)).a * min(dist_atten*4.0, 1.0); + sa *= noise; + col += da*sa*light_col[i].rgb*spec.rgb; + } + } + + out_col += col; + } + } + + fcol += out_col; + ++wght; + } + } + + if (wght <= 0) + { + discard; + } + + gl_FragColor.rgb = fcol/wght; + gl_FragColor.a = 0.0; + + +} diff --git a/indra/newview/app_settings/shaders/class1/deferred/multiSpotLightMSF.glsl b/indra/newview/app_settings/shaders/class1/deferred/multiSpotLightMSF.glsl new file mode 100644 index 0000000000..bdaa8e59c4 --- /dev/null +++ b/indra/newview/app_settings/shaders/class1/deferred/multiSpotLightMSF.glsl @@ -0,0 +1,233 @@ +/** + * @file multiSpotLightF.glsl + * + * $LicenseInfo:firstyear=2007&license=viewerlgpl$ + * $/LicenseInfo$ + */ + +#version 120 + +//class 1 -- no shadows + +#extension GL_ARB_texture_rectangle : enable +#extension GL_ARB_texture_multisample : enable + +uniform sampler2DMS diffuseRect; +uniform sampler2DMS specularRect; +uniform sampler2DMS depthMap; +uniform sampler2DMS normalMap; +uniform sampler2D noiseMap; +uniform sampler2D lightFunc; +uniform sampler2D projectionMap; + +uniform mat4 proj_mat; //screen space to light space +uniform float proj_near; //near clip for projection +uniform vec3 proj_p; //plane projection is emitting from (in screen space) +uniform vec3 proj_n; +uniform float proj_focus; //distance from plane to begin blurring +uniform float proj_lod; //(number of mips in proj map) +uniform float proj_range; //range between near clip and far clip plane of projection +uniform float proj_ambient_lod; +uniform float proj_ambiance; +uniform float near_clip; +uniform float far_clip; + +uniform vec3 proj_origin; //origin of projection to be used for angular attenuation +uniform float sun_wash; +uniform int proj_shadow_idx; +uniform float shadow_fade; + +varying vec4 vary_light; + +varying vec4 vary_fragcoord; +uniform vec2 screen_res; + +uniform mat4 inv_proj; + +vec4 texture2DLodSpecular(sampler2D projectionMap, vec2 tc, float lod) +{ + vec4 ret = texture2DLod(projectionMap, tc, lod); + + vec2 dist = tc-vec2(0.5); + + float det = max(1.0-lod/(proj_lod*0.5), 0.0); + + float d = dot(dist,dist); + + ret *= min(clamp((0.25-d)/0.25, 0.0, 1.0)+det, 1.0); + + return ret; +} + +vec4 texture2DLodDiffuse(sampler2D projectionMap, vec2 tc, float lod) +{ + vec4 ret = texture2DLod(projectionMap, tc, lod); + + vec2 dist = vec2(0.5) - abs(tc-vec2(0.5)); + + float det = min(lod/(proj_lod*0.5), 1.0); + + float d = min(dist.x, dist.y); + + float edge = 0.25*det; + + ret *= clamp(d/edge, 0.0, 1.0); + + return ret; +} + +vec4 texture2DLodAmbient(sampler2D projectionMap, vec2 tc, float lod) +{ + vec4 ret = texture2DLod(projectionMap, tc, lod); + + vec2 dist = tc-vec2(0.5); + + float d = dot(dist,dist); + + ret *= min(clamp((0.25-d)/0.25, 0.0, 1.0), 1.0); + + return ret; +} + + +vec4 getPosition(ivec2 pos_screen, int sample) +{ + float depth = texelFetch(depthMap, pos_screen, sample).r; + vec2 sc = vec2(pos_screen.xy)*2.0; + sc /= screen_res; + sc -= vec2(1.0,1.0); + vec4 ndc = vec4(sc.x, sc.y, 2.0*depth-1.0, 1.0); + vec4 pos = inv_proj * ndc; + pos /= pos.w; + pos.w = 1.0; + return pos; +} + +void main() +{ + int wght = 0; + + vec3 fcol = vec3(0,0,0); + + vec2 frag = (vary_fragcoord.xy*0.5+0.5)*screen_res; + + ivec2 itc = ivec2(frag.xy); + + for (int i = 0; i < samples; ++i) + { + vec3 pos = getPosition(itc, i).xyz; + vec3 lv = vary_light.xyz-pos.xyz; + float dist2 = dot(lv,lv); + dist2 /= vary_light.w; + if (dist2 <= 1.0) + { + vec3 norm = texelFetch(normalMap, itc, i).xyz*2.0-1.0; + + norm = normalize(norm); + float l_dist = -dot(lv, proj_n); + + vec4 proj_tc = (proj_mat * vec4(pos.xyz, 1.0)); + if (proj_tc.z >= 0.0) + { + proj_tc.xyz /= proj_tc.w; + + float fa = gl_Color.a+1.0; + float dist_atten = min(1.0-(dist2-1.0*(1.0-fa))/fa, 1.0); + if (dist_atten > 0.0) + { + lv = proj_origin-pos.xyz; + lv = normalize(lv); + float da = dot(norm, lv); + + vec3 col = vec3(0,0,0); + + vec3 diff_tex = texelFetch(diffuseRect, itc, i).rgb; + + float noise = texture2D(noiseMap, frag.xy/128.0).b; + if (proj_tc.z > 0.0 && + proj_tc.x < 1.0 && + proj_tc.y < 1.0 && + proj_tc.x > 0.0 && + proj_tc.y > 0.0) + { + float lit = 0.0; + float amb_da = proj_ambiance; + + if (da > 0.0) + { + float diff = clamp((l_dist-proj_focus)/proj_range, 0.0, 1.0); + float lod = diff * proj_lod; + + vec4 plcol = texture2DLodDiffuse(projectionMap, proj_tc.xy, lod); + + vec3 lcol = gl_Color.rgb * plcol.rgb * plcol.a; + + lit = da * dist_atten * noise; + + col = lcol*lit*diff_tex; + amb_da += (da*0.5)*proj_ambiance; + } + + //float diff = clamp((proj_range-proj_focus)/proj_range, 0.0, 1.0); + vec4 amb_plcol = texture2DLodAmbient(projectionMap, proj_tc.xy, proj_lod); + + amb_da += (da*da*0.5+0.5)*proj_ambiance; + + amb_da *= dist_atten * noise; + + amb_da = min(amb_da, 1.0-lit); + + col += amb_da*gl_Color.rgb*diff_tex.rgb*amb_plcol.rgb*amb_plcol.a; + } + + + vec4 spec = texelFetch(specularRect, itc, i); + if (spec.a > 0.0) + { + vec3 ref = reflect(normalize(pos), norm); + + //project from point pos in direction ref to plane proj_p, proj_n + vec3 pdelta = proj_p-pos; + float ds = dot(ref, proj_n); + + if (ds < 0.0) + { + vec3 pfinal = pos + ref * dot(pdelta, proj_n)/ds; + + vec4 stc = (proj_mat * vec4(pfinal.xyz, 1.0)); + + if (stc.z > 0.0) + { + stc.xy /= stc.w; + + float fatten = clamp(spec.a*spec.a+spec.a*0.5, 0.25, 1.0); + + stc.xy = (stc.xy - vec2(0.5)) * fatten + vec2(0.5); + + if (stc.x < 1.0 && + stc.y < 1.0 && + stc.x > 0.0 && + stc.y > 0.0) + { + vec4 scol = texture2DLodSpecular(projectionMap, stc.xy, proj_lod-spec.a*proj_lod); + col += dist_atten*scol.rgb*gl_Color.rgb*scol.a*spec.rgb; + } + } + } + } + + fcol += col; + ++wght; + } + } + } + } + + if (wght <= 0) + { + discard; + } + + gl_FragColor.rgb = fcol/wght; + gl_FragColor.a = 0.0; +} diff --git a/indra/newview/app_settings/shaders/class1/deferred/pointLightF.glsl b/indra/newview/app_settings/shaders/class1/deferred/pointLightF.glsl index 22ed9dcd40..0d771109fb 100644 --- a/indra/newview/app_settings/shaders/class1/deferred/pointLightF.glsl +++ b/indra/newview/app_settings/shaders/class1/deferred/pointLightF.glsl @@ -30,7 +30,7 @@ uniform vec4 viewport; vec4 getPosition(vec2 pos_screen) { - float depth = texture2DRect(depthMap, pos_screen.xy).a; + float depth = texture2DRect(depthMap, pos_screen.xy).r; vec2 sc = (pos_screen.xy-viewport.xy)*2.0; sc /= viewport.zw; sc -= vec2(1.0,1.0); diff --git a/indra/newview/app_settings/shaders/class1/deferred/pointLightMSF.glsl b/indra/newview/app_settings/shaders/class1/deferred/pointLightMSF.glsl index 22ed9dcd40..23b5e76735 100644 --- a/indra/newview/app_settings/shaders/class1/deferred/pointLightMSF.glsl +++ b/indra/newview/app_settings/shaders/class1/deferred/pointLightMSF.glsl @@ -8,14 +8,15 @@ #version 120 #extension GL_ARB_texture_rectangle : enable +#extension GL_ARB_texture_multisample : enable -uniform sampler2DRect diffuseRect; -uniform sampler2DRect specularRect; -uniform sampler2DRect normalMap; -uniform samplerCube environmentMap; +uniform sampler2DMS depthMap; +uniform sampler2DMS diffuseRect; +uniform sampler2DMS specularRect; +uniform sampler2DMS normalMap; uniform sampler2D noiseMap; uniform sampler2D lightFunc; -uniform sampler2DRect depthMap; + uniform vec3 env_mat[3]; uniform float sun_wash; @@ -28,10 +29,10 @@ uniform vec2 screen_res; uniform mat4 inv_proj; uniform vec4 viewport; -vec4 getPosition(vec2 pos_screen) +vec4 getPosition(ivec2 pos_screen, int sample) { - float depth = texture2DRect(depthMap, pos_screen.xy).a; - vec2 sc = (pos_screen.xy-viewport.xy)*2.0; + float depth = texelFetch(depthMap, pos_screen, sample).r; + vec2 sc = (vec2(pos_screen.xy)-viewport.xy)*2.0; sc /= viewport.zw; sc -= vec2(1.0,1.0); vec4 ndc = vec4(sc.x, sc.y, 2.0*depth-1.0, 1.0); @@ -48,53 +49,60 @@ void main() frag.xyz = frag.xyz*0.5+0.5; frag.xy *= screen_res; - vec3 pos = getPosition(frag.xy).xyz; - vec3 lv = vary_light.xyz-pos; - float dist2 = dot(lv,lv); - dist2 /= vary_light.w; - if (dist2 > 1.0) - { - discard; - } - - vec3 norm = texture2DRect(normalMap, frag.xy).xyz; - norm = vec3((norm.xy-0.5)*2.0,norm.z); // unpack norm - float da = dot(norm, lv); - if (da < 0.0) + ivec2 itc = ivec2(frag.xy); + + int wght = 0; + vec3 fcol = vec3(0,0,0); + + for (int s = 0; s < samples; ++s) { - discard; - } - - norm = normalize(norm); - lv = normalize(lv); - da = dot(norm, lv); + vec3 pos = getPosition(itc, s).xyz; + vec3 lv = vary_light.xyz-pos; + float dist2 = dot(lv,lv); + dist2 /= vary_light.w; + if (dist2 <= 1.0) + { + vec3 norm = texelFetch(normalMap, itc, s).xyz; + norm = vec3((norm.xy-0.5)*2.0,norm.z); // unpack norm + float da = dot(norm, lv); + if (da >= 0.0) + { + norm = normalize(norm); + lv = normalize(lv); + da = dot(norm, lv); - float noise = texture2D(noiseMap, frag.xy/128.0).b; + float noise = texture2D(noiseMap, frag.xy/128.0).b; - vec3 col = texture2DRect(diffuseRect, frag.xy).rgb; - float fa = gl_Color.a+1.0; - float dist_atten = clamp(1.0-(dist2-1.0*(1.0-fa))/fa, 0.0, 1.0); - float lit = da * dist_atten * noise; + vec3 col = texelFetch(diffuseRect, itc, s).rgb; + float fa = gl_Color.a+1.0; + float dist_atten = clamp(1.0-(dist2-1.0*(1.0-fa))/fa, 0.0, 1.0); + float lit = da * dist_atten * noise; - col = gl_Color.rgb*lit*col; + col = gl_Color.rgb*lit*col; - vec4 spec = texture2DRect(specularRect, frag.xy); - if (spec.a > 0.0) - { - float sa = dot(normalize(lv-normalize(pos)),norm); - if (sa > 0.0) - { - sa = texture2D(lightFunc, vec2(sa, spec.a)).a * min(dist_atten*4.0, 1.0); - sa *= noise; - col += da*sa*gl_Color.rgb*spec.rgb; + vec4 spec = texelFetch(specularRect, itc, s); + if (spec.a > 0.0) + { + float sa = dot(normalize(lv-normalize(pos)),norm); + if (sa > 0.0) + { + sa = texture2D(lightFunc, vec2(sa, spec.a)).a * min(dist_atten*4.0, 1.0); + sa *= noise; + col += da*sa*gl_Color.rgb*spec.rgb; + } + } + + fcol += col; + ++wght; + } } } - if (dot(col, col) <= 0.0) + if (wght <= 0) { discard; } - gl_FragColor.rgb = col; + gl_FragColor.rgb = fcol/wght; gl_FragColor.a = 0.0; } diff --git a/indra/newview/app_settings/shaders/class1/deferred/pointLightV.glsl b/indra/newview/app_settings/shaders/class1/deferred/pointLightV.glsl index 8e74feb615..2377947e7f 100644 --- a/indra/newview/app_settings/shaders/class1/deferred/pointLightV.glsl +++ b/indra/newview/app_settings/shaders/class1/deferred/pointLightV.glsl @@ -10,14 +10,9 @@ varying vec4 vary_light; varying vec4 vary_fragcoord; -uniform vec2 screen_res; -uniform float near_clip; - void main() { //transform vertex - gl_Position = gl_ModelViewProjectionMatrix * gl_Vertex; - vec4 pos = gl_ModelViewProjectionMatrix * gl_Vertex; vary_fragcoord = pos; @@ -25,6 +20,8 @@ void main() tex.w = 1.0; vary_light = gl_MultiTexCoord0; + + gl_Position = pos; gl_FrontColor = gl_Color; } diff --git a/indra/newview/app_settings/shaders/class1/deferred/postDeferredF.glsl b/indra/newview/app_settings/shaders/class1/deferred/postDeferredF.glsl index 77f1b2224c..d389add03b 100644 --- a/indra/newview/app_settings/shaders/class1/deferred/postDeferredF.glsl +++ b/indra/newview/app_settings/shaders/class1/deferred/postDeferredF.glsl @@ -29,7 +29,7 @@ varying vec2 vary_fragcoord; float getDepth(vec2 pos_screen) { - float z = texture2DRect(depthMap, pos_screen.xy).a; + float z = texture2DRect(depthMap, pos_screen.xy).r; z = z*2.0-1.0; vec4 ndc = vec4(0.0, 0.0, z, 1.0); vec4 p = inv_proj*ndc; diff --git a/indra/newview/app_settings/shaders/class1/deferred/postDeferredMSF.glsl b/indra/newview/app_settings/shaders/class1/deferred/postDeferredMSF.glsl index a34f882c39..711c6dc96d 100644 --- a/indra/newview/app_settings/shaders/class1/deferred/postDeferredMSF.glsl +++ b/indra/newview/app_settings/shaders/class1/deferred/postDeferredMSF.glsl @@ -31,12 +31,12 @@ varying vec2 vary_fragcoord; vec4 texture2DMS(sampler2DMS tex, ivec2 tc) { vec4 ret = vec4(0,0,0,0); - for (int i = 0; i < 4; ++i) + for (int i = 0; i < samples; ++i) { ret += texelFetch(tex, tc, i); } - return ret * 0.25; + return ret/samples; } float getDepth(ivec2 pos_screen) diff --git a/indra/newview/app_settings/shaders/class1/deferred/postDeferredNoDoFMSF.glsl b/indra/newview/app_settings/shaders/class1/deferred/postDeferredNoDoFMSF.glsl index 7b14974f8b..359b26b7d4 100644 --- a/indra/newview/app_settings/shaders/class1/deferred/postDeferredNoDoFMSF.glsl +++ b/indra/newview/app_settings/shaders/class1/deferred/postDeferredNoDoFMSF.glsl @@ -18,13 +18,14 @@ varying vec2 vary_fragcoord; vec4 texture2DMS(sampler2DMS tex, ivec2 tc) { - vec4 ret = texelFetch(tex,tc,0); - ret += texelFetch(tex,tc,1); - ret += texelFetch(tex,tc,2); - ret += texelFetch(tex,tc,3); - ret *= 0.25; + vec4 ret = vec4(0,0,0,0); - return ret; + for (int i = 0; i < samples; ++i) + { + ret += texelFetch(tex,tc,i); + } + + return ret/samples; } void main() diff --git a/indra/newview/app_settings/shaders/class1/deferred/softenLightF.glsl b/indra/newview/app_settings/shaders/class1/deferred/softenLightF.glsl index 29340c7e9f..c1147feff6 100644 --- a/indra/newview/app_settings/shaders/class1/deferred/softenLightF.glsl +++ b/indra/newview/app_settings/shaders/class1/deferred/softenLightF.glsl @@ -259,7 +259,7 @@ vec3 scaleSoftClip(vec3 light) void main() { vec2 tc = vary_fragcoord.xy; - float depth = texture2DRect(depthMap, tc.xy).a; + float depth = texture2DRect(depthMap, tc.xy).r; vec3 pos = getPosition_d(tc, depth).xyz; vec3 norm = texture2DRect(normalMap, tc).xyz; norm = vec3((norm.xy-0.5)*2.0,norm.z); // unpack norm diff --git a/indra/newview/app_settings/shaders/class1/deferred/softenLightMSF.glsl b/indra/newview/app_settings/shaders/class1/deferred/softenLightMSF.glsl index 4c38d91499..fcba4f57e6 100644 --- a/indra/newview/app_settings/shaders/class1/deferred/softenLightMSF.glsl +++ b/indra/newview/app_settings/shaders/class1/deferred/softenLightMSF.glsl @@ -251,18 +251,18 @@ vec3 scaleSoftClip(vec3 light) vec4 texture2DMS(sampler2DMS tex, ivec2 tc) { - vec4 ret = texelFetch(tex,tc,0); - ret += texelFetch(tex,tc,1); - ret += texelFetch(tex,tc,2); - ret += texelFetch(tex,tc,3); - ret *= 0.25; + vec4 ret = vec4(0,0,0,0); - return ret; + for (int i = 0; i < samples; ++i) + { + ret += texelFetch(tex,tc,i); + } + + return ret/samples; } void main() { - int samples = 4; vec2 tc = vary_fragcoord.xy; ivec2 itc = ivec2(tc); diff --git a/indra/newview/app_settings/shaders/class1/deferred/spotLightMSF.glsl b/indra/newview/app_settings/shaders/class1/deferred/spotLightMSF.glsl new file mode 100644 index 0000000000..3a9d9266bb --- /dev/null +++ b/indra/newview/app_settings/shaders/class1/deferred/spotLightMSF.glsl @@ -0,0 +1,234 @@ +/** + * @file multiSpotLightF.glsl + * + * $LicenseInfo:firstyear=2007&license=viewerlgpl$ + * $/LicenseInfo$ + */ + +#version 120 + +//class 1 -- no shadows + +#extension GL_ARB_texture_rectangle : enable +#extension GL_ARB_texture_multisample : enable + +uniform sampler2DMS diffuseRect; +uniform sampler2DMS specularRect; +uniform sampler2DMS depthMap; +uniform sampler2DMS normalMap; +uniform sampler2D noiseMap; +uniform sampler2D lightFunc; +uniform sampler2D projectionMap; + +uniform mat4 proj_mat; //screen space to light space +uniform float proj_near; //near clip for projection +uniform vec3 proj_p; //plane projection is emitting from (in screen space) +uniform vec3 proj_n; +uniform float proj_focus; //distance from plane to begin blurring +uniform float proj_lod; //(number of mips in proj map) +uniform float proj_range; //range between near clip and far clip plane of projection +uniform float proj_ambient_lod; +uniform float proj_ambiance; +uniform float near_clip; +uniform float far_clip; + +uniform vec3 proj_origin; //origin of projection to be used for angular attenuation +uniform float sun_wash; +uniform int proj_shadow_idx; +uniform float shadow_fade; + +varying vec4 vary_light; + +varying vec4 vary_fragcoord; +uniform vec2 screen_res; + +uniform mat4 inv_proj; + +vec4 texture2DLodSpecular(sampler2D projectionMap, vec2 tc, float lod) +{ + vec4 ret = texture2DLod(projectionMap, tc, lod); + + vec2 dist = tc-vec2(0.5); + + float det = max(1.0-lod/(proj_lod*0.5), 0.0); + + float d = dot(dist,dist); + + ret *= min(clamp((0.25-d)/0.25, 0.0, 1.0)+det, 1.0); + + return ret; +} + +vec4 texture2DLodDiffuse(sampler2D projectionMap, vec2 tc, float lod) +{ + vec4 ret = texture2DLod(projectionMap, tc, lod); + + vec2 dist = vec2(0.5) - abs(tc-vec2(0.5)); + + float det = min(lod/(proj_lod*0.5), 1.0); + + float d = min(dist.x, dist.y); + + float edge = 0.25*det; + + ret *= clamp(d/edge, 0.0, 1.0); + + return ret; +} + +vec4 texture2DLodAmbient(sampler2D projectionMap, vec2 tc, float lod) +{ + vec4 ret = texture2DLod(projectionMap, tc, lod); + + vec2 dist = tc-vec2(0.5); + + float d = dot(dist,dist); + + ret *= min(clamp((0.25-d)/0.25, 0.0, 1.0), 1.0); + + return ret; +} + + +vec4 getPosition(ivec2 pos_screen, int sample) +{ + float depth = texelFetch(depthMap, pos_screen, sample).r; + vec2 sc = vec2(pos_screen.xy)*2.0; + sc /= screen_res; + sc -= vec2(1.0,1.0); + vec4 ndc = vec4(sc.x, sc.y, 2.0*depth-1.0, 1.0); + vec4 pos = inv_proj * ndc; + pos /= pos.w; + pos.w = 1.0; + return pos; +} + +void main() +{ + vec4 frag = vary_fragcoord; + frag.xyz /= frag.w; + frag.xyz = frag.xyz*0.5+0.5; + frag.xy *= screen_res; + ivec2 itc = ivec2(frag.xy); + + vec3 fcol = vec3(0,0,0); + int wght = 0; + + for (int i = 0; i < samples; ++i) + { + vec3 pos = getPosition(itc, i).xyz; + vec3 lv = vary_light.xyz-pos.xyz; + float dist2 = dot(lv,lv); + dist2 /= vary_light.w; + if (dist2 <= 1.0) + { + vec3 norm = texelFetch(normalMap, itc, i).xyz*2.0-1.0; + + norm = normalize(norm); + float l_dist = -dot(lv, proj_n); + + vec4 proj_tc = (proj_mat * vec4(pos.xyz, 1.0)); + if (proj_tc.z >= 0.0) + { + proj_tc.xyz /= proj_tc.w; + + float fa = gl_Color.a+1.0; + float dist_atten = min(1.0-(dist2-1.0*(1.0-fa))/fa, 1.0); + if (dist_atten > 0.0) + { + lv = proj_origin-pos.xyz; + lv = normalize(lv); + float da = dot(norm, lv); + + vec3 col = vec3(0,0,0); + + vec3 diff_tex = texelFetch(diffuseRect, itc, i).rgb; + + float noise = texture2D(noiseMap, frag.xy/128.0).b; + if (proj_tc.z > 0.0 && + proj_tc.x < 1.0 && + proj_tc.y < 1.0 && + proj_tc.x > 0.0 && + proj_tc.y > 0.0) + { + float lit = 0.0; + float amb_da = proj_ambiance; + + if (da > 0.0) + { + float diff = clamp((l_dist-proj_focus)/proj_range, 0.0, 1.0); + float lod = diff * proj_lod; + + vec4 plcol = texture2DLodDiffuse(projectionMap, proj_tc.xy, lod); + + vec3 lcol = gl_Color.rgb * plcol.rgb * plcol.a; + + lit = da * dist_atten * noise; + + col = lcol*lit*diff_tex; + amb_da += (da*0.5)*proj_ambiance; + } + + //float diff = clamp((proj_range-proj_focus)/proj_range, 0.0, 1.0); + vec4 amb_plcol = texture2DLodAmbient(projectionMap, proj_tc.xy, proj_lod); + + amb_da += (da*da*0.5+0.5)*proj_ambiance; + + amb_da *= dist_atten * noise; + + amb_da = min(amb_da, 1.0-lit); + + col += amb_da*gl_Color.rgb*diff_tex.rgb*amb_plcol.rgb*amb_plcol.a; + } + + + vec4 spec = texelFetch(specularRect, itc, i); + if (spec.a > 0.0) + { + vec3 ref = reflect(normalize(pos), norm); + + //project from point pos in direction ref to plane proj_p, proj_n + vec3 pdelta = proj_p-pos; + float ds = dot(ref, proj_n); + + if (ds < 0.0) + { + vec3 pfinal = pos + ref * dot(pdelta, proj_n)/ds; + + vec4 stc = (proj_mat * vec4(pfinal.xyz, 1.0)); + + if (stc.z > 0.0) + { + stc.xy /= stc.w; + + float fatten = clamp(spec.a*spec.a+spec.a*0.5, 0.25, 1.0); + + stc.xy = (stc.xy - vec2(0.5)) * fatten + vec2(0.5); + + if (stc.x < 1.0 && + stc.y < 1.0 && + stc.x > 0.0 && + stc.y > 0.0) + { + vec4 scol = texture2DLodSpecular(projectionMap, stc.xy, proj_lod-spec.a*proj_lod); + col += dist_atten*scol.rgb*gl_Color.rgb*scol.a*spec.rgb; + } + } + } + } + + fcol += col; + ++wght; + } + } + } + } + + if (wght <= 0) + { + discard; + } + + gl_FragColor.rgb = fcol/wght; + gl_FragColor.a = 0.0; +} diff --git a/indra/newview/app_settings/shaders/class1/deferred/starsF.glsl b/indra/newview/app_settings/shaders/class1/deferred/starsF.glsl new file mode 100644 index 0000000000..b1c25b7fd3 --- /dev/null +++ b/indra/newview/app_settings/shaders/class1/deferred/starsF.glsl @@ -0,0 +1,19 @@ +/** + * @file starsF.glsl + * + * $LicenseInfo:firstyear=2007&license=viewerlgpl$ + * $/LicenseInfo$ + */ + +#version 120 + +uniform sampler2D diffuseMap; + +void main() +{ + vec4 col = gl_Color * texture2D(diffuseMap, gl_TexCoord[0].xy); + + gl_FragData[0] = col; + gl_FragData[1] = vec4(0,0,0,0); + gl_FragData[2] = vec4(0,0,1,0); +} diff --git a/indra/newview/app_settings/shaders/class1/deferred/starsV.glsl b/indra/newview/app_settings/shaders/class1/deferred/starsV.glsl new file mode 100644 index 0000000000..1f0376ce6b --- /dev/null +++ b/indra/newview/app_settings/shaders/class1/deferred/starsV.glsl @@ -0,0 +1,17 @@ +/** + * @file starsV.glsl + * + * $LicenseInfo:firstyear=2007&license=viewerlgpl$ + * $/LicenseInfo$ + */ + +#version 120 + + +void main() +{ + //transform vertex + gl_Position = gl_ModelViewProjectionMatrix * gl_Vertex; + gl_TexCoord[0] = gl_TextureMatrix[0] * gl_MultiTexCoord0; + gl_FrontColor = gl_Color; +} diff --git a/indra/newview/app_settings/shaders/class1/deferred/sunLightSSAOF.glsl b/indra/newview/app_settings/shaders/class1/deferred/sunLightSSAOF.glsl index cd91351ad4..d1e8359742 100644 --- a/indra/newview/app_settings/shaders/class1/deferred/sunLightSSAOF.glsl +++ b/indra/newview/app_settings/shaders/class1/deferred/sunLightSSAOF.glsl @@ -35,7 +35,7 @@ uniform float shadow_offset; vec4 getPosition(vec2 pos_screen) { - float depth = texture2DRect(depthMap, pos_screen.xy).a; + float depth = texture2DRect(depthMap, pos_screen.xy).r; vec2 sc = pos_screen.xy*2.0; sc /= screen_res; sc -= vec2(1.0,1.0); diff --git a/indra/newview/app_settings/shaders/class1/deferred/sunLightSSAOMSF.glsl b/indra/newview/app_settings/shaders/class1/deferred/sunLightSSAOMSF.glsl index 49b45dedd6..ce0ebc54f8 100644 --- a/indra/newview/app_settings/shaders/class1/deferred/sunLightSSAOMSF.glsl +++ b/indra/newview/app_settings/shaders/class1/deferred/sunLightSSAOMSF.glsl @@ -101,8 +101,6 @@ float calcAmbientOcclusion(vec4 pos, vec3 norm, int sample) void main() { - int samples = 4; - vec2 pos_screen = vary_fragcoord.xy; ivec2 itc = ivec2(pos_screen); diff --git a/indra/newview/app_settings/shaders/class1/effects/glowExtractMSF.glsl b/indra/newview/app_settings/shaders/class1/effects/glowExtractMSF.glsl new file mode 100644 index 0000000000..c03a0c61bd --- /dev/null +++ b/indra/newview/app_settings/shaders/class1/effects/glowExtractMSF.glsl @@ -0,0 +1,38 @@ +/** + * @file glowExtractF.glsl + * + * $LicenseInfo:firstyear=2007&license=viewerlgpl$ + * $/LicenseInfo$ + */ + +#version 120 + +#extension GL_ARB_texture_rectangle : enable +#extension GL_ARB_texture_multisample : enable + +uniform sampler2DMS diffuseMap; +uniform float minLuminance; +uniform float maxExtractAlpha; +uniform vec3 lumWeights; +uniform vec3 warmthWeights; +uniform float warmthAmount; + +void main() +{ + ivec2 itc = ivec2(gl_TexCoord[0].xy); + vec4 fcol = vec4(0,0,0,0); + + for (int i = 0; i < samples; i++) + { + vec4 col = texelFetch(diffuseMap, itc, i); + + /// CALCULATING LUMINANCE (Using NTSC lum weights) + /// http://en.wikipedia.org/wiki/Luma_%28video%29 + float lum = smoothstep(minLuminance, minLuminance+1.0, dot(col.rgb, lumWeights ) ); + float warmth = smoothstep(minLuminance, minLuminance+1.0, max(col.r * warmthWeights.r, max(col.g * warmthWeights.g, col.b * warmthWeights.b)) ); + + fcol += vec4(col.rgb, max(col.a, mix(lum, warmth, warmthAmount) * maxExtractAlpha)); + } + + gl_FragColor = fcol/samples; +} diff --git a/indra/newview/app_settings/shaders/class2/deferred/edgeF.glsl b/indra/newview/app_settings/shaders/class2/deferred/edgeF.glsl index 3155f3f929..30e1702e9f 100644 --- a/indra/newview/app_settings/shaders/class2/deferred/edgeF.glsl +++ b/indra/newview/app_settings/shaders/class2/deferred/edgeF.glsl @@ -22,7 +22,7 @@ uniform vec2 screen_res; float getDepth(vec2 pos_screen) { - float z = texture2DRect(depthMap, pos_screen.xy).a; + float z = texture2DRect(depthMap, pos_screen.xy).r; z = z*2.0-1.0; vec4 ndc = vec4(0.0, 0.0, z, 1.0); vec4 p = inv_proj*ndc; diff --git a/indra/newview/app_settings/shaders/class2/deferred/edgeMSF.glsl b/indra/newview/app_settings/shaders/class2/deferred/edgeMSF.glsl index a47ee678f6..38100d1523 100644 --- a/indra/newview/app_settings/shaders/class2/deferred/edgeMSF.glsl +++ b/indra/newview/app_settings/shaders/class2/deferred/edgeMSF.glsl @@ -21,8 +21,9 @@ uniform float norm_cutoff; uniform mat4 inv_proj; uniform vec2 screen_res; -float getDepth(float z) +float getDepth(ivec2 pos_screen, int sample) { + float z = texelFetch(depthMap, pos_screen, sample).r; z = z*2.0-1.0; vec4 ndc = vec4(0.0, 0.0, z, 1.0); vec4 p = inv_proj*ndc; @@ -31,7 +32,6 @@ float getDepth(float z) void main() { - float e = 0; ivec2 itc = ivec2(vary_fragcoord.xy); @@ -40,24 +40,24 @@ void main() { vec3 norm = texelFetch(normalMap, itc, i).xyz; norm = vec3((norm.xy-0.5)*2.0,norm.z); // unpack norm - float depth = getDepth(texelFetch(depthMap, itc, i).r); + float depth = getDepth(itc, i); vec2 tc = vary_fragcoord.xy; - float sc = 0.75; + int sc = 1; vec2 de; - de.x = (depth-getDepth(tc+vec2(sc, sc))) + (depth-getDepth(tc+vec2(-sc, -sc))); - de.y = (depth-getDepth(tc+vec2(-sc, sc))) + (depth-getDepth(tc+vec2(sc, -sc))); + de.x = (depth-getDepth(itc+ivec2(sc, sc),i)) + (depth-getDepth(itc+ivec2(-sc, -sc), i)); + de.y = (depth-getDepth(itc+ivec2(-sc, sc),i)) + (depth-getDepth(itc+ivec2(sc, -sc), i)); de /= depth; de *= de; de = step(depth_cutoff, de); vec2 ne; - vec3 nexnorm = texture2DRect(normalMap, tc+vec2(-sc,-sc)).rgb; + vec3 nexnorm = texelFetch(normalMap, itc+ivec2(-sc,-sc), i).rgb; nexnorm = vec3((nexnorm.xy-0.5)*2.0,nexnorm.z); // unpack norm ne.x = dot(nexnorm, norm); - vec3 neynorm = texture2DRect(normalMap, tc+vec2(sc,sc)).rgb; + vec3 neynorm = texelFetch(normalMap, itc+ivec2(sc,sc), i).rgb; neynorm = vec3((neynorm.xy-0.5)*2.0,neynorm.z); // unpack norm ne.y = dot(neynorm, norm); diff --git a/indra/newview/app_settings/shaders/class2/deferred/multiSpotLightF.glsl b/indra/newview/app_settings/shaders/class2/deferred/multiSpotLightF.glsl index d6cd984ebe..de987b1233 100644 --- a/indra/newview/app_settings/shaders/class2/deferred/multiSpotLightF.glsl +++ b/indra/newview/app_settings/shaders/class2/deferred/multiSpotLightF.glsl @@ -91,7 +91,7 @@ vec4 texture2DLodAmbient(sampler2D projectionMap, vec2 tc, float lod) vec4 getPosition(vec2 pos_screen) { - float depth = texture2DRect(depthMap, pos_screen.xy).a; + float depth = texture2DRect(depthMap, pos_screen.xy).r; vec2 sc = pos_screen.xy*2.0; sc /= screen_res; sc -= vec2(1.0,1.0); diff --git a/indra/newview/app_settings/shaders/class2/deferred/multiSpotLightMSF.glsl b/indra/newview/app_settings/shaders/class2/deferred/multiSpotLightMSF.glsl new file mode 100644 index 0000000000..c6f0ea2a5d --- /dev/null +++ b/indra/newview/app_settings/shaders/class2/deferred/multiSpotLightMSF.glsl @@ -0,0 +1,244 @@ +/** + * @file multiSpotLightF.glsl + * + * $LicenseInfo:firstyear=2007&license=viewerlgpl$ + * $/LicenseInfo$ + */ + +#version 120 + +#extension GL_ARB_texture_rectangle : enable +#extension GL_ARB_texture_multisample : enable + +uniform sampler2DMS diffuseRect; +uniform sampler2DMS specularRect; +uniform sampler2DMS depthMap; +uniform sampler2DMS normalMap; +uniform sampler2DMS lightMap; +uniform sampler2D noiseMap; +uniform sampler2D lightFunc; +uniform sampler2D projectionMap; + +uniform mat4 proj_mat; //screen space to light space +uniform float proj_near; //near clip for projection +uniform vec3 proj_p; //plane projection is emitting from (in screen space) +uniform vec3 proj_n; +uniform float proj_focus; //distance from plane to begin blurring +uniform float proj_lod; //(number of mips in proj map) +uniform float proj_range; //range between near clip and far clip plane of projection +uniform float proj_ambient_lod; +uniform float proj_ambiance; +uniform float near_clip; +uniform float far_clip; + +uniform vec3 proj_origin; //origin of projection to be used for angular attenuation +uniform float sun_wash; +uniform int proj_shadow_idx; +uniform float shadow_fade; + +varying vec4 vary_light; + +varying vec4 vary_fragcoord; +uniform vec2 screen_res; + +uniform mat4 inv_proj; + +vec4 texture2DLodSpecular(sampler2D projectionMap, vec2 tc, float lod) +{ + vec4 ret = texture2DLod(projectionMap, tc, lod); + + vec2 dist = tc-vec2(0.5); + + float det = max(1.0-lod/(proj_lod*0.5), 0.0); + + float d = dot(dist,dist); + + ret *= min(clamp((0.25-d)/0.25, 0.0, 1.0)+det, 1.0); + + return ret; +} + +vec4 texture2DLodDiffuse(sampler2D projectionMap, vec2 tc, float lod) +{ + vec4 ret = texture2DLod(projectionMap, tc, lod); + + vec2 dist = vec2(0.5) - abs(tc-vec2(0.5)); + + float det = min(lod/(proj_lod*0.5), 1.0); + + float d = min(dist.x, dist.y); + + float edge = 0.25*det; + + ret *= clamp(d/edge, 0.0, 1.0); + + return ret; +} + +vec4 texture2DLodAmbient(sampler2D projectionMap, vec2 tc, float lod) +{ + vec4 ret = texture2DLod(projectionMap, tc, lod); + + vec2 dist = tc-vec2(0.5); + + float d = dot(dist,dist); + + ret *= min(clamp((0.25-d)/0.25, 0.0, 1.0), 1.0); + + return ret; +} + + +vec4 getPosition(ivec2 pos_screen, int sample) +{ + float depth = texelFetch(depthMap, pos_screen, sample).r; + vec2 sc = vec2(pos_screen.xy)*2.0; + sc /= screen_res; + sc -= vec2(1.0,1.0); + vec4 ndc = vec4(sc.x, sc.y, 2.0*depth-1.0, 1.0); + vec4 pos = inv_proj * ndc; + pos /= pos.w; + pos.w = 1.0; + return pos; +} + +void main() +{ + int wght = 0; + + vec3 fcol = vec3(0,0,0); + + vec2 frag = (vary_fragcoord.xy*0.5+0.5)*screen_res; + + ivec2 itc = ivec2(frag.xy); + + for (int i = 0; i < samples; i++) + { + vec3 pos = getPosition(itc, i).xyz; + vec3 lv = vary_light.xyz-pos.xyz; + float dist2 = dot(lv,lv); + dist2 /= vary_light.w; + if (dist2 <= 1.0) + { + float shadow = 1.0; + + if (proj_shadow_idx >= 0) + { + vec4 shd = texelFetch(lightMap, itc, i); + float sh[2]; + sh[0] = shd.b; + sh[1] = shd.a; + shadow = min(sh[proj_shadow_idx]+shadow_fade, 1.0); + } + + vec3 norm = texelFetch(normalMap, itc, i).xyz; + norm = vec3((norm.xy-0.5)*2.0,norm.z); // unpack norm + + norm = normalize(norm); + float l_dist = -dot(lv, proj_n); + + vec4 proj_tc = (proj_mat * vec4(pos.xyz, 1.0)); + if (proj_tc.z >= 0.0) + { + proj_tc.xyz /= proj_tc.w; + + float fa = gl_Color.a+1.0; + float dist_atten = min(1.0-(dist2-1.0*(1.0-fa))/fa, 1.0); + if (dist_atten > 0.0) + { + lv = proj_origin-pos.xyz; + lv = normalize(lv); + float da = dot(norm, lv); + + vec3 col = vec3(0,0,0); + + vec3 diff_tex = texelFetch(diffuseRect, itc, i).rgb; + + float noise = texture2D(noiseMap, frag.xy/128.0).b; + if (proj_tc.z > 0.0 && + proj_tc.x < 1.0 && + proj_tc.y < 1.0 && + proj_tc.x > 0.0 && + proj_tc.y > 0.0) + { + float lit = 0.0; + float amb_da = proj_ambiance; + + if (da > 0.0) + { + float diff = clamp((l_dist-proj_focus)/proj_range, 0.0, 1.0); + float lod = diff * proj_lod; + + vec4 plcol = texture2DLodDiffuse(projectionMap, proj_tc.xy, lod); + + vec3 lcol = gl_Color.rgb * plcol.rgb * plcol.a; + + lit = da * dist_atten * noise; + + col = lcol*lit*diff_tex*shadow; + amb_da += (da*0.5)*(1.0-shadow)*proj_ambiance; + } + + //float diff = clamp((proj_range-proj_focus)/proj_range, 0.0, 1.0); + vec4 amb_plcol = texture2DLodAmbient(projectionMap, proj_tc.xy, proj_lod); + + amb_da += (da*da*0.5+0.5)*proj_ambiance; + + amb_da *= dist_atten * noise; + + amb_da = min(amb_da, 1.0-lit); + + col += amb_da*gl_Color.rgb*diff_tex.rgb*amb_plcol.rgb*amb_plcol.a; + } + + + vec4 spec = texelFetch(specularRect, itc, i); + if (spec.a > 0.0) + { + vec3 ref = reflect(normalize(pos), norm); + + //project from point pos in direction ref to plane proj_p, proj_n + vec3 pdelta = proj_p-pos; + float ds = dot(ref, proj_n); + + if (ds < 0.0) + { + vec3 pfinal = pos + ref * dot(pdelta, proj_n)/ds; + + vec4 stc = (proj_mat * vec4(pfinal.xyz, 1.0)); + + if (stc.z > 0.0) + { + stc.xy /= stc.w; + + float fatten = clamp(spec.a*spec.a+spec.a*0.5, 0.25, 1.0); + + stc.xy = (stc.xy - vec2(0.5)) * fatten + vec2(0.5); + + if (stc.x < 1.0 && + stc.y < 1.0 && + stc.x > 0.0 && + stc.y > 0.0) + { + vec4 scol = texture2DLodSpecular(projectionMap, stc.xy, proj_lod-spec.a*proj_lod); + col += dist_atten*scol.rgb*gl_Color.rgb*scol.a*spec.rgb*shadow; + } + } + } + } + + fcol += col; + wght++; + } + } + } + } + + if (wght <= 0) + { + discard; + } + + gl_FragColor.rgb = fcol/wght; + gl_FragColor.a = 0.0; +} diff --git a/indra/newview/app_settings/shaders/class2/deferred/softenLightF.glsl b/indra/newview/app_settings/shaders/class2/deferred/softenLightF.glsl index 0160e84278..dfa1964142 100644 --- a/indra/newview/app_settings/shaders/class2/deferred/softenLightF.glsl +++ b/indra/newview/app_settings/shaders/class2/deferred/softenLightF.glsl @@ -71,7 +71,7 @@ vec4 getPosition_d(vec2 pos_screen, float depth) vec4 getPosition(vec2 pos_screen) { //get position in screen space (world units) given window coordinate and depth map - float depth = texture2DRect(depthMap, pos_screen.xy).a; + float depth = texture2DRect(depthMap, pos_screen.xy).r; return getPosition_d(pos_screen, depth); } @@ -258,7 +258,7 @@ vec3 scaleSoftClip(vec3 light) void main() { vec2 tc = vary_fragcoord.xy; - float depth = texture2DRect(depthMap, tc.xy).a; + float depth = texture2DRect(depthMap, tc.xy).r; vec3 pos = getPosition_d(tc, depth).xyz; vec3 norm = texture2DRect(normalMap, tc).xyz; norm = vec3((norm.xy-0.5)*2.0,norm.z); // unpack norm @@ -288,54 +288,8 @@ void main() float sa = dot(refnormpersp, vary_light.xyz); vec3 dumbshiny = vary_SunlitColor*scol_ambocc.r*texture2D(lightFunc, vec2(sa, spec.a)).a; - /* - // screen-space cheap fakey reflection map - // - vec3 refnorm = normalize(reflect(vec3(0,0,-1), norm.xyz)); - depth -= 0.5; // unbias depth - // first figure out where we'll make our 2D guess from - vec2 ref2d = (0.25 * screen_res.y) * (refnorm.xy) * abs(refnorm.z) / depth; - // Offset the guess source a little according to a trivial - // checkerboard dither function and spec.a. - // This is meant to be similar to sampling a blurred version - // of the diffuse map. LOD would be better in that regard. - // The goal of the blur is to soften reflections in surfaces - // with low shinyness, and also to disguise our lameness. - float checkerboard = floor(mod(tc.x+tc.y, 2.0)); // 0.0, 1.0 - float checkoffset = (3.0 + (7.0*(1.0-spec.a)))*(checkerboard-0.5); - ref2d += vec2(checkoffset, checkoffset); - ref2d += tc.xy; // use as offset from destination - // Get attributes from the 2D guess point. - // We average two samples of diffuse (not of anything else) per - // pixel to try to reduce aliasing some more. - vec3 refcol = 0.5 * (texture2DRect(diffuseRect, ref2d + vec2(0.0, -checkoffset)).rgb + - texture2DRect(diffuseRect, ref2d + vec2(-checkoffset, 0.0)).rgb); - float refdepth = texture2DRect(depthMap, ref2d).a; - vec3 refpos = getPosition_d(ref2d, refdepth).xyz; - float refshad = texture2DRect(lightMap, ref2d).r; - vec3 refn = texture2DRect(normalMap, ref2d).rgb; - refn = vec3((refn.xy-0.5)*2.0,refn.z); // unpack norm - refn = normalize(refn); - // figure out how appropriate our guess actually was - float refapprop = max(0.0, dot(-refnorm, normalize(pos - refpos))); - // darken reflections from points which face away from the reflected ray - our guess was a back-face - //refapprop *= step(dot(refnorm, refn), 0.0); - refapprop = min(refapprop, max(0.0, -dot(refnorm, refn))); // more conservative variant - // get appropriate light strength for guess-point - // reflect light direction to increase the illusion that - // these are reflections. - vec3 reflight = reflect(lightnorm.xyz, norm.xyz); - float reflit = min(max(dot(refn, reflight.xyz), 0.0), refshad); - // apply sun color to guess-point, dampen according to inappropriateness of guess - float refmod = min(refapprop, reflit); - vec3 refprod = vary_SunlitColor * refcol.rgb * refmod; - vec3 ssshiny = (refprod * spec.a); - ssshiny *= 0.3; // dampen it even more - */ - vec3 ssshiny = vec3(0,0,0); - // add the two types of shiny together - col += (ssshiny + dumbshiny) * spec.rgb; + col += dumbshiny * spec.rgb; } col = atmosLighting(col); diff --git a/indra/newview/app_settings/shaders/class2/deferred/softenLightMSF.glsl b/indra/newview/app_settings/shaders/class2/deferred/softenLightMSF.glsl index 7d4548d6d9..71482d0c7b 100644 --- a/indra/newview/app_settings/shaders/class2/deferred/softenLightMSF.glsl +++ b/indra/newview/app_settings/shaders/class2/deferred/softenLightMSF.glsl @@ -252,7 +252,6 @@ vec3 scaleSoftClip(vec3 light) void main() { - int samples = 4; vec2 tc = vary_fragcoord.xy; ivec2 itc = ivec2(tc); diff --git a/indra/newview/app_settings/shaders/class2/deferred/spotLightMSF.glsl b/indra/newview/app_settings/shaders/class2/deferred/spotLightMSF.glsl new file mode 100644 index 0000000000..099a45718a --- /dev/null +++ b/indra/newview/app_settings/shaders/class2/deferred/spotLightMSF.glsl @@ -0,0 +1,245 @@ +/** + * @file multiSpotLightF.glsl + * + * $LicenseInfo:firstyear=2007&license=viewerlgpl$ + * $/LicenseInfo$ + */ + +#version 120 + +#extension GL_ARB_texture_rectangle : enable +#extension GL_ARB_texture_multisample : enable + +uniform sampler2DMS diffuseRect; +uniform sampler2DMS specularRect; +uniform sampler2DMS depthMap; +uniform sampler2DMS normalMap; +uniform sampler2DMS lightMap; +uniform sampler2D noiseMap; +uniform sampler2D lightFunc; +uniform sampler2D projectionMap; + +uniform mat4 proj_mat; //screen space to light space +uniform float proj_near; //near clip for projection +uniform vec3 proj_p; //plane projection is emitting from (in screen space) +uniform vec3 proj_n; +uniform float proj_focus; //distance from plane to begin blurring +uniform float proj_lod; //(number of mips in proj map) +uniform float proj_range; //range between near clip and far clip plane of projection +uniform float proj_ambient_lod; +uniform float proj_ambiance; +uniform float near_clip; +uniform float far_clip; + +uniform vec3 proj_origin; //origin of projection to be used for angular attenuation +uniform float sun_wash; +uniform int proj_shadow_idx; +uniform float shadow_fade; + +varying vec4 vary_light; + +varying vec4 vary_fragcoord; +uniform vec2 screen_res; + +uniform mat4 inv_proj; + +vec4 texture2DLodSpecular(sampler2D projectionMap, vec2 tc, float lod) +{ + vec4 ret = texture2DLod(projectionMap, tc, lod); + + vec2 dist = tc-vec2(0.5); + + float det = max(1.0-lod/(proj_lod*0.5), 0.0); + + float d = dot(dist,dist); + + ret *= min(clamp((0.25-d)/0.25, 0.0, 1.0)+det, 1.0); + + return ret; +} + +vec4 texture2DLodDiffuse(sampler2D projectionMap, vec2 tc, float lod) +{ + vec4 ret = texture2DLod(projectionMap, tc, lod); + + vec2 dist = vec2(0.5) - abs(tc-vec2(0.5)); + + float det = min(lod/(proj_lod*0.5), 1.0); + + float d = min(dist.x, dist.y); + + float edge = 0.25*det; + + ret *= clamp(d/edge, 0.0, 1.0); + + return ret; +} + +vec4 texture2DLodAmbient(sampler2D projectionMap, vec2 tc, float lod) +{ + vec4 ret = texture2DLod(projectionMap, tc, lod); + + vec2 dist = tc-vec2(0.5); + + float d = dot(dist,dist); + + ret *= min(clamp((0.25-d)/0.25, 0.0, 1.0), 1.0); + + return ret; +} + + +vec4 getPosition(ivec2 pos_screen, int sample) +{ + float depth = texelFetch(depthMap, pos_screen, sample).r; + vec2 sc = vec2(pos_screen.xy)*2.0; + sc /= screen_res; + sc -= vec2(1.0,1.0); + vec4 ndc = vec4(sc.x, sc.y, 2.0*depth-1.0, 1.0); + vec4 pos = inv_proj * ndc; + pos /= pos.w; + pos.w = 1.0; + return pos; +} + +void main() +{ + vec4 frag = vary_fragcoord; + frag.xyz /= frag.w; + frag.xyz = frag.xyz*0.5+0.5; + frag.xy *= screen_res; + ivec2 itc = ivec2(frag.xy); + + vec3 fcol = vec3(0,0,0); + int wght = 0; + + for (int i = 0; i < samples; i++) + { + vec3 pos = getPosition(itc, i).xyz; + vec3 lv = vary_light.xyz-pos.xyz; + float dist2 = dot(lv,lv); + dist2 /= vary_light.w; + if (dist2 <= 1.0) + { + float shadow = 1.0; + + if (proj_shadow_idx >= 0) + { + vec4 shd = texelFetch(lightMap, itc, i); + float sh[2]; + sh[0] = shd.b; + sh[1] = shd.a; + shadow = min(sh[proj_shadow_idx]+shadow_fade, 1.0); + } + + vec3 norm = texelFetch(normalMap, itc, i).xyz; + norm = vec3((norm.xy-0.5)*2.0,norm.z); // unpack norm + + norm = normalize(norm); + float l_dist = -dot(lv, proj_n); + + vec4 proj_tc = (proj_mat * vec4(pos.xyz, 1.0)); + if (proj_tc.z >= 0.0) + { + proj_tc.xyz /= proj_tc.w; + + float fa = gl_Color.a+1.0; + float dist_atten = min(1.0-(dist2-1.0*(1.0-fa))/fa, 1.0); + if (dist_atten > 0.0) + { + lv = proj_origin-pos.xyz; + lv = normalize(lv); + float da = dot(norm, lv); + + vec3 col = vec3(0,0,0); + + vec3 diff_tex = texelFetch(diffuseRect, itc, i).rgb; + + float noise = texture2D(noiseMap, frag.xy/128.0).b; + if (proj_tc.z > 0.0 && + proj_tc.x < 1.0 && + proj_tc.y < 1.0 && + proj_tc.x > 0.0 && + proj_tc.y > 0.0) + { + float lit = 0.0; + float amb_da = proj_ambiance; + + if (da > 0.0) + { + float diff = clamp((l_dist-proj_focus)/proj_range, 0.0, 1.0); + float lod = diff * proj_lod; + + vec4 plcol = texture2DLodDiffuse(projectionMap, proj_tc.xy, lod); + + vec3 lcol = gl_Color.rgb * plcol.rgb * plcol.a; + + lit = da * dist_atten * noise; + + col = lcol*lit*diff_tex*shadow; + amb_da += (da*0.5)*(1.0-shadow)*proj_ambiance; + } + + //float diff = clamp((proj_range-proj_focus)/proj_range, 0.0, 1.0); + vec4 amb_plcol = texture2DLodAmbient(projectionMap, proj_tc.xy, proj_lod); + + amb_da += (da*da*0.5+0.5)*proj_ambiance; + + amb_da *= dist_atten * noise; + + amb_da = min(amb_da, 1.0-lit); + + col += amb_da*gl_Color.rgb*diff_tex.rgb*amb_plcol.rgb*amb_plcol.a; + } + + + vec4 spec = texelFetch(specularRect, itc, i); + if (spec.a > 0.0) + { + vec3 ref = reflect(normalize(pos), norm); + + //project from point pos in direction ref to plane proj_p, proj_n + vec3 pdelta = proj_p-pos; + float ds = dot(ref, proj_n); + + if (ds < 0.0) + { + vec3 pfinal = pos + ref * dot(pdelta, proj_n)/ds; + + vec4 stc = (proj_mat * vec4(pfinal.xyz, 1.0)); + + if (stc.z > 0.0) + { + stc.xy /= stc.w; + + float fatten = clamp(spec.a*spec.a+spec.a*0.5, 0.25, 1.0); + + stc.xy = (stc.xy - vec2(0.5)) * fatten + vec2(0.5); + + if (stc.x < 1.0 && + stc.y < 1.0 && + stc.x > 0.0 && + stc.y > 0.0) + { + vec4 scol = texture2DLodSpecular(projectionMap, stc.xy, proj_lod-spec.a*proj_lod); + col += dist_atten*scol.rgb*gl_Color.rgb*scol.a*spec.rgb*shadow; + } + } + } + } + + fcol += col; + wght++; + } + } + } + } + + if (wght <= 0) + { + discard; + } + + gl_FragColor.rgb = fcol/wght; + gl_FragColor.a = 0.0; +} diff --git a/indra/newview/app_settings/shaders/class2/deferred/sunLightF.glsl b/indra/newview/app_settings/shaders/class2/deferred/sunLightF.glsl index 4369b3b34f..b724c134b9 100644 --- a/indra/newview/app_settings/shaders/class2/deferred/sunLightF.glsl +++ b/indra/newview/app_settings/shaders/class2/deferred/sunLightF.glsl @@ -45,7 +45,7 @@ uniform float spot_shadow_offset; vec4 getPosition(vec2 pos_screen) { - float depth = texture2DRect(depthMap, pos_screen.xy).a; + float depth = texture2DRect(depthMap, pos_screen.xy).r; vec2 sc = pos_screen.xy*2.0; sc /= screen_res; sc -= vec2(1.0,1.0); diff --git a/indra/newview/app_settings/shaders/class2/deferred/sunLightMSF.glsl b/indra/newview/app_settings/shaders/class2/deferred/sunLightMSF.glsl new file mode 100644 index 0000000000..dd6fa958c9 --- /dev/null +++ b/indra/newview/app_settings/shaders/class2/deferred/sunLightMSF.glsl @@ -0,0 +1,202 @@ +/** + * @file sunLightMSF.glsl + * + * $LicenseInfo:firstyear=2007&license=viewerlgpl$ + * $/LicenseInfo$ + */ + +#version 120 + +#extension GL_ARB_texture_rectangle : enable +#extension GL_ARB_texture_multisample : enable + +//class 2, shadows, no SSAO + +uniform sampler2DMS depthMap; +uniform sampler2DMS normalMap; +uniform sampler2DRectShadow shadowMap0; +uniform sampler2DRectShadow shadowMap1; +uniform sampler2DRectShadow shadowMap2; +uniform sampler2DRectShadow shadowMap3; +uniform sampler2DShadow shadowMap4; +uniform sampler2DShadow shadowMap5; + + +// Inputs +uniform mat4 shadow_matrix[6]; +uniform vec4 shadow_clip; +uniform float ssao_radius; +uniform float ssao_max_radius; +uniform float ssao_factor; +uniform float ssao_factor_inv; + +varying vec2 vary_fragcoord; +varying vec4 vary_light; + +uniform mat4 inv_proj; +uniform vec2 screen_res; +uniform vec2 shadow_res; +uniform vec2 proj_shadow_res; + +uniform float shadow_bias; +uniform float shadow_offset; + +uniform float spot_shadow_bias; +uniform float spot_shadow_offset; + +vec4 getPosition(ivec2 pos_screen, int sample) +{ + float depth = texelFetch(depthMap, pos_screen.xy, sample).r; + vec2 sc = vec2(pos_screen.xy)*2.0; + sc /= screen_res; + sc -= vec2(1.0,1.0); + vec4 ndc = vec4(sc.x, sc.y, 2.0*depth-1.0, 1.0); + vec4 pos = inv_proj * ndc; + pos /= pos.w; + pos.w = 1.0; + return pos; +} + +float pcfShadow(sampler2DRectShadow shadowMap, vec4 stc, float scl) +{ + stc.xyz /= stc.w; + stc.z += shadow_bias*scl; + + float cs = shadow2DRect(shadowMap, stc.xyz).x; + float shadow = cs; + + shadow += max(shadow2DRect(shadowMap, stc.xyz+vec3(1.5, 1.5, 0.0)).x, cs); + shadow += max(shadow2DRect(shadowMap, stc.xyz+vec3(1.5, -1.5, 0.0)).x, cs); + shadow += max(shadow2DRect(shadowMap, stc.xyz+vec3(-1.5, 1.5, 0.0)).x, cs); + shadow += max(shadow2DRect(shadowMap, stc.xyz+vec3(-1.5, -1.5, 0.0)).x, cs); + + return shadow/5.0; + + //return shadow; +} + +float pcfShadow(sampler2DShadow shadowMap, vec4 stc, float scl) +{ + stc.xyz /= stc.w; + stc.z += spot_shadow_bias*scl; + + float cs = shadow2D(shadowMap, stc.xyz).x; + float shadow = cs; + + vec2 off = 1.5/proj_shadow_res; + + shadow += max(shadow2D(shadowMap, stc.xyz+vec3(off.x, off.y, 0.0)).x, cs); + shadow += max(shadow2D(shadowMap, stc.xyz+vec3(off.x, -off.y, 0.0)).x, cs); + shadow += max(shadow2D(shadowMap, stc.xyz+vec3(-off.x, off.y, 0.0)).x, cs); + shadow += max(shadow2D(shadowMap, stc.xyz+vec3(-off.x, -off.y, 0.0)).x, cs); + + return shadow/5.0; + + //return shadow; +} + +void main() +{ + vec2 pos_screen = vary_fragcoord.xy; + ivec2 itc = ivec2(pos_screen); + + //try doing an unproject here + + vec4 fcol = vec4(0,0,0,0); + + for (int i = 0; i < samples; i++) + { + vec4 pos = getPosition(itc, i); + + vec4 nmap4 = texelFetch(normalMap, itc, i); + nmap4 = vec4((nmap4.xy-0.5)*2.0,nmap4.z,nmap4.w); // unpack norm + float displace = nmap4.w; + vec3 norm = nmap4.xyz; + + /*if (pos.z == 0.0) // do nothing for sky *FIX: REMOVE THIS IF/WHEN THE POSITION MAP IS BEING USED AS A STENCIL + { + gl_FragColor = vec4(0.0); // doesn't matter + return; + }*/ + + float shadow = 1.0; + float dp_directional_light = max(0.0, dot(norm, vary_light.xyz)); + + vec3 shadow_pos = pos.xyz + displace*norm; + vec3 offset = vary_light.xyz * (1.0-dp_directional_light); + + vec4 spos = vec4(shadow_pos+offset*shadow_offset, 1.0); + + if (spos.z > -shadow_clip.w) + { + if (dp_directional_light == 0.0) + { + // if we know this point is facing away from the sun then we know it's in shadow without having to do a squirrelly shadow-map lookup + shadow = 0.0; + } + else + { + vec4 lpos; + + if (spos.z < -shadow_clip.z) + { + lpos = shadow_matrix[3]*spos; + lpos.xy *= shadow_res; + shadow = pcfShadow(shadowMap3, lpos, 0.25); + shadow += max((pos.z+shadow_clip.z)/(shadow_clip.z-shadow_clip.w)*2.0-1.0, 0.0); + } + else if (spos.z < -shadow_clip.y) + { + lpos = shadow_matrix[2]*spos; + lpos.xy *= shadow_res; + shadow = pcfShadow(shadowMap2, lpos, 0.5); + } + else if (spos.z < -shadow_clip.x) + { + lpos = shadow_matrix[1]*spos; + lpos.xy *= shadow_res; + shadow = pcfShadow(shadowMap1, lpos, 0.75); + } + else + { + lpos = shadow_matrix[0]*spos; + lpos.xy *= shadow_res; + shadow = pcfShadow(shadowMap0, lpos, 1.0); + } + + // take the most-shadowed value out of these two: + // * the blurred sun shadow in the light (shadow) map + // * an unblurred dot product between the sun and this norm + // the goal is to err on the side of most-shadow to fill-in shadow holes and reduce artifacting + shadow = min(shadow, dp_directional_light); + + //lpos.xy /= lpos.w*32.0; + //if (fract(lpos.x) < 0.1 || fract(lpos.y) < 0.1) + //{ + // shadow = 0.0; + //} + + } + } + else + { + // more distant than the shadow map covers + shadow = 1.0; + } + + fcol[0] += shadow; + fcol[1] += 1.0; + + spos = vec4(shadow_pos+norm*spot_shadow_offset, 1.0); + + //spotlight shadow 1 + vec4 lpos = shadow_matrix[4]*spos; + fcol[2] += pcfShadow(shadowMap4, lpos, 0.8); + + //spotlight shadow 2 + lpos = shadow_matrix[5]*spos; + fcol[3] += pcfShadow(shadowMap5, lpos, 0.8); + } + + gl_FragColor = fcol/samples; +} diff --git a/indra/newview/app_settings/shaders/class2/deferred/sunLightSSAOF.glsl b/indra/newview/app_settings/shaders/class2/deferred/sunLightSSAOF.glsl index 847b36b1ac..68e18162f6 100644 --- a/indra/newview/app_settings/shaders/class2/deferred/sunLightSSAOF.glsl +++ b/indra/newview/app_settings/shaders/class2/deferred/sunLightSSAOF.glsl @@ -45,7 +45,7 @@ uniform float spot_shadow_offset; vec4 getPosition(vec2 pos_screen) { - float depth = texture2DRect(depthMap, pos_screen.xy).a; + float depth = texture2DRect(depthMap, pos_screen.xy).r; vec2 sc = pos_screen.xy*2.0; sc /= screen_res; sc -= vec2(1.0,1.0); diff --git a/indra/newview/app_settings/shaders/class2/deferred/sunLightSSAOMSF.glsl b/indra/newview/app_settings/shaders/class2/deferred/sunLightSSAOMSF.glsl new file mode 100644 index 0000000000..d28741f945 --- /dev/null +++ b/indra/newview/app_settings/shaders/class2/deferred/sunLightSSAOMSF.glsl @@ -0,0 +1,241 @@ +/** + * @file sunLightSSAOF.glsl + * + * Copyright (c) 2007-$CurrentYear$, Linden Research, Inc. + * $License$ + */ + +#version 120 + +#extension GL_ARB_texture_rectangle : enable +#extension GL_ARB_texture_multisample : enable + +//class 2 -- shadows and SSAO + +uniform sampler2DMS depthMap; +uniform sampler2DMS normalMap; +uniform sampler2DRectShadow shadowMap0; +uniform sampler2DRectShadow shadowMap1; +uniform sampler2DRectShadow shadowMap2; +uniform sampler2DRectShadow shadowMap3; +uniform sampler2DShadow shadowMap4; +uniform sampler2DShadow shadowMap5; +uniform sampler2D noiseMap; + +// Inputs +uniform mat4 shadow_matrix[6]; +uniform vec4 shadow_clip; +uniform float ssao_radius; +uniform float ssao_max_radius; +uniform float ssao_factor; +uniform float ssao_factor_inv; + +varying vec2 vary_fragcoord; +varying vec4 vary_light; + +uniform mat4 inv_proj; +uniform vec2 screen_res; +uniform vec2 shadow_res; +uniform vec2 proj_shadow_res; + +uniform float shadow_bias; +uniform float shadow_offset; + +uniform float spot_shadow_bias; +uniform float spot_shadow_offset; + +vec4 getPosition(ivec2 pos_screen, int sample) +{ + float depth = texelFetch(depthMap, pos_screen, sample).r; + vec2 sc = vec2(pos_screen.xy)*2.0; + sc /= screen_res; + sc -= vec2(1.0,1.0); + vec4 ndc = vec4(sc.x, sc.y, 2.0*depth-1.0, 1.0); + vec4 pos = inv_proj * ndc; + pos /= pos.w; + pos.w = 1.0; + return pos; +} + +//calculate decreases in ambient lighting when crowded out (SSAO) +float calcAmbientOcclusion(vec4 pos, vec3 norm, int sample) +{ + float ret = 1.0; + + vec2 kern[8]; + // exponentially (^2) distant occlusion samples spread around origin + kern[0] = vec2(-1.0, 0.0) * 0.125*0.125; + kern[1] = vec2(1.0, 0.0) * 0.250*0.250; + kern[2] = vec2(0.0, 1.0) * 0.375*0.375; + kern[3] = vec2(0.0, -1.0) * 0.500*0.500; + kern[4] = vec2(0.7071, 0.7071) * 0.625*0.625; + kern[5] = vec2(-0.7071, -0.7071) * 0.750*0.750; + kern[6] = vec2(-0.7071, 0.7071) * 0.875*0.875; + kern[7] = vec2(0.7071, -0.7071) * 1.000*1.000; + + vec2 pos_screen = vary_fragcoord.xy; + vec3 pos_world = pos.xyz; + vec2 noise_reflect = texture2D(noiseMap, vary_fragcoord.xy/128.0).xy; + + float angle_hidden = 0.0; + int points = 0; + + float scale = min(ssao_radius / -pos_world.z, ssao_max_radius); + + // it was found that keeping # of samples a constant was the fastest, probably due to compiler optimizations (unrolling?) + for (int i = 0; i < 8; i++) + { + ivec2 samppos_screen = ivec2(pos_screen + scale * reflect(kern[i], noise_reflect)); + vec3 samppos_world = getPosition(samppos_screen, sample).xyz; + + vec3 diff = pos_world - samppos_world; + float dist2 = dot(diff, diff); + + // assume each sample corresponds to an occluding sphere with constant radius, constant x-sectional area + // --> solid angle shrinking by the square of distance + //radius is somewhat arbitrary, can approx with just some constant k * 1 / dist^2 + //(k should vary inversely with # of samples, but this is taken care of later) + + angle_hidden = angle_hidden + float(dot((samppos_world - 0.05*norm - pos_world), norm) > 0.0) * min(1.0/dist2, ssao_factor_inv); + + // 'blocked' samples (significantly closer to camera relative to pos_world) are "no data", not "no occlusion" + points = points + int(diff.z > -1.0); + } + + angle_hidden = min(ssao_factor*angle_hidden/float(points), 1.0); + + ret = (1.0 - (float(points != 0) * angle_hidden)); + + return min(ret, 1.0); +} + +float pcfShadow(sampler2DRectShadow shadowMap, vec4 stc, float scl) +{ + stc.xyz /= stc.w; + stc.z += shadow_bias*scl; + + float cs = shadow2DRect(shadowMap, stc.xyz).x; + float shadow = cs; + + shadow += max(shadow2DRect(shadowMap, stc.xyz+vec3(1.5, 1.5, 0.0)).x, cs); + shadow += max(shadow2DRect(shadowMap, stc.xyz+vec3(1.5, -1.5, 0.0)).x, cs); + shadow += max(shadow2DRect(shadowMap, stc.xyz+vec3(-1.5, 1.5, 0.0)).x, cs); + shadow += max(shadow2DRect(shadowMap, stc.xyz+vec3(-1.5, -1.5, 0.0)).x, cs); + + return shadow/5.0; + + //return shadow; +} + +float pcfShadow(sampler2DShadow shadowMap, vec4 stc, float scl) +{ + stc.xyz /= stc.w; + stc.z += spot_shadow_bias*scl; + + float cs = shadow2D(shadowMap, stc.xyz).x; + float shadow = cs; + + vec2 off = 1.5/proj_shadow_res; + + shadow += max(shadow2D(shadowMap, stc.xyz+vec3(off.x, off.y, 0.0)).x, cs); + shadow += max(shadow2D(shadowMap, stc.xyz+vec3(off.x, -off.y, 0.0)).x, cs); + shadow += max(shadow2D(shadowMap, stc.xyz+vec3(-off.x, off.y, 0.0)).x, cs); + shadow += max(shadow2D(shadowMap, stc.xyz+vec3(-off.x, -off.y, 0.0)).x, cs); + + + return shadow/5.0; + + //return shadow; +} + +void main() +{ + vec2 pos_screen = vary_fragcoord.xy; + ivec2 itc = ivec2(pos_screen); + vec4 fcol = vec4(0,0,0,0); + + for (int i = 0; i < samples; i++) + { + vec4 pos = getPosition(itc, i); + + vec4 nmap4 = texelFetch(normalMap, itc, i); + nmap4 = vec4((nmap4.xy-0.5)*2.0,nmap4.z,nmap4.w); // unpack norm + float displace = nmap4.w; + vec3 norm = nmap4.xyz; + + float shadow = 1.0; + float dp_directional_light = max(0.0, dot(norm, vary_light.xyz)); + + vec3 shadow_pos = pos.xyz + displace*norm; + vec3 offset = vary_light.xyz * (1.0-dp_directional_light); + + vec4 spos = vec4(shadow_pos+offset*shadow_offset, 1.0); + + if (spos.z > -shadow_clip.w) + { + if (dp_directional_light == 0.0) + { + // if we know this point is facing away from the sun then we know it's in shadow without having to do a squirrelly shadow-map lookup + shadow = 0.0; + } + else + { + vec4 lpos; + + if (spos.z < -shadow_clip.z) + { + lpos = shadow_matrix[3]*spos; + lpos.xy *= shadow_res; + shadow = pcfShadow(shadowMap3, lpos, 0.25); + shadow += max((pos.z+shadow_clip.z)/(shadow_clip.z-shadow_clip.w)*2.0-1.0, 0.0); + } + else if (spos.z < -shadow_clip.y) + { + lpos = shadow_matrix[2]*spos; + lpos.xy *= shadow_res; + shadow = pcfShadow(shadowMap2, lpos, 0.5); + } + else if (spos.z < -shadow_clip.x) + { + lpos = shadow_matrix[1]*spos; + lpos.xy *= shadow_res; + shadow = pcfShadow(shadowMap1, lpos, 0.75); + } + else + { + lpos = shadow_matrix[0]*spos; + lpos.xy *= shadow_res; + shadow = pcfShadow(shadowMap0, lpos, 1.0); + } + + // take the most-shadowed value out of these two: + // * the blurred sun shadow in the light (shadow) map + // * an unblurred dot product between the sun and this norm + // the goal is to err on the side of most-shadow to fill-in shadow holes and reduce artifacting + shadow = min(shadow, dp_directional_light); + + } + } + else + { + // more distant than the shadow map covers + shadow = 1.0; + } + + + fcol[0] += shadow; + fcol[1] += calcAmbientOcclusion(pos, norm, i); + + spos.xyz = shadow_pos+offset*spot_shadow_offset; + + //spotlight shadow 1 + vec4 lpos = shadow_matrix[4]*spos; + fcol[2] += pcfShadow(shadowMap4, lpos, 0.8); + + //spotlight shadow 2 + lpos = shadow_matrix[5]*spos; + fcol[3] += pcfShadow(shadowMap5, lpos, 0.8); + } + + gl_FragColor = fcol / samples; +} diff --git a/indra/newview/lldrawpoolwlsky.cpp b/indra/newview/lldrawpoolwlsky.cpp index 66b834ab11..409b18d522 100644 --- a/indra/newview/lldrawpoolwlsky.cpp +++ b/indra/newview/lldrawpoolwlsky.cpp @@ -278,11 +278,13 @@ void LLDrawPoolWLSky::renderDeferred(S32 pass) renderSkyHaze(camHeightLocal); - /*LLVector3 const & origin = LLViewerCamera::getInstance()->getOrigin(); + LLVector3 const & origin = LLViewerCamera::getInstance()->getOrigin(); glPushMatrix(); + glTranslatef(origin.mV[0], origin.mV[1], origin.mV[2]); + gDeferredStarProgram.bind(); // *NOTE: have to bind a texture here since register combiners blending in // renderStars() requires something to be bound and we might as well only // bind the moon's texture once. @@ -292,8 +294,9 @@ void LLDrawPoolWLSky::renderDeferred(S32 pass) renderStars(); + gDeferredStarProgram.unbind(); - glPopMatrix();*/ + glPopMatrix(); renderSkyClouds(camHeightLocal); 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)); diff --git a/indra/newview/llviewershadermgr.cpp b/indra/newview/llviewershadermgr.cpp index 84fb1eb5d0..911a8bb65f 100644 --- a/indra/newview/llviewershadermgr.cpp +++ b/indra/newview/llviewershadermgr.cpp @@ -140,6 +140,7 @@ LLGLSLShader gDeferredPostProgram; LLGLSLShader gDeferredPostNoDoFProgram; LLGLSLShader gDeferredWLSkyProgram; LLGLSLShader gDeferredWLCloudProgram; +LLGLSLShader gDeferredStarProgram; LLGLSLShader gLuminanceGatherProgram; @@ -193,6 +194,7 @@ LLViewerShaderMgr::LLViewerShaderMgr() : mShaderList.push_back(&gDeferredAvatarAlphaProgram); mShaderList.push_back(&gDeferredWLSkyProgram); mShaderList.push_back(&gDeferredWLCloudProgram); + mShaderList.push_back(&gDeferredStarProgram); } LLViewerShaderMgr::~LLViewerShaderMgr() @@ -350,6 +352,9 @@ void LLViewerShaderMgr::setShaders() return; } + //setup preprocessor definitions + LLShaderMgr::instance()->mDefinitions["samples"] = llformat("%d", gSavedSettings.getU32("RenderFSAASamples")); + reentrance = true; // Make sure the compiled shader map is cleared before we recompile shaders. @@ -836,6 +841,8 @@ BOOL LLViewerShaderMgr::loadShadersEffects() { BOOL success = TRUE; + bool multisample = gSavedSettings.getU32("RenderFSAASamples") > 0 && gGLManager.mHasTextureMultisample; + if (mVertexShaderLevel[SHADER_EFFECT] == 0) { gGlowProgram.unload(); @@ -861,10 +868,21 @@ BOOL LLViewerShaderMgr::loadShadersEffects() if (success) { + std::string fragment; + + if (gSavedSettings.getU32("RenderFSAASamples") > 0 && LLRenderTarget::sUseFBO && gGLManager.mHasTextureMultisample) + { + fragment = "effects/glowExtractMSF.glsl"; + } + else + { + fragment = "effects/glowExtractF.glsl"; + } + gGlowExtractProgram.mName = "Glow Extract Shader (Post)"; gGlowExtractProgram.mShaderFiles.clear(); gGlowExtractProgram.mShaderFiles.push_back(make_pair("effects/glowExtractV.glsl", GL_VERTEX_SHADER_ARB)); - gGlowExtractProgram.mShaderFiles.push_back(make_pair("effects/glowExtractF.glsl", GL_FRAGMENT_SHADER_ARB)); + gGlowExtractProgram.mShaderFiles.push_back(make_pair(fragment, GL_FRAGMENT_SHADER_ARB)); gGlowExtractProgram.mShaderLevel = mVertexShaderLevel[SHADER_EFFECT]; success = gGlowExtractProgram.createShader(NULL, &mGlowExtractUniforms); if (!success) @@ -957,6 +975,7 @@ BOOL LLViewerShaderMgr::loadShadersDeferred() gDeferredWaterProgram.unload(); gDeferredWLSkyProgram.unload(); gDeferredWLCloudProgram.unload(); + gDeferredStarProgram.unload(); return TRUE; } @@ -1046,40 +1065,83 @@ BOOL LLViewerShaderMgr::loadShadersDeferred() if (success) { + std::string fragment; + + if (multisample) + { + fragment = "deferred/pointLightMSF.glsl"; + } + else + { + fragment = "deferred/pointLightF.glsl"; + } + gDeferredLightProgram.mName = "Deferred Light Shader"; gDeferredLightProgram.mShaderFiles.clear(); gDeferredLightProgram.mShaderFiles.push_back(make_pair("deferred/pointLightV.glsl", GL_VERTEX_SHADER_ARB)); - gDeferredLightProgram.mShaderFiles.push_back(make_pair("deferred/pointLightF.glsl", GL_FRAGMENT_SHADER_ARB)); + gDeferredLightProgram.mShaderFiles.push_back(make_pair(fragment, GL_FRAGMENT_SHADER_ARB)); gDeferredLightProgram.mShaderLevel = mVertexShaderLevel[SHADER_DEFERRED]; success = gDeferredLightProgram.createShader(NULL, NULL); } if (success) { + std::string fragment; + if (multisample) + { + fragment = "deferred/multiPointLightMSF.glsl"; + } + else + { + fragment = "deferred/multiPointLightF.glsl"; + } + gDeferredMultiLightProgram.mName = "Deferred MultiLight Shader"; gDeferredMultiLightProgram.mShaderFiles.clear(); gDeferredMultiLightProgram.mShaderFiles.push_back(make_pair("deferred/multiPointLightV.glsl", GL_VERTEX_SHADER_ARB)); - gDeferredMultiLightProgram.mShaderFiles.push_back(make_pair("deferred/multiPointLightF.glsl", GL_FRAGMENT_SHADER_ARB)); + gDeferredMultiLightProgram.mShaderFiles.push_back(make_pair(fragment, GL_FRAGMENT_SHADER_ARB)); gDeferredMultiLightProgram.mShaderLevel = mVertexShaderLevel[SHADER_DEFERRED]; success = gDeferredMultiLightProgram.createShader(NULL, NULL); } if (success) { + std::string fragment; + + if (multisample) + { + fragment = "deferred/spotLightMSF.glsl"; + } + else + { + fragment = "deferred/multiSpotLightF.glsl"; + } + gDeferredSpotLightProgram.mName = "Deferred SpotLight Shader"; gDeferredSpotLightProgram.mShaderFiles.clear(); gDeferredSpotLightProgram.mShaderFiles.push_back(make_pair("deferred/pointLightV.glsl", GL_VERTEX_SHADER_ARB)); - gDeferredSpotLightProgram.mShaderFiles.push_back(make_pair("deferred/multiSpotLightF.glsl", GL_FRAGMENT_SHADER_ARB)); + gDeferredSpotLightProgram.mShaderFiles.push_back(make_pair(fragment, GL_FRAGMENT_SHADER_ARB)); gDeferredSpotLightProgram.mShaderLevel = mVertexShaderLevel[SHADER_DEFERRED]; success = gDeferredSpotLightProgram.createShader(NULL, NULL); } if (success) { + std::string fragment; + + if (multisample) + { + fragment = "deferred/multiSpotLightMSF.glsl"; + } + else + { + fragment = "deferred/multiSpotLightF.glsl"; + } + gDeferredMultiSpotLightProgram.mName = "Deferred MultiSpotLight Shader"; gDeferredMultiSpotLightProgram.mShaderFiles.clear(); gDeferredMultiSpotLightProgram.mShaderFiles.push_back(make_pair("deferred/pointLightV.glsl", GL_VERTEX_SHADER_ARB)); - gDeferredMultiSpotLightProgram.mShaderFiles.push_back(make_pair("deferred/multiSpotLightF.glsl", GL_FRAGMENT_SHADER_ARB)); + gDeferredMultiSpotLightProgram.mShaderFiles.push_back(make_pair(fragment, GL_FRAGMENT_SHADER_ARB)); gDeferredMultiSpotLightProgram.mShaderLevel = mVertexShaderLevel[SHADER_DEFERRED]; success = gDeferredMultiSpotLightProgram.createShader(NULL, NULL); } @@ -1343,6 +1405,17 @@ BOOL LLViewerShaderMgr::loadShadersDeferred() success = gDeferredWLCloudProgram.createShader(NULL, &mWLUniforms); } + if (success) + { + gDeferredStarProgram.mName = "Deferred Star Program"; + gDeferredStarProgram.mShaderFiles.clear(); + gDeferredStarProgram.mShaderFiles.push_back(make_pair("deferred/starsV.glsl", GL_VERTEX_SHADER_ARB)); + gDeferredStarProgram.mShaderFiles.push_back(make_pair("deferred/starsF.glsl", GL_FRAGMENT_SHADER_ARB)); + gDeferredStarProgram.mShaderLevel = mVertexShaderLevel[SHADER_DEFERRED]; + gDeferredStarProgram.mShaderGroup = LLGLSLShader::SG_SKY; + success = gDeferredStarProgram.createShader(NULL, &mWLUniforms); + } + if (mVertexShaderLevel[SHADER_DEFERRED] > 1) { if (success) diff --git a/indra/newview/llviewershadermgr.h b/indra/newview/llviewershadermgr.h index b13775742b..f5371f0619 100644 --- a/indra/newview/llviewershadermgr.h +++ b/indra/newview/llviewershadermgr.h @@ -376,7 +376,7 @@ extern LLGLSLShader gDeferredFullbrightProgram; extern LLGLSLShader gDeferredAvatarAlphaProgram; extern LLGLSLShader gDeferredWLSkyProgram; extern LLGLSLShader gDeferredWLCloudProgram; - +extern LLGLSLShader gDeferredStarProgram; extern LLGLSLShader gLuminanceGatherProgram; //current avatar shader parameter pointer diff --git a/indra/newview/pipeline.cpp b/indra/newview/pipeline.cpp index ef3cbe5278..ffe3b4b1a4 100644 --- a/indra/newview/pipeline.cpp +++ b/indra/newview/pipeline.cpp @@ -589,8 +589,8 @@ void LLPipeline::allocateScreenBuffer(U32 resX, U32 resY) mScreenWidth = resX; mScreenHeight = resY; - //never use more than 4 samples for render targets - U32 samples = llmin(gSavedSettings.getU32("RenderFSAASamples"), (U32) 4); + U32 samples = llmin(gSavedSettings.getU32("RenderFSAASamples"), (U32) 16); + if (gGLManager.mIsATI) { //disable multisampling of render targets where ATI is involved samples = 0; @@ -6170,11 +6170,8 @@ void LLPipeline::renderBloom(BOOL for_snapshot, F32 zoom_factor, int subfield) gGL.setAlphaRejectSettings(LLRender::CF_DEFAULT); gGL.setSceneBlendType(LLRender::BT_ADD_WITH_ALPHA); - gGL.getTexUnit(0)->unbind(LLTexUnit::TT_TEXTURE); - gGL.getTexUnit(0)->disable(); - gGL.getTexUnit(0)->enable(LLTexUnit::TT_RECT_TEXTURE); - gGL.getTexUnit(0)->bind(&mScreen); - + mScreen.bindTexture(0, 0); + gGL.color4f(1,1,1,1); gPipeline.enableLightsFullbright(LLColor4(1,1,1,1)); gGL.begin(LLRender::TRIANGLE_STRIP); @@ -6189,7 +6186,7 @@ void LLPipeline::renderBloom(BOOL for_snapshot, F32 zoom_factor, int subfield) gGL.end(); - gGL.getTexUnit(0)->enable(LLTexUnit::TT_TEXTURE); + gGL.getTexUnit(0)->unbind(mScreen.getUsage()); mGlow[2].flush(); } @@ -6217,7 +6214,6 @@ void LLPipeline::renderBloom(BOOL for_snapshot, F32 zoom_factor, int subfield) for (S32 i = 0; i < kernel; i++) { - gGL.getTexUnit(0)->unbind(LLTexUnit::TT_TEXTURE); { LLFastTimer ftm(FTM_RENDER_BLOOM_FBO); mGlow[i%2].bindTarget(); -- 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/llrender/llimagegl.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