diff options
Diffstat (limited to 'indra/llrender')
-rw-r--r-- | indra/llrender/llcubemap.cpp | 19 | ||||
-rw-r--r-- | indra/llrender/llgl.cpp | 320 | ||||
-rw-r--r-- | indra/llrender/llgl.h | 14 | ||||
-rw-r--r-- | indra/llrender/llglheaders.h | 52 | ||||
-rw-r--r-- | indra/llrender/llglslshader.cpp | 28 | ||||
-rw-r--r-- | indra/llrender/llglslshader.h | 6 | ||||
-rw-r--r-- | indra/llrender/llimagegl.cpp | 108 | ||||
-rw-r--r-- | indra/llrender/llpostprocess.cpp | 22 | ||||
-rw-r--r-- | indra/llrender/llrender.cpp | 700 | ||||
-rw-r--r-- | indra/llrender/llrender.h | 55 | ||||
-rw-r--r-- | indra/llrender/llrendersphere.cpp | 97 | ||||
-rw-r--r-- | indra/llrender/llrendersphere.h | 5 | ||||
-rw-r--r-- | indra/llrender/llrendertarget.cpp | 112 | ||||
-rw-r--r-- | indra/llrender/llrendertarget.h | 2 | ||||
-rw-r--r-- | indra/llrender/llshadermgr.cpp | 114 | ||||
-rw-r--r-- | indra/llrender/llvertexbuffer.cpp | 1001 | ||||
-rw-r--r-- | indra/llrender/llvertexbuffer.h | 46 |
17 files changed, 1979 insertions, 722 deletions
diff --git a/indra/llrender/llcubemap.cpp b/indra/llrender/llcubemap.cpp index 1b10354c22..a75f01c708 100644 --- a/indra/llrender/llcubemap.cpp +++ b/indra/llrender/llcubemap.cpp @@ -36,6 +36,7 @@ #include "m4math.h" #include "llrender.h" +#include "llglslshader.h" #include "llglheaders.h" @@ -195,7 +196,7 @@ void LLCubeMap::enableTexture(S32 stage) void LLCubeMap::enableTextureCoords(S32 stage) { mTextureCoordStage = stage; - if (gGLManager.mHasCubeMap && stage >= 0 && LLCubeMap::sUseCubeMaps) + if (!LLGLSLShader::sNoFixedFunction && gGLManager.mHasCubeMap && stage >= 0 && LLCubeMap::sUseCubeMaps) { if (stage > 0) { @@ -237,7 +238,7 @@ void LLCubeMap::disableTexture(void) void LLCubeMap::disableTextureCoords(void) { - if (gGLManager.mHasCubeMap && mTextureCoordStage >= 0 && LLCubeMap::sUseCubeMaps) + if (!LLGLSLShader::sNoFixedFunction && gGLManager.mHasCubeMap && mTextureCoordStage >= 0 && LLCubeMap::sUseCubeMaps) { if (mTextureCoordStage > 0) { @@ -273,10 +274,10 @@ void LLCubeMap::setMatrix(S32 stage) LLMatrix4 trans(mat3); trans.transpose(); - glMatrixMode(GL_TEXTURE); - glPushMatrix(); - glLoadMatrixf((F32 *)trans.mMatrix); - glMatrixMode(GL_MODELVIEW); + gGL.matrixMode(LLRender::MM_TEXTURE); + gGL.pushMatrix(); + gGL.loadMatrix((F32 *)trans.mMatrix); + gGL.matrixMode(LLRender::MM_MODELVIEW); /*if (stage > 0) { @@ -292,9 +293,9 @@ void LLCubeMap::restoreMatrix() { gGL.getTexUnit(mMatrixStage)->activate(); } - glMatrixMode(GL_TEXTURE); - glPopMatrix(); - glMatrixMode(GL_MODELVIEW); + gGL.matrixMode(LLRender::MM_TEXTURE); + gGL.popMatrix(); + gGL.matrixMode(LLRender::MM_MODELVIEW); /*if (mMatrixStage > 0) { diff --git a/indra/llrender/llgl.cpp b/indra/llrender/llgl.cpp index 4e3cfb9c8a..c0d15fe6f3 100644 --- a/indra/llrender/llgl.cpp +++ b/indra/llrender/llgl.cpp @@ -67,6 +67,36 @@ static const std::string HEADLESS_VERSION_STRING("1.0"); std::ofstream gFailLog; +#if GL_ARB_debug_output + +#ifndef APIENTRY +#define APIENTRY +#endif + +void APIENTRY gl_debug_callback(GLenum source, + GLenum type, + GLuint id, + GLenum severity, + GLsizei length, + const GLchar* message, + GLvoid* userParam) +{ + if (severity == GL_DEBUG_SEVERITY_HIGH_ARB) + { + llwarns << "----- GL ERROR --------" << llendl; + } + else + { + llwarns << "----- GL WARNING -------" << llendl; + } + llwarns << "Type: " << std::hex << type << llendl; + llwarns << "ID: " << std::hex << id << llendl; + llwarns << "Severity: " << std::hex << severity << llendl; + llwarns << "Message: " << message << llendl; + llwarns << "-----------------------" << llendl; +} +#endif + void ll_init_fail_log(std::string filename) { gFailLog.open(filename.c_str()); @@ -110,6 +140,11 @@ std::list<LLGLUpdate*> LLGLUpdate::sGLQ; #if (LL_WINDOWS || LL_LINUX || LL_SOLARIS) && !LL_MESA_HEADLESS // ATI prototypes + +#if LL_WINDOWS +PFNGLGETSTRINGIPROC glGetStringi = NULL; +#endif + // vertex blending prototypes PFNGLWEIGHTPOINTERARBPROC glWeightPointerARB = NULL; PFNGLVERTEXBLENDARBPROC glVertexBlendARB = NULL; @@ -128,6 +163,12 @@ PFNGLUNMAPBUFFERARBPROC glUnmapBufferARB = NULL; PFNGLGETBUFFERPARAMETERIVARBPROC glGetBufferParameterivARB = NULL; PFNGLGETBUFFERPOINTERVARBPROC glGetBufferPointervARB = NULL; +//GL_ARB_vertex_array_object +PFNGLBINDVERTEXARRAYPROC glBindVertexArray = NULL; +PFNGLDELETEVERTEXARRAYSPROC glDeleteVertexArrays = NULL; +PFNGLGENVERTEXARRAYSPROC glGenVertexArrays = NULL; +PFNGLISVERTEXARRAYPROC glIsVertexArray = NULL; + // GL_ARB_map_buffer_range PFNGLMAPBUFFERRANGEPROC glMapBufferRange = NULL; PFNGLFLUSHMAPPEDBUFFERRANGEPROC glFlushMappedBufferRange = NULL; @@ -197,10 +238,16 @@ PFNGLRENDERBUFFERSTORAGEMULTISAMPLEPROC glRenderbufferStorageMultisample = NULL; PFNGLFRAMEBUFFERTEXTURELAYERPROC glFramebufferTextureLayer = NULL; //GL_ARB_texture_multisample -PFNGLTEXIMAGE2DMULTISAMPLEPROC glTexImage2DMultisample; -PFNGLTEXIMAGE3DMULTISAMPLEPROC glTexImage3DMultisample; -PFNGLGETMULTISAMPLEFVPROC glGetMultisamplefv; -PFNGLSAMPLEMASKIPROC glSampleMaski; +PFNGLTEXIMAGE2DMULTISAMPLEPROC glTexImage2DMultisample = NULL; +PFNGLTEXIMAGE3DMULTISAMPLEPROC glTexImage3DMultisample = NULL; +PFNGLGETMULTISAMPLEFVPROC glGetMultisamplefv = NULL; +PFNGLSAMPLEMASKIPROC glSampleMaski = NULL; + +//GL_ARB_debug_output +PFNGLDEBUGMESSAGECONTROLARBPROC glDebugMessageControlARB = NULL; +PFNGLDEBUGMESSAGEINSERTARBPROC glDebugMessageInsertARB = NULL; +PFNGLDEBUGMESSAGECALLBACKARBPROC glDebugMessageCallbackARB = NULL; +PFNGLGETDEBUGMESSAGELOGARBPROC glGetDebugMessageLogARB = NULL; // GL_EXT_blend_func_separate PFNGLBLENDFUNCSEPARATEEXTPROC glBlendFuncSeparateEXT = NULL; @@ -249,6 +296,10 @@ PFNGLGETUNIFORMFVARBPROC glGetUniformfvARB = NULL; PFNGLGETUNIFORMIVARBPROC glGetUniformivARB = NULL; PFNGLGETSHADERSOURCEARBPROC glGetShaderSourceARB = NULL; +#if LL_WINDOWS +PFNWGLCREATECONTEXTATTRIBSARBPROC wglCreateContextAttribsARB = NULL; +#endif + // vertex shader prototypes #if LL_LINUX || LL_SOLARIS PFNGLVERTEXATTRIB1DARBPROC glVertexAttrib1dARB = NULL; @@ -349,6 +400,7 @@ LLGLManager::LLGLManager() : mHasBlendFuncSeparate(FALSE), mHasSync(FALSE), mHasVertexBufferObject(FALSE), + mHasVertexArrayObject(FALSE), mHasMapBufferRange(FALSE), mHasFlushBufferRange(FALSE), mHasPBuffer(FALSE), @@ -370,6 +422,7 @@ LLGLManager::LLGLManager() : mHasAnisotropic(FALSE), mHasARBEnvCombine(FALSE), mHasCubeMap(FALSE), + mHasDebugOutput(FALSE), mIsATI(FALSE), mIsNVIDIA(FALSE), @@ -409,6 +462,15 @@ void LLGLManager::initWGL() LL_WARNS("RenderInit") << "No ARB pixel format extensions" << LL_ENDL; } + if (ExtensionExists("WGL_ARB_create_context",gGLHExts.mSysExts)) + { + GLH_EXT_NAME(wglCreateContextAttribsARB) = (PFNWGLCREATECONTEXTATTRIBSARBPROC)GLH_EXT_GET_PROC_ADDRESS("wglCreateContextAttribsARB"); + } + else + { + LL_WARNS("RenderInit") << "No ARB create context extensions" << LL_ENDL; + } + if (ExtensionExists("WGL_EXT_swap_control", gGLHExts.mSysExts)) { GLH_EXT_NAME(wglSwapIntervalEXT) = (PFNWGLSWAPINTERVALEXTPROC)GLH_EXT_GET_PROC_ADDRESS("wglSwapIntervalEXT"); @@ -438,13 +500,45 @@ bool LLGLManager::initGL() LL_ERRS("RenderInit") << "Calling init on LLGLManager after already initialized!" << LL_ENDL; } - GLint alpha_bits; - glGetIntegerv( GL_ALPHA_BITS, &alpha_bits ); - if( 8 != alpha_bits ) + stop_glerror(); + +#if LL_WINDOWS + if (!glGetStringi) { - LL_WARNS("RenderInit") << "Frame buffer has less than 8 bits of alpha. Avatar texture compositing will fail." << LL_ENDL; + glGetStringi = (PFNGLGETSTRINGIPROC) GLH_EXT_GET_PROC_ADDRESS("glGetStringi"); } + //reload extensions string (may have changed after using wglCreateContextAttrib) + if (glGetStringi) + { + std::stringstream str; + + GLint count = 0; + glGetIntegerv(GL_NUM_EXTENSIONS, &count); + for (GLint i = 0; i < count; ++i) + { + std::string ext((const char*) glGetStringi(GL_EXTENSIONS, i)); + str << ext << " "; + LL_DEBUGS("GLExtensions") << ext << llendl; + } + + { + PFNWGLGETEXTENSIONSSTRINGARBPROC wglGetExtensionsStringARB = 0; + wglGetExtensionsStringARB = (PFNWGLGETEXTENSIONSSTRINGARBPROC)wglGetProcAddress("wglGetExtensionsStringARB"); + if(wglGetExtensionsStringARB) + { + str << (const char*) wglGetExtensionsStringARB(wglGetCurrentDC()); + } + } + + free(gGLHExts.mSysExts); + std::string extensions = str.str(); + gGLHExts.mSysExts = strdup(extensions.c_str()); + } +#endif + + stop_glerror(); + // Extract video card strings and convert to upper case to // work around driver-to-driver variation in capitalization. mGLVendor = std::string((const char *)glGetString(GL_VENDOR)); @@ -531,8 +625,12 @@ bool LLGLManager::initGL() mGLVendorShort = "MISC"; } + stop_glerror(); // This is called here because it depends on the setting of mIsGF2or4MX, and sets up mHasMultitexture. initExtensions(); + stop_glerror(); + + S32 old_vram = mVRAM; if (mHasATIMemInfo) { //ask the gl how much vram is free at startup and attempt to use no more than half of that @@ -548,7 +646,27 @@ bool LLGLManager::initGL() mVRAM = dedicated_memory/1024; } - if (mHasMultitexture) + if (mVRAM < 256) + { //something likely went wrong using the above extensions, fall back to old method + mVRAM = old_vram; + } + + stop_glerror(); + + stop_glerror(); + + 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 (LLRender::sGLCoreProfile) + { + mNumTextureUnits = llmin(mNumTextureImageUnits, MAX_GL_TEXTURE_UNITS); + } + else if (mHasMultitexture) { GLint num_tex_units; glGetIntegerv(GL_MAX_TEXTURE_UNITS_ARB, &num_tex_units); @@ -567,12 +685,7 @@ 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); - } + stop_glerror(); if (mHasTextureMultisample) { @@ -582,6 +695,21 @@ bool LLGLManager::initGL() glGetIntegerv(GL_MAX_SAMPLE_MASK_WORDS, &mMaxSampleMaskWords); } + stop_glerror(); + +#if LL_WINDOWS + if (mHasDebugOutput && gDebugGL) + { //setup debug output callback + //glDebugMessageControlARB(GL_DONT_CARE, GL_DONT_CARE, GL_DEBUG_SEVERITY_LOW_ARB, 0, NULL, GL_TRUE); + glDebugMessageCallbackARB((GLDEBUGPROCARB) gl_debug_callback, NULL); + glEnable(GL_DEBUG_OUTPUT_SYNCHRONOUS_ARB); + } +#endif + + stop_glerror(); + + //HACK always disable texture multisample, use FXAA instead + mHasTextureMultisample = FALSE; #if LL_WINDOWS if (mIsATI) { //using multisample textures on ATI results in black screen for some reason @@ -593,10 +721,17 @@ bool LLGLManager::initGL() { glGetIntegerv(GL_MAX_SAMPLES, &mMaxSamples); } + + stop_glerror(); setToDebugGPU(); + stop_glerror(); + initGLStates(); + + stop_glerror(); + return true; } @@ -774,7 +909,7 @@ void LLGLManager::initExtensions() mHasVertexShader = FALSE; mHasFragmentShader = FALSE; mHasTextureRectangle = FALSE; -#else // LL_MESA_HEADLESS +#else // LL_MESA_HEADLESS //important, gGLHExts.mSysExts is uninitialized until after glh_init_extensions is called mHasMultitexture = glh_init_extensions("GL_ARB_multitexture"); mHasATIMemInfo = ExtensionExists("GL_ATI_meminfo", gGLHExts.mSysExts); mHasNVXMemInfo = ExtensionExists("GL_NVX_gpu_memory_info", gGLHExts.mSysExts); @@ -788,6 +923,7 @@ 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); + mHasVertexArrayObject = ExtensionExists("GL_ARB_vertex_array_object", gGLHExts.mSysExts); mHasSync = ExtensionExists("GL_ARB_sync", gGLHExts.mSysExts); mHasMapBufferRange = ExtensionExists("GL_ARB_map_buffer_range", gGLHExts.mSysExts); mHasFlushBufferRange = ExtensionExists("GL_APPLE_flush_buffer_range", gGLHExts.mSysExts); @@ -806,13 +942,14 @@ void LLGLManager::initExtensions() 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); + mHasDebugOutput = ExtensionExists("GL_ARB_debug_output", gGLHExts.mSysExts); #if !LL_DARWIN mHasPointParameters = !mIsATI && ExtensionExists("GL_ARB_point_parameters", gGLHExts.mSysExts); #endif - mHasShaderObjects = ExtensionExists("GL_ARB_shader_objects", gGLHExts.mSysExts) && ExtensionExists("GL_ARB_shading_language_100", gGLHExts.mSysExts); + mHasShaderObjects = ExtensionExists("GL_ARB_shader_objects", gGLHExts.mSysExts) && (LLRender::sGLCoreProfile || ExtensionExists("GL_ARB_shading_language_100", gGLHExts.mSysExts)); mHasVertexShader = ExtensionExists("GL_ARB_vertex_program", gGLHExts.mSysExts) && ExtensionExists("GL_ARB_vertex_shader", gGLHExts.mSysExts) - && ExtensionExists("GL_ARB_shading_language_100", gGLHExts.mSysExts); - mHasFragmentShader = ExtensionExists("GL_ARB_fragment_shader", gGLHExts.mSysExts) && ExtensionExists("GL_ARB_shading_language_100", gGLHExts.mSysExts); + && (LLRender::sGLCoreProfile || ExtensionExists("GL_ARB_shading_language_100", gGLHExts.mSysExts)); + mHasFragmentShader = ExtensionExists("GL_ARB_fragment_shader", gGLHExts.mSysExts) && (LLRender::sGLCoreProfile || ExtensionExists("GL_ARB_shading_language_100", gGLHExts.mSysExts)); #endif #if LL_LINUX || LL_SOLARIS @@ -985,6 +1122,13 @@ void LLGLManager::initExtensions() mHasVertexBufferObject = FALSE; } } + if (mHasVertexArrayObject) + { + glBindVertexArray = (PFNGLBINDVERTEXARRAYPROC) GLH_EXT_GET_PROC_ADDRESS("glBindVertexArray"); + glDeleteVertexArrays = (PFNGLDELETEVERTEXARRAYSPROC) GLH_EXT_GET_PROC_ADDRESS("glDeleteVertexArrays"); + glGenVertexArrays = (PFNGLGENVERTEXARRAYSPROC) GLH_EXT_GET_PROC_ADDRESS("glGenVertexArrays"); + glIsVertexArray = (PFNGLISVERTEXARRAYPROC) GLH_EXT_GET_PROC_ADDRESS("glIsVertexArray"); + } if (mHasSync) { glFenceSync = (PFNGLFENCESYNCPROC) GLH_EXT_GET_PROC_ADDRESS("glFenceSync"); @@ -1039,6 +1183,13 @@ void LLGLManager::initExtensions() glGetMultisamplefv = (PFNGLGETMULTISAMPLEFVPROC) GLH_EXT_GET_PROC_ADDRESS("glGetMultisamplefv"); glSampleMaski = (PFNGLSAMPLEMASKIPROC) GLH_EXT_GET_PROC_ADDRESS("glSampleMaski"); } + if (mHasDebugOutput) + { + glDebugMessageControlARB = (PFNGLDEBUGMESSAGECONTROLARBPROC) GLH_EXT_GET_PROC_ADDRESS("glDebugMessageControlARB"); + glDebugMessageInsertARB = (PFNGLDEBUGMESSAGEINSERTARBPROC) GLH_EXT_GET_PROC_ADDRESS("glDebugMessageInsertARB"); + glDebugMessageCallbackARB = (PFNGLDEBUGMESSAGECALLBACKARBPROC) GLH_EXT_GET_PROC_ADDRESS("glDebugMessageCallbackARB"); + glGetDebugMessageLogARB = (PFNGLGETDEBUGMESSAGELOGARBPROC) GLH_EXT_GET_PROC_ADDRESS("glGetDebugMessageLogARB"); + } #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"); @@ -1193,7 +1344,7 @@ void rotate_quat(LLQuaternion& rotation) { F32 angle_radians, x, y, z; rotation.getAngleAxis(&angle_radians, &x, &y, &z); - glRotatef(angle_radians * RAD_TO_DEG, x, y, z); + gGL.rotatef(angle_radians * RAD_TO_DEG, x, y, z); } void flush_glerror() @@ -1230,10 +1381,6 @@ void log_glerror() void do_assert_glerror() { - if (LL_UNLIKELY(!gGLManager.mInited)) - { - LL_ERRS("RenderInit") << "GL not initialized" << LL_ENDL; - } // Create or update texture to be used with this data GLenum error; error = glGetError(); @@ -1326,11 +1473,6 @@ void LLGLState::initClass() //make sure multisample defaults to disabled sStateMap[GL_MULTISAMPLE_ARB] = GL_FALSE; glDisable(GL_MULTISAMPLE_ARB); - - sStateMap[GL_MULTISAMPLE_ARB] = GL_FALSE; - glDisable(GL_MULTISAMPLE_ARB); - - glEnableClientState(GL_VERTEX_ARRAY); } //static @@ -1604,7 +1746,7 @@ void LLGLState::checkTextureChannels(const std::string& msg) void LLGLState::checkClientArrays(const std::string& msg, U32 data_mask) { - if (!gDebugGL) + if (!gDebugGL || LLGLSLShader::sNoFixedFunction) { return; } @@ -1661,7 +1803,7 @@ void LLGLState::checkClientArrays(const std::string& msg, U32 data_mask) }; - for (S32 j = 0; j < 4; j++) + for (S32 j = 1; j < 4; j++) { if (glIsEnabled(value[j])) { @@ -1783,17 +1925,26 @@ LLGLState::LLGLState(LLGLenum state, S32 enabled) : mState(state), mWasEnabled(FALSE), mIsEnabled(FALSE) { if (LLGLSLShader::sNoFixedFunction) - { //always disable state that's deprecated post GL 3.0 + { //always ignore state that's deprecated post GL 3.0 switch (state) { case GL_ALPHA_TEST: - enabled = 0; + case GL_NORMALIZE: + case GL_TEXTURE_GEN_R: + case GL_TEXTURE_GEN_S: + case GL_TEXTURE_GEN_T: + case GL_TEXTURE_GEN_Q: + case GL_LIGHTING: + case GL_COLOR_MATERIAL: + case GL_FOG: + case GL_LINE_STIPPLE: + mState = 0; break; } } stop_glerror(); - if (state) + if (mState) { mWasEnabled = sStateMap[state]; llassert(mWasEnabled == glIsEnabled(state)); @@ -1875,79 +2026,6 @@ void LLGLManager::initGLStates() //////////////////////////////////////////////////////////////////////////////// -void enable_vertex_weighting(const S32 index) -{ -#if GL_ARB_vertex_program - if (index > 0) glEnableVertexAttribArrayARB(index); // vertex weights -#endif -} - -void disable_vertex_weighting(const S32 index) -{ -#if GL_ARB_vertex_program - if (index > 0) glDisableVertexAttribArrayARB(index); // vertex weights -#endif -} - -void enable_binormals(const S32 index) -{ -#if GL_ARB_vertex_program - if (index > 0) - { - glEnableVertexAttribArrayARB(index); // binormals - } -#endif -} - -void disable_binormals(const S32 index) -{ -#if GL_ARB_vertex_program - if (index > 0) - { - glDisableVertexAttribArrayARB(index); // binormals - } -#endif -} - - -void enable_cloth_weights(const S32 index) -{ -#if GL_ARB_vertex_program - if (index > 0) glEnableVertexAttribArrayARB(index); -#endif -} - -void disable_cloth_weights(const S32 index) -{ -#if GL_ARB_vertex_program - if (index > 0) glDisableVertexAttribArrayARB(index); -#endif -} - -void set_vertex_weights(const S32 index, const U32 stride, const F32 *weights) -{ -#if GL_ARB_vertex_program - if (index > 0) glVertexAttribPointerARB(index, 1, GL_FLOAT, FALSE, stride, weights); - stop_glerror(); -#endif -} - -void set_vertex_clothing_weights(const S32 index, const U32 stride, const LLVector4 *weights) -{ -#if GL_ARB_vertex_program - if (index > 0) glVertexAttribPointerARB(index, 4, GL_FLOAT, TRUE, stride, weights); - stop_glerror(); -#endif -} - -void set_binormals(const S32 index, const U32 stride,const LLVector3 *binormals) -{ -#if GL_ARB_vertex_program - if (index > 0) glVertexAttribPointerARB(index, 3, GL_FLOAT, FALSE, stride, binormals); - stop_glerror(); -#endif -} - void parse_gl_version( S32* major, S32* minor, S32* release, std::string* vendor_specific ) { // GL_VERSION returns a null-terminated string with the format: @@ -2060,20 +2138,20 @@ void LLGLUserClipPlane::setPlane(F32 a, F32 b, F32 c, F32 d) glh::matrix4f suffix; suffix.set_row(2, cplane); glh::matrix4f newP = suffix * P; - glMatrixMode(GL_PROJECTION); - glPushMatrix(); - glLoadMatrixf(newP.m); + gGL.matrixMode(LLRender::MM_PROJECTION); + gGL.pushMatrix(); + gGL.loadMatrix(newP.m); gGLObliqueProjectionInverse = LLMatrix4(newP.inverse().transpose().m); - glMatrixMode(GL_MODELVIEW); + gGL.matrixMode(LLRender::MM_MODELVIEW); } LLGLUserClipPlane::~LLGLUserClipPlane() { if (mApply) { - glMatrixMode(GL_PROJECTION); - glPopMatrix(); - glMatrixMode(GL_MODELVIEW); + gGL.matrixMode(LLRender::MM_PROJECTION); + gGL.popMatrix(); + gGL.matrixMode(LLRender::MM_MODELVIEW); } } @@ -2263,16 +2341,16 @@ LLGLSquashToFarClip::LLGLSquashToFarClip(glh::matrix4f P, U32 layer) P.element(2, i) = P.element(3, i) * depth; } - glMatrixMode(GL_PROJECTION); - glPushMatrix(); - glLoadMatrixf(P.m); - glMatrixMode(GL_MODELVIEW); + gGL.matrixMode(LLRender::MM_PROJECTION); + gGL.pushMatrix(); + gGL.loadMatrix(P.m); + gGL.matrixMode(LLRender::MM_MODELVIEW); } LLGLSquashToFarClip::~LLGLSquashToFarClip() { - glMatrixMode(GL_PROJECTION); - glPopMatrix(); - glMatrixMode(GL_MODELVIEW); + gGL.matrixMode(LLRender::MM_PROJECTION); + gGL.popMatrix(); + gGL.matrixMode(LLRender::MM_MODELVIEW); } diff --git a/indra/llrender/llgl.h b/indra/llrender/llgl.h index d736133f3f..dee7ec0739 100644 --- a/indra/llrender/llgl.h +++ b/indra/llrender/llgl.h @@ -88,6 +88,7 @@ public: // ARB Extensions BOOL mHasVertexBufferObject; + BOOL mHasVertexArrayObject; BOOL mHasSync; BOOL mHasMapBufferRange; BOOL mHasFlushBufferRange; @@ -112,6 +113,7 @@ public: BOOL mHasAnisotropic; BOOL mHasARBEnvCombine; BOOL mHasCubeMap; + BOOL mHasDebugOutput; // Vendor-specific extensions BOOL mIsATI; @@ -252,7 +254,7 @@ public: static void dumpStates(); static void checkStates(const std::string& msg = ""); static void checkTextureChannels(const std::string& msg = ""); - static void checkClientArrays(const std::string& msg = "", U32 data_mask = 0x0001); + static void checkClientArrays(const std::string& msg = "", U32 data_mask = 0); protected: static boost::unordered_map<LLGLenum, LLGLboolean> sStateMap; @@ -419,15 +421,7 @@ extern LLMatrix4 gGLObliqueProjectionInverse; #include "llglstates.h" void init_glstates(); -void enable_vertex_weighting(const S32 index); -void disable_vertex_weighting(const S32 index); -void enable_binormals(const S32 index); -void disable_binormals(const S32 index); -void enable_cloth_weights(const S32 index); -void disable_cloth_weights(const S32 index); -void set_vertex_weights(const S32 index, const U32 stride, const F32 *weights); -void set_vertex_clothing_weights(const S32 index, const U32 stride, const LLVector4 *weights); -void set_binormals(const S32 index, const U32 stride, const LLVector3 *binormals); + void parse_gl_version( S32* major, S32* minor, S32* release, std::string* vendor_specific ); extern BOOL gClothRipple; diff --git a/indra/llrender/llglheaders.h b/indra/llrender/llglheaders.h index 851a75629e..10aad202e1 100644 --- a/indra/llrender/llglheaders.h +++ b/indra/llrender/llglheaders.h @@ -68,6 +68,12 @@ extern PFNGLUNMAPBUFFERARBPROC glUnmapBufferARB; extern PFNGLGETBUFFERPARAMETERIVARBPROC glGetBufferParameterivARB; extern PFNGLGETBUFFERPOINTERVARBPROC glGetBufferPointervARB; +// GL_ARB_vertex_array_object +extern PFNGLBINDVERTEXARRAYPROC glBindVertexArray; +extern PFNGLDELETEVERTEXARRAYSPROC glDeleteVertexArrays; +extern PFNGLGENVERTEXARRAYSPROC glGenVertexArrays; +extern PFNGLISVERTEXARRAYPROC glIsVertexArray; + // GL_ARB_sync extern PFNGLFENCESYNCPROC glFenceSync; extern PFNGLISSYNCPROC glIsSync; @@ -310,6 +316,12 @@ extern PFNGLCLIENTACTIVETEXTUREARBPROC glClientActiveTextureARB; extern PFNGLDRAWRANGEELEMENTSPROC glDrawRangeElements; #endif // LL_LINUX_NV_GL_HEADERS +// GL_ARB_vertex_array_object +extern PFNGLBINDVERTEXARRAYPROC glBindVertexArray; +extern PFNGLDELETEVERTEXARRAYSPROC glDeleteVertexArrays; +extern PFNGLGENVERTEXARRAYSPROC glGenVertexArrays; +extern PFNGLISVERTEXARRAYPROC glIsVertexArray; + // GL_ARB_vertex_buffer_object extern PFNGLBINDBUFFERARBPROC glBindBufferARB; extern PFNGLDELETEBUFFERSARBPROC glDeleteBuffersARB; @@ -531,6 +543,9 @@ extern PFNGLSAMPLEMASKIPROC glSampleMaski; #include "GL/glext.h" #include "GL/glh_extensions.h" +// WGL_ARB_create_context +extern PFNWGLCREATECONTEXTATTRIBSARBPROC wglCreateContextAttribsARB; +extern PFNGLGETSTRINGIPROC glGetStringi; // GL_ARB_vertex_buffer_object extern PFNGLBINDBUFFERARBPROC glBindBufferARB; @@ -545,6 +560,12 @@ extern PFNGLUNMAPBUFFERARBPROC glUnmapBufferARB; extern PFNGLGETBUFFERPARAMETERIVARBPROC glGetBufferParameterivARB; extern PFNGLGETBUFFERPOINTERVARBPROC glGetBufferPointervARB; +// GL_ARB_vertex_array_object +extern PFNGLBINDVERTEXARRAYPROC glBindVertexArray; +extern PFNGLDELETEVERTEXARRAYSPROC glDeleteVertexArrays; +extern PFNGLGENVERTEXARRAYSPROC glGenVertexArrays; +extern PFNGLISVERTEXARRAYPROC glIsVertexArray; + // GL_ARB_sync extern PFNGLFENCESYNCPROC glFenceSync; extern PFNGLISSYNCPROC glIsSync; @@ -735,6 +756,12 @@ extern PFNGLTEXIMAGE3DMULTISAMPLEPROC glTexImage3DMultisample; extern PFNGLGETMULTISAMPLEFVPROC glGetMultisamplefv; extern PFNGLSAMPLEMASKIPROC glSampleMaski; +//GL_ARB_debug_output +extern PFNGLDEBUGMESSAGECONTROLARBPROC glDebugMessageControlARB; +extern PFNGLDEBUGMESSAGEINSERTARBPROC glDebugMessageInsertARB; +extern PFNGLDEBUGMESSAGECALLBACKARBPROC glDebugMessageCallbackARB; +extern PFNGLGETDEBUGMESSAGELOGARBPROC glGetDebugMessageLogARB; + #elif LL_DARWIN //---------------------------------------------------------------------------- // LL_DARWIN @@ -899,6 +926,31 @@ extern void glGetBufferPointervARB (GLenum, GLenum, GLvoid* *); #endif /* GL_GLEXT_FUNCTION_POINTERS */ #endif +#ifndef GL_ARB_texture_rg +#define GL_RG 0x8227 +#define GL_RG_INTEGER 0x8228 +#define GL_R8 0x8229 +#define GL_R16 0x822A +#define GL_RG8 0x822B +#define GL_RG16 0x822C +#define GL_R16F 0x822D +#define GL_R32F 0x822E +#define GL_RG16F 0x822F +#define GL_RG32F 0x8230 +#define GL_R8I 0x8231 +#define GL_R8UI 0x8232 +#define GL_R16I 0x8233 +#define GL_R16UI 0x8234 +#define GL_R32I 0x8235 +#define GL_R32UI 0x8236 +#define GL_RG8I 0x8237 +#define GL_RG8UI 0x8238 +#define GL_RG16I 0x8239 +#define GL_RG16UI 0x823A +#define GL_RG32I 0x823B +#define GL_RG32UI 0x823C +#endif + // May be needed for DARWIN... // #ifndef GL_ARB_compressed_tex_image // #define GL_ARB_compressed_tex_image 1 diff --git a/indra/llrender/llglslshader.cpp b/indra/llrender/llglslshader.cpp index f51d83abe4..ddadf07d73 100644 --- a/indra/llrender/llglslshader.cpp +++ b/indra/llrender/llglslshader.cpp @@ -31,6 +31,7 @@ #include "llshadermgr.h" #include "llfile.h" #include "llrender.h" +#include "llvertexbuffer.h" #if LL_DARWIN #include "OpenGL/OpenGL.h" @@ -50,6 +51,7 @@ using std::string; GLhandleARB LLGLSLShader::sCurBoundShader = 0; LLGLSLShader* LLGLSLShader::sCurBoundShaderPtr = NULL; +S32 LLGLSLShader::sIndexedTextureChannels = 0; bool LLGLSLShader::sNoFixedFunction = false; //UI shader -- declared here so llui_libtest will link properly @@ -75,6 +77,7 @@ hasAlphaMask(false) LLGLSLShader::LLGLSLShader() : mProgramObject(0), mActiveTextureChannels(0), mShaderLevel(0), mShaderGroup(SG_DEFAULT), mUniformsDirty(FALSE) { + } void LLGLSLShader::unload() @@ -110,17 +113,19 @@ void LLGLSLShader::unload() BOOL LLGLSLShader::createShader(vector<string> * attributes, vector<string> * uniforms) { + //reloading, reset matrix hash values + for (U32 i = 0; i < LLRender::NUM_MATRIX_MODES; ++i) + { + mMatHash[i] = 0xFFFFFFFF; + } + mLightHash = 0xFFFFFFFF; + llassert_always(!mShaderFiles.empty()); BOOL success = TRUE; // Create program mProgramObject = glCreateProgramObjectARB(); - if (gGLManager.mGLVersion < 3.1f) - { //force indexed texture channels to 1 if GL version is old (performance improvement for drivers with poor branching shader model support) - mFeatures.mIndexedTextureChannels = llmin(mFeatures.mIndexedTextureChannels, 1); - } - //compile new source vector< pair<string,GLenum> >::iterator fileIter = mShaderFiles.begin(); for ( ; fileIter != mShaderFiles.end(); fileIter++ ) @@ -235,6 +240,13 @@ void LLGLSLShader::attachObjects(GLhandleARB* objects, S32 count) BOOL LLGLSLShader::mapAttributes(const vector<string> * attributes) { + //before linking, make sure reserved attributes always have consistent locations + for (U32 i = 0; i < LLShaderMgr::instance()->mReservedAttribs.size(); i++) + { + const char* name = LLShaderMgr::instance()->mReservedAttribs[i].c_str(); + glBindAttribLocationARB(mProgramObject, i, (const GLcharARB *) name); + } + //link the program BOOL res = link(); @@ -386,6 +398,7 @@ void LLGLSLShader::bind() gGL.flush(); if (gGLManager.mHasShaderObjects) { + LLVertexBuffer::unbind(); glUseProgramObjectARB(mProgramObject); sCurBoundShader = mProgramObject; sCurBoundShaderPtr = this; @@ -411,6 +424,7 @@ void LLGLSLShader::unbind() stop_glerror(); } } + LLVertexBuffer::unbind(); glUseProgramObjectARB(0); sCurBoundShader = 0; sCurBoundShaderPtr = NULL; @@ -420,6 +434,7 @@ void LLGLSLShader::unbind() void LLGLSLShader::bindNoShader(void) { + LLVertexBuffer::unbind(); glUseProgramObjectARB(0); sCurBoundShader = 0; sCurBoundShaderPtr = NULL; @@ -930,7 +945,9 @@ void LLGLSLShader::uniform4fv(const string& uniform, U32 count, const GLfloat* v std::map<GLint, LLVector4>::iterator iter = mValue.find(location); if (iter == mValue.end() || shouldChange(iter->second,vec) || count != 1) { + stop_glerror(); glUniform4fvARB(location, count, v); + stop_glerror(); mValue[location] = vec; } } @@ -987,6 +1004,7 @@ void LLGLSLShader::vertexAttrib4fv(U32 index, GLfloat* v) void LLGLSLShader::setAlphaRange(F32 minimum, F32 maximum) { + gGL.flush(); uniform1f("minimum_alpha", minimum); uniform1f("maximum_alpha", maximum); } diff --git a/indra/llrender/llglslshader.h b/indra/llrender/llglslshader.h index 558ea66b50..beef57796d 100644 --- a/indra/llrender/llglslshader.h +++ b/indra/llrender/llglslshader.h @@ -69,7 +69,7 @@ public: static GLhandleARB sCurBoundShader; static LLGLSLShader* sCurBoundShaderPtr; - + static S32 sIndexedTextureChannels; static bool sNoFixedFunction; void unload(); @@ -117,7 +117,6 @@ public: GLint getAttribLocation(U32 attrib); GLint mapUniformTextureChannel(GLint location, GLenum type); - //enable/disable texture channel for specified uniform //if given texture uniform is active in the shader, //the corresponding channel will be active upon return @@ -132,6 +131,9 @@ public: // Unbinds any previously bound shader by explicitly binding no shader. static void bindNoShader(void); + U32 mMatHash[LLRender::NUM_MATRIX_MODES]; + U32 mLightHash; + GLhandleARB mProgramObject; std::vector<GLint> mAttribute; //lookup table of attribute enum to attribute channel std::vector<GLint> mUniform; //lookup table of uniform enum to uniform location diff --git a/indra/llrender/llimagegl.cpp b/indra/llrender/llimagegl.cpp index 7188b0fa44..3d3c94ef3e 100644 --- a/indra/llrender/llimagegl.cpp +++ b/indra/llrender/llimagegl.cpp @@ -36,7 +36,9 @@ #include "llmath.h" #include "llgl.h" +#include "llglslshader.h" #include "llrender.h" + //---------------------------------------------------------------------------- const F32 MIN_TEXTURE_LIFETIME = 10.f; @@ -725,7 +727,10 @@ void LLImageGL::setImage(const U8* data_in, BOOL data_hasmips) { if (mAutoGenMips) { - glTexParameteri(LLTexUnit::getInternalType(mBindTarget), GL_GENERATE_MIPMAP_SGIS, TRUE); + if (!gGLManager.mHasFramebufferObject) + { + glTexParameteri(LLTexUnit::getInternalType(mBindTarget), GL_GENERATE_MIPMAP_SGIS, TRUE); + } stop_glerror(); { // LLFastTimer t2(FTM_TEMP4); @@ -754,6 +759,11 @@ void LLImageGL::setImage(const U8* data_in, BOOL data_hasmips) stop_glerror(); } } + + if (gGLManager.mHasFramebufferObject) + { + glGenerateMipmap(LLTexUnit::getInternalType(mBindTarget)); + } } else { @@ -875,6 +885,9 @@ void LLImageGL::setImage(const U8* data_in, BOOL data_hasmips) BOOL LLImageGL::preAddToAtlas(S32 discard_level, const LLImageRaw* raw_image) { + //not compatible with core GL profile + llassert(!LLRender::sGLCoreProfile); + if (gGLManager.mIsDisabled) { llwarns << "Trying to create a texture while GL is disabled!" << llendl; @@ -901,29 +914,29 @@ BOOL LLImageGL::preAddToAtlas(S32 discard_level, const LLImageRaw* raw_image) { switch (mComponents) { - case 1: + case 1: // Use luminance alpha (for fonts) mFormatInternal = GL_LUMINANCE8; mFormatPrimary = GL_LUMINANCE; mFormatType = GL_UNSIGNED_BYTE; break; - case 2: + case 2: // Use luminance alpha (for fonts) mFormatInternal = GL_LUMINANCE8_ALPHA8; mFormatPrimary = GL_LUMINANCE_ALPHA; mFormatType = GL_UNSIGNED_BYTE; break; - case 3: + case 3: mFormatInternal = GL_RGB8; mFormatPrimary = GL_RGB; mFormatType = GL_UNSIGNED_BYTE; break; - case 4: + case 4: mFormatInternal = GL_RGBA8; mFormatPrimary = GL_RGBA; mFormatType = GL_UNSIGNED_BYTE; break; - default: + default: llerrs << "Bad number of components for texture: " << (U32)getComponents() << llendl; } } @@ -1099,8 +1112,75 @@ void LLImageGL::deleteTextures(S32 numTextures, U32 *textures, bool immediate) // static void LLImageGL::setManualImage(U32 target, S32 miplevel, S32 intformat, S32 width, S32 height, U32 pixformat, U32 pixtype, const void *pixels) { - glTexImage2D(target, miplevel, intformat, width, height, 0, pixformat, pixtype, pixels); + bool use_scratch = false; + U32* scratch = NULL; + if (LLRender::sGLCoreProfile) + { + if (intformat == GL_ALPHA8 && pixformat == GL_ALPHA && pixtype == GL_UNSIGNED_BYTE) + { //GL_ALPHA is deprecated, convert to RGBA + use_scratch = true; + scratch = new U32[width*height]; + + U32 pixel_count = (U32) (width*height); + for (U32 i = 0; i < pixel_count; i++) + { + U8* pix = (U8*) &scratch[i]; + pix[0] = pix[1] = pix[2] = 0; + pix[3] = ((U8*) pixels)[i]; + } + + pixformat = GL_RGBA; + intformat = GL_RGBA8; + } + + if (intformat == GL_LUMINANCE8_ALPHA8 && pixformat == GL_LUMINANCE_ALPHA && pixtype == GL_UNSIGNED_BYTE) + { //GL_LUMINANCE_ALPHA is deprecated, convert to RGBA + use_scratch = true; + scratch = new U32[width*height]; + + U32 pixel_count = (U32) (width*height); + for (U32 i = 0; i < pixel_count; i++) + { + U8 lum = ((U8*) pixels)[i*2+0]; + U8 alpha = ((U8*) pixels)[i*2+1]; + + U8* pix = (U8*) &scratch[i]; + pix[0] = pix[1] = pix[2] = lum; + pix[3] = alpha; + } + + pixformat = GL_RGBA; + intformat = GL_RGBA8; + } + + if (intformat == GL_LUMINANCE8 && pixformat == GL_LUMINANCE && pixtype == GL_UNSIGNED_BYTE) + { //GL_LUMINANCE_ALPHA is deprecated, convert to RGB + use_scratch = true; + scratch = new U32[width*height]; + + U32 pixel_count = (U32) (width*height); + for (U32 i = 0; i < pixel_count; i++) + { + U8 lum = ((U8*) pixels)[i]; + + U8* pix = (U8*) &scratch[i]; + pix[0] = pix[1] = pix[2] = lum; + pix[3] = 255; + } + + pixformat = GL_RGBA; + intformat = GL_RGB8; + } + } + stop_glerror(); + glTexImage2D(target, miplevel, intformat, width, height, 0, pixformat, pixtype, use_scratch ? scratch : pixels); + stop_glerror(); + + if (use_scratch) + { + delete [] scratch; + } } //create an empty GL texture: just create a texture name @@ -1167,29 +1247,29 @@ BOOL LLImageGL::createGLTexture(S32 discard_level, const LLImageRaw* imageraw, S { switch (mComponents) { - case 1: + case 1: // Use luminance alpha (for fonts) mFormatInternal = GL_LUMINANCE8; mFormatPrimary = GL_LUMINANCE; mFormatType = GL_UNSIGNED_BYTE; break; - case 2: + case 2: // Use luminance alpha (for fonts) mFormatInternal = GL_LUMINANCE8_ALPHA8; mFormatPrimary = GL_LUMINANCE_ALPHA; mFormatType = GL_UNSIGNED_BYTE; break; - case 3: + case 3: mFormatInternal = GL_RGB8; mFormatPrimary = GL_RGB; mFormatType = GL_UNSIGNED_BYTE; break; - case 4: + case 4: mFormatInternal = GL_RGBA8; mFormatPrimary = GL_RGBA; mFormatType = GL_UNSIGNED_BYTE; break; - default: + default: llerrs << "Bad number of components for texture: " << (U32)getComponents() << llendl; } @@ -1212,6 +1292,7 @@ BOOL LLImageGL::createGLTexture(S32 discard_level, const LLImageRaw* imageraw, S BOOL LLImageGL::createGLTexture(S32 discard_level, const U8* data_in, BOOL data_hasmips, S32 usename) { llassert(data_in); + stop_glerror(); if (discard_level < 0) { @@ -1240,8 +1321,11 @@ BOOL LLImageGL::createGLTexture(S32 discard_level, const U8* data_in, BOOL data_ stop_glerror(); { llverify(gGL.getTexUnit(0)->bind(this)); + stop_glerror(); glTexParameteri(LLTexUnit::getInternalType(mBindTarget), GL_TEXTURE_BASE_LEVEL, 0); + stop_glerror(); glTexParameteri(LLTexUnit::getInternalType(mBindTarget), GL_TEXTURE_MAX_LEVEL, mMaxDiscardLevel-discard_level); + stop_glerror(); } } if (!mTexName) diff --git a/indra/llrender/llpostprocess.cpp b/indra/llrender/llpostprocess.cpp index d76b2d9004..c0045c8044 100644 --- a/indra/llrender/llpostprocess.cpp +++ b/indra/llrender/llpostprocess.cpp @@ -466,21 +466,21 @@ void LLPostProcess::drawOrthoQuad(unsigned int width, unsigned int height, QuadT void LLPostProcess::viewOrthogonal(unsigned int width, unsigned int height) { - glMatrixMode(GL_PROJECTION); - glPushMatrix(); - glLoadIdentity(); - glOrtho( 0.f, (GLdouble) width , (GLdouble) height , 0.f, -1.f, 1.f ); - glMatrixMode(GL_MODELVIEW); - glPushMatrix(); - glLoadIdentity(); + gGL.matrixMode(LLRender::MM_PROJECTION); + gGL.pushMatrix(); + gGL.loadIdentity(); + gGL.ortho( 0.f, (GLdouble) width , (GLdouble) height , 0.f, -1.f, 1.f ); + gGL.matrixMode(LLRender::MM_MODELVIEW); + gGL.pushMatrix(); + gGL.loadIdentity(); } void LLPostProcess::viewPerspective(void) { - glMatrixMode( GL_PROJECTION ); - glPopMatrix(); - glMatrixMode( GL_MODELVIEW ); - glPopMatrix(); + gGL.matrixMode(LLRender::MM_PROJECTION); + gGL.popMatrix(); + gGL.matrixMode(LLRender::MM_MODELVIEW); + gGL.popMatrix(); } void LLPostProcess::changeOrthogonal(unsigned int width, unsigned int height) diff --git a/indra/llrender/llrender.cpp b/indra/llrender/llrender.cpp index edcc47aa14..bbdd0a7a60 100644 --- a/indra/llrender/llrender.cpp +++ b/indra/llrender/llrender.cpp @@ -1,4 +1,4 @@ -/** + /** * @file llrender.cpp * @brief LLRender implementation * @@ -47,6 +47,7 @@ S32 gGLViewport[4]; U32 LLRender::sUICalls = 0; U32 LLRender::sUIVerts = 0; U32 LLTexUnit::sWhiteTexture = 0; +bool LLRender::sGLCoreProfile = false; static const U32 LL_NUM_TEXTURE_LAYERS = 32; static const U32 LL_NUM_LIGHT_UNITS = 8; @@ -178,10 +179,13 @@ void LLTexUnit::enable(eTextureType type) if ( (mCurrTexType != type || gGL.mDirty) && (type != TT_NONE) ) { + stop_glerror(); activate(); + stop_glerror(); if (mCurrTexType != TT_NONE && !gGL.mDirty) { disable(); // Force a disable of a previous texture type if it's enabled. + stop_glerror(); } mCurrTexType = type; @@ -190,7 +194,9 @@ void LLTexUnit::enable(eTextureType type) type != LLTexUnit::TT_MULTISAMPLE_TEXTURE && mIndex < gGLManager.mNumTextureUnits) { + stop_glerror(); glEnable(sGLTextureType[type]); + stop_glerror(); } } } @@ -286,26 +292,35 @@ bool LLTexUnit::bind(LLImageGL* texture, bool for_rendering, bool forceBind) { return bind(LLImageGL::sDefaultGLTexture) ; } + stop_glerror(); return false ; } if ((mCurrTexture != texture->getTexName()) || forceBind) { gGL.flush(); + stop_glerror(); activate(); + stop_glerror(); enable(texture->getTarget()); + stop_glerror(); mCurrTexture = texture->getTexName(); glBindTexture(sGLTextureType[texture->getTarget()], mCurrTexture); + stop_glerror(); texture->updateBindStats(texture->mTextureMemory); mHasMipMaps = texture->mHasMipMaps; if (texture->mTexOptionsDirty) { + stop_glerror(); texture->mTexOptionsDirty = false; setTextureAddressMode(texture->mAddressMode); setTextureFilteringOption(texture->mFilterOption); + stop_glerror(); } } + stop_glerror(); + return true; } @@ -814,14 +829,16 @@ LLLightState::LLLightState(S32 index) mAmbient.set(0,0,0,1); mPosition.set(0,0,1,0); mSpotDirection.set(0,0,-1); - } void LLLightState::enable() { if (!mEnabled) { - glEnable(GL_LIGHT0+mIndex); + if (!LLGLSLShader::sNoFixedFunction) + { + glEnable(GL_LIGHT0+mIndex); + } mEnabled = true; } } @@ -830,7 +847,10 @@ void LLLightState::disable() { if (mEnabled) { - glDisable(GL_LIGHT0+mIndex); + if (!LLGLSLShader::sNoFixedFunction) + { + glDisable(GL_LIGHT0+mIndex); + } mEnabled = false; } } @@ -839,8 +859,12 @@ void LLLightState::setDiffuse(const LLColor4& diffuse) { if (mDiffuse != diffuse) { + ++gGL.mLightHash; mDiffuse = diffuse; - glLightfv(GL_LIGHT0+mIndex, GL_DIFFUSE, mDiffuse.mV); + if (!LLGLSLShader::sNoFixedFunction) + { + glLightfv(GL_LIGHT0+mIndex, GL_DIFFUSE, mDiffuse.mV); + } } } @@ -848,8 +872,12 @@ void LLLightState::setAmbient(const LLColor4& ambient) { if (mAmbient != ambient) { + ++gGL.mLightHash; mAmbient = ambient; - glLightfv(GL_LIGHT0+mIndex, GL_AMBIENT, mAmbient.mV); + if (!LLGLSLShader::sNoFixedFunction) + { + glLightfv(GL_LIGHT0+mIndex, GL_AMBIENT, mAmbient.mV); + } } } @@ -857,16 +885,34 @@ void LLLightState::setSpecular(const LLColor4& specular) { if (mSpecular != specular) { + ++gGL.mLightHash; mSpecular = specular; - glLightfv(GL_LIGHT0+mIndex, GL_SPECULAR, mSpecular.mV); + if (!LLGLSLShader::sNoFixedFunction) + { + glLightfv(GL_LIGHT0+mIndex, GL_SPECULAR, mSpecular.mV); + } } } void LLLightState::setPosition(const LLVector4& position) { //always set position because modelview matrix may have changed + ++gGL.mLightHash; mPosition = position; - glLightfv(GL_LIGHT0+mIndex, GL_POSITION, mPosition.mV); + if (!LLGLSLShader::sNoFixedFunction) + { + glLightfv(GL_LIGHT0+mIndex, GL_POSITION, mPosition.mV); + } + else + { //transform position by current modelview matrix + glh::vec4f pos(position.mV); + + const glh::matrix4f& mat = gGL.getModelviewMatrix(); + mat.mult_matrix_vec(pos); + + mPosition.set(pos.v); + } + } void LLLightState::setConstantAttenuation(const F32& atten) @@ -874,7 +920,11 @@ void LLLightState::setConstantAttenuation(const F32& atten) if (mConstantAtten != atten) { mConstantAtten = atten; - glLightf(GL_LIGHT0+mIndex, GL_CONSTANT_ATTENUATION, atten); + ++gGL.mLightHash; + if (!LLGLSLShader::sNoFixedFunction) + { + glLightf(GL_LIGHT0+mIndex, GL_CONSTANT_ATTENUATION, atten); + } } } @@ -882,8 +932,12 @@ void LLLightState::setLinearAttenuation(const F32& atten) { if (mLinearAtten != atten) { + ++gGL.mLightHash; mLinearAtten = atten; - glLightf(GL_LIGHT0+mIndex, GL_LINEAR_ATTENUATION, atten); + if (!LLGLSLShader::sNoFixedFunction) + { + glLightf(GL_LIGHT0+mIndex, GL_LINEAR_ATTENUATION, atten); + } } } @@ -891,8 +945,12 @@ void LLLightState::setQuadraticAttenuation(const F32& atten) { if (mQuadraticAtten != atten) { + ++gGL.mLightHash; mQuadraticAtten = atten; - glLightf(GL_LIGHT0+mIndex, GL_QUADRATIC_ATTENUATION, atten); + if (!LLGLSLShader::sNoFixedFunction) + { + glLightf(GL_LIGHT0+mIndex, GL_QUADRATIC_ATTENUATION, atten); + } } } @@ -900,8 +958,12 @@ void LLLightState::setSpotExponent(const F32& exponent) { if (mSpotExponent != exponent) { + ++gGL.mLightHash; mSpotExponent = exponent; - glLightf(GL_LIGHT0+mIndex, GL_SPOT_EXPONENT, exponent); + if (!LLGLSLShader::sNoFixedFunction) + { + glLightf(GL_LIGHT0+mIndex, GL_SPOT_EXPONENT, exponent); + } } } @@ -909,21 +971,39 @@ void LLLightState::setSpotCutoff(const F32& cutoff) { if (mSpotCutoff != cutoff) { + ++gGL.mLightHash; mSpotCutoff = cutoff; - glLightf(GL_LIGHT0+mIndex, GL_SPOT_CUTOFF, cutoff); + if (!LLGLSLShader::sNoFixedFunction) + { + glLightf(GL_LIGHT0+mIndex, GL_SPOT_CUTOFF, cutoff); + } } } void LLLightState::setSpotDirection(const LLVector3& direction) { //always set direction because modelview matrix may have changed + ++gGL.mLightHash; mSpotDirection = direction; - glLightfv(GL_LIGHT0+mIndex, GL_SPOT_DIRECTION, direction.mV); + if (!LLGLSLShader::sNoFixedFunction) + { + glLightfv(GL_LIGHT0+mIndex, GL_SPOT_DIRECTION, direction.mV); + } + else + { //transform direction by current modelview matrix + glh::vec3f dir(direction.mV); + + const glh::matrix4f& mat = gGL.getModelviewMatrix(); + mat.mult_matrix_dir(dir); + + mSpotDirection.set(direction); + } } LLRender::LLRender() : mDirty(false), mCount(0), + mQuadCycle(0), mMode(LLRender::TRIANGLES), mCurrTextureUnitIndex(0), mMaxAnisotropy(0.f) @@ -951,6 +1031,17 @@ LLRender::LLRender() mCurrBlendAlphaSFactor = BF_UNDEF; mCurrBlendColorDFactor = BF_UNDEF; mCurrBlendAlphaDFactor = BF_UNDEF; + + mMatrixMode = LLRender::MM_MODELVIEW; + + for (U32 i = 0; i < NUM_MATRIX_MODES; ++i) + { + mMatIdx[i] = 0; + mMatHash[i] = 0; + mCurMatHash[i] = 0xFFFFFFFF; + } + + mLightHash = 0; } LLRender::~LLRender() @@ -961,12 +1052,13 @@ LLRender::~LLRender() void LLRender::init() { llassert_always(mBuffer.isNull()) ; - + stop_glerror(); mBuffer = new LLVertexBuffer(immediate_mask, 0); mBuffer->allocateBuffer(4096, 0, TRUE); mBuffer->getVertexStrider(mVerticesp); mBuffer->getTexCoord0Strider(mTexcoordsp); mBuffer->getColorStrider(mColorsp); + stop_glerror(); } void LLRender::shutdown() @@ -1007,28 +1099,310 @@ void LLRender::refreshState(void) mDirty = false; } +void LLRender::syncLightState() +{ + LLGLSLShader* shader = LLGLSLShader::sCurBoundShaderPtr; + + if (!shader) + { + return; + } + + if (shader->mLightHash != mLightHash) + { + shader->mLightHash = mLightHash; + + LLVector4 position[8]; + LLVector3 direction[8]; + LLVector3 attenuation[8]; + LLVector3 diffuse[8]; + + for (U32 i = 0; i < 8; i++) + { + LLLightState* light = mLightState[i]; + + position[i] = light->mPosition; + direction[i] = light->mSpotDirection; + attenuation[i].set(light->mLinearAtten, light->mQuadraticAtten, light->mSpecular.mV[3]); + diffuse[i].set(light->mDiffuse.mV); + } + + shader->uniform4fv("light_position", 8, position[0].mV); + shader->uniform3fv("light_direction", 8, direction[0].mV); + shader->uniform3fv("light_attenuation", 8, attenuation[0].mV); + shader->uniform3fv("light_diffuse", 8, diffuse[0].mV); + shader->uniform4fv("light_ambient", 1, mAmbientLightColor.mV); + //HACK -- duplicate sunlight color for compatibility with drivers that can't deal with multiple shader objects referencing the same uniform + shader->uniform4fv("sunlight_color", 1, diffuse[0].mV); + } +} + +void LLRender::syncMatrices() +{ + stop_glerror(); + + GLenum mode[] = + { + GL_MODELVIEW, + GL_PROJECTION, + GL_TEXTURE, + GL_TEXTURE, + GL_TEXTURE, + GL_TEXTURE, + }; + + std::string name[] = + { + "modelview_matrix", + "projection_matrix", + "texture_matrix0", + "texture_matrix1", + "texture_matrix2", + "texture_matrix3", + }; + + LLGLSLShader* shader = LLGLSLShader::sCurBoundShaderPtr; + + if (shader) + { + + llassert(shader); + + bool do_normal = false; + bool do_mvp = false; + + for (U32 i = 0; i < NUM_MATRIX_MODES; ++i) + { + if (mMatHash[i] != shader->mMatHash[i]) + { + shader->uniformMatrix4fv(name[i], 1, GL_FALSE, mMatrix[i][mMatIdx[i]].m); + shader->mMatHash[i] = mMatHash[i]; + + if (i == MM_MODELVIEW) + { + do_normal = true; + do_mvp = true; + } + else if (i == MM_PROJECTION) + { + do_mvp = true; + } + } + } + + if (do_normal) + { + S32 loc = shader->getUniformLocation("normal_matrix"); + if (loc > -1) + { + U32 i = MM_MODELVIEW; + + glh::matrix4f norm = mMatrix[i][mMatIdx[i]].inverse().transpose(); + + F32 norm_mat[] = + { + norm.m[0], norm.m[1], norm.m[2], + norm.m[4], norm.m[5], norm.m[6], + norm.m[8], norm.m[9], norm.m[10] + }; + + shader->uniformMatrix3fv("normal_matrix", 1, GL_FALSE, norm_mat); + } + } + + if (do_mvp) + { + S32 loc = shader->getUniformLocation("modelview_projection_matrix"); + if (loc > -1) + { + U32 mv = MM_MODELVIEW; + U32 proj = MM_PROJECTION; + + glh::matrix4f mvp = mMatrix[mv][mMatIdx[mv]]; + mvp.mult_left(mMatrix[proj][mMatIdx[proj]]); + + shader->uniformMatrix4fv("modelview_projection_matrix", 1, GL_FALSE, mvp.m); + } + } + + if (shader->mFeatures.hasLighting || shader->mFeatures.calculatesLighting) + { //also sync light state + syncLightState(); + } + } + else if (!LLGLSLShader::sNoFixedFunction) + { + for (U32 i = 0; i < 2; ++i) + { + if (mMatHash[i] != mCurMatHash[i]) + { + glMatrixMode(mode[i]); + glLoadMatrixf(mMatrix[i][mMatIdx[i]].m); + mCurMatHash[i] = mMatHash[i]; + } + } + + for (U32 i = 2; i < NUM_MATRIX_MODES; ++i) + { + if (mMatHash[i] != mCurMatHash[i]) + { + gGL.getTexUnit(i-2)->activate(); + glMatrixMode(mode[i]); + glLoadMatrixf(mMatrix[i][mMatIdx[i]].m); + mCurMatHash[i] = mMatHash[i]; + } + } + } + + stop_glerror(); +} + void LLRender::translatef(const GLfloat& x, const GLfloat& y, const GLfloat& z) { flush(); - glTranslatef(x,y,z); + + glh::matrix4f trans_mat(1,0,0,x, + 0,1,0,y, + 0,0,1,z, + 0,0,0,1); + + mMatrix[mMatrixMode][mMatIdx[mMatrixMode]].mult_right(trans_mat); + mMatHash[mMatrixMode]++; } void LLRender::scalef(const GLfloat& x, const GLfloat& y, const GLfloat& z) { flush(); - glScalef(x,y,z); + + glh::matrix4f scale_mat(x,0,0,0, + 0,y,0,0, + 0,0,z,0, + 0,0,0,1); + + mMatrix[mMatrixMode][mMatIdx[mMatrixMode]].mult_right(scale_mat); + mMatHash[mMatrixMode]++; +} + +void LLRender::ortho(F32 left, F32 right, F32 bottom, F32 top, F32 zNear, F32 zFar) +{ + flush(); + + glh::matrix4f ortho_mat(2.f/(right-left),0,0, -(right+left)/(right-left), + 0,2.f/(top-bottom),0, -(top+bottom)/(top-bottom), + 0,0,-2.f/(zFar-zNear), -(zFar+zNear)/(zFar-zNear), + 0,0,0,1); + + mMatrix[mMatrixMode][mMatIdx[mMatrixMode]].mult_right(ortho_mat); + mMatHash[mMatrixMode]++; +} + +void LLRender::rotatef(const GLfloat& a, const GLfloat& x, const GLfloat& y, const GLfloat& z) +{ + flush(); + + F32 r = a * DEG_TO_RAD; + + F32 c = cosf(r); + F32 s = sinf(r); + + F32 ic = 1.f-c; + + glh::matrix4f rot_mat(x*x*ic+c, x*y*ic-z*s, x*z*ic+y*s, 0, + x*y*ic+z*s, y*y*ic+c, y*z*ic-x*s, 0, + x*z*ic-y*s, y*z*ic+x*s, z*z*ic+c, 0, + 0,0,0,1); + + mMatrix[mMatrixMode][mMatIdx[mMatrixMode]].mult_right(rot_mat); + mMatHash[mMatrixMode]++; } void LLRender::pushMatrix() { flush(); - glPushMatrix(); + + if (mMatIdx[mMatrixMode] < LL_MATRIX_STACK_DEPTH-1) + { + mMatrix[mMatrixMode][mMatIdx[mMatrixMode]+1] = mMatrix[mMatrixMode][mMatIdx[mMatrixMode]]; + ++mMatIdx[mMatrixMode]; + } + else + { + llwarns << "Matrix stack overflow." << llendl; + } } void LLRender::popMatrix() { flush(); - glPopMatrix(); + if (mMatIdx[mMatrixMode] > 0) + { + --mMatIdx[mMatrixMode]; + mMatHash[mMatrixMode]++; + } + else + { + llwarns << "Matrix stack underflow." << llendl; + } +} + +void LLRender::loadMatrix(const GLfloat* m) +{ + mMatrix[mMatrixMode][mMatIdx[mMatrixMode]].set_value((GLfloat*) m); + mMatHash[mMatrixMode]++; +} + +void LLRender::loadMatrix(const GLdouble* dm) +{ + F32 m[16]; + for (U32 i = 0; i < 16; i++) + { + m[i] = (F32) dm[i]; + } + + loadMatrix(m); +} + +void LLRender::multMatrix(const GLfloat* m) +{ + flush(); + + glh::matrix4f mat((GLfloat*) m); + + mMatrix[mMatrixMode][mMatIdx[mMatrixMode]].mult_right(mat); + mMatHash[mMatrixMode]++; +} + +void LLRender::matrixMode(U32 mode) +{ + if (mode == MM_TEXTURE) + { + mode = MM_TEXTURE0 + gGL.getCurrentTexUnitIndex(); + } + + llassert(mode < NUM_MATRIX_MODES); + mMatrixMode = mode; +} + +void LLRender::multMatrix(const GLdouble* dm) +{ + F32 m[16]; + for (U32 i = 0; i < 16; i++) + { + m[i] = (F32) dm[i]; + } + + multMatrix(m); +} + +void LLRender::loadIdentity() +{ + mMatrix[mMatrixMode][mMatIdx[mMatrixMode]].make_identity(); + mMatHash[mMatrixMode]++; +} + +const glh::matrix4f& LLRender::getModelviewMatrix() +{ + return mMatrix[MM_MODELVIEW][mMatIdx[MM_MODELVIEW]]; } void LLRender::translateUI(F32 x, F32 y, F32 z) @@ -1284,6 +1658,19 @@ LLLightState* LLRender::getLight(U32 index) return NULL; } +void LLRender::setAmbientLightColor(const LLColor4& color) +{ + if (color != mAmbientLightColor) + { + ++mLightHash; + mAmbientLightColor = color; + if (!LLGLSLShader::sNoFixedFunction) + { + glLightModelfv(GL_LIGHT_MODEL_AMBIENT, color.mV); + } + } +} + bool LLRender::verifyTexUnitActive(U32 unitToVerify) { if (mCurrTextureUnitIndex == unitToVerify) @@ -1309,6 +1696,11 @@ void LLRender::begin(const GLuint& mode) { if (mode != mMode) { + if (mode == LLRender::QUADS) + { + mQuadCycle = 1; + } + if (mMode == LLRender::QUADS || mMode == LLRender::LINES || mMode == LLRender::TRIANGLES || @@ -1396,7 +1788,7 @@ void LLRender::flush() if (gDebugGL) { - if (mMode == LLRender::QUADS) + if (mMode == LLRender::QUADS && !sGLCoreProfile) { if (mCount%4 != 0) { @@ -1421,12 +1813,34 @@ void LLRender::flush() } } + //store mCount in a local variable to avoid re-entrance (drawArrays may call flush) + U32 count = mCount; + mCount = 0; + + if (mBuffer->useVBOs() && !mBuffer->isLocked()) + { //hack to only flush the part of the buffer that was updated (relies on stream draw using buffersubdata) + mBuffer->getVertexStrider(mVerticesp, 0, count); + mBuffer->getTexCoord0Strider(mTexcoordsp, 0, count); + mBuffer->getColorStrider(mColorsp, 0, count); + } + + mBuffer->flush(); mBuffer->setBuffer(immediate_mask); - mBuffer->drawArrays(mMode, 0, mCount); + + if (mMode == LLRender::QUADS && sGLCoreProfile) + { + mBuffer->drawArrays(LLRender::TRIANGLES, 0, count); + mQuadCycle = 1; + } + else + { + mBuffer->drawArrays(mMode, 0, count); + } + + mVerticesp[0] = mVerticesp[count]; + mTexcoordsp[0] = mTexcoordsp[count]; + mColorsp[0] = mColorsp[count]; - mVerticesp[0] = mVerticesp[mCount]; - mTexcoordsp[0] = mTexcoordsp[mCount]; - mColorsp[0] = mColorsp[mCount]; mCount = 0; } } @@ -1450,10 +1864,29 @@ void LLRender::vertex3f(const GLfloat& x, const GLfloat& y, const GLfloat& z) mVerticesp[mCount] = vert; } + if (mMode == LLRender::QUADS && LLRender::sGLCoreProfile) + { + mQuadCycle++; + if (mQuadCycle == 4) + { //copy two vertices so fourth quad element will add a triangle + mQuadCycle = 0; + + mCount++; + mVerticesp[mCount] = mVerticesp[mCount-3]; + mColorsp[mCount] = mColorsp[mCount-3]; + mTexcoordsp[mCount] = mTexcoordsp[mCount-3]; + + mCount++; + mVerticesp[mCount] = mVerticesp[mCount-2]; + mColorsp[mCount] = mColorsp[mCount-2]; + mTexcoordsp[mCount] = mTexcoordsp[mCount-2]; + } + } + mCount++; mVerticesp[mCount] = mVerticesp[mCount-1]; mColorsp[mCount] = mColorsp[mCount-1]; - mTexcoordsp[mCount] = mTexcoordsp[mCount-1]; + mTexcoordsp[mCount] = mTexcoordsp[mCount-1]; } void LLRender::vertexBatchPreTransformed(LLVector3* verts, S32 vert_count) @@ -1464,13 +1897,50 @@ void LLRender::vertexBatchPreTransformed(LLVector3* verts, S32 vert_count) return; } - for (S32 i = 0; i < vert_count; i++) + if (sGLCoreProfile && mMode == LLRender::QUADS) + { //quads are deprecated, convert to triangle list + S32 i = 0; + + while (i < vert_count) + { + //read first three + mVerticesp[mCount++] = verts[i++]; + mTexcoordsp[mCount] = mTexcoordsp[mCount-1]; + mColorsp[mCount] = mColorsp[mCount-1]; + + mVerticesp[mCount++] = verts[i++]; + mTexcoordsp[mCount] = mTexcoordsp[mCount-1]; + mColorsp[mCount] = mColorsp[mCount-1]; + + mVerticesp[mCount++] = verts[i++]; + mTexcoordsp[mCount] = mTexcoordsp[mCount-1]; + mColorsp[mCount] = mColorsp[mCount-1]; + + //copy two + mVerticesp[mCount++] = verts[i-3]; + mTexcoordsp[mCount] = mTexcoordsp[mCount-1]; + mColorsp[mCount] = mColorsp[mCount-1]; + + mVerticesp[mCount++] = verts[i-1]; + mTexcoordsp[mCount] = mTexcoordsp[mCount-1]; + mColorsp[mCount] = mColorsp[mCount-1]; + + //copy last one + mVerticesp[mCount++] = verts[i++]; + mTexcoordsp[mCount] = mTexcoordsp[mCount-1]; + mColorsp[mCount] = mColorsp[mCount-1]; + } + } + else { - mVerticesp[mCount] = verts[i]; + for (S32 i = 0; i < vert_count; i++) + { + mVerticesp[mCount] = verts[i]; - mCount++; - mTexcoordsp[mCount] = mTexcoordsp[mCount-1]; - mColorsp[mCount] = mColorsp[mCount-1]; + mCount++; + mTexcoordsp[mCount] = mTexcoordsp[mCount-1]; + mColorsp[mCount] = mColorsp[mCount-1]; + } } mVerticesp[mCount] = mVerticesp[mCount-1]; @@ -1484,13 +1954,50 @@ void LLRender::vertexBatchPreTransformed(LLVector3* verts, LLVector2* uvs, S32 v return; } - for (S32 i = 0; i < vert_count; i++) + if (sGLCoreProfile && mMode == LLRender::QUADS) + { //quads are deprecated, convert to triangle list + S32 i = 0; + + while (i < vert_count) + { + //read first three + mVerticesp[mCount] = verts[i]; + mTexcoordsp[mCount++] = uvs[i++]; + mColorsp[mCount] = mColorsp[mCount-1]; + + mVerticesp[mCount] = verts[i]; + mTexcoordsp[mCount++] = uvs[i++]; + mColorsp[mCount] = mColorsp[mCount-1]; + + mVerticesp[mCount] = verts[i]; + mTexcoordsp[mCount++] = uvs[i++]; + mColorsp[mCount] = mColorsp[mCount-1]; + + //copy last two + mVerticesp[mCount] = verts[i-3]; + mTexcoordsp[mCount++] = uvs[i-3]; + mColorsp[mCount] = mColorsp[mCount-1]; + + mVerticesp[mCount] = verts[i-1]; + mTexcoordsp[mCount++] = uvs[i-1]; + mColorsp[mCount] = mColorsp[mCount-1]; + + //copy last one + mVerticesp[mCount] = verts[i]; + mTexcoordsp[mCount++] = uvs[i++]; + mColorsp[mCount] = mColorsp[mCount-1]; + } + } + else { - mVerticesp[mCount] = verts[i]; - mTexcoordsp[mCount] = uvs[i]; + for (S32 i = 0; i < vert_count; i++) + { + mVerticesp[mCount] = verts[i]; + mTexcoordsp[mCount] = uvs[i]; - mCount++; - mColorsp[mCount] = mColorsp[mCount-1]; + mCount++; + mColorsp[mCount] = mColorsp[mCount-1]; + } } mVerticesp[mCount] = mVerticesp[mCount-1]; @@ -1505,13 +2012,51 @@ void LLRender::vertexBatchPreTransformed(LLVector3* verts, LLVector2* uvs, LLCol return; } - for (S32 i = 0; i < vert_count; i++) + + if (sGLCoreProfile && mMode == LLRender::QUADS) + { //quads are deprecated, convert to triangle list + S32 i = 0; + + while (i < vert_count) + { + //read first three + mVerticesp[mCount] = verts[i]; + mTexcoordsp[mCount] = uvs[i]; + mColorsp[mCount++] = colors[i++]; + + mVerticesp[mCount] = verts[i]; + mTexcoordsp[mCount] = uvs[i]; + mColorsp[mCount++] = colors[i++]; + + mVerticesp[mCount] = verts[i]; + mTexcoordsp[mCount] = uvs[i]; + mColorsp[mCount++] = colors[i++]; + + //copy last two + mVerticesp[mCount] = verts[i-3]; + mTexcoordsp[mCount] = uvs[i-3]; + mColorsp[mCount++] = colors[i-3]; + + mVerticesp[mCount] = verts[i-1]; + mTexcoordsp[mCount] = uvs[i-1]; + mColorsp[mCount++] = colors[i-1]; + + //copy last one + mVerticesp[mCount] = verts[i]; + mTexcoordsp[mCount] = uvs[i]; + mColorsp[mCount++] = colors[i++]; + } + } + else { - mVerticesp[mCount] = verts[i]; - mTexcoordsp[mCount] = uvs[i]; - mColorsp[mCount] = colors[i]; + for (S32 i = 0; i < vert_count; i++) + { + mVerticesp[mCount] = verts[i]; + mTexcoordsp[mCount] = uvs[i]; + mColorsp[mCount] = colors[i]; - mCount++; + mCount++; + } } mVerticesp[mCount] = mVerticesp[mCount-1]; @@ -1586,6 +2131,81 @@ void LLRender::color3fv(const GLfloat* c) color4f(c[0],c[1],c[2],1); } +void LLRender::diffuseColor3f(F32 r, F32 g, F32 b) +{ + LLGLSLShader* shader = LLGLSLShader::sCurBoundShaderPtr; + llassert(!LLGLSLShader::sNoFixedFunction || shader != NULL); + + if (shader) + { + shader->uniform4f("color", r,g,b,1.f); + } + else + { + glColor3f(r,g,b); + } +} + +void LLRender::diffuseColor3fv(const F32* c) +{ + LLGLSLShader* shader = LLGLSLShader::sCurBoundShaderPtr; + llassert(!LLGLSLShader::sNoFixedFunction || shader != NULL); + + if (shader) + { + shader->uniform4f("color", c[0], c[1], c[2], 1.f); + } + else + { + glColor3fv(c); + } +} + +void LLRender::diffuseColor4f(F32 r, F32 g, F32 b, F32 a) +{ + LLGLSLShader* shader = LLGLSLShader::sCurBoundShaderPtr; + llassert(!LLGLSLShader::sNoFixedFunction || shader != NULL); + + if (shader) + { + shader->uniform4f("color", r,g,b,a); + } + else + { + glColor4f(r,g,b,a); + } +} + +void LLRender::diffuseColor4fv(const F32* c) +{ + LLGLSLShader* shader = LLGLSLShader::sCurBoundShaderPtr; + llassert(!LLGLSLShader::sNoFixedFunction || shader != NULL); + + if (shader) + { + shader->uniform4fv("color", 1, c); + } + else + { + glColor4fv(c); + } +} + +void LLRender::diffuseColor4ubv(const U8* c) +{ + LLGLSLShader* shader = LLGLSLShader::sCurBoundShaderPtr; + llassert(!LLGLSLShader::sNoFixedFunction || shader != NULL); + + if (shader) + { + shader->uniform4f("color", c[0]/255.f, c[1]/255.f, c[2]/255.f, c[3]/255.f); + } + else + { + glColor4ubv(c); + } +} + void LLRender::debugTexUnits(void) { LL_INFOS("TextureUnit") << "Active TexUnit: " << mCurrTextureUnitIndex << LL_ENDL; diff --git a/indra/llrender/llrender.h b/indra/llrender/llrender.h index 8f7ee30d87..61e503d384 100644 --- a/indra/llrender/llrender.h +++ b/indra/llrender/llrender.h @@ -41,6 +41,8 @@ #include "llstrider.h" #include "llpointer.h" #include "llglheaders.h" +#include "llmatrix4a.h" +#include "glh/glh_linear.h" class LLVertexBuffer; class LLCubeMap; @@ -48,6 +50,8 @@ class LLImageGL; class LLRenderTarget; class LLTexture ; +#define LL_MATRIX_STACK_DEPTH 32 + class LLTexUnit { friend class LLRender; @@ -235,6 +239,8 @@ public: void setSpotDirection(const LLVector3& direction); protected: + friend class LLRender; + S32 mIndex; bool mEnabled; LLColor4 mDiffuse; @@ -308,6 +314,18 @@ public: BF_UNDEF } eBlendFactor; + typedef enum + { + MM_MODELVIEW = 0, + MM_PROJECTION, + MM_TEXTURE0, + MM_TEXTURE1, + MM_TEXTURE2, + MM_TEXTURE3, + NUM_MATRIX_MODES, + MM_TEXTURE + } eMatrixMode; + LLRender(); ~LLRender(); void init() ; @@ -319,8 +337,22 @@ public: void translatef(const GLfloat& x, const GLfloat& y, const GLfloat& z); void scalef(const GLfloat& x, const GLfloat& y, const GLfloat& z); + void rotatef(const GLfloat& a, const GLfloat& x, const GLfloat& y, const GLfloat& z); + void ortho(F32 left, F32 right, F32 bottom, F32 top, F32 zNear, F32 zFar); + void pushMatrix(); void popMatrix(); + void loadMatrix(const GLfloat* m); + void loadMatrix(const GLdouble* m); + void loadIdentity(); + void multMatrix(const GLfloat* m); + void multMatrix(const GLdouble* m); + void matrixMode(U32 mode); + + const glh::matrix4f& getModelviewMatrix(); + + void syncMatrices(); + void syncLightState(); void translateUI(F32 x, F32 y, F32 z); void scaleUI(F32 x, F32 y, F32 z); @@ -351,6 +383,12 @@ public: void color3fv(const GLfloat* c); void color4ubv(const GLubyte* c); + void diffuseColor3f(F32 r, F32 g, F32 b); + void diffuseColor3fv(const F32* c); + void diffuseColor4f(F32 r, F32 g, F32 b, F32 a); + void diffuseColor4fv(const F32* c); + void diffuseColor4ubv(const U8* c); + void vertexBatchPreTransformed(LLVector3* verts, S32 vert_count); void vertexBatchPreTransformed(LLVector3* verts, LLVector2* uvs, S32 vert_count); void vertexBatchPreTransformed(LLVector3* verts, LLVector2* uvs, LLColor4U*, S32 vert_count); @@ -368,7 +406,8 @@ public: eBlendFactor alpha_sfactor, eBlendFactor alpha_dfactor); LLLightState* getLight(U32 index); - + void setAmbientLightColor(const LLColor4& color); + LLTexUnit* getTexUnit(U32 index); U32 getCurrentTexUnitIndex(void) const { return mCurrTextureUnitIndex; } @@ -389,9 +428,21 @@ public: public: static U32 sUICalls; static U32 sUIVerts; + static bool sGLCoreProfile; private: - bool mDirty; + friend class LLLightState; + + U32 mMatrixMode; + U32 mMatIdx[NUM_MATRIX_MODES]; + U32 mMatHash[NUM_MATRIX_MODES]; + glh::matrix4f mMatrix[NUM_MATRIX_MODES][LL_MATRIX_STACK_DEPTH]; + U32 mCurMatHash[NUM_MATRIX_MODES]; + U32 mLightHash; + LLColor4 mAmbientLightColor; + + bool mDirty; + U32 mQuadCycle; U32 mCount; U32 mMode; U32 mCurrTextureUnitIndex; diff --git a/indra/llrender/llrendersphere.cpp b/indra/llrender/llrendersphere.cpp index a5cd70445f..e7e07a1ab2 100644 --- a/indra/llrender/llrendersphere.cpp +++ b/indra/llrender/llrendersphere.cpp @@ -35,106 +35,11 @@ #include "llglheaders.h" -GLUquadricObj *gQuadObj2 = NULL; LLRenderSphere gSphere; -void drawSolidSphere(GLdouble radius, GLint slices, GLint stacks); - -void drawSolidSphere(GLdouble radius, GLint slices, GLint stacks) -{ - if (!gQuadObj2) - { - gQuadObj2 = gluNewQuadric(); - if (!gQuadObj2) - { - llwarns << "drawSolidSphere couldn't allocate quadric" << llendl; - return; - } - } - - gluQuadricDrawStyle(gQuadObj2, GLU_FILL); - gluQuadricNormals(gQuadObj2, GLU_SMOOTH); - // If we ever changed/used the texture or orientation state - // of quadObj, we'd need to change it to the defaults here - // with gluQuadricTexture and/or gluQuadricOrientation. - gluQuadricTexture(gQuadObj2, GL_TRUE); - gluSphere(gQuadObj2, radius, slices, stacks); -} - - -// A couple thoughts on sphere drawing: -// 1) You need more slices than stacks, but little less than 2:1 -// 2) At low LOD, setting stacks to an odd number avoids a "band" around the equator, making things look smoother -void LLRenderSphere::prerender() -{ - // Create a series of display lists for different LODs - mDList[0] = glGenLists(1); - glNewList(mDList[0], GL_COMPILE); - drawSolidSphere(1.0, 30, 20); - glEndList(); - - mDList[1] = glGenLists(1); - glNewList(mDList[1], GL_COMPILE); - drawSolidSphere(1.0, 20, 15); - glEndList(); - - mDList[2] = glGenLists(1); - glNewList(mDList[2], GL_COMPILE); - drawSolidSphere(1.0, 12, 8); - glEndList(); - - mDList[3] = glGenLists(1); - glNewList(mDList[3], GL_COMPILE); - drawSolidSphere(1.0, 8, 5); - glEndList(); -} - -void LLRenderSphere::cleanupGL() -{ - for (S32 detail = 0; detail < 4; detail++) - { - glDeleteLists(mDList[detail], 1); - mDList[detail] = 0; - } - - if (gQuadObj2) - { - gluDeleteQuadric(gQuadObj2); - gQuadObj2 = NULL; - } -} - -// Constants here are empirically derived from my eyeballs, JNC -// -// The toughest adjustment is the cutoff for the lowest LOD -// Maybe we should have more LODs at the low end? -void LLRenderSphere::render(F32 pixel_area) -{ - S32 level_of_detail; - - if (pixel_area > 10000.f) - { - level_of_detail = 0; - } - else if (pixel_area > 800.f) - { - level_of_detail = 1; - } - else if (pixel_area > 100.f) - { - level_of_detail = 2; - } - else - { - level_of_detail = 3; - } - glCallList(mDList[level_of_detail]); -} - - void LLRenderSphere::render() { - glCallList(mDList[0]); + renderGGL(); } inline LLVector3 polar_to_cart(F32 latitude, F32 longitude) diff --git a/indra/llrender/llrendersphere.h b/indra/llrender/llrendersphere.h index 96a6bec80c..f8e9e86e7f 100644 --- a/indra/llrender/llrendersphere.h +++ b/indra/llrender/llrendersphere.h @@ -40,11 +40,6 @@ void lat2xyz(LLVector3 * result, F32 lat, F32 lon); // utility routine class LLRenderSphere { public: - LLGLuint mDList[5]; - - void prerender(); - void cleanupGL(); - void render(F32 pixel_area); // of a box of size 1.0 at that position void render(); // render at highest LOD void renderGGL(); // render using LLRender diff --git a/indra/llrender/llrendertarget.cpp b/indra/llrender/llrendertarget.cpp index 8c0d3592df..1aa12614ea 100644 --- a/indra/llrender/llrendertarget.cpp +++ b/indra/llrender/llrendertarget.cpp @@ -31,8 +31,7 @@ #include "llgl.h" LLRenderTarget* LLRenderTarget::sBoundTarget = NULL; - - +U32 LLRenderTarget::sBytesAllocated = 0; void check_framebuffer_status() { @@ -62,8 +61,7 @@ LLRenderTarget::LLRenderTarget() : mStencil(0), mUseDepth(false), mRenderDepth(false), - mUsage(LLTexUnit::TT_TEXTURE), - mSamples(0) + mUsage(LLTexUnit::TT_TEXTURE) { } @@ -84,20 +82,6 @@ bool LLRenderTarget::allocate(U32 resx, U32 resy, U32 color_fmt, bool depth, boo mStencil = stencil; mUsage = usage; mUseDepth = depth; - mSamples = samples; - - 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) { @@ -157,21 +141,6 @@ bool LLRenderTarget::addColorAttachment(U32 color_fmt) stop_glerror(); -#ifdef GL_ARB_texture_multisample - if (mSamples > 1) - { - clear_glerror(); - glTexImage2DMultisample(LLTexUnit::getInternalType(mUsage), mSamples, color_fmt, mResX, mResY, GL_TRUE); - if (glGetError() != GL_NO_ERROR) - { - llwarns << "Could not allocate multisample color buffer for render target." << llendl; - return false; - } - } - else -#else - llassert_always(mSamples <= 1); -#endif { clear_glerror(); LLImageGL::setManualImage(LLTexUnit::getInternalType(mUsage), 0, color_fmt, mResX, mResY, GL_RGBA, GL_UNSIGNED_BYTE, NULL); @@ -182,32 +151,32 @@ bool LLRenderTarget::addColorAttachment(U32 color_fmt) } } + sBytesAllocated += mResX*mResY*4; + 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 (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 (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) @@ -250,26 +219,16 @@ bool LLRenderTarget::allocateDepth() { LLImageGL::generateTextures(1, &mDepth); gGL.getTexUnit(0)->bindManual(mUsage, mDepth); - if (mSamples == 0) - { - U32 internal_type = LLTexUnit::getInternalType(mUsage); - gGL.getTexUnit(0)->setTextureFilteringOption(LLTexUnit::TFO_POINT); - stop_glerror(); - clear_glerror(); - LLImageGL::setManualImage(internal_type, 0, GL_DEPTH_COMPONENT32, mResX, mResY, GL_DEPTH_COMPONENT, GL_UNSIGNED_INT, NULL); - } -#ifdef GL_ARB_texture_multisample - else - { - stop_glerror(); - clear_glerror(); - glTexImage2DMultisample(LLTexUnit::getInternalType(mUsage), mSamples, GL_DEPTH_COMPONENT32, mResX, mResY, GL_TRUE); - } -#else - llassert_always(mSamples <= 1); -#endif + + U32 internal_type = LLTexUnit::getInternalType(mUsage); + stop_glerror(); + clear_glerror(); + LLImageGL::setManualImage(internal_type, 0, GL_DEPTH_COMPONENT24, mResX, mResY, GL_DEPTH_COMPONENT, GL_UNSIGNED_INT, NULL); + gGL.getTexUnit(0)->setTextureFilteringOption(LLTexUnit::TFO_POINT); } + sBytesAllocated += mResX*mResY*4; + if (glGetError() != GL_NO_ERROR) { llwarns << "Unable to allocate depth buffer for render target." << llendl; @@ -339,14 +298,16 @@ void LLRenderTarget::release() stop_glerror(); } mDepth = 0; + + sBytesAllocated -= mResX*mResY*4; } else if (mUseDepth && mFBO) { //detach shared depth buffer glBindFramebuffer(GL_FRAMEBUFFER, mFBO); if (mStencil) { //attached as a renderbuffer - glFramebufferRenderbuffer(GL_FRAMEBUFFER, GL_DEPTH_ATTACHMENT, GL_RENDERBUFFER, 0); glFramebufferRenderbuffer(GL_FRAMEBUFFER, GL_STENCIL_ATTACHMENT, GL_RENDERBUFFER, 0); + glFramebufferRenderbuffer(GL_FRAMEBUFFER, GL_DEPTH_ATTACHMENT, GL_RENDERBUFFER, 0); mStencil = false; } else @@ -364,6 +325,7 @@ void LLRenderTarget::release() if (mTex.size() > 0) { + sBytesAllocated -= mResX*mResY*4*mTex.size(); LLImageGL::deleteTextures(mTex.size(), &mTex[0], true); mTex.clear(); } diff --git a/indra/llrender/llrendertarget.h b/indra/llrender/llrendertarget.h index dea1de12d8..2735ab21c5 100644 --- a/indra/llrender/llrendertarget.h +++ b/indra/llrender/llrendertarget.h @@ -64,6 +64,7 @@ class LLRenderTarget public: //whether or not to use FBO implementation static bool sUseFBO; + static U32 sBytesAllocated; LLRenderTarget(); ~LLRenderTarget(); @@ -147,7 +148,6 @@ protected: bool mUseDepth; bool mRenderDepth; LLTexUnit::eTextureType mUsage; - U32 mSamples; static LLRenderTarget* sBoundTarget; }; diff --git a/indra/llrender/llshadermgr.cpp b/indra/llrender/llshadermgr.cpp index 986c1f2774..16180c6831 100644 --- a/indra/llrender/llshadermgr.cpp +++ b/indra/llrender/llshadermgr.cpp @@ -81,7 +81,14 @@ BOOL LLShaderMgr::attachShaderFeatures(LLGLSLShader * shader) // NOTE order of shader object attaching is VERY IMPORTANT!!! if (features->calculatesAtmospherics) { - if (!shader->attachObject("windlight/atmosphericsVarsV.glsl")) + if (features->hasWaterFog) + { + if (!shader->attachObject("windlight/atmosphericsVarsWaterV.glsl")) + { + return FALSE; + } + } + else if (!shader->attachObject("windlight/atmosphericsVarsV.glsl")) { return FALSE; } @@ -161,7 +168,14 @@ BOOL LLShaderMgr::attachShaderFeatures(LLGLSLShader * shader) if(features->calculatesAtmospherics) { - if (!shader->attachObject("windlight/atmosphericsVarsF.glsl")) + if (features->hasWaterFog) + { + if (!shader->attachObject("windlight/atmosphericsVarsWaterF.glsl")) + { + return FALSE; + } + } + else if (!shader->attachObject("windlight/atmosphericsVarsF.glsl")) { return FALSE; } @@ -241,7 +255,7 @@ BOOL LLShaderMgr::attachShaderFeatures(LLGLSLShader * shader) return FALSE; } } - shader->mFeatures.mIndexedTextureChannels = gGLManager.mNumTextureImageUnits-1; + shader->mFeatures.mIndexedTextureChannels = llmax(LLGLSLShader::sIndexedTextureChannels-1, 1); } } @@ -280,7 +294,7 @@ BOOL LLShaderMgr::attachShaderFeatures(LLGLSLShader * shader) return FALSE; } } - shader->mFeatures.mIndexedTextureChannels = gGLManager.mNumTextureImageUnits-1; + shader->mFeatures.mIndexedTextureChannels = llmax(LLGLSLShader::sIndexedTextureChannels-1, 1); } } } @@ -304,7 +318,7 @@ BOOL LLShaderMgr::attachShaderFeatures(LLGLSLShader * shader) { return FALSE; } - shader->mFeatures.mIndexedTextureChannels = gGLManager.mNumTextureImageUnits-1; + shader->mFeatures.mIndexedTextureChannels = llmax(LLGLSLShader::sIndexedTextureChannels-1, 1); } } else if (features->hasWaterFog) @@ -336,7 +350,7 @@ BOOL LLShaderMgr::attachShaderFeatures(LLGLSLShader * shader) { return FALSE; } - shader->mFeatures.mIndexedTextureChannels = gGLManager.mNumTextureImageUnits-1; + shader->mFeatures.mIndexedTextureChannels = llmax(LLGLSLShader::sIndexedTextureChannels-1, 1); } } @@ -355,7 +369,7 @@ BOOL LLShaderMgr::attachShaderFeatures(LLGLSLShader * shader) { return FALSE; } - shader->mFeatures.mIndexedTextureChannels = gGLManager.mNumTextureImageUnits-1; + shader->mFeatures.mIndexedTextureChannels = llmax(LLGLSLShader::sIndexedTextureChannels-1, 1); } } @@ -395,7 +409,7 @@ BOOL LLShaderMgr::attachShaderFeatures(LLGLSLShader * shader) return FALSE; } } - shader->mFeatures.mIndexedTextureChannels = gGLManager.mNumTextureImageUnits-1; + shader->mFeatures.mIndexedTextureChannels = llmax(LLGLSLShader::sIndexedTextureChannels-1, 1); } } } @@ -419,7 +433,7 @@ BOOL LLShaderMgr::attachShaderFeatures(LLGLSLShader * shader) { return FALSE; } - shader->mFeatures.mIndexedTextureChannels = gGLManager.mNumTextureImageUnits-1; + shader->mFeatures.mIndexedTextureChannels = llmax(LLGLSLShader::sIndexedTextureChannels-1, 1); } } @@ -438,10 +452,26 @@ BOOL LLShaderMgr::attachShaderFeatures(LLGLSLShader * shader) { return FALSE; } - shader->mFeatures.mIndexedTextureChannels = gGLManager.mNumTextureImageUnits-1; + shader->mFeatures.mIndexedTextureChannels = llmax(LLGLSLShader::sIndexedTextureChannels-1, 1); } } } + + if (features->mIndexedTextureChannels <= 1) + { + if (!shader->attachObject("objects/nonindexedTextureV.glsl")) + { + return FALSE; + } + } + else + { + if (!shader->attachObject("objects/indexedTextureV.glsl")) + { + return FALSE; + } + } + return TRUE; } @@ -477,7 +507,7 @@ void LLShaderMgr::dumpObjectLog(GLhandleARB ret, BOOL warns) } else { - LL_DEBUGS("ShaderLoading") << log << LL_ENDL; + LL_INFOS("ShaderLoading") << log << LL_ENDL; } } } @@ -531,23 +561,57 @@ GLhandleARB LLShaderMgr::loadShaderFile(const std::string& filename, S32 & shade } //we can't have any lines longer than 1024 characters - //or any shaders longer than 1024 lines... deal - DaveP + //or any shaders longer than 4096 lines... deal - DaveP GLcharARB buff[1024]; - GLcharARB* text[1024]; + GLcharARB* text[4096]; GLuint count = 0; if (gGLManager.mGLVersion < 2.1f) { text[count++] = strdup("#version 110\n"); + text[count++] = strdup("#define ATTRIBUTE attribute\n"); + text[count++] = strdup("#define VARYING varying\n"); } else if (gGLManager.mGLVersion < 3.f) { //set version to 1.20 text[count++] = strdup("#version 120\n"); + text[count++] = strdup("#define FXAA_GLSL_120 1\n"); + text[count++] = strdup("#define FXAA_FAST_PIXEL_OFFSET 0\n"); + text[count++] = strdup("#define ATTRIBUTE attribute\n"); + text[count++] = strdup("#define VARYING varying\n"); } else - { //set version to 1.30 - text[count++] = strdup("#version 130\n"); + { + if (gGLManager.mGLVersion < 4.f) + { + //set version to 1.30 + text[count++] = strdup("#version 130\n"); + } + else + { //set version to 400 + text[count++] = strdup("#version 400\n"); + } + + text[count++] = strdup("#define DEFINE_GL_FRAGCOLOR 1\n"); + text[count++] = strdup("#define FXAA_GLSL_130 1\n"); + + text[count++] = strdup("#define ATTRIBUTE in\n"); + + if (type == GL_VERTEX_SHADER_ARB) + { //"varying" state is "out" in a vertex program, "in" in a fragment program + // ("varying" is deprecated after version 1.20) + text[count++] = strdup("#define VARYING out\n"); + } + else + { + text[count++] = strdup("#define VARYING in\n"); + } + + //backwards compatibility with legacy texture lookup syntax + text[count++] = strdup("#define textureCube texture\n"); + text[count++] = strdup("#define texture2DLod textureLod\n"); + text[count++] = strdup("#define shadow2D texture\n"); } //copy preprocessor definitions into buffer @@ -571,7 +635,7 @@ GLhandleARB LLShaderMgr::loadShaderFile(const std::string& filename, S32 & shade . uniform sampler2D texN; - varying float vary_texture_index; + VARYING float vary_texture_index; vec4 diffuseLookup(vec2 texcoord) { @@ -597,7 +661,11 @@ GLhandleARB LLShaderMgr::loadShaderFile(const std::string& filename, S32 & shade text[count++] = strdup(decl.c_str()); } - text[count++] = strdup("varying float vary_texture_index;\n"); + if (texture_index_channels > 1) + { + text[count++] = strdup("VARYING float vary_texture_index;\n"); + } + text[count++] = strdup("vec4 diffuseLookup(vec2 texcoord)\n"); text[count++] = strdup("{\n"); @@ -649,7 +717,7 @@ GLhandleARB LLShaderMgr::loadShaderFile(const std::string& filename, S32 & shade } //copy file into memory - while( fgets((char *)buff, 1024, file) != NULL && count < LL_ARRAY_SIZE(buff) ) + while( fgets((char *)buff, 1024, file) != NULL && count < LL_ARRAY_SIZE(text) ) { text[count++] = (GLcharARB *)strdup((char *)buff); } @@ -704,14 +772,24 @@ GLhandleARB LLShaderMgr::loadShaderFile(const std::string& filename, S32 & shade LL_WARNS("ShaderLoading") << "GLSL Compilation Error: (" << error << ") in " << filename << LL_ENDL; dumpObjectLog(ret); +#if LL_WINDOWS std::stringstream ostr; //dump shader source for debugging for (GLuint i = 0; i < count; i++) { ostr << i << ": " << text[i]; + + if (i % 128 == 0) + { //dump every 128 lines + + LL_WARNS("ShaderLoading") << "\n" << ostr.str() << llendl; + ostr = std::stringstream(); + } + } LL_WARNS("ShaderLoading") << "\n" << ostr.str() << llendl; +#endif // LL_WINDOWS ret = 0; } diff --git a/indra/llrender/llvertexbuffer.cpp b/indra/llrender/llvertexbuffer.cpp index 8fd1193780..4484a880cc 100644 --- a/indra/llrender/llvertexbuffer.cpp +++ b/indra/llrender/llvertexbuffer.cpp @@ -34,6 +34,7 @@ #include "llmemtype.h" #include "llrender.h" #include "llvector4a.h" +#include "llshadermgr.h" #include "llglslshader.h" #include "llmemory.h" @@ -55,6 +56,7 @@ S32 LLVertexBuffer::sMappedCount = 0; BOOL LLVertexBuffer::sDisableVBOMapping = FALSE ; BOOL LLVertexBuffer::sEnableVBOs = TRUE; U32 LLVertexBuffer::sGLRenderBuffer = 0; +U32 LLVertexBuffer::sGLRenderArray = 0; U32 LLVertexBuffer::sGLRenderIndices = 0; U32 LLVertexBuffer::sLastMask = 0; BOOL LLVertexBuffer::sVBOActive = FALSE; @@ -62,12 +64,10 @@ BOOL LLVertexBuffer::sIBOActive = FALSE; U32 LLVertexBuffer::sAllocatedBytes = 0; BOOL LLVertexBuffer::sMapped = FALSE; BOOL LLVertexBuffer::sUseStreamDraw = TRUE; +BOOL LLVertexBuffer::sUseVAO = FALSE; BOOL LLVertexBuffer::sPreferStreamDraw = FALSE; -S32 LLVertexBuffer::sWeight4Loc = -1; - std::vector<U32> LLVertexBuffer::sDeleteList; - const U32 FENCE_WAIT_TIME_NANOSECONDS = 10000; //1 ms class LLGLSyncFence : public LLGLFence @@ -122,6 +122,7 @@ public: }; +//NOTE: each component must be AT LEAST 4 bytes in size to avoid a performance penalty on AMD hardware S32 LLVertexBuffer::sTypeSize[LLVertexBuffer::TYPE_MAX] = { sizeof(LLVector4), // TYPE_VERTEX, @@ -131,10 +132,12 @@ S32 LLVertexBuffer::sTypeSize[LLVertexBuffer::TYPE_MAX] = sizeof(LLVector2), // TYPE_TEXCOORD2, sizeof(LLVector2), // TYPE_TEXCOORD3, sizeof(LLColor4U), // TYPE_COLOR, + sizeof(LLColor4U), // TYPE_EMISSIVE, only alpha is used currently sizeof(LLVector4), // TYPE_BINORMAL, sizeof(F32), // TYPE_WEIGHT, sizeof(LLVector4), // TYPE_WEIGHT4, sizeof(LLVector4), // TYPE_CLOTHWEIGHT, + sizeof(LLVector4), // TYPE_TEXTURE_INDEX (actually exists as position.w), no extra data, but stride is 16 bytes }; U32 LLVertexBuffer::sGLMode[LLRender::NUM_MODES] = @@ -149,146 +152,147 @@ U32 LLVertexBuffer::sGLMode[LLRender::NUM_MODES] = GL_LINE_LOOP, }; + //static void LLVertexBuffer::setupClientArrays(U32 data_mask) { - /*if (LLGLImmediate::sStarted) - { - llerrs << "Cannot use LLGLImmediate and LLVertexBuffer simultaneously!" << llendl; - }*/ - if (sLastMask != data_mask) { - U32 mask[] = + BOOL error = FALSE; + + if (LLGLSLShader::sNoFixedFunction) { - MAP_VERTEX, - MAP_NORMAL, - MAP_TEXCOORD0, - MAP_COLOR, - }; - - GLenum array[] = + for (U32 i = 0; i < TYPE_MAX; ++i) + { + S32 loc = i; + + U32 mask = 1 << i; + + if (sLastMask & (1 << i)) + { //was enabled + if (!(data_mask & mask)) + { //needs to be disabled + glDisableVertexAttribArrayARB(loc); + } + } + else + { //was disabled + if (data_mask & mask) + { //needs to be enabled + glEnableVertexAttribArrayARB(loc); + } + } + } + } + else { - GL_VERTEX_ARRAY, - GL_NORMAL_ARRAY, - GL_TEXTURE_COORD_ARRAY, - GL_COLOR_ARRAY, - }; - BOOL error = FALSE; - for (U32 i = 0; i < 4; ++i) - { - if (sLastMask & mask[i]) - { //was enabled - if (!(data_mask & mask[i]) && i > 0) - { //needs to be disabled - glDisableClientState(array[i]); + GLenum array[] = + { + GL_VERTEX_ARRAY, + GL_NORMAL_ARRAY, + GL_TEXTURE_COORD_ARRAY, + GL_COLOR_ARRAY, + }; + + GLenum mask[] = + { + MAP_VERTEX, + MAP_NORMAL, + MAP_TEXCOORD0, + MAP_COLOR + }; + + + + for (U32 i = 0; i < 4; ++i) + { + if (sLastMask & mask[i]) + { //was enabled + if (!(data_mask & mask[i])) + { //needs to be disabled + glDisableClientState(array[i]); + } + else if (gDebugGL) + { //needs to be enabled, make sure it was (DEBUG) + if (!glIsEnabled(array[i])) + { + if (gDebugSession) + { + error = TRUE; + gFailLog << "Bad client state! " << array[i] << " disabled." << std::endl; + } + else + { + llerrs << "Bad client state! " << array[i] << " disabled." << llendl; + } + } + } } - else if (gDebugGL) - { //needs to be enabled, make sure it was (DEBUG TEMPORARY) - if (i > 0 && !glIsEnabled(array[i])) - { + else + { //was disabled + if (data_mask & mask[i]) + { //needs to be enabled + glEnableClientState(array[i]); + } + else if (gDebugGL && glIsEnabled(array[i])) + { //needs to be disabled, make sure it was (DEBUG TEMPORARY) if (gDebugSession) { error = TRUE; - gFailLog << "Bad client state! " << array[i] << " disabled." << std::endl; + gFailLog << "Bad client state! " << array[i] << " enabled." << std::endl; } else { - llerrs << "Bad client state! " << array[i] << " disabled." << llendl; + llerrs << "Bad client state! " << array[i] << " enabled." << llendl; } } } } - else - { //was disabled - if (data_mask & mask[i] && i > 0) - { //needs to be enabled - glEnableClientState(array[i]); - } - else if (gDebugGL && i > 0 && glIsEnabled(array[i])) - { //needs to be disabled, make sure it was (DEBUG TEMPORARY) - if (gDebugSession) - { - error = TRUE; - gFailLog << "Bad client state! " << array[i] << " enabled." << std::endl; - } - else - { - llerrs << "Bad client state! " << array[i] << " enabled." << llendl; + + U32 map_tc[] = + { + MAP_TEXCOORD1, + MAP_TEXCOORD2, + MAP_TEXCOORD3 + }; + + for (U32 i = 0; i < 3; i++) + { + if (sLastMask & map_tc[i]) + { + if (!(data_mask & map_tc[i])) + { //disable + glClientActiveTextureARB(GL_TEXTURE1_ARB+i); + glDisableClientState(GL_TEXTURE_COORD_ARRAY); + glClientActiveTextureARB(GL_TEXTURE0_ARB); } } + else if (data_mask & map_tc[i]) + { + glClientActiveTextureARB(GL_TEXTURE1_ARB+i); + glEnableClientState(GL_TEXTURE_COORD_ARRAY); + glClientActiveTextureARB(GL_TEXTURE0_ARB); + } } - } - - if (error) - { - ll_fail("LLVertexBuffer::setupClientArrays failed"); - } - - U32 map_tc[] = - { - MAP_TEXCOORD1, - MAP_TEXCOORD2, - MAP_TEXCOORD3 - }; - for (U32 i = 0; i < 3; i++) - { - if (sLastMask & map_tc[i]) + if (sLastMask & MAP_BINORMAL) { - if (!(data_mask & map_tc[i])) + if (!(data_mask & MAP_BINORMAL)) { - glClientActiveTextureARB(GL_TEXTURE1_ARB+i); + glClientActiveTextureARB(GL_TEXTURE2_ARB); glDisableClientState(GL_TEXTURE_COORD_ARRAY); glClientActiveTextureARB(GL_TEXTURE0_ARB); } } - else if (data_mask & map_tc[i]) - { - glClientActiveTextureARB(GL_TEXTURE1_ARB+i); - glEnableClientState(GL_TEXTURE_COORD_ARRAY); - glClientActiveTextureARB(GL_TEXTURE0_ARB); - } - } - - if (sLastMask & MAP_BINORMAL) - { - if (!(data_mask & MAP_BINORMAL)) + else if (data_mask & MAP_BINORMAL) { glClientActiveTextureARB(GL_TEXTURE2_ARB); - glDisableClientState(GL_TEXTURE_COORD_ARRAY); + glEnableClientState(GL_TEXTURE_COORD_ARRAY); glClientActiveTextureARB(GL_TEXTURE0_ARB); } } - else if (data_mask & MAP_BINORMAL) - { - glClientActiveTextureARB(GL_TEXTURE2_ARB); - glEnableClientState(GL_TEXTURE_COORD_ARRAY); - glClientActiveTextureARB(GL_TEXTURE0_ARB); - } - - if (sLastMask & MAP_WEIGHT4) - { - if (sWeight4Loc < 0) - { - llerrs << "Weighting disabled but vertex buffer still bound!" << llendl; - } - - if (!(data_mask & MAP_WEIGHT4)) - { //disable 4-component skin weight - glDisableVertexAttribArrayARB(sWeight4Loc); - } - } - else if (data_mask & MAP_WEIGHT4) - { - if (sWeight4Loc >= 0) - { //enable 4-component skin weight - glEnableVertexAttribArrayARB(sWeight4Loc); - } - } - sLastMask = data_mask; } } @@ -296,6 +300,9 @@ void LLVertexBuffer::setupClientArrays(U32 data_mask) //static void LLVertexBuffer::drawArrays(U32 mode, const std::vector<LLVector3>& pos, const std::vector<LLVector3>& norm) { + llassert(!LLGLSLShader::sNoFixedFunction || LLGLSLShader::sCurBoundShaderPtr != NULL); + gGL.syncMatrices(); + U32 count = pos.size(); llassert_always(norm.size() >= pos.size()); llassert_always(count > 0) ; @@ -304,12 +311,67 @@ void LLVertexBuffer::drawArrays(U32 mode, const std::vector<LLVector3>& pos, con setupClientArrays(MAP_VERTEX | MAP_NORMAL); - glVertexPointer(3, GL_FLOAT, 0, pos[0].mV); - glNormalPointer(GL_FLOAT, 0, norm[0].mV); + LLGLSLShader* shader = LLGLSLShader::sCurBoundShaderPtr; + + if (shader) + { + S32 loc = LLVertexBuffer::TYPE_VERTEX; + if (loc > -1) + { + glVertexAttribPointerARB(loc, 3, GL_FLOAT, GL_FALSE, 0, pos[0].mV); + } + loc = LLVertexBuffer::TYPE_NORMAL; + if (loc > -1) + { + glVertexAttribPointerARB(loc, 3, GL_FLOAT, GL_FALSE, 0, norm[0].mV); + } + } + else + { + glVertexPointer(3, GL_FLOAT, 0, pos[0].mV); + glNormalPointer(GL_FLOAT, 0, norm[0].mV); + } glDrawArrays(sGLMode[mode], 0, count); } +//static +void LLVertexBuffer::drawElements(U32 mode, const LLVector4a* pos, const LLVector2* tc, S32 num_indices, const U16* indicesp) +{ + llassert(!LLGLSLShader::sNoFixedFunction || LLGLSLShader::sCurBoundShaderPtr != NULL); + + gGL.syncMatrices(); + + U32 mask = LLVertexBuffer::MAP_VERTEX; + if (tc) + { + mask = mask | LLVertexBuffer::MAP_TEXCOORD0; + } + + unbind(); + + setupClientArrays(mask); + + if (LLGLSLShader::sNoFixedFunction) + { + S32 loc = LLVertexBuffer::TYPE_VERTEX; + glVertexAttribPointerARB(loc, 3, GL_FLOAT, GL_FALSE, 16, pos); + + if (tc) + { + loc = LLVertexBuffer::TYPE_TEXCOORD0; + glVertexAttribPointerARB(loc, 2, GL_FLOAT, GL_FALSE, 0, tc); + } + } + else + { + glTexCoordPointer(2, GL_FLOAT, 0, tc); + glVertexPointer(3, GL_FLOAT, 16, pos); + } + + glDrawElements(sGLMode[mode], num_indices, GL_UNSIGNED_SHORT, indicesp); +} + void LLVertexBuffer::validateRange(U32 start, U32 end, U32 count, U32 indices_offset) const { if (start >= (U32) mRequestedNumVerts || @@ -343,16 +405,40 @@ void LLVertexBuffer::drawRange(U32 mode, U32 start, U32 end, U32 count, U32 indi { validateRange(start, end, count, indices_offset); + gGL.syncMatrices(); + llassert(mRequestedNumVerts >= 0); + llassert(!LLGLSLShader::sNoFixedFunction || LLGLSLShader::sCurBoundShaderPtr != NULL); - if (mGLIndices != sGLRenderIndices) + if (mGLArray) + { + if (mGLArray != sGLRenderArray) + { + llerrs << "Wrong vertex array bound." << llendl; + } + } + else { - llerrs << "Wrong index buffer bound." << llendl; + if (mGLIndices != sGLRenderIndices) + { + llerrs << "Wrong index buffer bound." << llendl; + } + + if (mGLBuffer != sGLRenderBuffer) + { + llerrs << "Wrong vertex buffer bound." << llendl; + } } - if (mGLBuffer != sGLRenderBuffer) + if (gDebugGL && !mGLArray && useVBOs()) { - llerrs << "Wrong vertex buffer bound." << llendl; + GLint elem = 0; + glGetIntegerv(GL_ELEMENT_ARRAY_BUFFER_BINDING_ARB, &elem); + + if (elem != mGLIndices) + { + llerrs << "Wrong index buffer bound!" << llendl; + } } if (mode >= LLRender::NUM_MODES) @@ -372,6 +458,10 @@ void LLVertexBuffer::drawRange(U32 mode, U32 start, U32 end, U32 count, U32 indi void LLVertexBuffer::draw(U32 mode, U32 count, U32 indices_offset) const { + llassert(!LLGLSLShader::sNoFixedFunction || LLGLSLShader::sCurBoundShaderPtr != NULL); + + gGL.syncMatrices(); + llassert(mRequestedNumIndices >= 0); if (indices_offset >= (U32) mRequestedNumIndices || indices_offset + count > (U32) mRequestedNumIndices) @@ -379,14 +469,24 @@ void LLVertexBuffer::draw(U32 mode, U32 count, U32 indices_offset) const llerrs << "Bad index buffer draw range: [" << indices_offset << ", " << indices_offset+count << "]" << llendl; } - if (mGLIndices != sGLRenderIndices) + if (mGLArray) { - llerrs << "Wrong index buffer bound." << llendl; + if (mGLArray != sGLRenderArray) + { + llerrs << "Wrong vertex array bound." << llendl; + } } - - if (mGLBuffer != sGLRenderBuffer) + else { - llerrs << "Wrong vertex buffer bound." << llendl; + if (mGLIndices != sGLRenderIndices) + { + llerrs << "Wrong index buffer bound." << llendl; + } + + if (mGLBuffer != sGLRenderBuffer) + { + llerrs << "Wrong vertex buffer bound." << llendl; + } } if (mode >= LLRender::NUM_MODES) @@ -404,6 +504,10 @@ void LLVertexBuffer::draw(U32 mode, U32 count, U32 indices_offset) const void LLVertexBuffer::drawArrays(U32 mode, U32 first, U32 count) const { + llassert(!LLGLSLShader::sNoFixedFunction || LLGLSLShader::sCurBoundShaderPtr != NULL); + + gGL.syncMatrices(); + llassert(mRequestedNumVerts >= 0); if (first >= (U32) mRequestedNumVerts || first + count > (U32) mRequestedNumVerts) @@ -411,9 +515,19 @@ void LLVertexBuffer::drawArrays(U32 mode, U32 first, U32 count) const llerrs << "Bad vertex buffer draw range: [" << first << ", " << first+count << "]" << llendl; } - if (mGLBuffer != sGLRenderBuffer || useVBOs() != sVBOActive) + if (mGLArray) + { + if (mGLArray != sGLRenderArray) + { + llerrs << "Wrong vertex array bound." << llendl; + } + } + else { - llerrs << "Wrong vertex buffer bound." << llendl; + if (mGLBuffer != sGLRenderBuffer || useVBOs() != sVBOActive) + { + llerrs << "Wrong vertex buffer bound." << llendl; + } } if (mode >= LLRender::NUM_MODES) @@ -446,7 +560,7 @@ void LLVertexBuffer::initClass(bool use_vbo, bool no_vbo_mapping) sDisableVBOMapping = sEnableVBOs && no_vbo_mapping ; if(!sPrivatePoolp) - { + { sPrivatePoolp = LLPrivateMemoryPoolManager::getInstance()->newPool(LLPrivateMemoryPool::STATIC) ; } } @@ -454,6 +568,16 @@ void LLVertexBuffer::initClass(bool use_vbo, bool no_vbo_mapping) //static void LLVertexBuffer::unbind() { + if (sGLRenderArray) + { +#if GL_ARB_vertex_array_object + glBindVertexArray(0); +#endif + sGLRenderArray = 0; + sGLRenderIndices = 0; + sIBOActive = FALSE; + } + if (sVBOActive) { glBindBufferARB(GL_ARRAY_BUFFER_ARB, 0); @@ -505,6 +629,7 @@ LLVertexBuffer::LLVertexBuffer(U32 typemask, S32 usage) : mRequestedNumIndices(-1), mUsage(usage), mGLBuffer(0), + mGLArray(0), mGLIndices(0), mMappedData(NULL), mMappedIndexData(NULL), @@ -534,6 +659,11 @@ LLVertexBuffer::LLVertexBuffer(U32 typemask, S32 usage) : mUsage = GL_STREAM_DRAW_ARB; } + if (mUsage == 0 && LLRender::sGLCoreProfile) + { //MUST use VBOs for all rendering + mUsage = GL_STREAM_DRAW_ARB; + } + //zero out offsets for (U32 i = 0; i < TYPE_MAX; i++) { @@ -552,12 +682,12 @@ LLVertexBuffer::LLVertexBuffer(U32 typemask, S32 usage) : S32 LLVertexBuffer::calcOffsets(const U32& typemask, S32* offsets, S32 num_vertices) { S32 offset = 0; - for (S32 i=0; i<TYPE_MAX; i++) + for (S32 i=0; i<TYPE_TEXTURE_INDEX; i++) { U32 mask = 1<<i; if (typemask & mask) { - if (offsets) + if (offsets && LLVertexBuffer::sTypeSize[i]) { offsets[i] = offset; offset += LLVertexBuffer::sTypeSize[i]*num_vertices; @@ -566,6 +696,8 @@ S32 LLVertexBuffer::calcOffsets(const U32& typemask, S32* offsets, S32 num_verti } } + offsets[TYPE_TEXTURE_INDEX] = offsets[TYPE_VERTEX] + 12; + return offset+16; } @@ -573,7 +705,7 @@ S32 LLVertexBuffer::calcOffsets(const U32& typemask, S32* offsets, S32 num_verti S32 LLVertexBuffer::calcVertexSize(const U32& typemask) { S32 size = 0; - for (S32 i = 0; i < TYPE_MAX; i++) + for (S32 i = 0; i < TYPE_TEXTURE_INDEX; i++) { U32 mask = 1<<i; if (typemask & mask) @@ -597,6 +729,14 @@ LLVertexBuffer::~LLVertexBuffer() LLMemType mt2(LLMemType::MTYPE_VERTEX_DESTRUCTOR); destroyGLBuffer(); destroyGLIndices(); + + if (mGLArray) + { +#if GL_ARB_vertex_array_object + glDeleteVertexArrays(1, &mGLArray); +#endif + } + sCount--; if (mFence) @@ -888,7 +1028,9 @@ void LLVertexBuffer::updateNumIndices(S32 nindices) void LLVertexBuffer::allocateBuffer(S32 nverts, S32 nindices, bool create) { LLMemType mt2(LLMemType::MTYPE_VERTEX_ALLOCATE_BUFFER); - + + stop_glerror(); + if (nverts < 0 || nindices < 0 || nverts > 65536) { @@ -906,11 +1048,104 @@ void LLVertexBuffer::allocateBuffer(S32 nverts, S32 nindices, bool create) { createGLBuffer(); createGLIndices(); + + //actually allocate space for the vertex buffer if using VBO mapping + flush(); + + if (gGLManager.mHasVertexArrayObject && useVBOs() && (LLRender::sGLCoreProfile || sUseVAO)) + { +#if GL_ARB_vertex_array_object + glGenVertexArrays(1, &mGLArray); +#endif + setupVertexArray(); + } } sAllocatedBytes += getSize() + getIndicesSize(); } +void LLVertexBuffer::setupVertexArray() +{ + if (!mGLArray) + { + return; + } + +#if GL_ARB_vertex_array_object + glBindVertexArray(mGLArray); +#endif + sGLRenderArray = mGLArray; + + U32 attrib_size[] = + { + 3, //TYPE_VERTEX, + 3, //TYPE_NORMAL, + 2, //TYPE_TEXCOORD0, + 2, //TYPE_TEXCOORD1, + 2, //TYPE_TEXCOORD2, + 2, //TYPE_TEXCOORD3, + 4, //TYPE_COLOR, + 4, //TYPE_EMISSIVE, + 3, //TYPE_BINORMAL, + 1, //TYPE_WEIGHT, + 4, //TYPE_WEIGHT4, + 4, //TYPE_CLOTHWEIGHT, + 1, //TYPE_TEXTURE_INDEX + }; + + U32 attrib_type[] = + { + GL_FLOAT, //TYPE_VERTEX, + GL_FLOAT, //TYPE_NORMAL, + GL_FLOAT, //TYPE_TEXCOORD0, + GL_FLOAT, //TYPE_TEXCOORD1, + GL_FLOAT, //TYPE_TEXCOORD2, + GL_FLOAT, //TYPE_TEXCOORD3, + GL_UNSIGNED_BYTE, //TYPE_COLOR, + GL_UNSIGNED_BYTE, //TYPE_EMISSIVE, + GL_FLOAT, //TYPE_BINORMAL, + GL_FLOAT, //TYPE_WEIGHT, + GL_FLOAT, //TYPE_WEIGHT4, + GL_FLOAT, //TYPE_CLOTHWEIGHT, + GL_FLOAT, //TYPE_TEXTURE_INDEX + }; + + U32 attrib_normalized[] = + { + GL_FALSE, //TYPE_VERTEX, + GL_FALSE, //TYPE_NORMAL, + GL_FALSE, //TYPE_TEXCOORD0, + GL_FALSE, //TYPE_TEXCOORD1, + GL_FALSE, //TYPE_TEXCOORD2, + GL_FALSE, //TYPE_TEXCOORD3, + GL_TRUE, //TYPE_COLOR, + GL_TRUE, //TYPE_EMISSIVE, + GL_FALSE, //TYPE_BINORMAL, + GL_FALSE, //TYPE_WEIGHT, + GL_FALSE, //TYPE_WEIGHT4, + GL_FALSE, //TYPE_CLOTHWEIGHT, + GL_FALSE, //TYPE_TEXTURE_INDEX + }; + + bindGLBuffer(true); + bindGLIndices(true); + + for (U32 i = 0; i < TYPE_MAX; ++i) + { + if (mTypeMask & (1 << i)) + { + glEnableVertexAttribArrayARB(i); + glVertexAttribPointerARB(i, attrib_size[i], attrib_type[i], attrib_normalized[i], sTypeSize[i], (void*) mOffsets[i]); + } + else + { + glDisableVertexAttribArrayARB(i); + } + } + + unbind(); +} + void LLVertexBuffer::resizeBuffer(S32 newnverts, S32 newnindices) { llassert(newnverts >= 0); @@ -986,8 +1221,13 @@ void LLVertexBuffer::resizeBuffer(S32 newnverts, S32 newnindices) if (mResized && useVBOs()) { - freeClientBuffer() ; - setBuffer(0); + freeClientBuffer(); + flush(); + + if (mGLArray) + { //if size changed, offsets changed + setupVertexArray(); + } } } @@ -1052,6 +1292,7 @@ bool expand_region(LLVertexBuffer::MappedRegion& region, S32 index, S32 count) // Map for data access U8* LLVertexBuffer::mapVertexBuffer(S32 type, S32 index, S32 count, bool map_range) { + bindGLBuffer(true); LLMemType mt2(LLMemType::MTYPE_VERTEX_MAP_BUFFER); if (mFinal) { @@ -1102,7 +1343,6 @@ U8* LLVertexBuffer::mapVertexBuffer(S32 type, S32 index, S32 count, bool map_ran if (!mVertexLocked) { LLMemType mt_v(LLMemType::MTYPE_VERTEX_MAP_BUFFER_VERTICES); - setBuffer(0, type); mVertexLocked = TRUE; sMappedCount++; stop_glerror(); @@ -1132,6 +1372,18 @@ U8* LLVertexBuffer::mapVertexBuffer(S32 type, S32 index, S32 count, bool map_ran else { #ifdef GL_ARB_map_buffer_range + + if (gDebugGL) + { + GLint size = 0; + glGetBufferParameterivARB(GL_ARRAY_BUFFER_ARB, GL_BUFFER_SIZE_ARB, &size); + + if (size < mSize) + { + llerrs << "Invalid buffer size." << llendl; + } + } + src = (U8*) glMapBufferRange(GL_ARRAY_BUFFER_ARB, 0, mSize, GL_MAP_WRITE_BIT | GL_MAP_FLUSH_EXPLICIT_BIT); @@ -1217,6 +1469,7 @@ U8* LLVertexBuffer::mapVertexBuffer(S32 type, S32 index, S32 count, bool map_ran U8* LLVertexBuffer::mapIndexBuffer(S32 index, S32 count, bool map_range) { LLMemType mt2(LLMemType::MTYPE_VERTEX_MAP_BUFFER); + bindGLIndices(true); if (mFinal) { llerrs << "LLVertexBuffer::mapIndexBuffer() called on a finalized buffer." << llendl; @@ -1264,11 +1517,21 @@ U8* LLVertexBuffer::mapIndexBuffer(S32 index, S32 count, bool map_range) { LLMemType mt_v(LLMemType::MTYPE_VERTEX_MAP_BUFFER_INDICES); - setBuffer(0, TYPE_INDEX); mIndexLocked = TRUE; sMappedCount++; stop_glerror(); + if (gDebugGL && useVBOs()) + { + GLint elem = 0; + glGetIntegerv(GL_ELEMENT_ARRAY_BUFFER_BINDING_ARB, &elem); + + if (elem != mGLIndices) + { + llerrs << "Wrong index buffer bound!" << llendl; + } + } + if(sDisableVBOMapping) { map_range = false; @@ -1365,19 +1628,20 @@ U8* LLVertexBuffer::mapIndexBuffer(S32 index, S32 count, bool map_range) } } -void LLVertexBuffer::unmapBuffer(S32 type) +void LLVertexBuffer::unmapBuffer() { LLMemType mt2(LLMemType::MTYPE_VERTEX_UNMAP_BUFFER); - if (!useVBOs() || type == -2) + if (!useVBOs()) { return ; //nothing to unmap } bool updated_all = false ; - if (mMappedData && mVertexLocked && type != TYPE_INDEX) + if (mMappedData && mVertexLocked) { - updated_all = (mIndexLocked && type < 0) ; //both vertex and index buffers done updating + bindGLBuffer(true); + updated_all = mIndexLocked; //both vertex and index buffers done updating if(sDisableVBOMapping) { @@ -1441,8 +1705,9 @@ void LLVertexBuffer::unmapBuffer(S32 type) sMappedCount--; } - if (mMappedIndexData && mIndexLocked && (type < 0 || type == TYPE_INDEX)) + if (mMappedIndexData && mIndexLocked) { + bindGLIndices(); if(sDisableVBOMapping) { if (!mMappedIndexRegions.empty()) @@ -1576,6 +1841,10 @@ bool LLVertexBuffer::getVertexStrider(LLStrider<LLVector3>& strider, S32 index, { return VertexBufferStrider<LLVector3,TYPE_VERTEX>::get(*this, strider, index, count, map_range); } +bool LLVertexBuffer::getVertexStrider(LLStrider<LLVector4a>& strider, S32 index, S32 count, bool map_range) +{ + return VertexBufferStrider<LLVector4a,TYPE_VERTEX>::get(*this, strider, index, count, map_range); +} bool LLVertexBuffer::getIndexStrider(LLStrider<U16>& strider, S32 index, S32 count, bool map_range) { return VertexBufferStrider<U16,TYPE_INDEX>::get(*this, strider, index, count, map_range); @@ -1601,6 +1870,10 @@ bool LLVertexBuffer::getColorStrider(LLStrider<LLColor4U>& strider, S32 index, S { return VertexBufferStrider<LLColor4U,TYPE_COLOR>::get(*this, strider, index, count, map_range); } +bool LLVertexBuffer::getEmissiveStrider(LLStrider<LLColor4U>& strider, S32 index, S32 count, bool map_range) +{ + return VertexBufferStrider<LLColor4U,TYPE_EMISSIVE>::get(*this, strider, index, count, map_range); +} bool LLVertexBuffer::getWeightStrider(LLStrider<F32>& strider, S32 index, S32 count, bool map_range) { return VertexBufferStrider<F32,TYPE_WEIGHT>::get(*this, strider, index, count, map_range); @@ -1618,43 +1891,162 @@ bool LLVertexBuffer::getClothWeightStrider(LLStrider<LLVector4>& strider, S32 in //---------------------------------------------------------------------------- +bool LLVertexBuffer::bindGLArray() +{ + if (mGLArray && sGLRenderArray != mGLArray) + { +#if GL_ARB_vertex_array_object + glBindVertexArray(mGLArray); +#endif + sGLRenderArray = mGLArray; + + //really shouldn't be necessary, but some drivers don't properly restore the + //state of GL_ELEMENT_ARRAY_BUFFER_BINDING + bindGLIndices(); + + return true; + } + + return false; +} + +bool LLVertexBuffer::bindGLBuffer(bool force_bind) +{ + bindGLArray(); + + bool ret = false; + + if (useVBOs() && (force_bind || (mGLBuffer && (mGLBuffer != sGLRenderBuffer || !sVBOActive)))) + { + /*if (sMapped) + { + llerrs << "VBO bound while another VBO mapped!" << llendl; + }*/ + glBindBufferARB(GL_ARRAY_BUFFER_ARB, mGLBuffer); + sGLRenderBuffer = mGLBuffer; + sBindCount++; + sVBOActive = TRUE; + + if (mGLArray) + { + llassert(sGLRenderArray == mGLArray); + //mCachedRenderBuffer = mGLBuffer; + } + + ret = true; + } + + return ret; +} + +bool LLVertexBuffer::bindGLIndices(bool force_bind) +{ + bindGLArray(); + + bool ret = false; + if (useVBOs() && (force_bind || (mGLIndices && (mGLIndices != sGLRenderIndices || !sIBOActive)))) + { + /*if (sMapped) + { + llerrs << "VBO bound while another VBO mapped!" << llendl; + }*/ + glBindBufferARB(GL_ELEMENT_ARRAY_BUFFER_ARB, mGLIndices); + sGLRenderIndices = mGLIndices; + stop_glerror(); + sBindCount++; + sIBOActive = TRUE; + ret = true; + } + + return ret; +} + +void LLVertexBuffer::flush() +{ + if (useVBOs()) + { + if (mResized) + { + if (mGLBuffer) + { + stop_glerror(); + bindGLBuffer(true); + glBufferDataARB(GL_ARRAY_BUFFER_ARB, getSize(), NULL, mUsage); + stop_glerror(); + } + if (mGLIndices) + { + stop_glerror(); + bindGLIndices(true); + glBufferDataARB(GL_ELEMENT_ARRAY_BUFFER_ARB, getIndicesSize(), NULL, mUsage); + stop_glerror(); + } + + mEmpty = TRUE; + mResized = FALSE; + } + + unmapBuffer(); + } +} + // Set for rendering -void LLVertexBuffer::setBuffer(U32 data_mask, S32 type) +void LLVertexBuffer::setBuffer(U32 data_mask) { + flush(); + LLMemType mt2(LLMemType::MTYPE_VERTEX_SET_BUFFER); //set up pointers if the data mask is different ... BOOL setup = (sLastMask != data_mask); + if (gDebugGL && data_mask != 0) + { //make sure data requirements are fulfilled + LLGLSLShader* shader = LLGLSLShader::sCurBoundShaderPtr; + if (shader) + { + U32 required_mask = 0; + for (U32 i = 0; i < LLVertexBuffer::TYPE_TEXTURE_INDEX; ++i) + { + if (shader->getAttribLocation(i) > -1) + { + U32 required = 1 << i; + if ((data_mask & required) == 0) + { + llwarns << "Missing attribute: " << LLShaderMgr::instance()->mReservedAttribs[i] << llendl; + } + + required_mask |= required; + } + } + + if ((data_mask & required_mask) != required_mask) + { + llerrs << "Shader consumption mismatches data provision." << llendl; + } + } + } + if (useVBOs()) { - if (mGLBuffer && (mGLBuffer != sGLRenderBuffer || !sVBOActive)) + if (mGLArray) { - /*if (sMapped) - { - llerrs << "VBO bound while another VBO mapped!" << llendl; - }*/ - stop_glerror(); - glBindBufferARB(GL_ARRAY_BUFFER_ARB, mGLBuffer); - stop_glerror(); - sBindCount++; - sVBOActive = TRUE; - setup = TRUE; // ... or the bound buffer changed + bindGLArray(); + setup = FALSE; //do NOT perform pointer setup if using VAO } - if (mGLIndices && (mGLIndices != sGLRenderIndices || !sIBOActive)) + else { - /*if (sMapped) + if (bindGLBuffer()) { - llerrs << "VBO bound while another VBO mapped!" << llendl; - }*/ - stop_glerror(); - glBindBufferARB(GL_ELEMENT_ARRAY_BUFFER_ARB, mGLIndices); - stop_glerror(); - sBindCount++; - sIBOActive = TRUE; + setup = TRUE; + } + if (bindGLIndices()) + { + setup = TRUE; + } } - + BOOL error = FALSE; - if (gDebugGL) + if (gDebugGL && !mGLArray) { GLint buff; glGetIntegerv(GL_ARRAY_BUFFER_BINDING_ARB, &buff); @@ -1689,81 +2081,20 @@ void LLVertexBuffer::setBuffer(U32 data_mask, S32 type) } } - if (mResized) + + } + else + { + if (sGLRenderArray) { - if (gDebugGL) - { - GLint buff; - glGetIntegerv(GL_ARRAY_BUFFER_BINDING_ARB, &buff); - if ((GLuint)buff != mGLBuffer) - { - if (gDebugSession) - { - error = TRUE; - gFailLog << "Invalid GL vertex buffer bound: " << std::endl; - } - else - { - llerrs << "Invalid GL vertex buffer bound: " << buff << llendl; - } - } - - if (mGLIndices != 0) - { - glGetIntegerv(GL_ELEMENT_ARRAY_BUFFER_BINDING_ARB, &buff); - if ((GLuint)buff != mGLIndices) - { - if (gDebugSession) - { - error = TRUE; - gFailLog << "Invalid GL index buffer bound: "<< std::endl; - } - else - { - llerrs << "Invalid GL index buffer bound: " << buff << llendl; - } - } - } - } - - if (mGLBuffer) - { - stop_glerror(); - glBufferDataARB(GL_ARRAY_BUFFER_ARB, getSize(), NULL, mUsage); - stop_glerror(); - } - if (mGLIndices) - { - stop_glerror(); - glBufferDataARB(GL_ELEMENT_ARRAY_BUFFER_ARB, getIndicesSize(), NULL, mUsage); - stop_glerror(); - } - - mEmpty = TRUE; - mResized = FALSE; - - if (data_mask != 0) - { - if (gDebugSession) - { - error = TRUE; - gFailLog << "Buffer set for rendering before being filled after resize." << std::endl; - } - else - { - llerrs << "Buffer set for rendering before being filled after resize." << llendl; - } - } +#if GL_ARB_vertex_array_object + glBindVertexArray(0); +#endif + sGLRenderArray = 0; + sGLRenderIndices = 0; + sIBOActive = FALSE; } - if (error) - { - ll_fail("LLVertexBuffer::mapBuffer failed"); - } - unmapBuffer(type); - } - else - { if (mGLBuffer) { if (sVBOActive) @@ -1775,30 +2106,30 @@ void LLVertexBuffer::setBuffer(U32 data_mask, S32 type) } if (sGLRenderBuffer != mGLBuffer) { + sGLRenderBuffer = mGLBuffer; setup = TRUE; // ... or a client memory pointer changed } } - if (mGLIndices && sIBOActive) + if (mGLIndices) { - /*if (sMapped) + if (sIBOActive) { - llerrs << "VBO unbound while potentially mapped!" << llendl; - }*/ - glBindBufferARB(GL_ELEMENT_ARRAY_BUFFER_ARB, 0); - sBindCount++; - sIBOActive = FALSE; + glBindBufferARB(GL_ELEMENT_ARRAY_BUFFER_ARB, 0); + sBindCount++; + sIBOActive = FALSE; + } + + sGLRenderIndices = mGLIndices; } } - setupClientArrays(data_mask); - - if (mGLIndices) + if (!mGLArray) { - sGLRenderIndices = mGLIndices; + setupClientArrays(data_mask); } + if (mGLBuffer) { - sGLRenderBuffer = mGLBuffer; if (data_mask && setup) { setupVertexBuffer(data_mask); // subclass specific setup (virtual function) @@ -1808,80 +2139,150 @@ void LLVertexBuffer::setBuffer(U32 data_mask, S32 type) } // virtual (default) -void LLVertexBuffer::setupVertexBuffer(U32 data_mask) const +void LLVertexBuffer::setupVertexBuffer(U32 data_mask) { LLMemType mt2(LLMemType::MTYPE_VERTEX_SETUP_VERTEX_BUFFER); stop_glerror(); U8* base = useVBOs() ? (U8*) mAlignedOffset : mMappedData; - if ((data_mask & mTypeMask) != data_mask) + /*if ((data_mask & mTypeMask) != data_mask) { llerrs << "LLVertexBuffer::setupVertexBuffer missing required components for supplied data mask." << llendl; - } - - if (data_mask & MAP_NORMAL) - { - glNormalPointer(GL_FLOAT, LLVertexBuffer::sTypeSize[TYPE_NORMAL], (void*)(base + mOffsets[TYPE_NORMAL])); - } - if (data_mask & MAP_TEXCOORD3) - { - glClientActiveTextureARB(GL_TEXTURE3_ARB); - glTexCoordPointer(2,GL_FLOAT, LLVertexBuffer::sTypeSize[TYPE_TEXCOORD3], (void*)(base + mOffsets[TYPE_TEXCOORD3])); - glClientActiveTextureARB(GL_TEXTURE0_ARB); - } - if (data_mask & MAP_TEXCOORD2) - { - glClientActiveTextureARB(GL_TEXTURE2_ARB); - glTexCoordPointer(2,GL_FLOAT, LLVertexBuffer::sTypeSize[TYPE_TEXCOORD2], (void*)(base + mOffsets[TYPE_TEXCOORD2])); - glClientActiveTextureARB(GL_TEXTURE0_ARB); - } - if (data_mask & MAP_TEXCOORD1) - { - glClientActiveTextureARB(GL_TEXTURE1_ARB); - glTexCoordPointer(2,GL_FLOAT, LLVertexBuffer::sTypeSize[TYPE_TEXCOORD1], (void*)(base + mOffsets[TYPE_TEXCOORD1])); - glClientActiveTextureARB(GL_TEXTURE0_ARB); - } - if (data_mask & MAP_BINORMAL) - { - glClientActiveTextureARB(GL_TEXTURE2_ARB); - glTexCoordPointer(3,GL_FLOAT, LLVertexBuffer::sTypeSize[TYPE_BINORMAL], (void*)(base + mOffsets[TYPE_BINORMAL])); - glClientActiveTextureARB(GL_TEXTURE0_ARB); - } - if (data_mask & MAP_TEXCOORD0) - { - glTexCoordPointer(2,GL_FLOAT, LLVertexBuffer::sTypeSize[TYPE_TEXCOORD0], (void*)(base + mOffsets[TYPE_TEXCOORD0])); - } - if (data_mask & MAP_COLOR) - { - glColorPointer(4, GL_UNSIGNED_BYTE, LLVertexBuffer::sTypeSize[TYPE_COLOR], (void*)(base + mOffsets[TYPE_COLOR])); - } - - if (data_mask & MAP_WEIGHT) - { - glVertexAttribPointerARB(1, 1, GL_FLOAT, FALSE, LLVertexBuffer::sTypeSize[TYPE_WEIGHT], (void*)(base + mOffsets[TYPE_WEIGHT])); - } - - if (data_mask & MAP_WEIGHT4 && sWeight4Loc != -1) - { - glVertexAttribPointerARB(sWeight4Loc, 4, GL_FLOAT, FALSE, LLVertexBuffer::sTypeSize[TYPE_WEIGHT4], (void*)(base+mOffsets[TYPE_WEIGHT4])); - } + }*/ - if (data_mask & MAP_CLOTHWEIGHT) - { - glVertexAttribPointerARB(4, 4, GL_FLOAT, TRUE, LLVertexBuffer::sTypeSize[TYPE_CLOTHWEIGHT], (void*)(base + mOffsets[TYPE_CLOTHWEIGHT])); - } - if (data_mask & MAP_VERTEX) + if (LLGLSLShader::sNoFixedFunction) { + if (data_mask & MAP_NORMAL) + { + S32 loc = TYPE_NORMAL; + void* ptr = (void*)(base + mOffsets[TYPE_NORMAL]); + glVertexAttribPointerARB(loc, 3, GL_FLOAT, GL_FALSE, LLVertexBuffer::sTypeSize[TYPE_NORMAL], ptr); + } + if (data_mask & MAP_TEXCOORD3) + { + S32 loc = TYPE_TEXCOORD3; + void* ptr = (void*)(base + mOffsets[TYPE_TEXCOORD3]); + glVertexAttribPointerARB(loc,2,GL_FLOAT, GL_FALSE, LLVertexBuffer::sTypeSize[TYPE_TEXCOORD3], ptr); + } + if (data_mask & MAP_TEXCOORD2) + { + S32 loc = TYPE_TEXCOORD2; + void* ptr = (void*)(base + mOffsets[TYPE_TEXCOORD2]); + glVertexAttribPointerARB(loc,2,GL_FLOAT, GL_FALSE, LLVertexBuffer::sTypeSize[TYPE_TEXCOORD2], ptr); + } + if (data_mask & MAP_TEXCOORD1) + { + S32 loc = TYPE_TEXCOORD1; + void* ptr = (void*)(base + mOffsets[TYPE_TEXCOORD1]); + glVertexAttribPointerARB(loc,2,GL_FLOAT, GL_FALSE, LLVertexBuffer::sTypeSize[TYPE_TEXCOORD1], ptr); + } + if (data_mask & MAP_BINORMAL) + { + S32 loc = TYPE_BINORMAL; + void* ptr = (void*)(base + mOffsets[TYPE_BINORMAL]); + glVertexAttribPointerARB(loc, 3,GL_FLOAT, GL_FALSE, LLVertexBuffer::sTypeSize[TYPE_BINORMAL], ptr); + } + if (data_mask & MAP_TEXCOORD0) + { + S32 loc = TYPE_TEXCOORD0; + void* ptr = (void*)(base + mOffsets[TYPE_TEXCOORD0]); + glVertexAttribPointerARB(loc,2,GL_FLOAT, GL_FALSE, LLVertexBuffer::sTypeSize[TYPE_TEXCOORD0], ptr); + } + if (data_mask & MAP_COLOR) + { + S32 loc = TYPE_COLOR; + void* ptr = (void*)(base + mOffsets[TYPE_COLOR]); + glVertexAttribPointerARB(loc, 4, GL_UNSIGNED_BYTE, GL_TRUE, LLVertexBuffer::sTypeSize[TYPE_COLOR], ptr); + } + if (data_mask & MAP_EMISSIVE) + { + S32 loc = TYPE_EMISSIVE; + void* ptr = (void*)(base + mOffsets[TYPE_EMISSIVE]); + glVertexAttribPointerARB(loc, 4, GL_UNSIGNED_BYTE, GL_TRUE, LLVertexBuffer::sTypeSize[TYPE_EMISSIVE], ptr); + } + if (data_mask & MAP_WEIGHT) + { + S32 loc = TYPE_WEIGHT; + void* ptr = (void*)(base + mOffsets[TYPE_WEIGHT]); + glVertexAttribPointerARB(loc, 1, GL_FLOAT, FALSE, LLVertexBuffer::sTypeSize[TYPE_WEIGHT], ptr); + } + if (data_mask & MAP_WEIGHT4) + { + S32 loc = TYPE_WEIGHT4; + void* ptr = (void*)(base+mOffsets[TYPE_WEIGHT4]); + glVertexAttribPointerARB(loc, 4, GL_FLOAT, FALSE, LLVertexBuffer::sTypeSize[TYPE_WEIGHT4], ptr); + } + if (data_mask & MAP_CLOTHWEIGHT) + { + S32 loc = TYPE_CLOTHWEIGHT; + void* ptr = (void*)(base + mOffsets[TYPE_CLOTHWEIGHT]); + glVertexAttribPointerARB(loc, 4, GL_FLOAT, TRUE, LLVertexBuffer::sTypeSize[TYPE_CLOTHWEIGHT], ptr); + } if (data_mask & MAP_TEXTURE_INDEX) { - glVertexPointer(4,GL_FLOAT, LLVertexBuffer::sTypeSize[TYPE_VERTEX], (void*)(base + 0)); + S32 loc = TYPE_TEXTURE_INDEX; + void *ptr = (void*) (base + mOffsets[TYPE_VERTEX] + 12); + glVertexAttribPointerARB(loc, 1, GL_FLOAT, GL_FALSE, LLVertexBuffer::sTypeSize[TYPE_VERTEX], ptr); } - else + if (data_mask & MAP_VERTEX) { - glVertexPointer(3,GL_FLOAT, LLVertexBuffer::sTypeSize[TYPE_VERTEX], (void*)(base + 0)); + S32 loc = TYPE_VERTEX; + void* ptr = (void*)(base + mOffsets[TYPE_VERTEX]); + glVertexAttribPointerARB(loc, 3,GL_FLOAT, GL_FALSE, LLVertexBuffer::sTypeSize[TYPE_VERTEX], ptr); + } + } + else + { + if (data_mask & MAP_NORMAL) + { + glNormalPointer(GL_FLOAT, LLVertexBuffer::sTypeSize[TYPE_NORMAL], (void*)(base + mOffsets[TYPE_NORMAL])); + } + if (data_mask & MAP_TEXCOORD3) + { + glClientActiveTextureARB(GL_TEXTURE3_ARB); + glTexCoordPointer(2,GL_FLOAT, LLVertexBuffer::sTypeSize[TYPE_TEXCOORD3], (void*)(base + mOffsets[TYPE_TEXCOORD3])); + glClientActiveTextureARB(GL_TEXTURE0_ARB); + } + if (data_mask & MAP_TEXCOORD2) + { + glClientActiveTextureARB(GL_TEXTURE2_ARB); + glTexCoordPointer(2,GL_FLOAT, LLVertexBuffer::sTypeSize[TYPE_TEXCOORD2], (void*)(base + mOffsets[TYPE_TEXCOORD2])); + glClientActiveTextureARB(GL_TEXTURE0_ARB); + } + if (data_mask & MAP_TEXCOORD1) + { + glClientActiveTextureARB(GL_TEXTURE1_ARB); + glTexCoordPointer(2,GL_FLOAT, LLVertexBuffer::sTypeSize[TYPE_TEXCOORD1], (void*)(base + mOffsets[TYPE_TEXCOORD1])); + glClientActiveTextureARB(GL_TEXTURE0_ARB); + } + if (data_mask & MAP_BINORMAL) + { + glClientActiveTextureARB(GL_TEXTURE2_ARB); + glTexCoordPointer(3,GL_FLOAT, LLVertexBuffer::sTypeSize[TYPE_BINORMAL], (void*)(base + mOffsets[TYPE_BINORMAL])); + glClientActiveTextureARB(GL_TEXTURE0_ARB); + } + if (data_mask & MAP_TEXCOORD0) + { + glTexCoordPointer(2,GL_FLOAT, LLVertexBuffer::sTypeSize[TYPE_TEXCOORD0], (void*)(base + mOffsets[TYPE_TEXCOORD0])); } + if (data_mask & MAP_COLOR) + { + glColorPointer(4, GL_UNSIGNED_BYTE, LLVertexBuffer::sTypeSize[TYPE_COLOR], (void*)(base + mOffsets[TYPE_COLOR])); + } + if (data_mask & MAP_VERTEX) + { + glVertexPointer(3,GL_FLOAT, LLVertexBuffer::sTypeSize[TYPE_VERTEX], (void*)(base + 0)); + } } llglassertok(); } +LLVertexBuffer::MappedRegion::MappedRegion(S32 type, S32 index, S32 count) +: mType(type), mIndex(index), mCount(count) +{ + llassert(mType == LLVertexBuffer::TYPE_INDEX || + mType < LLVertexBuffer::TYPE_TEXTURE_INDEX); +} + + diff --git a/indra/llrender/llvertexbuffer.h b/indra/llrender/llvertexbuffer.h index 578cec3885..b50c409c49 100644 --- a/indra/llrender/llvertexbuffer.h +++ b/indra/llrender/llvertexbuffer.h @@ -38,6 +38,8 @@ #include <vector> #include <list> +#define LL_MAX_VERTEX_ATTRIB_LOCATION 64 + //============================================================================ // NOTES // Threading: @@ -49,7 +51,6 @@ //============================================================================ // gl name pools for dynamic and streaming buffers - class LLVBOPool : public LLGLNamePool { protected: @@ -90,9 +91,7 @@ public: S32 mIndex; S32 mCount; - MappedRegion(S32 type, S32 index, S32 count) - : mType(type), mIndex(index), mCount(count) - { } + MappedRegion(S32 type, S32 index, S32 count); }; LLVertexBuffer(const LLVertexBuffer& rhs) @@ -111,15 +110,15 @@ public: static LLVBOPool sStreamIBOPool; static LLVBOPool sDynamicIBOPool; - static S32 sWeight4Loc; - static BOOL sUseStreamDraw; + static BOOL sUseVAO; static BOOL sPreferStreamDraw; static void initClass(bool use_vbo, bool no_vbo_mapping); static void cleanupClass(); static void setupClientArrays(U32 data_mask); static void drawArrays(U32 mode, const std::vector<LLVector3>& pos, const std::vector<LLVector3>& norm); + static void drawElements(U32 mode, const LLVector4a* pos, const LLVector2* tc, S32 num_indices, const U16* indicesp); static void clientCopy(F64 max_time = 0.005); //copy data from client to GL static void unbind(); //unbind any bound vertex buffer @@ -133,24 +132,29 @@ public: static S32 calcOffsets(const U32& typemask, S32* offsets, S32 num_vertices); + //WARNING -- when updating these enums you MUST + // 1 - update LLVertexBuffer::sTypeSize + // 2 - add a strider accessor + // 3 - modify LLVertexBuffer::setupVertexBuffer + // 4 - modify LLVertexBuffer::setupClientArray + // 5 - modify LLViewerShaderMgr::mReservedAttribs + // 6 - update LLVertexBuffer::setupVertexArray enum { - TYPE_VERTEX, + TYPE_VERTEX = 0, TYPE_NORMAL, TYPE_TEXCOORD0, TYPE_TEXCOORD1, TYPE_TEXCOORD2, TYPE_TEXCOORD3, TYPE_COLOR, - // These use VertexAttribPointer and should possibly be made generic + TYPE_EMISSIVE, TYPE_BINORMAL, TYPE_WEIGHT, TYPE_WEIGHT4, TYPE_CLOTHWEIGHT, - TYPE_MAX, - TYPE_INDEX, - - //no actual additional data, but indicates position.w is texture index TYPE_TEXTURE_INDEX, + TYPE_MAX, + TYPE_INDEX, }; enum { MAP_VERTEX = (1<<TYPE_VERTEX), @@ -160,6 +164,7 @@ public: MAP_TEXCOORD2 = (1<<TYPE_TEXCOORD2), MAP_TEXCOORD3 = (1<<TYPE_TEXCOORD3), MAP_COLOR = (1<<TYPE_COLOR), + MAP_EMISSIVE = (1<<TYPE_EMISSIVE), // These use VertexAttribPointer and should possibly be made generic MAP_BINORMAL = (1<<TYPE_BINORMAL), MAP_WEIGHT = (1<<TYPE_WEIGHT), @@ -173,10 +178,14 @@ protected: virtual ~LLVertexBuffer(); // use unref() - virtual void setupVertexBuffer(U32 data_mask) const; // pure virtual, called from mapBuffer() + virtual void setupVertexBuffer(U32 data_mask); // pure virtual, called from mapBuffer() + void setupVertexArray(); void genBuffer(); void genIndices(); + bool bindGLBuffer(bool force_bind = false); + bool bindGLIndices(bool force_bind = false); + bool bindGLArray(); void releaseBuffer(); void releaseIndices(); void createGLBuffer(); @@ -186,7 +195,7 @@ protected: void updateNumVerts(S32 nverts); void updateNumIndices(S32 nindices); virtual BOOL useVBOs() const; - void unmapBuffer(S32 type); + void unmapBuffer(); void freeClientBuffer() ; void allocateClientVertexBuffer() ; void allocateClientIndexBuffer() ; @@ -199,7 +208,8 @@ public: 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 + virtual void setBuffer(U32 data_mask); // calls setupVertexBuffer() if data_mask is not 0 + void flush(); //flush pending data to GL memory // allocate buffer void allocateBuffer(S32 nverts, S32 nindices, bool create); virtual void resizeBuffer(S32 newnverts, S32 newnindices); @@ -212,16 +222,19 @@ public: // setVertsNorms(verts, norms); // vb->unmapBuffer(); bool getVertexStrider(LLStrider<LLVector3>& strider, S32 index=0, S32 count = -1, bool map_range = false); + bool getVertexStrider(LLStrider<LLVector4a>& 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 getEmissiveStrider(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; } S32 getNumVerts() const { return mNumVerts; } @@ -262,6 +275,8 @@ protected: S32 mUsage; // GL usage U32 mGLBuffer; // GL VBO handle U32 mGLIndices; // GL IBO handle + U32 mGLArray; // GL VAO handle + U8* mMappedData; // pointer to currently mapped data (NULL if unmapped) U8* mMappedIndexData; // pointer to currently mapped indices (NULL if unmapped) BOOL mVertexLocked; // if TRUE, vertex buffer is being or has been written to in client memory @@ -298,6 +313,7 @@ public: static S32 sTypeSize[TYPE_MAX]; static U32 sGLMode[LLRender::NUM_MODES]; static U32 sGLRenderBuffer; + static U32 sGLRenderArray; static U32 sGLRenderIndices; static BOOL sVBOActive; static BOOL sIBOActive; |