diff options
Diffstat (limited to 'indra/llrender')
-rw-r--r-- | indra/llrender/llfontgl.cpp | 2 | ||||
-rw-r--r-- | indra/llrender/llgl.cpp | 173 | ||||
-rw-r--r-- | indra/llrender/llgl.h | 11 | ||||
-rw-r--r-- | indra/llrender/llglheaders.h | 66 | ||||
-rw-r--r-- | indra/llrender/llglslshader.cpp | 49 | ||||
-rw-r--r-- | indra/llrender/llglslshader.h | 4 | ||||
-rw-r--r-- | indra/llrender/llimagegl.cpp | 15 | ||||
-rw-r--r-- | indra/llrender/llimagegl.h | 2 | ||||
-rw-r--r-- | indra/llrender/llrender.cpp | 43 | ||||
-rw-r--r-- | indra/llrender/llrender.h | 1 | ||||
-rw-r--r-- | indra/llrender/llrendertarget.cpp | 424 | ||||
-rw-r--r-- | indra/llrender/llrendertarget.h | 26 | ||||
-rw-r--r-- | indra/llrender/llshadermgr.cpp | 226 | ||||
-rw-r--r-- | indra/llrender/llshadermgr.h | 5 | ||||
-rw-r--r-- | indra/llrender/llvertexbuffer.cpp | 416 | ||||
-rw-r--r-- | indra/llrender/llvertexbuffer.h | 58 |
16 files changed, 987 insertions, 534 deletions
diff --git a/indra/llrender/llfontgl.cpp b/indra/llrender/llfontgl.cpp index d6a31dc862..180ae4dfa6 100644 --- a/indra/llrender/llfontgl.cpp +++ b/indra/llrender/llfontgl.cpp @@ -554,7 +554,7 @@ S32 LLFontGL::maxDrawableChars(const llwchar* wchars, F32 max_pixels, S32 max_ch BOOL in_word = FALSE; // avoid S32 overflow when max_pixels == S32_MAX by staying in floating point - F32 scaled_max_pixels = ceil(max_pixels * sScaleX); + F32 scaled_max_pixels = max_pixels * sScaleX; F32 width_padding = 0.f; LLFontGlyphInfo* next_glyph = NULL; diff --git a/indra/llrender/llgl.cpp b/indra/llrender/llgl.cpp index f29ee0e57e..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; @@ -178,6 +183,12 @@ PFNGLBLITFRAMEBUFFERPROC glBlitFramebuffer = NULL; PFNGLRENDERBUFFERSTORAGEMULTISAMPLEPROC glRenderbufferStorageMultisample = NULL; PFNGLFRAMEBUFFERTEXTURELAYERPROC glFramebufferTextureLayer = NULL; +//GL_ARB_texture_multisample +PFNGLTEXIMAGE2DMULTISAMPLEPROC glTexImage2DMultisample; +PFNGLTEXIMAGE3DMULTISAMPLEPROC glTexImage3DMultisample; +PFNGLGETMULTISAMPLEFVPROC glGetMultisamplefv; +PFNGLSAMPLEMASKIPROC glSampleMaski; + // GL_EXT_blend_func_separate PFNGLBLENDFUNCSEPARATEEXTPROC glBlendFuncSeparateEXT = NULL; @@ -321,18 +332,26 @@ LLGLManager::LLGLManager() : mHasMipMapGeneration(FALSE), mHasCompressedTextures(FALSE), mHasFramebufferObject(FALSE), + mMaxSamples(0), mHasBlendFuncSeparate(FALSE), mHasVertexBufferObject(FALSE), + mHasMapBufferRange(FALSE), mHasPBuffer(FALSE), mHasShaderObjects(FALSE), mHasVertexShader(FALSE), mHasFragmentShader(FALSE), + mNumTextureImageUnits(0), mHasOcclusionQuery(FALSE), mHasOcclusionQuery2(FALSE), mHasPointParameters(FALSE), mHasDrawBuffers(FALSE), mHasTextureRectangle(FALSE), + mHasTextureMultisample(FALSE), + mMaxSampleMaskWords(0), + mMaxColorTextureSamples(0), + mMaxDepthTextureSamples(0), + mMaxIntegerSamples(0), mHasAnisotropic(FALSE), mHasARBEnvCombine(FALSE), @@ -534,6 +553,26 @@ bool LLGLManager::initGL() return false; } + if (mHasFragmentShader) + { + GLint num_tex_image_units; + glGetIntegerv(GL_MAX_TEXTURE_IMAGE_UNITS_ARB, &num_tex_image_units); + mNumTextureImageUnits = llmin(num_tex_image_units, 32); + } + + if (mHasTextureMultisample) + { + glGetIntegerv(GL_MAX_COLOR_TEXTURE_SAMPLES, &mMaxColorTextureSamples); + glGetIntegerv(GL_MAX_DEPTH_TEXTURE_SAMPLES, &mMaxDepthTextureSamples); + glGetIntegerv(GL_MAX_INTEGER_SAMPLES, &mMaxIntegerSamples); + glGetIntegerv(GL_MAX_SAMPLE_MASK_WORDS, &mMaxSampleMaskWords); + } + + if (mHasFramebufferObject) + { + glGetIntegerv(GL_MAX_SAMPLES, &mMaxSamples); + } + setToDebugGPU(); initGLStates(); @@ -640,6 +679,14 @@ std::string LLGLManager::getRawGLString() return gl_string; } +U32 LLGLManager::getNumFBOFSAASamples(U32 samples) +{ + samples = llmin(samples, (U32) mMaxColorTextureSamples); + samples = llmin(samples, (U32) mMaxDepthTextureSamples); + samples = llmin(samples, (U32) 4); + return samples; +} + void LLGLManager::shutdownGL() { if (mInited) @@ -720,6 +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 @@ -734,6 +782,7 @@ void LLGLManager::initExtensions() mHasDrawBuffers = ExtensionExists("GL_ARB_draw_buffers", gGLHExts.mSysExts); mHasBlendFuncSeparate = ExtensionExists("GL_EXT_blend_func_separate", gGLHExts.mSysExts); mHasTextureRectangle = ExtensionExists("GL_ARB_texture_rectangle", gGLHExts.mSysExts); + mHasTextureMultisample = ExtensionExists("GL_ARB_texture_multisample", gGLHExts.mSysExts); #if !LL_DARWIN mHasPointParameters = !mIsATI && ExtensionExists("GL_ARB_point_parameters", gGLHExts.mSysExts); #endif @@ -878,11 +927,13 @@ void LLGLManager::initExtensions() LL_INFOS("RenderInit") << "Disabling mip-map generation for Intel GPUs" << LL_ENDL; mHasMipMapGeneration = FALSE; } +#if !LL_DARWIN if (mIsATI && mHasMipMapGeneration) { LL_INFOS("RenderInit") << "Disabling mip-map generation for ATI GPUs (performance opt)" << LL_ENDL; mHasMipMapGeneration = FALSE; } +#endif // Misc glGetIntegerv(GL_MAX_ELEMENTS_VERTICES, (GLint*) &mGLMaxVertexRange); @@ -911,6 +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; @@ -943,6 +999,13 @@ void LLGLManager::initExtensions() { glBlendFuncSeparateEXT = (PFNGLBLENDFUNCSEPARATEEXTPROC) GLH_EXT_GET_PROC_ADDRESS("glBlendFuncSeparateEXT"); } + if (mHasTextureMultisample) + { + glTexImage2DMultisample = (PFNGLTEXIMAGE2DMULTISAMPLEPROC) GLH_EXT_GET_PROC_ADDRESS("glTexImage2DMultisample"); + glTexImage3DMultisample = (PFNGLTEXIMAGE3DMULTISAMPLEPROC) GLH_EXT_GET_PROC_ADDRESS("glTexImage3DMultisample"); + glGetMultisamplefv = (PFNGLGETMULTISAMPLEFVPROC) GLH_EXT_GET_PROC_ADDRESS("glGetMultisamplefv"); + glSampleMaski = (PFNGLSAMPLEMASKIPROC) GLH_EXT_GET_PROC_ADDRESS("glSampleMaski"); + } #if (!LL_LINUX && !LL_SOLARIS) || LL_LINUX_NV_GL_HEADERS // This is expected to be a static symbol on Linux GL implementations, except if we use the nvidia headers - bah glDrawRangeElements = (PFNGLDRAWRANGEELEMENTSPROC)GLH_EXT_GET_PROC_ADDRESS("glDrawRangeElements"); @@ -1360,10 +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", @@ -1374,7 +1433,8 @@ void LLGLState::checkTextureChannels(const std::string& msg) "GL_TEXTURE_GEN_T", "GL_TEXTURE_GEN_Q", "GL_TEXTURE_GEN_R", - "GL_TEXTURE_RECTANGLE_ARB" + "GL_TEXTURE_RECTANGLE_ARB", + "GL_TEXTURE_2D_MULTISAMPLE" }; static GLint value[] = @@ -1387,7 +1447,8 @@ void LLGLState::checkTextureChannels(const std::string& msg) GL_TEXTURE_GEN_T, GL_TEXTURE_GEN_Q, GL_TEXTURE_GEN_R, - GL_TEXTURE_RECTANGLE_ARB + GL_TEXTURE_RECTANGLE_ARB, + GL_TEXTURE_2D_MULTISAMPLE }; GLint stackDepth = 0; @@ -1396,68 +1457,96 @@ 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) + { + 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++) { - gFailLog << "Texture matrix in channel " << i << " corrupt." << std::endl; + if (j == 8 && !gGLManager.mHasTextureRectangle || + j == 9 && !gGLManager.mHasTextureMultisample) + { + continue; + } + + 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(); } - } - - for (S32 j = (i == 0 ? 1 : 0); - j < (gGLManager.mHasTextureRectangle ? 9 : 8); j++) - { - if (glIsEnabled(value[j])) + 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; + } } } } + stop_glerror(); gGL.getTexUnit(0)->activate(); glClientActiveTextureARB(GL_TEXTURE0_ARB); stop_glerror(); diff --git a/indra/llrender/llgl.h b/indra/llrender/llgl.h index 3d002fd8c4..d1bee00161 100644 --- a/indra/llrender/llgl.h +++ b/indra/llrender/llgl.h @@ -83,20 +83,28 @@ public: BOOL mHasMipMapGeneration; BOOL mHasCompressedTextures; BOOL mHasFramebufferObject; + S32 mMaxSamples; BOOL mHasBlendFuncSeparate; - + // ARB Extensions BOOL mHasVertexBufferObject; + BOOL mHasMapBufferRange; BOOL mHasPBuffer; BOOL mHasShaderObjects; BOOL mHasVertexShader; BOOL mHasFragmentShader; + S32 mNumTextureImageUnits; BOOL mHasOcclusionQuery; BOOL mHasOcclusionQuery2; BOOL mHasPointParameters; BOOL mHasDrawBuffers; BOOL mHasDepthClamp; BOOL mHasTextureRectangle; + BOOL mHasTextureMultisample; + S32 mMaxSampleMaskWords; + S32 mMaxColorTextureSamples; + S32 mMaxDepthTextureSamples; + S32 mMaxIntegerSamples; // Other extensions. BOOL mHasAnisotropic; @@ -138,6 +146,7 @@ public: void printGLInfoString(); void getGLInfo(LLSD& info); + U32 getNumFBOFSAASamples(U32 desired_samples = 32); // In ALL CAPS std::string mGLVendor; std::string mGLVendorShort; diff --git a/indra/llrender/llglheaders.h b/indra/llrender/llglheaders.h index d8140a124d..f35f329f00 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; @@ -474,6 +482,11 @@ extern PFNGLFRAMEBUFFERTEXTURELAYERPROC glFramebufferTextureLayer; //GL_ARB_draw_buffers extern PFNGLDRAWBUFFERSARBPROC glDrawBuffersARB; +//GL_ARB_texture_multisample +extern PFNGLTEXIMAGE2DMULTISAMPLEPROC glTexImage2DMultisample; +extern PFNGLTEXIMAGE3DMULTISAMPLEPROC glTexImage3DMultisample; +extern PFNGLGETMULTISAMPLEFVPROC glGetMultisamplefv; +extern PFNGLSAMPLEMASKIPROC glSampleMaski; #elif LL_WINDOWS //---------------------------------------------------------------------------- @@ -506,6 +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; @@ -673,6 +690,11 @@ extern PFNGLFRAMEBUFFERTEXTURELAYERPROC glFramebufferTextureLayer; //GL_ARB_draw_buffers extern PFNGLDRAWBUFFERSARBPROC glDrawBuffersARB; +//GL_ARB_texture_multisample +extern PFNGLTEXIMAGE2DMULTISAMPLEPROC glTexImage2DMultisample; +extern PFNGLTEXIMAGE3DMULTISAMPLEPROC glTexImage3DMultisample; +extern PFNGLGETMULTISAMPLEFVPROC glGetMultisamplefv; +extern PFNGLSAMPLEMASKIPROC glSampleMaski; #elif LL_DARWIN //---------------------------------------------------------------------------- @@ -714,13 +736,55 @@ extern void glGenerateMipmapEXT(GLenum target) AVAILABLE_MAC_OS_X_VERSION_10_4_A #ifndef GL_ARB_framebuffer_object #define glGenerateMipmap glGenerateMipmapEXT +#define GL_MAX_SAMPLES 0x8D57 #endif + // GL_ARB_draw_buffers extern void glDrawBuffersARB(GLsizei n, const GLenum* bufs) AVAILABLE_MAC_OS_X_VERSION_10_4_AND_LATER; #ifdef __cplusplus extern "C" { #endif + +// +// Define map buffer range headers on Mac +// +#ifndef GL_ARB_map_buffer_range +#define GL_MAP_READ_BIT 0x0001 +#define GL_MAP_WRITE_BIT 0x0002 +#define GL_MAP_INVALIDATE_RANGE_BIT 0x0004 +#define GL_MAP_INVALIDATE_BUFFER_BIT 0x0008 +#define GL_MAP_FLUSH_EXPLICIT_BIT 0x0010 +#define GL_MAP_UNSYNCHRONIZED_BIT 0x0020 +#endif + +// +// Define multisample headers on Mac +// +#ifndef GL_ARB_texture_multisample +#define GL_SAMPLE_POSITION 0x8E50 +#define GL_SAMPLE_MASK 0x8E51 +#define GL_SAMPLE_MASK_VALUE 0x8E52 +#define GL_MAX_SAMPLE_MASK_WORDS 0x8E59 +#define GL_TEXTURE_2D_MULTISAMPLE 0x9100 +#define GL_PROXY_TEXTURE_2D_MULTISAMPLE 0x9101 +#define GL_TEXTURE_2D_MULTISAMPLE_ARRAY 0x9102 +#define GL_PROXY_TEXTURE_2D_MULTISAMPLE_ARRAY 0x9103 +#define GL_TEXTURE_BINDING_2D_MULTISAMPLE 0x9104 +#define GL_TEXTURE_BINDING_2D_MULTISAMPLE_ARRAY 0x9105 +#define GL_TEXTURE_SAMPLES 0x9106 +#define GL_TEXTURE_FIXED_SAMPLE_LOCATIONS 0x9107 +#define GL_SAMPLER_2D_MULTISAMPLE 0x9108 +#define GL_INT_SAMPLER_2D_MULTISAMPLE 0x9109 +#define GL_UNSIGNED_INT_SAMPLER_2D_MULTISAMPLE 0x910A +#define GL_SAMPLER_2D_MULTISAMPLE_ARRAY 0x910B +#define GL_INT_SAMPLER_2D_MULTISAMPLE_ARRAY 0x910C +#define GL_UNSIGNED_INT_SAMPLER_2D_MULTISAMPLE_ARRAY 0x910D +#define GL_MAX_COLOR_TEXTURE_SAMPLES 0x910E +#define GL_MAX_DEPTH_TEXTURE_SAMPLES 0x910F +#define GL_MAX_INTEGER_SAMPLES 0x9110 +#endif + // // Define vertex buffer object headers on Mac // @@ -757,7 +821,7 @@ extern "C" { #define GL_DYNAMIC_READ_ARB 0x88E9 #define GL_DYNAMIC_COPY_ARB 0x88EA #endif - + #ifndef GL_ARB_vertex_buffer_object diff --git a/indra/llrender/llglslshader.cpp b/indra/llrender/llglslshader.cpp index 257bcd9380..8e99f62de6 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; +GLhandleARB LLGLSLShader::sCurBoundShader = 0; + BOOL shouldChange(const LLVector4& v1, const LLVector4& v2) { return v1 != v2; @@ -56,7 +58,7 @@ BOOL shouldChange(const LLVector4& v1, const LLVector4& v2) LLShaderFeatures::LLShaderFeatures() : calculatesLighting(false), isShiny(false), isFullbright(false), hasWaterFog(false), hasTransport(false), hasSkinning(false), hasObjectSkinning(false), hasAtmospherics(false), isSpecular(false), -hasGamma(false), hasLighting(false), calculatesAtmospherics(false) +hasGamma(false), hasLighting(false), calculatesAtmospherics(false), mIndexedTextureChannels(0), disableTextureIndex(false) { } @@ -107,16 +109,11 @@ BOOL LLGLSLShader::createShader(vector<string> * attributes, // Create program mProgramObject = glCreateProgramObjectARB(); - // Attach existing objects - if (!LLShaderMgr::instance()->attachShaderFeatures(this)) - { - return FALSE; - } - + //compile new source vector< pair<string,GLenum> >::iterator fileIter = mShaderFiles.begin(); for ( ; fileIter != mShaderFiles.end(); fileIter++ ) { - GLhandleARB shaderhandle = LLShaderMgr::instance()->loadShaderFile((*fileIter).first, mShaderLevel, (*fileIter).second); + GLhandleARB shaderhandle = LLShaderMgr::instance()->loadShaderFile((*fileIter).first, mShaderLevel, (*fileIter).second, mFeatures.mIndexedTextureChannels); LL_DEBUGS("ShaderLoading") << "SHADER FILE: " << (*fileIter).first << " mShaderLevel=" << mShaderLevel << LL_ENDL; if (shaderhandle > 0) { @@ -128,6 +125,12 @@ BOOL LLGLSLShader::createShader(vector<string> * attributes, } } + // Attach existing objects + if (!LLShaderMgr::instance()->attachShaderFeatures(this)) + { + return FALSE; + } + // Map attributes and uniforms if (success) { @@ -149,6 +152,29 @@ BOOL LLGLSLShader::createShader(vector<string> * attributes, return createShader(attributes,uniforms); } } + else if (mFeatures.mIndexedTextureChannels > 0) + { //override texture channels for indexed texture rendering + bind(); + S32 channel_count = mFeatures.mIndexedTextureChannels; + + for (S32 i = 0; i < channel_count; i++) + { + uniform1i(llformat("tex%d", i), i); + } + + S32 cur_tex = channel_count; //adjust any texture channels that might have been overwritten + for (U32 i = 0; i < mTexture.size(); i++) + { + if (mTexture[i] > -1 && mTexture[i] < channel_count) + { + llassert(cur_tex < gGLManager.mNumTextureImageUnits); + uniform1i(i, cur_tex); + mTexture[i] = cur_tex++; + } + } + unbind(); + } + return success; } @@ -293,7 +319,8 @@ void LLGLSLShader::mapUniform(GLint index, const vector<string> * uniforms) GLint LLGLSLShader::mapUniformTextureChannel(GLint location, GLenum type) { - if (type >= GL_SAMPLER_1D_ARB && type <= GL_SAMPLER_2D_RECT_SHADOW_ARB) + if (type >= GL_SAMPLER_1D_ARB && type <= GL_SAMPLER_2D_RECT_SHADOW_ARB || + type == GL_SAMPLER_2D_MULTISAMPLE) { //this here is a texture glUniform1iARB(location, mActiveTextureChannels); LL_DEBUGS("ShaderLoading") << "Assigned to texture channel " << mActiveTextureChannels << LL_ENDL; @@ -342,7 +369,7 @@ void LLGLSLShader::bind() if (gGLManager.mHasShaderObjects) { glUseProgramObjectARB(mProgramObject); - + sCurBoundShader = mProgramObject; if (mUniformsDirty) { LLShaderMgr::instance()->updateShaderUniforms(this); @@ -365,6 +392,7 @@ void LLGLSLShader::unbind() } } glUseProgramObjectARB(0); + sCurBoundShader = 0; stop_glerror(); } } @@ -372,6 +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 d46ddbbe18..4922eb6d67 100644 --- a/indra/llrender/llglslshader.h +++ b/indra/llrender/llglslshader.h @@ -45,6 +45,8 @@ public: bool hasObjectSkinning; bool hasAtmospherics; bool hasGamma; + S32 mIndexedTextureChannels; + bool disableTextureIndex; // char numLights; @@ -64,6 +66,8 @@ public: LLGLSLShader(); + static GLhandleARB sCurBoundShader; + void unload(); BOOL createShader(std::vector<std::string> * attributes, std::vector<std::string> * uniforms); diff --git a/indra/llrender/llimagegl.cpp b/indra/llrender/llimagegl.cpp index d408077c68..60a5962234 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 @@ -1413,11 +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(LLTexUnit::TT_TEXTURE); + tex_unit->unbind(tex_unit->getCurrType()); stop_glerror(); } } 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/llrender.cpp b/indra/llrender/llrender.cpp index 49e10c4790..6a3f186531 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" @@ -46,14 +47,15 @@ S32 gGLViewport[4]; U32 LLRender::sUICalls = 0; U32 LLRender::sUIVerts = 0; -static const U32 LL_NUM_TEXTURE_LAYERS = 16; +static const U32 LL_NUM_TEXTURE_LAYERS = 32; static const U32 LL_NUM_LIGHT_UNITS = 8; static GLenum sGLTextureType[] = { GL_TEXTURE_2D, GL_TEXTURE_RECTANGLE_ARB, - GL_TEXTURE_CUBE_MAP_ARB + GL_TEXTURE_CUBE_MAP_ARB, + GL_TEXTURE_2D_MULTISAMPLE }; static GLint sGLAddressMode[] = @@ -119,14 +121,29 @@ void LLTexUnit::refreshState(void) gGL.flush(); glActiveTextureARB(GL_TEXTURE0_ARB + mIndex); + + // + // Per apple spec, don't call glEnable/glDisable when index exceeds max texture units + // http://www.mailinglistarchive.com/html/mac-opengl@lists.apple.com/2008-07/msg00653.html + // + bool enableDisable = (mIndex < gGLManager.mNumTextureUnits) && mCurrTexType != LLTexUnit::TT_MULTISAMPLE_TEXTURE; + if (mCurrTexType != TT_NONE) { - glEnable(sGLTextureType[mCurrTexType]); + if (enableDisable) + { + glEnable(sGLTextureType[mCurrTexType]); + } + glBindTexture(sGLTextureType[mCurrTexType], mCurrTexture); } else { - glDisable(GL_TEXTURE_2D); + if (enableDisable) + { + glDisable(GL_TEXTURE_2D); + } + glBindTexture(GL_TEXTURE_2D, 0); } @@ -167,7 +184,11 @@ void LLTexUnit::enable(eTextureType type) mCurrTexType = type; gGL.flush(); - glEnable(sGLTextureType[type]); + if (type != LLTexUnit::TT_MULTISAMPLE_TEXTURE && + mIndex < gGLManager.mNumTextureUnits) + { + glEnable(sGLTextureType[type]); + } } } @@ -180,7 +201,12 @@ void LLTexUnit::disable(void) activate(); unbind(mCurrTexType); gGL.flush(); - glDisable(sGLTextureType[mCurrTexType]); + if (mCurrTexType != LLTexUnit::TT_MULTISAMPLE_TEXTURE && + mIndex < gGLManager.mNumTextureUnits) + { + glDisable(sGLTextureType[mCurrTexType]); + } + mCurrTexType = TT_NONE; } } @@ -268,7 +294,7 @@ 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); @@ -378,6 +404,7 @@ void LLTexUnit::unbind(eTextureType type) activate(); mCurrTexture = 0; glBindTexture(sGLTextureType[type], 0); + stop_glerror(); } } @@ -399,7 +426,7 @@ void LLTexUnit::setTextureAddressMode(eTextureAddressMode mode) void LLTexUnit::setTextureFilteringOption(LLTexUnit::eTextureFilterOptions option) { - if (mIndex < 0 || mCurrTexture == 0) return; + if (mIndex < 0 || mCurrTexture == 0 || mCurrTexType == LLTexUnit::TT_MULTISAMPLE_TEXTURE) return; gGL.flush(); diff --git a/indra/llrender/llrender.h b/indra/llrender/llrender.h index 7ba14f7b40..41e7b35341 100644 --- a/indra/llrender/llrender.h +++ b/indra/llrender/llrender.h @@ -57,6 +57,7 @@ public: TT_TEXTURE = 0, // Standard 2D Texture TT_RECT_TEXTURE, // Non power of 2 texture TT_CUBE_MAP, // 6-sided cube map texture + TT_MULTISAMPLE_TEXTURE, // see GL_ARB_texture_multisample TT_NONE // No texture type is currently enabled } eTextureType; diff --git a/indra/llrender/llrendertarget.cpp b/indra/llrender/llrendertarget.cpp index cd2556d435..b6463309e1 100644 --- a/indra/llrender/llrendertarget.cpp +++ b/indra/llrender/llrendertarget.cpp @@ -44,6 +44,7 @@ void check_framebuffer_status() case GL_FRAMEBUFFER_COMPLETE: break; default: + llwarns << "check_framebuffer_status failed -- " << std::hex << status << llendl; ll_fail("check_framebuffer_status failed"); break; } @@ -62,8 +63,7 @@ LLRenderTarget::LLRenderTarget() : mUseDepth(false), mRenderDepth(false), mUsage(LLTexUnit::TT_TEXTURE), - mSamples(0), - mSampleBuffer(NULL) + mSamples(0) { } @@ -72,23 +72,32 @@ LLRenderTarget::~LLRenderTarget() release(); } - -void LLRenderTarget::setSampleBuffer(LLMultisampleBuffer* buffer) -{ - mSampleBuffer = buffer; -} - -void LLRenderTarget::allocate(U32 resx, U32 resy, U32 color_fmt, bool depth, bool stencil, LLTexUnit::eTextureType usage, bool use_fbo) +void LLRenderTarget::allocate(U32 resx, U32 resy, U32 color_fmt, bool depth, bool stencil, LLTexUnit::eTextureType usage, bool use_fbo, S32 samples) { stop_glerror(); + + release(); + mResX = resx; mResY = resy; mStencil = stencil; mUsage = usage; mUseDepth = depth; + mSamples = samples; - release(); + mSamples = gGLManager.getNumFBOFSAASamples(mSamples); + + if (mSamples > 1 && gGLManager.mHasTextureMultisample) + { + mUsage = LLTexUnit::TT_MULTISAMPLE_TEXTURE; + //no support for multisampled stencil targets yet + mStencil = false; + } + else + { + mSamples = 0; + } if ((sUseFBO || use_fbo) && gGLManager.mHasFramebufferObject) { @@ -145,29 +154,51 @@ void LLRenderTarget::addColorAttachment(U32 color_fmt) stop_glerror(); - LLImageGL::setManualImage(LLTexUnit::getInternalType(mUsage), 0, color_fmt, mResX, mResY, GL_RGBA, GL_UNSIGNED_BYTE, NULL); - - stop_glerror(); - if (offset == 0) +#ifdef GL_ARB_texture_multisample + if (mSamples > 1) { - gGL.getTexUnit(0)->setTextureFilteringOption(LLTexUnit::TFO_BILINEAR); + glTexImage2DMultisample(LLTexUnit::getInternalType(mUsage), mSamples, color_fmt, mResX, mResY, GL_TRUE); } else - { //don't filter data attachments - gGL.getTexUnit(0)->setTextureFilteringOption(LLTexUnit::TFO_POINT); - } - if (mUsage != LLTexUnit::TT_RECT_TEXTURE) +#else + llassert_always(mSamples <= 1); +#endif { - gGL.getTexUnit(0)->setTextureAddressMode(LLTexUnit::TAM_MIRROR); + LLImageGL::setManualImage(LLTexUnit::getInternalType(mUsage), 0, color_fmt, mResX, mResY, GL_RGBA, GL_UNSIGNED_BYTE, NULL); } - else - { - // ATI doesn't support mirrored repeat for rectangular textures. - gGL.getTexUnit(0)->setTextureAddressMode(LLTexUnit::TAM_CLAMP); + + stop_glerror(); + + if (mSamples == 0) + { + if (offset == 0) + { //use bilinear filtering on single texture render targets that aren't multisampled + gGL.getTexUnit(0)->setTextureFilteringOption(LLTexUnit::TFO_BILINEAR); + stop_glerror(); + } + else + { //don't filter data attachments + gGL.getTexUnit(0)->setTextureFilteringOption(LLTexUnit::TFO_POINT); + stop_glerror(); + } + + if (mUsage != LLTexUnit::TT_RECT_TEXTURE) + { + gGL.getTexUnit(0)->setTextureAddressMode(LLTexUnit::TAM_MIRROR); + stop_glerror(); + } + else + { + // ATI doesn't support mirrored repeat for rectangular textures. + gGL.getTexUnit(0)->setTextureAddressMode(LLTexUnit::TAM_CLAMP); + stop_glerror(); + } } + if (mFBO) { + stop_glerror(); glBindFramebuffer(GL_FRAMEBUFFER, mFBO); glFramebufferTexture2D(GL_FRAMEBUFFER, GL_COLOR_ATTACHMENT0+offset, LLTexUnit::getInternalType(mUsage), tex, 0); @@ -180,6 +211,12 @@ void LLRenderTarget::addColorAttachment(U32 color_fmt) mTex.push_back(tex); + if (gDebugGL) + { //bind and unbind to validate target + bindTarget(); + flush(); + } + } void LLRenderTarget::allocateDepth() @@ -196,9 +233,20 @@ void LLRenderTarget::allocateDepth() { LLImageGL::generateTextures(1, &mDepth); gGL.getTexUnit(0)->bindManual(mUsage, mDepth); - U32 internal_type = LLTexUnit::getInternalType(mUsage); - gGL.getTexUnit(0)->setTextureFilteringOption(LLTexUnit::TFO_POINT); - LLImageGL::setManualImage(internal_type, 0, GL_DEPTH_COMPONENT32, mResX, mResY, GL_DEPTH_COMPONENT, GL_UNSIGNED_INT, NULL); + if (mSamples == 0) + { + U32 internal_type = LLTexUnit::getInternalType(mUsage); + gGL.getTexUnit(0)->setTextureFilteringOption(LLTexUnit::TFO_POINT); + LLImageGL::setManualImage(internal_type, 0, GL_DEPTH_COMPONENT32, mResX, mResY, GL_DEPTH_COMPONENT, GL_UNSIGNED_INT, NULL); + } +#ifdef GL_ARB_texture_multisample + else + { + glTexImage2DMultisample(LLTexUnit::getInternalType(mUsage), mSamples, GL_DEPTH_COMPONENT32, mResX, mResY, GL_TRUE); + } +#else + llassert_always(mSamples <= 1); +#endif } } @@ -238,6 +286,9 @@ void LLRenderTarget::shareDepthBuffer(LLRenderTarget& target) glFramebufferTexture2D(GL_FRAMEBUFFER, GL_DEPTH_ATTACHMENT, LLTexUnit::getInternalType(mUsage), mDepth, 0); stop_glerror(); } + + check_framebuffer_status(); + glBindFramebuffer(GL_FRAMEBUFFER, 0); target.mUseDepth = true; @@ -255,7 +306,7 @@ void LLRenderTarget::release() } else { - LLImageGL::deleteTextures(1, &mDepth); + LLImageGL::deleteTextures(1, &mDepth, true); stop_glerror(); } mDepth = 0; @@ -284,11 +335,12 @@ void LLRenderTarget::release() if (mTex.size() > 0) { - LLImageGL::deleteTextures(mTex.size(), &mTex[0]); + LLImageGL::deleteTextures(mTex.size(), &mTex[0], true); mTex.clear(); } + + mResX = mResY = 0; - mSampleBuffer = NULL; sBoundTarget = NULL; } @@ -297,34 +349,27 @@ void LLRenderTarget::bindTarget() if (mFBO) { stop_glerror(); - if (mSampleBuffer) - { - mSampleBuffer->bindTarget(this); - stop_glerror(); + + glBindFramebuffer(GL_FRAMEBUFFER, mFBO); + stop_glerror(); + if (gGLManager.mHasDrawBuffers) + { //setup multiple render targets + GLenum drawbuffers[] = {GL_COLOR_ATTACHMENT0, + GL_COLOR_ATTACHMENT1, + GL_COLOR_ATTACHMENT2, + GL_COLOR_ATTACHMENT3}; + glDrawBuffersARB(mTex.size(), drawbuffers); } - else - { - glBindFramebuffer(GL_FRAMEBUFFER, mFBO); - stop_glerror(); - if (gGLManager.mHasDrawBuffers) - { //setup multiple render targets - GLenum drawbuffers[] = {GL_COLOR_ATTACHMENT0, - GL_COLOR_ATTACHMENT1, - GL_COLOR_ATTACHMENT2, - GL_COLOR_ATTACHMENT3}; - glDrawBuffersARB(mTex.size(), drawbuffers); - } - if (mTex.empty()) - { //no color buffer to draw to - glDrawBuffer(GL_NONE); - glReadBuffer(GL_NONE); - } + if (mTex.empty()) + { //no color buffer to draw to + glDrawBuffer(GL_NONE); + glReadBuffer(GL_NONE); + } - check_framebuffer_status(); + check_framebuffer_status(); - stop_glerror(); - } + stop_glerror(); } glViewport(0, 0, mResX, mResY); @@ -406,50 +451,8 @@ void LLRenderTarget::flush(bool fetch_depth) else { stop_glerror(); - glBindFramebuffer(GL_FRAMEBUFFER, 0); - stop_glerror(); - - if (mSampleBuffer) - { - LLGLEnable multisample(GL_MULTISAMPLE); - stop_glerror(); - glBindFramebuffer(GL_FRAMEBUFFER, mFBO); - stop_glerror(); - check_framebuffer_status(); - glBindFramebuffer(GL_READ_FRAMEBUFFER, mSampleBuffer->mFBO); - check_framebuffer_status(); - - stop_glerror(); - glBlitFramebuffer(0, 0, mResX, mResY, 0, 0, mResX, mResY, GL_COLOR_BUFFER_BIT | GL_DEPTH_BUFFER_BIT | GL_STENCIL_BUFFER_BIT, GL_NEAREST); - stop_glerror(); - - if (mTex.size() > 1) - { - for (U32 i = 1; i < mTex.size(); ++i) - { - glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, - LLTexUnit::getInternalType(mUsage), mTex[i], 0); - stop_glerror(); - glFramebufferRenderbuffer(GL_READ_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_RENDERBUFFER, mSampleBuffer->mTex[i]); - stop_glerror(); - glBlitFramebuffer(0, 0, mResX, mResY, 0, 0, mResX, mResY, GL_COLOR_BUFFER_BIT, GL_NEAREST); - stop_glerror(); - } - - for (U32 i = 0; i < mTex.size(); ++i) - { - glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0+i, - LLTexUnit::getInternalType(mUsage), mTex[i], 0); - stop_glerror(); - glFramebufferRenderbuffer(GL_READ_FRAMEBUFFER, GL_COLOR_ATTACHMENT0+i, GL_RENDERBUFFER, mSampleBuffer->mTex[i]); - stop_glerror(); - } - } - } - - glBindFramebuffer(GL_FRAMEBUFFER, 0); } } @@ -466,37 +469,36 @@ void LLRenderTarget::copyContents(LLRenderTarget& source, S32 srcX0, S32 srcY0, llerrs << "Cannot copy framebuffer contents for non FBO render targets." << llendl; } - if (mSampleBuffer) + + if (mask == GL_DEPTH_BUFFER_BIT && source.mStencil != mStencil) { - mSampleBuffer->copyContents(source, srcX0, srcY0, srcX1, srcY1, dstX0, dstY0, dstX1, dstY1, mask, filter); + stop_glerror(); + + glBindFramebuffer(GL_FRAMEBUFFER, source.mFBO); + check_framebuffer_status(); + gGL.getTexUnit(0)->bind(this, true); + stop_glerror(); + glCopyTexSubImage2D(LLTexUnit::getInternalType(mUsage), 0, srcX0, srcY0, dstX0, dstY0, dstX1, dstY1); + stop_glerror(); + glBindFramebuffer(GL_FRAMEBUFFER, 0); + stop_glerror(); } else { - if (mask == GL_DEPTH_BUFFER_BIT && source.mStencil != mStencil) - { - stop_glerror(); - - glBindFramebuffer(GL_FRAMEBUFFER, source.mFBO); - gGL.getTexUnit(0)->bind(this, true); - stop_glerror(); - glCopyTexSubImage2D(LLTexUnit::getInternalType(mUsage), 0, srcX0, srcY0, dstX0, dstY0, dstX1, dstY1); - stop_glerror(); - glBindFramebuffer(GL_FRAMEBUFFER, 0); - stop_glerror(); - } - else - { - glBindFramebuffer(GL_READ_FRAMEBUFFER, source.mFBO); - stop_glerror(); - glBindFramebuffer(GL_DRAW_FRAMEBUFFER, mFBO); - stop_glerror(); - check_framebuffer_status(); - stop_glerror(); - glBlitFramebuffer(srcX0, srcY0, srcX1, srcY1, dstX0, dstY0, dstX1, dstY1, mask, filter); - stop_glerror(); - glBindFramebuffer(GL_FRAMEBUFFER, 0); - stop_glerror(); - } + glBindFramebuffer(GL_READ_FRAMEBUFFER, source.mFBO); + stop_glerror(); + glBindFramebuffer(GL_DRAW_FRAMEBUFFER, mFBO); + stop_glerror(); + check_framebuffer_status(); + stop_glerror(); + glBlitFramebuffer(srcX0, srcY0, srcX1, srcY1, dstX0, dstY0, dstX1, dstY1, mask, filter); + stop_glerror(); + glBindFramebuffer(GL_READ_FRAMEBUFFER, 0); + stop_glerror(); + glBindFramebuffer(GL_DRAW_FRAMEBUFFER, 0); + stop_glerror(); + glBindFramebuffer(GL_FRAMEBUFFER, 0); + stop_glerror(); } } @@ -539,179 +541,3 @@ void LLRenderTarget::getViewport(S32* viewport) viewport[3] = mResY; } -//================================================== -// LLMultisampleBuffer implementation -//================================================== -LLMultisampleBuffer::LLMultisampleBuffer() -{ - -} - -LLMultisampleBuffer::~LLMultisampleBuffer() -{ - release(); -} - -void LLMultisampleBuffer::release() -{ - if (mFBO) - { - glDeleteFramebuffers(1, (GLuint *) &mFBO); - mFBO = 0; - } - - if (mTex.size() > 0) - { - glDeleteRenderbuffers(mTex.size(), (GLuint *) &mTex[0]); - mTex.clear(); - } - - if (mDepth) - { - glDeleteRenderbuffers(1, (GLuint *) &mDepth); - mDepth = 0; - } -} - -void LLMultisampleBuffer::bindTarget() -{ - bindTarget(this); -} - -void LLMultisampleBuffer::bindTarget(LLRenderTarget* ref) -{ - if (!ref) - { - ref = this; - } - - glBindFramebuffer(GL_FRAMEBUFFER, mFBO); - if (gGLManager.mHasDrawBuffers) - { //setup multiple render targets - GLenum drawbuffers[] = {GL_COLOR_ATTACHMENT0, - GL_COLOR_ATTACHMENT1, - GL_COLOR_ATTACHMENT2, - GL_COLOR_ATTACHMENT3}; - glDrawBuffersARB(ref->mTex.size(), drawbuffers); - } - - check_framebuffer_status(); - - glViewport(0, 0, mResX, mResY); - - sBoundTarget = this; -} - -void LLMultisampleBuffer::allocate(U32 resx, U32 resy, U32 color_fmt, bool depth, bool stencil, LLTexUnit::eTextureType usage, bool use_fbo ) -{ - allocate(resx,resy,color_fmt,depth,stencil,usage,use_fbo,2); -} - -void LLMultisampleBuffer::allocate(U32 resx, U32 resy, U32 color_fmt, bool depth, bool stencil, LLTexUnit::eTextureType usage, bool use_fbo, U32 samples ) -{ - stop_glerror(); - mResX = resx; - mResY = resy; - - mUsage = usage; - mUseDepth = depth; - mStencil = stencil; - - release(); - - mSamples = samples; - - if (mSamples <= 1) - { - llerrs << "Cannot create a multisample buffer with less than 2 samples." << llendl; - } - - stop_glerror(); - - if ((sUseFBO || use_fbo) && gGLManager.mHasFramebufferObject) - { - - if (depth) - { - stop_glerror(); - allocateDepth(); - stop_glerror(); - } - - glGenFramebuffers(1, (GLuint *) &mFBO); - - glBindFramebuffer(GL_FRAMEBUFFER, mFBO); - - if (mDepth) - { - glFramebufferRenderbuffer(GL_FRAMEBUFFER, GL_DEPTH_ATTACHMENT, GL_RENDERBUFFER, mDepth); - if (mStencil) - { - glFramebufferRenderbuffer(GL_FRAMEBUFFER, GL_STENCIL_ATTACHMENT, GL_RENDERBUFFER, mDepth); - } - } - - stop_glerror(); - glBindFramebuffer(GL_FRAMEBUFFER, 0); - stop_glerror(); - } - - addColorAttachment(color_fmt); -} - -void LLMultisampleBuffer::addColorAttachment(U32 color_fmt) -{ - if (color_fmt == 0) - { - return; - } - - U32 offset = mTex.size(); - if (offset >= 4 || - (offset > 0 && (mFBO == 0 || !gGLManager.mHasDrawBuffers))) - { - llerrs << "Too many color attachments!" << llendl; - } - - U32 tex; - glGenRenderbuffers(1, &tex); - - glBindRenderbuffer(GL_RENDERBUFFER, tex); - glRenderbufferStorageMultisample(GL_RENDERBUFFER, mSamples, color_fmt, mResX, mResY); - stop_glerror(); - - if (mFBO) - { - glBindFramebuffer(GL_FRAMEBUFFER, mFBO); - glFramebufferRenderbuffer(GL_FRAMEBUFFER, GL_COLOR_ATTACHMENT0+offset, GL_RENDERBUFFER, tex); - stop_glerror(); - GLenum status = glCheckFramebufferStatus(GL_FRAMEBUFFER); - switch (status) - { - case GL_FRAMEBUFFER_COMPLETE: - break; - default: - llerrs << "WTF? " << std::hex << status << llendl; - break; - } - - glBindFramebuffer(GL_FRAMEBUFFER, 0); - } - - mTex.push_back(tex); -} - -void LLMultisampleBuffer::allocateDepth() -{ - glGenRenderbuffers(1, (GLuint* ) &mDepth); - glBindRenderbuffer(GL_RENDERBUFFER, mDepth); - if (mStencil) - { - glRenderbufferStorageMultisample(GL_RENDERBUFFER, mSamples, GL_DEPTH24_STENCIL8, mResX, mResY); - } - else - { - glRenderbufferStorageMultisample(GL_RENDERBUFFER, mSamples, GL_DEPTH_COMPONENT16, mResX, mResY); - } -} - diff --git a/indra/llrender/llrendertarget.h b/indra/llrender/llrendertarget.h index 12dd1c8b90..094b58b562 100644 --- a/indra/llrender/llrendertarget.h +++ b/indra/llrender/llrendertarget.h @@ -71,10 +71,7 @@ public: //allocate resources for rendering //must be called before use //multiple calls will release previously allocated resources - void allocate(U32 resx, U32 resy, U32 color_fmt, bool depth, bool stencil, LLTexUnit::eTextureType usage = LLTexUnit::TT_TEXTURE, bool use_fbo = FALSE); - - //provide this render target with a multisample resource. - void setSampleBuffer(LLMultisampleBuffer* buffer); + void allocate(U32 resx, U32 resy, U32 color_fmt, bool depth, bool stencil, LLTexUnit::eTextureType usage = LLTexUnit::TT_TEXTURE, bool use_fbo = false, S32 samples = 0); //add color buffer attachment //limit of 4 color attachments per render target @@ -141,7 +138,6 @@ public: static LLRenderTarget* getCurrentBoundTarget() { return sBoundTarget; } protected: - friend class LLMultisampleBuffer; U32 mResX; U32 mResY; std::vector<U32> mTex; @@ -152,26 +148,8 @@ protected: bool mRenderDepth; LLTexUnit::eTextureType mUsage; U32 mSamples; - LLMultisampleBuffer* mSampleBuffer; - - static LLRenderTarget* sBoundTarget; -}; - -class LLMultisampleBuffer : public LLRenderTarget -{ -public: - LLMultisampleBuffer(); - virtual ~LLMultisampleBuffer(); - - virtual void release(); - - virtual void bindTarget(); - void bindTarget(LLRenderTarget* ref); - virtual void allocate(U32 resx, U32 resy, U32 color_fmt, bool depth, bool stencil, LLTexUnit::eTextureType usage, bool use_fbo); - void allocate(U32 resx, U32 resy, U32 color_fmt, bool depth, bool stencil, LLTexUnit::eTextureType usage, bool use_fbo, U32 samples); - virtual void addColorAttachment(U32 color_fmt); - virtual void allocateDepth(); + static LLRenderTarget* sBoundTarget; }; #endif //!LL_MESA_HEADLESS diff --git a/indra/llrender/llshadermgr.cpp b/indra/llrender/llshadermgr.cpp index 98a0a93084..bdc103b917 100644 --- a/indra/llrender/llshadermgr.cpp +++ b/indra/llrender/llshadermgr.cpp @@ -209,17 +209,39 @@ BOOL LLShaderMgr::attachShaderFeatures(LLGLSLShader * shader) if (features->hasWaterFog) { - if (!shader->attachObject("lighting/lightWaterF.glsl")) + if (features->disableTextureIndex) { - return FALSE; + if (!shader->attachObject("lighting/lightWaterNonIndexedF.glsl")) + { + return FALSE; + } + } + else + { + if (!shader->attachObject("lighting/lightWaterF.glsl")) + { + return FALSE; + } + shader->mFeatures.mIndexedTextureChannels = gGLManager.mNumTextureImageUnits-1; } } else { - if (!shader->attachObject("lighting/lightF.glsl")) + if (features->disableTextureIndex) { - return FALSE; + if (!shader->attachObject("lighting/lightNonIndexedF.glsl")) + { + return FALSE; + } + } + else + { + if (!shader->attachObject("lighting/lightF.glsl")) + { + return FALSE; + } + shader->mFeatures.mIndexedTextureChannels = gGLManager.mNumTextureImageUnits-1; } } } @@ -230,32 +252,76 @@ BOOL LLShaderMgr::attachShaderFeatures(LLGLSLShader * shader) if (features->isShiny && features->hasWaterFog) { - if (!shader->attachObject("lighting/lightFullbrightShinyWaterF.glsl")) + if (features->disableTextureIndex) { - return FALSE; + if (!shader->attachObject("lighting/lightFullbrightShinyWaterNonIndexedF.glsl")) + { + return FALSE; + } + } + else + { + if (!shader->attachObject("lighting/lightFullbrightShinyWaterF.glsl")) + { + return FALSE; + } + shader->mFeatures.mIndexedTextureChannels = gGLManager.mNumTextureImageUnits-1; } } else if (features->hasWaterFog) { - if (!shader->attachObject("lighting/lightFullbrightWaterF.glsl")) + if (features->disableTextureIndex) { - return FALSE; + if (!shader->attachObject("lighting/lightFullbrightWaterNonIndexedF.glsl")) + { + return FALSE; + } + } + else + { + if (!shader->attachObject("lighting/lightFullbrightWaterF.glsl")) + { + return FALSE; + } + shader->mFeatures.mIndexedTextureChannels = gGLManager.mNumTextureImageUnits-1; } } else if (features->isShiny) { - if (!shader->attachObject("lighting/lightFullbrightShinyF.glsl")) + if (features->disableTextureIndex) { - return FALSE; + if (!shader->attachObject("lighting/lightFullbrightShinyNonIndexedF.glsl")) + { + return FALSE; + } + } + else + { + if (!shader->attachObject("lighting/lightFullbrightShinyF.glsl")) + { + return FALSE; + } + shader->mFeatures.mIndexedTextureChannels = gGLManager.mNumTextureImageUnits-1; } } else { - if (!shader->attachObject("lighting/lightFullbrightF.glsl")) + if (features->disableTextureIndex) { - return FALSE; + if (!shader->attachObject("lighting/lightFullbrightNonIndexedF.glsl")) + { + return FALSE; + } + } + else + { + if (!shader->attachObject("lighting/lightFullbrightF.glsl")) + { + return FALSE; + } + shader->mFeatures.mIndexedTextureChannels = gGLManager.mNumTextureImageUnits-1; } } } @@ -266,17 +332,39 @@ BOOL LLShaderMgr::attachShaderFeatures(LLGLSLShader * shader) if (features->hasWaterFog) { - if (!shader->attachObject("lighting/lightShinyWaterF.glsl")) + if (features->disableTextureIndex) { - return FALSE; + if (!shader->attachObject("lighting/lightShinyWaterNonIndexedF.glsl")) + { + return FALSE; + } + } + else + { + if (!shader->attachObject("lighting/lightShinyWaterF.glsl")) + { + return FALSE; + } + shader->mFeatures.mIndexedTextureChannels = gGLManager.mNumTextureImageUnits-1; } } else { - if (!shader->attachObject("lighting/lightShinyF.glsl")) + if (features->disableTextureIndex) { - return FALSE; + if (!shader->attachObject("lighting/lightShinyNonIndexedF.glsl")) + { + return FALSE; + } + } + else + { + if (!shader->attachObject("lighting/lightShinyF.glsl")) + { + return FALSE; + } + shader->mFeatures.mIndexedTextureChannels = gGLManager.mNumTextureImageUnits-1; } } } @@ -315,12 +403,12 @@ void LLShaderMgr::dumpObjectLog(GLhandleARB ret, BOOL warns) } else { - LL_INFOS("ShaderLoading") << log << LL_ENDL; + LL_DEBUGS("ShaderLoading") << log << LL_ENDL; } } } -GLhandleARB LLShaderMgr::loadShaderFile(const std::string& filename, S32 & shader_level, GLenum type) +GLhandleARB LLShaderMgr::loadShaderFile(const std::string& filename, S32 & shader_level, GLenum type, S32 texture_index_channels) { GLenum error = GL_NO_ERROR; if (gDebugGL) @@ -374,6 +462,106 @@ GLhandleARB LLShaderMgr::loadShaderFile(const std::string& filename, S32 & shade GLcharARB* text[1024]; GLuint count = 0; + if (gGLManager.mGLVersion < 3.f) + { + //set version to 1.20 + text[count++] = strdup("#version 120\n"); + } + else + { //set version to 1.30 + text[count++] = strdup("#version 130\n"); + } + + //copy preprocessor definitions into buffer + for (std::map<std::string,std::string>::iterator iter = mDefinitions.begin(); iter != mDefinitions.end(); ++iter) + { + std::string define = "#define " + iter->first + " " + iter->second + "\n"; + text[count++] = (GLcharARB *) strdup(define.c_str()); + } + + if (texture_index_channels > 0 && type == GL_FRAGMENT_SHADER_ARB) + { + //use specified number of texture channels for indexed texture rendering + + /* prepend shader code that looks like this: + + uniform sampler2D tex0; + uniform sampler2D tex1; + uniform sampler2D tex2; + . + . + . + uniform sampler2D texN; + + varying float vary_texture_index; + + vec4 diffuseLookup(vec2 texcoord) + { + switch (int(vary_texture_index+0.25)) + { + case 0: return texture2D(tex0, texcoord); + case 1: return texture2D(tex1, texcoord); + case 2: return texture2D(tex2, texcoord); + . + . + . + case N: return texture2D(texN, texcoord); + } + + return vec4(0,0,0,0); + } + */ + + //uniform declartion + for (S32 i = 0; i < texture_index_channels; ++i) + { + std::string decl = llformat("uniform sampler2D tex%d;\n", i); + text[count++] = strdup(decl.c_str()); + } + + text[count++] = strdup("varying float vary_texture_index;\n"); + text[count++] = strdup("vec4 diffuseLookup(vec2 texcoord)\n"); + text[count++] = strdup("{\n"); + + + if (gGLManager.mGLVersion >= 3.f) + { + text[count++] = strdup("\tswitch (int(vary_texture_index+0.25))\n"); + text[count++] = strdup("\t{\n"); + + //switch body + for (S32 i = 0; i < texture_index_channels; ++i) + { + std::string case_str = llformat("\t\tcase %d: return texture2D(tex%d, texcoord);\n", i, i); + text[count++] = strdup(case_str.c_str()); + } + + text[count++] = strdup("\t}\n"); + } + else + { + //switches aren't supported, make block that looks like: + /* + int ti = int(vary_texture_index+0.25); + if (ti == 0) return texture2D(tex0, texcoord); + if (ti == 1) return texture2D(tex1, texcoord); + . + . + . + if (ti == N) return texture2D(texN, texcoord); + */ + + text[count++] = strdup("int ti = int(vary_texture_index+0.25);\n"); + for (S32 i = 0; i < texture_index_channels; ++i) + { + std::string if_str = llformat("if (ti == %d) return texture2D(tex%d, texcoord);\n", i, i); + text[count++] = strdup(if_str.c_str()); + } + } + + text[count++] = strdup("\treturn vec4(0,0,0,0);\n"); + text[count++] = strdup("}\n"); + } //copy file into memory while( fgets((char *)buff, 1024, file) != NULL && count < LL_ARRAY_SIZE(buff) ) @@ -457,7 +645,7 @@ GLhandleARB LLShaderMgr::loadShaderFile(const std::string& filename, S32 & shade if (shader_level > 1) { shader_level--; - return loadShaderFile(filename,shader_level,type); + return loadShaderFile(filename,shader_level,type,texture_index_channels); } LL_WARNS("ShaderLoading") << "Failed to load " << filename << LL_ENDL; } diff --git a/indra/llrender/llshadermgr.h b/indra/llrender/llshadermgr.h index c54c4608d7..2f30103811 100644 --- a/indra/llrender/llshadermgr.h +++ b/indra/llrender/llshadermgr.h @@ -43,7 +43,7 @@ public: void dumpObjectLog(GLhandleARB ret, BOOL warns = TRUE); BOOL linkProgramObject(GLhandleARB obj, BOOL suppress_errors = FALSE); BOOL validateProgramObject(GLhandleARB obj); - GLhandleARB loadShaderFile(const std::string& filename, S32 & shader_level, GLenum type); + GLhandleARB loadShaderFile(const std::string& filename, S32 & shader_level, GLenum type, S32 texture_index_channels = -1); // Implemented in the application to actually point to the shader directory. virtual std::string getShaderDirPrefix(void) = 0; // Pure Virtual @@ -60,6 +60,9 @@ public: std::vector<std::string> mReservedUniforms; + //preprocessor definitions (name/value) + std::map<std::string, std::string> mDefinitions; + protected: // our parameter manager singleton instance diff --git a/indra/llrender/llvertexbuffer.cpp b/indra/llrender/llvertexbuffer.cpp index 8c9171ccf4..4a0b964e61 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 +965,45 @@ 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-index; + } + + 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; + break; + } + } + } + + if (!mapped) + { + //not already mapped, map new region + MappedRegion region(type, !sDisableVBOMapping && 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 +1012,95 @@ 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; +#ifdef GL_ARB_map_buffer_range + 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 +#else + llassert_always(!gGLManager.mHasMapBufferRange); +#endif + { + map_range = false; + src = (U8*) glMapBufferARB(GL_ARRAY_BUFFER_ARB, GL_WRITE_ONLY_ARB); + } + mMappedData = LL_NEXT_ALIGNED_ADDRESS<U8>(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 && !sDisableVBOMapping) + { + 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 +1112,41 @@ 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-index; + } + + 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; + break; + } + } + + if (!mapped) + { + //not already mapped, map new region + MappedRegion region(TYPE_INDEX, !sDisableVBOMapping && 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 +1156,36 @@ 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<U8>(src); + U8* src = NULL; +#ifdef GL_ARB_map_buffer_range + 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 +#else + llassert_always(!gGLManager.mHasMapBufferRange); +#endif + { + map_range = false; + src = (U8*) glMapBufferARB(GL_ELEMENT_ARRAY_BUFFER_ARB, GL_WRITE_ONLY_ARB); + } + + mMappedIndexData = src; //LL_NEXT_ALIGNED_ADDRESS<U8>(src); mAlignedIndexOffset = mMappedIndexData - src; stop_glerror(); } @@ -1068,31 +1214,81 @@ U8* LLVertexBuffer::mapIndexBuffer(S32 access) sMappedCount++; } + else + { + map_range = false; + } - return mMappedIndexData ; + if (map_range && !sDisableVBOMapping) + { + 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 if(sDisableVBOMapping) { - stop_glerror(); - glBufferSubDataARB(GL_ARRAY_BUFFER_ARB, 0, getSize(), mMappedData); - stop_glerror(); + 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; + glBufferSubDataARB(GL_ARRAY_BUFFER_ARB, offset, length, mMappedData+offset); + stop_glerror(); + } + + mMappedVertexRegions.clear(); + } + else + { + stop_glerror(); + glBufferSubDataARB(GL_ARRAY_BUFFER_ARB, 0, getSize(), mMappedData); + stop_glerror(); + } } else { +#ifdef GL_ARB_map_buffer_range + 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(); + } + } +#else + llassert_always(!gGLManager.mHasMapBufferRange); +#endif stop_glerror(); glUnmapBufferARB(GL_ARRAY_BUFFER_ARB); stop_glerror(); @@ -1103,17 +1299,53 @@ 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) { - stop_glerror(); - glBufferSubDataARB(GL_ELEMENT_ARRAY_BUFFER_ARB, 0, getIndicesSize(), mMappedIndexData); - stop_glerror(); + 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; + glBufferSubDataARB(GL_ELEMENT_ARRAY_BUFFER_ARB, offset, length, mMappedIndexData+offset); + stop_glerror(); + } + + mMappedIndexRegions.clear(); + } + else + { + stop_glerror(); + glBufferSubDataARB(GL_ELEMENT_ARRAY_BUFFER_ARB, 0, getIndicesSize(), mMappedIndexData); + stop_glerror(); + } } else { +#ifdef GL_ARB_map_buffer_range + 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(); + } + } +#else + llassert_always(!gGLManager.mHasMapBufferRange); +#endif stop_glerror(); glUnmapBufferARB(GL_ELEMENT_ARRAY_BUFFER_ARB); stop_glerror(); @@ -1152,19 +1384,19 @@ template <class T,S32 type> struct VertexBufferStrider typedef LLStrider<T> 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); + U8* ptr = vbo.mapIndexBuffer(index, count, map_range); - if (vbo.mapIndexBuffer() == NULL) + 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 +1404,15 @@ template <class T,S32 type> 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 +1424,48 @@ template <class T,S32 type> struct VertexBufferStrider } }; -bool LLVertexBuffer::getVertexStrider(LLStrider<LLVector3>& strider, S32 index) -{ - return VertexBufferStrider<LLVector3,TYPE_VERTEX>::get(*this, strider, index); -} -bool LLVertexBuffer::getIndexStrider(LLStrider<U16>& strider, S32 index) +bool LLVertexBuffer::getVertexStrider(LLStrider<LLVector3>& strider, S32 index, S32 count, bool map_range) { - return VertexBufferStrider<U16,TYPE_INDEX>::get(*this, strider, index); + return VertexBufferStrider<LLVector3,TYPE_VERTEX>::get(*this, strider, index, count, map_range); } -bool LLVertexBuffer::getTexCoord0Strider(LLStrider<LLVector2>& strider, S32 index) +bool LLVertexBuffer::getIndexStrider(LLStrider<U16>& strider, S32 index, S32 count, bool map_range) { - return VertexBufferStrider<LLVector2,TYPE_TEXCOORD0>::get(*this, strider, index); + return VertexBufferStrider<U16,TYPE_INDEX>::get(*this, strider, index, count, map_range); } -bool LLVertexBuffer::getTexCoord1Strider(LLStrider<LLVector2>& strider, S32 index) +bool LLVertexBuffer::getTexCoord0Strider(LLStrider<LLVector2>& strider, S32 index, S32 count, bool map_range) { - return VertexBufferStrider<LLVector2,TYPE_TEXCOORD1>::get(*this, strider, index); + return VertexBufferStrider<LLVector2,TYPE_TEXCOORD0>::get(*this, strider, index, count, map_range); } -/*bool LLVertexBuffer::getTexCoord2Strider(LLStrider<LLVector2>& strider, S32 index) +bool LLVertexBuffer::getTexCoord1Strider(LLStrider<LLVector2>& strider, S32 index, S32 count, bool map_range) { - return VertexBufferStrider<LLVector2,TYPE_TEXCOORD2>::get(*this, strider, index); + return VertexBufferStrider<LLVector2,TYPE_TEXCOORD1>::get(*this, strider, index, count, map_range); } -bool LLVertexBuffer::getTexCoord3Strider(LLStrider<LLVector2>& strider, S32 index) -{ - return VertexBufferStrider<LLVector2,TYPE_TEXCOORD3>::get(*this, strider, index); -}*/ -bool LLVertexBuffer::getNormalStrider(LLStrider<LLVector3>& strider, S32 index) + +bool LLVertexBuffer::getNormalStrider(LLStrider<LLVector3>& strider, S32 index, S32 count, bool map_range) { - return VertexBufferStrider<LLVector3,TYPE_NORMAL>::get(*this, strider, index); + return VertexBufferStrider<LLVector3,TYPE_NORMAL>::get(*this, strider, index, count, map_range); } -bool LLVertexBuffer::getBinormalStrider(LLStrider<LLVector3>& strider, S32 index) +bool LLVertexBuffer::getBinormalStrider(LLStrider<LLVector3>& strider, S32 index, S32 count, bool map_range) { - return VertexBufferStrider<LLVector3,TYPE_BINORMAL>::get(*this, strider, index); + return VertexBufferStrider<LLVector3,TYPE_BINORMAL>::get(*this, strider, index, count, map_range); } -bool LLVertexBuffer::getColorStrider(LLStrider<LLColor4U>& strider, S32 index) +bool LLVertexBuffer::getColorStrider(LLStrider<LLColor4U>& strider, S32 index, S32 count, bool map_range) { - return VertexBufferStrider<LLColor4U,TYPE_COLOR>::get(*this, strider, index); + return VertexBufferStrider<LLColor4U,TYPE_COLOR>::get(*this, strider, index, count, map_range); } -bool LLVertexBuffer::getWeightStrider(LLStrider<F32>& strider, S32 index) +bool LLVertexBuffer::getWeightStrider(LLStrider<F32>& strider, S32 index, S32 count, bool map_range) { - return VertexBufferStrider<F32,TYPE_WEIGHT>::get(*this, strider, index); + return VertexBufferStrider<F32,TYPE_WEIGHT>::get(*this, strider, index, count, map_range); } -bool LLVertexBuffer::getWeight4Strider(LLStrider<LLVector4>& strider, S32 index) +bool LLVertexBuffer::getWeight4Strider(LLStrider<LLVector4>& strider, S32 index, S32 count, bool map_range) { - return VertexBufferStrider<LLVector4,TYPE_WEIGHT4>::get(*this, strider, index); + return VertexBufferStrider<LLVector4,TYPE_WEIGHT4>::get(*this, strider, index, count, map_range); } -bool LLVertexBuffer::getClothWeightStrider(LLStrider<LLVector4>& strider, S32 index) +bool LLVertexBuffer::getClothWeightStrider(LLStrider<LLVector4>& strider, S32 index, S32 count, bool map_range) { - return VertexBufferStrider<LLVector4,TYPE_CLOTHWEIGHT>::get(*this, strider, index); + return VertexBufferStrider<LLVector4,TYPE_CLOTHWEIGHT>::get(*this, strider, index, count, map_range); } //---------------------------------------------------------------------------- @@ -1497,17 +1724,16 @@ void LLVertexBuffer::setupVertexBuffer(U32 data_mask) const } if (data_mask & MAP_VERTEX) { - glVertexPointer(3,GL_FLOAT, LLVertexBuffer::sTypeSize[TYPE_VERTEX], (void*)(base + 0)); + if (data_mask & MAP_TEXTURE_INDEX) + { + glVertexPointer(4,GL_FLOAT, LLVertexBuffer::sTypeSize[TYPE_VERTEX], (void*)(base + 0)); + } + else + { + glVertexPointer(3,GL_FLOAT, LLVertexBuffer::sTypeSize[TYPE_VERTEX], (void*)(base + 0)); + } } 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 a9f22193f8..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; @@ -130,6 +142,9 @@ public: TYPE_CLOTHWEIGHT, TYPE_MAX, TYPE_INDEX, + + //no actual additional data, but indicates position.w is texture index + TYPE_TEXTURE_INDEX, }; enum { MAP_VERTEX = (1<<TYPE_VERTEX), @@ -144,6 +159,7 @@ public: MAP_WEIGHT = (1<<TYPE_WEIGHT), MAP_WEIGHT4 = (1<<TYPE_WEIGHT4), MAP_CLOTHWEIGHT = (1<<TYPE_CLOTHWEIGHT), + MAP_TEXTURE_INDEX = (1<<TYPE_TEXTURE_INDEX), }; protected: @@ -173,8 +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 @@ -189,16 +205,16 @@ public: // vb->getNormalStrider(norms); // setVertsNorms(verts, norms); // vb->unmapBuffer(); - bool getVertexStrider(LLStrider<LLVector3>& strider, S32 index=0); - bool getIndexStrider(LLStrider<U16>& strider, S32 index=0); - bool getTexCoord0Strider(LLStrider<LLVector2>& strider, S32 index=0); - bool getTexCoord1Strider(LLStrider<LLVector2>& strider, S32 index=0); - bool getNormalStrider(LLStrider<LLVector3>& strider, S32 index=0); - bool getBinormalStrider(LLStrider<LLVector3>& strider, S32 index=0); - bool getColorStrider(LLStrider<LLColor4U>& strider, S32 index=0); - bool getWeightStrider(LLStrider<F32>& strider, S32 index=0); - bool getWeight4Strider(LLStrider<LLVector4>& strider, S32 index=0); - bool getClothWeightStrider(LLStrider<LLVector4>& strider, S32 index=0); + bool getVertexStrider(LLStrider<LLVector3>& strider, S32 index=0, S32 count = -1, bool map_range = false); + bool getIndexStrider(LLStrider<U16>& strider, S32 index=0, S32 count = -1, bool map_range = false); + bool getTexCoord0Strider(LLStrider<LLVector2>& strider, S32 index=0, S32 count = -1, bool map_range = false); + bool getTexCoord1Strider(LLStrider<LLVector2>& strider, S32 index=0, S32 count = -1, bool map_range = false); + bool getNormalStrider(LLStrider<LLVector3>& strider, S32 index=0, S32 count = -1, bool map_range = false); + bool getBinormalStrider(LLStrider<LLVector3>& strider, S32 index=0, S32 count = -1, bool map_range = false); + bool getColorStrider(LLStrider<LLColor4U>& strider, S32 index=0, S32 count = -1, bool map_range = false); + bool getWeightStrider(LLStrider<F32>& strider, S32 index=0, S32 count = -1, bool map_range = false); + bool getWeight4Strider(LLStrider<LLVector4>& strider, S32 index=0, S32 count = -1, bool map_range = false); + bool getClothWeightStrider(LLStrider<LLVector4>& strider, S32 index=0, S32 count = -1, bool map_range = false); BOOL isEmpty() const { return mEmpty; } BOOL isLocked() const { return mVertexLocked || mIndexLocked; } @@ -218,8 +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; @@ -253,20 +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<DirtyRegion> mDirtyRegions; //vector of dirty regions to rebuild + std::vector<MappedRegion> mMappedVertexRegions; + std::vector<MappedRegion> mMappedIndexRegions; public: static S32 sCount; |