diff options
Diffstat (limited to 'indra/llrender')
25 files changed, 1474 insertions, 798 deletions
diff --git a/indra/llrender/CMakeLists.txt b/indra/llrender/CMakeLists.txt index 5c13df9f81..516af93316 100644 --- a/indra/llrender/CMakeLists.txt +++ b/indra/llrender/CMakeLists.txt @@ -36,6 +36,7 @@ set(llrender_SOURCE_FILES llglslshader.cpp llimagegl.cpp llpostprocess.cpp + llrendernavprim.cpp llrendersphere.cpp llshadermgr.cpp lltexture.cpp @@ -59,6 +60,7 @@ set(llrender_HEADER_FILES llimagegl.h llpostprocess.h llrender.h + llrendernavprim.h llrendersphere.h llshadermgr.h lltexture.h diff --git a/indra/llrender/llcubemap.cpp b/indra/llrender/llcubemap.cpp index 45a3b18179..362452d837 100644 --- a/indra/llrender/llcubemap.cpp +++ b/indra/llrender/llcubemap.cpp @@ -81,7 +81,7 @@ void LLCubeMap::initGL() { U32 texname = 0; - LLImageGL::generateTextures(1, &texname); + LLImageGL::generateTextures(LLTexUnit::TT_CUBE_MAP, GL_RGB8, 1, &texname); for (int i = 0; i < 6; i++) { diff --git a/indra/llrender/llfontfreetype.cpp b/indra/llrender/llfontfreetype.cpp index 91c8a37022..66d4ad2d87 100644 --- a/indra/llrender/llfontfreetype.cpp +++ b/indra/llrender/llfontfreetype.cpp @@ -55,7 +55,10 @@ FT_Library gFTLibrary = NULL; //static void LLFontManager::initClass() { - gFontManagerp = new LLFontManager; + if (!gFontManagerp) + { + gFontManagerp = new LLFontManager; + } } //static @@ -136,7 +139,7 @@ BOOL LLFontFreetype::loadFace(const std::string& filename, F32 point_size, F32 v FT_Done_Face(mFTFace); mFTFace = NULL; } - + int error; error = FT_New_Face( gFTLibrary, diff --git a/indra/llrender/llfontgl.cpp b/indra/llrender/llfontgl.cpp index 607473d416..8772779645 100644 --- a/indra/llrender/llfontgl.cpp +++ b/indra/llrender/llfontgl.cpp @@ -56,8 +56,9 @@ std::string LLFontGL::sAppDir; LLColor4 LLFontGL::sShadowColor(0.f, 0.f, 0.f, 1.f); LLFontRegistry* LLFontGL::sFontRegistry = NULL; -LLCoordFont LLFontGL::sCurOrigin; -std::vector<LLCoordFont> LLFontGL::sOriginStack; +LLCoordGL LLFontGL::sCurOrigin; +F32 LLFontGL::sCurDepth; +std::vector<std::pair<LLCoordGL, F32> > LLFontGL::sOriginStack; const F32 EXT_X_BEARING = 1.f; const F32 EXT_Y_BEARING = 0.f; @@ -68,20 +69,6 @@ const F32 PIXEL_CORRECTION_DISTANCE = 0.01f; const F32 PAD_UVY = 0.5f; // half of vertical padding between glyphs in the glyph texture const F32 DROP_SHADOW_SOFT_STRENGTH = 0.3f; -static F32 llfont_round_x(F32 x) -{ - //return llfloor((x-LLFontGL::sCurOrigin.mX)/LLFontGL::sScaleX+0.5f)*LLFontGL::sScaleX+LLFontGL::sCurOrigin.mX; - //return llfloor(x/LLFontGL::sScaleX+0.5f)*LLFontGL::sScaleY; - return x; -} - -static F32 llfont_round_y(F32 y) -{ - //return llfloor((y-LLFontGL::sCurOrigin.mY)/LLFontGL::sScaleY+0.5f)*LLFontGL::sScaleY+LLFontGL::sCurOrigin.mY; - //return llfloor(y+0.5f); - return y; -} - LLFontGL::LLFontGL() { } @@ -115,23 +102,23 @@ static LLFastTimer::DeclareTimer FTM_RENDER_FONTS("Fonts"); S32 LLFontGL::render(const LLWString &wstr, S32 begin_offset, const LLRect& rect, const LLColor4 &color, HAlign halign, VAlign valign, U8 style, ShadowType shadow, S32 max_chars, F32* right_x, BOOL use_ellipses) const { - F32 x = rect.mLeft; + F32 x = (F32)rect.mLeft; F32 y = 0.f; switch(valign) { case TOP: - y = rect.mTop; + y = (F32)rect.mTop; break; case VCENTER: - y = rect.getCenterY(); + y = (F32)rect.getCenterY(); break; case BASELINE: case BOTTOM: - y = rect.mBottom; + y = (F32)rect.mBottom; break; default: - y = rect.mBottom; + y = (F32)rect.mBottom; break; } return render(wstr, begin_offset, x, y, color, halign, valign, style, shadow, max_chars, rect.getWidth(), right_x, use_ellipses); @@ -177,21 +164,11 @@ S32 LLFontGL::render(const LLWString &wstr, S32 begin_offset, F32 x, F32 y, cons gGL.loadUIIdentity(); - //gGL.translateUI(floorf(sCurOrigin.mX*sScaleX), floorf(sCurOrigin.mY*sScaleY), sCurOrigin.mZ); - - // this code snaps the text origin to a pixel grid to start with - //F32 pixel_offset_x = llround((F32)sCurOrigin.mX) - (sCurOrigin.mX); - //F32 pixel_offset_y = llround((F32)sCurOrigin.mY) - (sCurOrigin.mY); - //gGL.translateUI(-pixel_offset_x, -pixel_offset_y, 0.f); - LLVector2 origin(floorf(sCurOrigin.mX*sScaleX), floorf(sCurOrigin.mY*sScaleY)); - // snap the text origin to a pixel grid to start with - origin.mV[VX] -= llround((F32)sCurOrigin.mX) - (sCurOrigin.mX); - origin.mV[VY] -= llround((F32)sCurOrigin.mY) - (sCurOrigin.mY); - // Depth translation, so that floating text appears 'inworld' - // and is correclty occluded. - gGL.translatef(0.f,0.f,sCurOrigin.mZ); + // Depth translation, so that floating text appears 'in-world' + // and is correctly occluded. + gGL.translatef(0.f,0.f,sCurDepth); S32 chars_drawn = 0; S32 i; @@ -215,16 +192,17 @@ S32 LLFontGL::render(const LLWString &wstr, S32 begin_offset, F32 x, F32 y, cons cur_y = ((F32)y * sScaleY) + origin.mV[VY]; // Offset y by vertical alignment. + // use unscaled font metrics here switch (valign) { case TOP: - cur_y -= mFontFreetype->getAscenderHeight(); + cur_y -= llceil(mFontFreetype->getAscenderHeight()); break; case BOTTOM: - cur_y += mFontFreetype->getDescenderHeight(); + cur_y += llceil(mFontFreetype->getDescenderHeight()); break; case VCENTER: - cur_y -= (mFontFreetype->getAscenderHeight() - mFontFreetype->getDescenderHeight()) / 2.f; + cur_y -= llceil((llceil(mFontFreetype->getAscenderHeight()) - llceil(mFontFreetype->getDescenderHeight())) / 2.f); break; case BASELINE: // Baseline, do nothing. @@ -250,7 +228,7 @@ S32 LLFontGL::render(const LLWString &wstr, S32 begin_offset, F32 x, F32 y, cons cur_render_y = cur_y; cur_render_x = cur_x; - F32 start_x = llround(cur_x); + F32 start_x = (F32)llround(cur_x); const LLFontBitmapCache* font_bitmap_cache = mFontFreetype->getFontBitmapCache(); @@ -304,6 +282,18 @@ S32 LLFontGL::render(const LLWString &wstr, S32 begin_offset, F32 x, F32 y, cons S32 next_bitmap_num = fgi->mBitmapNum; if (next_bitmap_num != bitmap_num) { + // Actually draw the queued glyphs before switching their texture; + // otherwise the queued glyphs will be taken from wrong textures. + if (glyph_count > 0) + { + gGL.begin(LLRender::QUADS); + { + gGL.vertexBatchPreTransformed(vertices, uvs, colors, glyph_count * 4); + } + gGL.end(); + glyph_count = 0; + } + bitmap_num = next_bitmap_num; LLImageGL *font_image = font_bitmap_cache->getImageGL(bitmap_num); gGL.getTexUnit(0)->bind(font_image); @@ -322,10 +312,10 @@ S32 LLFontGL::render(const LLWString &wstr, S32 begin_offset, F32 x, F32 y, cons (fgi->mXBitmapOffset + fgi->mWidth) * inv_width, (fgi->mYBitmapOffset - PAD_UVY) * inv_height); // snap glyph origin to whole screen pixel - LLRectf screen_rect(llround(cur_render_x + (F32)fgi->mXBearing), - llround(cur_render_y + (F32)fgi->mYBearing), - llround(cur_render_x + (F32)fgi->mXBearing) + (F32)fgi->mWidth, - llround(cur_render_y + (F32)fgi->mYBearing) - (F32)fgi->mHeight); + LLRectf screen_rect((F32)llround(cur_render_x + (F32)fgi->mXBearing), + (F32)llround(cur_render_y + (F32)fgi->mYBearing), + (F32)llround(cur_render_x + (F32)fgi->mXBearing) + (F32)fgi->mWidth, + (F32)llround(cur_render_y + (F32)fgi->mYBearing) - (F32)fgi->mHeight); if (glyph_count >= GLYPH_BATCH_SIZE) { @@ -378,12 +368,12 @@ S32 LLFontGL::render(const LLWString &wstr, S32 begin_offset, F32 x, F32 y, cons //FIXME: add underline as glyph? if (style_to_add & UNDERLINE) { - F32 descender = mFontFreetype->getDescenderHeight(); + F32 descender = (F32)llfloor(mFontFreetype->getDescenderHeight()); gGL.getTexUnit(0)->unbind(LLTexUnit::TT_TEXTURE); gGL.begin(LLRender::LINES); - gGL.vertex2f(start_x, cur_y - (descender)); - gGL.vertex2f(cur_x, cur_y - (descender)); + gGL.vertex2f(start_x, cur_y - descender); + gGL.vertex2f(cur_x, cur_y - descender); gGL.end(); } @@ -432,19 +422,19 @@ S32 LLFontGL::renderUTF8(const std::string &text, S32 begin_offset, S32 x, S32 y } // font metrics - override for LLFontFreetype that returns units of virtual pixels -F32 LLFontGL::getLineHeight() const +F32 LLFontGL::getAscenderHeight() const { - return (F32)llround(mFontFreetype->getLineHeight() / sScaleY); + return mFontFreetype->getAscenderHeight() / sScaleY; } -F32 LLFontGL::getAscenderHeight() const +F32 LLFontGL::getDescenderHeight() const { - return (F32)llround(mFontFreetype->getAscenderHeight() / sScaleY); + return mFontFreetype->getDescenderHeight() / sScaleY; } -F32 LLFontGL::getDescenderHeight() const +S32 LLFontGL::getLineHeight() const { - return (F32)llround(mFontFreetype->getDescenderHeight() / sScaleY); + return llceil(mFontFreetype->getAscenderHeight() / sScaleY) + llceil(mFontFreetype->getDescenderHeight() / sScaleY); } S32 LLFontGL::getWidth(const std::string& utf8text) const @@ -609,6 +599,11 @@ S32 LLFontGL::maxDrawableChars(const llwchar* wchars, F32 max_pixels, S32 max_ch if(!fgi) { fgi = mFontFreetype->getGlyphInfo(wch); + + if (NULL == fgi) + { + return 0; + } } // account for glyphs that run beyond the starting point for the next glyphs @@ -633,7 +628,7 @@ S32 LLFontGL::maxDrawableChars(const llwchar* wchars, F32 max_pixels, S32 max_ch } // Round after kerning. - cur_x = llround(cur_x); + cur_x = (F32)llround(cur_x); drawn_x = cur_x; } @@ -704,7 +699,7 @@ S32 LLFontGL::firstDrawableChar(const llwchar* wchars, F32 max_pixels, S32 text_ } // Round after kerning. - total_width = llround(total_width); + total_width = (F32)llround(total_width); } if (drawable_chars == 0) @@ -787,7 +782,7 @@ S32 LLFontGL::charFromPixelOffset(const llwchar* wchars, S32 begin_offset, F32 t // Round after kerning. - cur_x = llround(cur_x); + cur_x = (F32)llround(cur_x); } return llmin(max_chars, pos - begin_offset); @@ -799,7 +794,7 @@ const LLFontDescriptor& LLFontGL::getFontDesc() const } // static -void LLFontGL::initClass(F32 screen_dpi, F32 x_scale, F32 y_scale, const std::string& app_dir, const std::vector<std::string>& xui_paths, bool create_gl_textures) +void LLFontGL::initClass(F32 screen_dpi, F32 x_scale, F32 y_scale, const std::string& app_dir, bool create_gl_textures) { sVertDPI = (F32)llfloor(screen_dpi * y_scale); sHorizDPI = (F32)llfloor(screen_dpi * x_scale); @@ -810,7 +805,7 @@ void LLFontGL::initClass(F32 screen_dpi, F32 x_scale, F32 y_scale, const std::st // Font registry init if (!sFontRegistry) { - sFontRegistry = new LLFontRegistry(xui_paths, create_gl_textures); + sFontRegistry = new LLFontRegistry(create_gl_textures); sFontRegistry->parseFontInfo("fonts.xml"); } else @@ -1134,22 +1129,22 @@ void LLFontGL::renderQuad(LLVector3* vertex_out, LLVector2* uv_out, LLColor4U* c { S32 index = 0; - vertex_out[index] = LLVector3(llfont_round_x(screen_rect.mRight), llfont_round_y(screen_rect.mTop), 0.f); + vertex_out[index] = LLVector3(screen_rect.mRight, screen_rect.mTop, 0.f); uv_out[index] = LLVector2(uv_rect.mRight, uv_rect.mTop); colors_out[index] = color; index++; - vertex_out[index] = LLVector3(llfont_round_x(screen_rect.mLeft), llfont_round_y(screen_rect.mTop), 0.f); + vertex_out[index] = LLVector3(screen_rect.mLeft, screen_rect.mTop, 0.f); uv_out[index] = LLVector2(uv_rect.mLeft, uv_rect.mTop); colors_out[index] = color; index++; - vertex_out[index] = LLVector3(llfont_round_x(screen_rect.mLeft), llfont_round_y(screen_rect.mBottom), 0.f); + vertex_out[index] = LLVector3(screen_rect.mLeft, screen_rect.mBottom, 0.f); uv_out[index] = LLVector2(uv_rect.mLeft, uv_rect.mBottom); colors_out[index] = color; index++; - vertex_out[index] = LLVector3(llfont_round_x(screen_rect.mRight), llfont_round_y(screen_rect.mBottom), 0.f); + vertex_out[index] = LLVector3(screen_rect.mRight, screen_rect.mBottom, 0.f); uv_out[index] = LLVector2(uv_rect.mRight, uv_rect.mBottom); colors_out[index] = color; } diff --git a/indra/llrender/llfontgl.h b/indra/llrender/llfontgl.h index dc8d848ed2..0988e99deb 100644 --- a/indra/llrender/llfontgl.h +++ b/indra/llrender/llfontgl.h @@ -115,9 +115,9 @@ public: S32 renderUTF8(const std::string &text, S32 begin_offset, S32 x, S32 y, const LLColor4 &color, HAlign halign, VAlign valign, U8 style = NORMAL, ShadowType shadow = NO_SHADOW) const; // font metrics - override for LLFontFreetype that returns units of virtual pixels - F32 getLineHeight() const; F32 getAscenderHeight() const; F32 getDescenderHeight() const; + S32 getLineHeight() const; S32 getWidth(const std::string& utf8text) const; S32 getWidth(const llwchar* wchars) const; @@ -150,7 +150,7 @@ public: const LLFontDescriptor& getFontDesc() const; - static void initClass(F32 screen_dpi, F32 x_scale, F32 y_scale, const std::string& app_dir, const std::vector<std::string>& xui_paths, bool create_gl_textures = true); + static void initClass(F32 screen_dpi, F32 x_scale, F32 y_scale, const std::string& app_dir, bool create_gl_textures = true); // Load sans-serif, sans-serif-small, etc. // Slow, requires multiple seconds to load fonts. @@ -188,8 +188,9 @@ public: static std::string getFontPathLocal(); static std::string getFontPathSystem(); - static LLCoordFont sCurOrigin; - static std::vector<LLCoordFont> sOriginStack; + static LLCoordGL sCurOrigin; + static F32 sCurDepth; + static std::vector<std::pair<LLCoordGL, F32> > sOriginStack; static LLColor4 sShadowColor; diff --git a/indra/llrender/llfontregistry.cpp b/indra/llrender/llfontregistry.cpp index 4d22eba3d9..b5bdba996f 100644 --- a/indra/llrender/llfontregistry.cpp +++ b/indra/llrender/llfontregistry.cpp @@ -163,14 +163,9 @@ LLFontDescriptor LLFontDescriptor::normalize() const return LLFontDescriptor(new_name,new_size,new_style,getFileNames()); } -LLFontRegistry::LLFontRegistry(const string_vec_t& xui_paths, - bool create_gl_textures) +LLFontRegistry::LLFontRegistry(bool create_gl_textures) : mCreateGLTextures(create_gl_textures) { - // Propagate this down from LLUICtrlFactory so LLRender doesn't - // need an upstream dependency on LLUI. - mXUIPaths = xui_paths; - // This is potentially a slow directory traversal, so we want to // cache the result. mUltimateFallbackList = LLWindow::getDynamicFallbackFontList(); @@ -183,27 +178,30 @@ LLFontRegistry::~LLFontRegistry() bool LLFontRegistry::parseFontInfo(const std::string& xml_filename) { - bool success = false; // Succeed if we find at least one XUI file - const string_vec_t& xml_paths = mXUIPaths; + bool success = false; // Succeed if we find and read at least one XUI file + const string_vec_t xml_paths = gDirUtilp->findSkinnedFilenames(LLDir::XUI, xml_filename); + if (xml_paths.empty()) + { + // We didn't even find one single XUI file + return false; + } + for (string_vec_t::const_iterator path_it = xml_paths.begin(); path_it != xml_paths.end(); ++path_it) { - LLXMLNodePtr root; - std::string full_filename = gDirUtilp->findSkinnedFilename(*path_it, xml_filename); - bool parsed_file = LLXMLNode::parseFile(full_filename, root, NULL); + bool parsed_file = LLXMLNode::parseFile(*path_it, root, NULL); if (!parsed_file) continue; - + if ( root.isNull() || ! root->hasName( "fonts" ) ) { - llwarns << "Bad font info file: " - << full_filename << llendl; + llwarns << "Bad font info file: " << *path_it << llendl; continue; } - + std::string root_name; root->getAttributeString("name",root_name); if (root->hasName("fonts")) @@ -215,7 +213,7 @@ bool LLFontRegistry::parseFontInfo(const std::string& xml_filename) } //if (success) // dump(); - + return success; } diff --git a/indra/llrender/llfontregistry.h b/indra/llrender/llfontregistry.h index 8b06191c56..059248fbbd 100644 --- a/indra/llrender/llfontregistry.h +++ b/indra/llrender/llfontregistry.h @@ -67,8 +67,7 @@ class LLFontRegistry public: // create_gl_textures - set to false for test apps with no OpenGL window, // such as llui_libtest - LLFontRegistry(const string_vec_t& xui_paths, - bool create_gl_textures); + LLFontRegistry(bool create_gl_textures); ~LLFontRegistry(); // Load standard font info from XML file(s). @@ -105,7 +104,6 @@ private: font_size_map_t mFontSizes; string_vec_t mUltimateFallbackList; - string_vec_t mXUIPaths; bool mCreateGLTextures; }; diff --git a/indra/llrender/llgl.cpp b/indra/llrender/llgl.cpp index 946e602fee..9e4857b6bc 100644 --- a/indra/llrender/llgl.cpp +++ b/indra/llrender/llgl.cpp @@ -44,7 +44,6 @@ #include "llmath.h" #include "m4math.h" #include "llstring.h" -#include "llmemtype.h" #include "llstacktrace.h" #include "llglheaders.h" @@ -94,9 +93,15 @@ void APIENTRY gl_debug_callback(GLenum source, llwarns << "Severity: " << std::hex << severity << llendl; llwarns << "Message: " << message << llendl; llwarns << "-----------------------" << llendl; + if (severity == GL_DEBUG_SEVERITY_HIGH_ARB) + { + llerrs << "Halting on GL Error" << llendl; + } } #endif +void parse_glsl_version(S32& major, S32& minor); + void ll_init_fail_log(std::string filename) { gFailLog.open(filename.c_str()); @@ -243,6 +248,12 @@ PFNGLTEXIMAGE3DMULTISAMPLEPROC glTexImage3DMultisample = NULL; PFNGLGETMULTISAMPLEFVPROC glGetMultisamplefv = NULL; PFNGLSAMPLEMASKIPROC glSampleMaski = NULL; +//transform feedback (4.0 core) +PFNGLBEGINTRANSFORMFEEDBACKPROC glBeginTransformFeedback = NULL; +PFNGLENDTRANSFORMFEEDBACKPROC glEndTransformFeedback = NULL; +PFNGLTRANSFORMFEEDBACKVARYINGSPROC glTransformFeedbackVaryings = NULL; +PFNGLBINDBUFFERRANGEPROC glBindBufferRange = NULL; + //GL_ARB_debug_output PFNGLDEBUGMESSAGECONTROLARBPROC glDebugMessageControlARB = NULL; PFNGLDEBUGMESSAGEINSERTARBPROC glDebugMessageInsertARB = NULL; @@ -295,6 +306,7 @@ PFNGLGETACTIVEUNIFORMARBPROC glGetActiveUniformARB = NULL; PFNGLGETUNIFORMFVARBPROC glGetUniformfvARB = NULL; PFNGLGETUNIFORMIVARBPROC glGetUniformivARB = NULL; PFNGLGETSHADERSOURCEARBPROC glGetShaderSourceARB = NULL; +PFNGLVERTEXATTRIBIPOINTERPROC glVertexAttribIPointer = NULL; #if LL_WINDOWS PFNWGLCREATECONTEXTATTRIBSARBPROC wglCreateContextAttribsARB = NULL; @@ -414,6 +426,7 @@ LLGLManager::LLGLManager() : mHasDrawBuffers(FALSE), mHasTextureRectangle(FALSE), mHasTextureMultisample(FALSE), + mHasTransformFeedback(FALSE), mMaxSampleMaskWords(0), mMaxColorTextureSamples(0), mMaxDepthTextureSamples(0), @@ -443,7 +456,8 @@ LLGLManager::LLGLManager() : mDriverVersionMinor(0), mDriverVersionRelease(0), mGLVersion(1.0f), - + mGLSLVersionMajor(0), + mGLSLVersionMinor(0), mVRAM(0), mGLMaxVertexRange(0), mGLMaxIndexRange(0) @@ -550,10 +564,34 @@ bool LLGLManager::initGL() parse_gl_version( &mDriverVersionMajor, &mDriverVersionMinor, &mDriverVersionRelease, - &mDriverVersionVendorString ); + &mDriverVersionVendorString, + &mGLVersionString); mGLVersion = mDriverVersionMajor + mDriverVersionMinor * .1f; + if (mGLVersion >= 2.f) + { + parse_glsl_version(mGLSLVersionMajor, mGLSLVersionMinor); + +#if LL_DARWIN + //never use GLSL greater than 1.20 on OSX + if (mGLSLVersionMajor > 1 || mGLSLVersionMinor >= 30) + { + mGLSLVersionMajor = 1; + mGLSLVersionMinor = 20; + } +#endif + } + + if (mGLVersion >= 2.1f && LLImageGL::sCompressTextures) + { //use texture compression + glHint(GL_TEXTURE_COMPRESSION_HINT, GL_NICEST); + } + else + { //GL version is < 3.0, always disable texture compression + LLImageGL::sCompressTextures = false; + } + // Trailing space necessary to keep "nVidia Corpor_ati_on" cards // from being recognized as ATI. if (mGLVendor.substr(0,4) == "ATI ") @@ -574,11 +612,8 @@ bool LLGLManager::initGL() #endif // LL_WINDOWS #if (LL_WINDOWS || LL_LINUX) && !LL_MESA_HEADLESS - // release 7277 is a point at which we verify that ATI OpenGL - // drivers get pretty stable with SL, ~Catalyst 8.2, - // for both Win32 and Linux. - if (mDriverVersionRelease < 7277 && - mDriverVersionRelease != 0) // 0 == Undetectable driver version - these get to pretend to be new ATI drivers, though that decision may be revisited. + // count any pre OpenGL 3.0 implementation as an old driver + if (mGLVersion < 3.f) { mATIOldDriver = TRUE; } @@ -717,6 +752,11 @@ bool LLGLManager::initGL() } #endif + if (mIsIntel && mGLVersion <= 3.f) + { //never try to use framebuffer objects on older intel drivers (crashy) + mHasFramebufferObject = FALSE; + } + if (mHasFramebufferObject) { glGetIntegerv(GL_MAX_SAMPLES, &mMaxSamples); @@ -905,7 +945,6 @@ void LLGLManager::initExtensions() mHasMultitexture = glh_init_extensions("GL_ARB_multitexture"); mHasATIMemInfo = ExtensionExists("GL_ATI_meminfo", gGLHExts.mSysExts); mHasNVXMemInfo = ExtensionExists("GL_NVX_gpu_memory_info", gGLHExts.mSysExts); - mHasMipMapGeneration = glh_init_extensions("GL_SGIS_generate_mipmap"); mHasSeparateSpecularColor = glh_init_extensions("GL_EXT_separate_specular_color"); mHasAnisotropic = glh_init_extensions("GL_EXT_texture_filter_anisotropic"); glh_init_extensions("GL_ARB_texture_cube_map"); @@ -930,11 +969,14 @@ void LLGLManager::initExtensions() ExtensionExists("GL_EXT_packed_depth_stencil", gGLHExts.mSysExts); #endif + mHasMipMapGeneration = mHasFramebufferObject || mGLVersion >= 1.4f; + mHasDrawBuffers = ExtensionExists("GL_ARB_draw_buffers", gGLHExts.mSysExts); mHasBlendFuncSeparate = ExtensionExists("GL_EXT_blend_func_separate", gGLHExts.mSysExts); mHasTextureRectangle = ExtensionExists("GL_ARB_texture_rectangle", gGLHExts.mSysExts); mHasTextureMultisample = ExtensionExists("GL_ARB_texture_multisample", gGLHExts.mSysExts); mHasDebugOutput = ExtensionExists("GL_ARB_debug_output", gGLHExts.mSysExts); + mHasTransformFeedback = mGLVersion >= 4.f ? TRUE : FALSE; #if !LL_DARWIN mHasPointParameters = !mIsATI && ExtensionExists("GL_ARB_point_parameters", gGLHExts.mSysExts); #endif @@ -1174,7 +1216,14 @@ void LLGLManager::initExtensions() glTexImage3DMultisample = (PFNGLTEXIMAGE3DMULTISAMPLEPROC) GLH_EXT_GET_PROC_ADDRESS("glTexImage3DMultisample"); glGetMultisamplefv = (PFNGLGETMULTISAMPLEFVPROC) GLH_EXT_GET_PROC_ADDRESS("glGetMultisamplefv"); glSampleMaski = (PFNGLSAMPLEMASKIPROC) GLH_EXT_GET_PROC_ADDRESS("glSampleMaski"); - } + } + if (mHasTransformFeedback) + { + glBeginTransformFeedback = (PFNGLBEGINTRANSFORMFEEDBACKPROC) GLH_EXT_GET_PROC_ADDRESS("glBeginTransformFeedback"); + glEndTransformFeedback = (PFNGLENDTRANSFORMFEEDBACKPROC) GLH_EXT_GET_PROC_ADDRESS("glEndTransformFeedback"); + glTransformFeedbackVaryings = (PFNGLTRANSFORMFEEDBACKVARYINGSPROC) GLH_EXT_GET_PROC_ADDRESS("glTransformFeedbackVaryings"); + glBindBufferRange = (PFNGLBINDBUFFERRANGEPROC) GLH_EXT_GET_PROC_ADDRESS("glBindBufferRange"); + } if (mHasDebugOutput) { glDebugMessageControlARB = (PFNGLDEBUGMESSAGECONTROLARBPROC) GLH_EXT_GET_PROC_ADDRESS("glDebugMessageControlARB"); @@ -1300,6 +1349,7 @@ void LLGLManager::initExtensions() glVertexAttrib4uivARB = (PFNGLVERTEXATTRIB4UIVARBPROC) GLH_EXT_GET_PROC_ADDRESS("glVertexAttrib4uivARB"); glVertexAttrib4usvARB = (PFNGLVERTEXATTRIB4USVARBPROC) GLH_EXT_GET_PROC_ADDRESS("glVertexAttrib4usvARB"); glVertexAttribPointerARB = (PFNGLVERTEXATTRIBPOINTERARBPROC) GLH_EXT_GET_PROC_ADDRESS("glVertexAttribPointerARB"); + glVertexAttribIPointer = (PFNGLVERTEXATTRIBIPOINTERPROC) GLH_EXT_GET_PROC_ADDRESS("glVertexAttribIPointer"); glEnableVertexAttribArrayARB = (PFNGLENABLEVERTEXATTRIBARRAYARBPROC) GLH_EXT_GET_PROC_ADDRESS("glEnableVertexAttribArrayARB"); glDisableVertexAttribArrayARB = (PFNGLDISABLEVERTEXATTRIBARRAYARBPROC) GLH_EXT_GET_PROC_ADDRESS("glDisableVertexAttribArrayARB"); glProgramStringARB = (PFNGLPROGRAMSTRINGARBPROC) GLH_EXT_GET_PROC_ADDRESS("glProgramStringARB"); @@ -1878,7 +1928,7 @@ void LLGLState::checkClientArrays(const std::string& msg, U32 data_mask) glClientActiveTextureARB(GL_TEXTURE0_ARB); gGL.getTexUnit(0)->activate(); - if (gGLManager.mHasVertexShader) + if (gGLManager.mHasVertexShader && LLGLSLShader::sNoFixedFunction) { //make sure vertex attribs are all disabled GLint count; glGetIntegerv(GL_MAX_VERTEX_ATTRIBS_ARB, &count); @@ -1930,6 +1980,7 @@ LLGLState::LLGLState(LLGLenum state, S32 enabled) : case GL_COLOR_MATERIAL: case GL_FOG: case GL_LINE_STIPPLE: + case GL_POLYGON_STIPPLE: mState = 0; break; } @@ -2018,7 +2069,7 @@ void LLGLManager::initGLStates() //////////////////////////////////////////////////////////////////////////////// -void parse_gl_version( S32* major, S32* minor, S32* release, std::string* vendor_specific ) +void parse_gl_version( S32* major, S32* minor, S32* release, std::string* vendor_specific, std::string* version_string ) { // GL_VERSION returns a null-terminated string with the format: // <major>.<minor>[.<release>] [<vendor specific>] @@ -2034,6 +2085,8 @@ void parse_gl_version( S32* major, S32* minor, S32* release, std::string* vendor return; } + version_string->assign(version); + std::string ver_copy( version ); S32 len = (S32)strlen( version ); /* Flawfinder: ignore */ S32 i = 0; @@ -2098,6 +2151,55 @@ void parse_gl_version( S32* major, S32* minor, S32* release, std::string* vendor } } + +void parse_glsl_version(S32& major, S32& minor) +{ + // GL_SHADING_LANGUAGE_VERSION returns a null-terminated string with the format: + // <major>.<minor>[.<release>] [<vendor specific>] + + const char* version = (const char*) glGetString(GL_SHADING_LANGUAGE_VERSION); + major = 0; + minor = 0; + + if( !version ) + { + return; + } + + std::string ver_copy( version ); + S32 len = (S32)strlen( version ); /* Flawfinder: ignore */ + S32 i = 0; + S32 start; + // Find the major version + start = i; + for( ; i < len; i++ ) + { + if( '.' == version[i] ) + { + break; + } + } + std::string major_str = ver_copy.substr(start,i-start); + LLStringUtil::convertToS32(major_str, major); + + if( '.' == version[i] ) + { + i++; + } + + // Find the minor version + start = i; + for( ; i < len; i++ ) + { + if( ('.' == version[i]) || isspace(version[i]) ) + { + break; + } + } + std::string minor_str = ver_copy.substr(start,i-start); + LLStringUtil::convertToS32(minor_str, minor); +} + LLGLUserClipPlane::LLGLUserClipPlane(const LLPlane& p, const glh::matrix4f& modelview, const glh::matrix4f& projection, bool apply) { mApply = apply; @@ -2220,7 +2322,6 @@ void LLGLNamePool::release(GLuint name) //static void LLGLNamePool::upkeepPools() { - LLMemType mt(LLMemType::MTYPE_UPKEEP_POOLS); for (tracker_t::instance_iter iter = beginInstances(); iter != endInstances(); ++iter) { LLGLNamePool & pool = *iter; @@ -2346,3 +2447,65 @@ LLGLSquashToFarClip::~LLGLSquashToFarClip() gGL.matrixMode(LLRender::MM_MODELVIEW); } + + +LLGLSyncFence::LLGLSyncFence() +{ +#ifdef GL_ARB_sync + mSync = 0; +#endif +} + +LLGLSyncFence::~LLGLSyncFence() +{ +#ifdef GL_ARB_sync + if (mSync) + { + glDeleteSync(mSync); + } +#endif +} + +void LLGLSyncFence::placeFence() +{ +#ifdef GL_ARB_sync + if (mSync) + { + glDeleteSync(mSync); + } + mSync = glFenceSync(GL_SYNC_GPU_COMMANDS_COMPLETE, 0); +#endif +} + +bool LLGLSyncFence::isCompleted() +{ + bool ret = true; +#ifdef GL_ARB_sync + if (mSync) + { + GLenum status = glClientWaitSync(mSync, 0, 1); + if (status == GL_TIMEOUT_EXPIRED) + { + ret = false; + } + } +#endif + return ret; +} + +void LLGLSyncFence::wait() +{ +#ifdef GL_ARB_sync + if (mSync) + { + while (glClientWaitSync(mSync, 0, FENCE_WAIT_TIME_NANOSECONDS) == GL_TIMEOUT_EXPIRED) + { //track the number of times we've waited here + static S32 waits = 0; + waits++; + } + } +#endif +} + + + diff --git a/indra/llrender/llgl.h b/indra/llrender/llgl.h index 6a147b8e19..d70e764769 100644 --- a/indra/llrender/llgl.h +++ b/indra/llrender/llgl.h @@ -104,6 +104,7 @@ public: BOOL mHasDepthClamp; BOOL mHasTextureRectangle; BOOL mHasTextureMultisample; + BOOL mHasTransformFeedback; S32 mMaxSampleMaskWords; S32 mMaxColorTextureSamples; S32 mMaxDepthTextureSamples; @@ -138,7 +139,10 @@ public: S32 mDriverVersionMinor; S32 mDriverVersionRelease; F32 mGLVersion; // e.g = 1.4 + S32 mGLSLVersionMajor; + S32 mGLSLVersionMinor; std::string mDriverVersionVendorString; + std::string mGLVersionString; S32 mVRAM; // VRAM in MB S32 mGLMaxVertexRange; @@ -415,13 +419,42 @@ public: virtual void updateGL() = 0; }; +const U32 FENCE_WAIT_TIME_NANOSECONDS = 1000; //1 ms + +class LLGLFence +{ +public: + virtual ~LLGLFence() + { + } + + virtual void placeFence() = 0; + virtual bool isCompleted() = 0; + virtual void wait() = 0; +}; + +class LLGLSyncFence : public LLGLFence +{ +public: +#ifdef GL_ARB_sync + GLsync mSync; +#endif + + LLGLSyncFence(); + virtual ~LLGLSyncFence(); + + void placeFence(); + bool isCompleted(); + void wait(); +}; + extern LLMatrix4 gGLObliqueProjectionInverse; #include "llglstates.h" void init_glstates(); -void parse_gl_version( S32* major, S32* minor, S32* release, std::string* vendor_specific ); +void parse_gl_version( S32* major, S32* minor, S32* release, std::string* vendor_specific, std::string* version_string ); extern BOOL gClothRipple; extern BOOL gHeadlessClient; diff --git a/indra/llrender/llglheaders.h b/indra/llrender/llglheaders.h index 10aad202e1..509de51f4d 100644 --- a/indra/llrender/llglheaders.h +++ b/indra/llrender/llglheaders.h @@ -199,6 +199,7 @@ extern PFNGLVERTEXATTRIB4UBVARBPROC glVertexAttrib4ubvARB; extern PFNGLVERTEXATTRIB4UIVARBPROC glVertexAttrib4uivARB; extern PFNGLVERTEXATTRIB4USVARBPROC glVertexAttrib4usvARB; extern PFNGLVERTEXATTRIBPOINTERARBPROC glVertexAttribPointerARB; +extern PFNGLVERTEXATTRIBIPOINTERPROC glVertexAttribIPointer; extern PFNGLENABLEVERTEXATTRIBARRAYARBPROC glEnableVertexAttribArrayARB; extern PFNGLDISABLEVERTEXATTRIBARRAYARBPROC glDisableVertexAttribArrayARB; extern PFNGLPROGRAMSTRINGARBPROC glProgramStringARB; @@ -460,6 +461,7 @@ extern PFNGLVERTEXATTRIB4UBVARBPROC glVertexAttrib4ubvARB; extern PFNGLVERTEXATTRIB4UIVARBPROC glVertexAttrib4uivARB; extern PFNGLVERTEXATTRIB4USVARBPROC glVertexAttrib4usvARB; extern PFNGLVERTEXATTRIBPOINTERARBPROC glVertexAttribPointerARB; +extern PFNGLVERTEXATTRIBIPOINTERPROC glVertexAttribIPointer; extern PFNGLENABLEVERTEXATTRIBARRAYARBPROC glEnableVertexAttribArrayARB; extern PFNGLDISABLEVERTEXATTRIBARRAYARBPROC glDisableVertexAttribArrayARB; extern PFNGLPROGRAMSTRINGARBPROC glProgramStringARB; @@ -526,6 +528,13 @@ extern PFNGLTEXIMAGE3DMULTISAMPLEPROC glTexImage3DMultisample; extern PFNGLGETMULTISAMPLEFVPROC glGetMultisamplefv; extern PFNGLSAMPLEMASKIPROC glSampleMaski; +//transform feedback (4.0 core) +extern PFNGLBEGINTRANSFORMFEEDBACKPROC glBeginTransformFeedback; +extern PFNGLENDTRANSFORMFEEDBACKPROC glEndTransformFeedback; +extern PFNGLTRANSFORMFEEDBACKVARYINGSPROC glTransformFeedbackVaryings; +extern PFNGLBINDBUFFERRANGEPROC glBindBufferRange; + + #elif LL_WINDOWS //---------------------------------------------------------------------------- // LL_WINDOWS @@ -693,6 +702,7 @@ extern PFNGLVERTEXATTRIB4UBVARBPROC glVertexAttrib4ubvARB; extern PFNGLVERTEXATTRIB4UIVARBPROC glVertexAttrib4uivARB; extern PFNGLVERTEXATTRIB4USVARBPROC glVertexAttrib4usvARB; extern PFNGLVERTEXATTRIBPOINTERARBPROC glVertexAttribPointerARB; +extern PFNGLVERTEXATTRIBIPOINTERPROC glVertexAttribIPointer; extern PFNGLENABLEVERTEXATTRIBARRAYARBPROC glEnableVertexAttribArrayARB; extern PFNGLDISABLEVERTEXATTRIBARRAYARBPROC glDisableVertexAttribArrayARB; extern PFNGLPROGRAMSTRINGARBPROC glProgramStringARB; @@ -756,6 +766,12 @@ extern PFNGLTEXIMAGE3DMULTISAMPLEPROC glTexImage3DMultisample; extern PFNGLGETMULTISAMPLEFVPROC glGetMultisamplefv; extern PFNGLSAMPLEMASKIPROC glSampleMaski; +//transform feedback (4.0 core) +extern PFNGLBEGINTRANSFORMFEEDBACKPROC glBeginTransformFeedback; +extern PFNGLENDTRANSFORMFEEDBACKPROC glEndTransformFeedback; +extern PFNGLTRANSFORMFEEDBACKVARYINGSPROC glTransformFeedbackVaryings; +extern PFNGLBINDBUFFERRANGEPROC glBindBufferRange; + //GL_ARB_debug_output extern PFNGLDEBUGMESSAGECONTROLARBPROC glDebugMessageControlARB; extern PFNGLDEBUGMESSAGEINSERTARBPROC glDebugMessageInsertARB; @@ -977,7 +993,12 @@ extern void glGetBufferPointervARB (GLenum, GLenum, GLvoid* *); } #endif +#if __MAC_OS_X_VERSION_MAX_ALLOWED <= 1070 +#include <OpenGL/gl.h> +#else #include <AGL/gl.h> +#endif + #endif // LL_MESA / LL_WINDOWS / LL_DARWIN diff --git a/indra/llrender/llglslshader.cpp b/indra/llrender/llglslshader.cpp index 5a6f3d8292..7cbf39096e 100644 --- a/indra/llrender/llglslshader.cpp +++ b/indra/llrender/llglslshader.cpp @@ -64,10 +64,23 @@ BOOL shouldChange(const LLVector4& v1, const LLVector4& v2) } LLShaderFeatures::LLShaderFeatures() -: calculatesLighting(false), isShiny(false), isFullbright(false), hasWaterFog(false), -hasTransport(false), hasSkinning(false), hasObjectSkinning(false), hasAtmospherics(false), isSpecular(false), -hasGamma(false), hasLighting(false), calculatesAtmospherics(false), mIndexedTextureChannels(0), disableTextureIndex(false), -hasAlphaMask(false) + : atmosphericHelpers(false) + , calculatesLighting(false) + , calculatesAtmospherics(false) + , hasLighting(false) + , isAlphaLighting(false) + , isShiny(false) + , isFullbright(false) + , isSpecular(false) + , hasWaterFog(false) + , hasTransport(false) + , hasSkinning(false) + , hasObjectSkinning(false) + , hasAtmospherics(false) + , hasGamma(false) + , mIndexedTextureChannels(0) + , disableTextureIndex(false) + , hasAlphaMask(false) { } @@ -96,7 +109,12 @@ void LLGLSLShader::unload() glGetAttachedObjectsARB(mProgramObject, 1024, &count, obj); for (GLsizei i = 0; i < count; i++) { - glDeleteObjectARB(obj[i]); +#if !LL_DARWIN + if (glIsProgramARB(obj[i])) +#endif + { + glDeleteObjectARB(obj[i]); + } } glDeleteObjectARB(mProgramObject); @@ -111,7 +129,9 @@ void LLGLSLShader::unload() } BOOL LLGLSLShader::createShader(vector<string> * attributes, - vector<string> * uniforms) + vector<string> * uniforms, + U32 varying_count, + const char** varyings) { //reloading, reset matrix hash values for (U32 i = 0; i < LLRender::NUM_MATRIX_MODES; ++i) @@ -148,11 +168,19 @@ BOOL LLGLSLShader::createShader(vector<string> * attributes, return FALSE; } - if (gGLManager.mGLVersion < 3.1f) - { //attachShaderFeatures may have set the number of indexed texture channels, so set to 1 again + if (gGLManager.mGLSLVersionMajor < 2 && gGLManager.mGLSLVersionMinor < 3) + { //indexed texture rendering requires GLSL 1.3 or later + //attachShaderFeatures may have set the number of indexed texture channels, so set to 1 again mFeatures.mIndexedTextureChannels = llmin(mFeatures.mIndexedTextureChannels, 1); } +#ifdef GL_INTERLEAVED_ATTRIBS + if (varying_count > 0 && varyings) + { + glTransformFeedbackVaryings(mProgramObject, varying_count, varyings, GL_INTERLEAVED_ATTRIBS); + } +#endif + // Map attributes and uniforms if (success) { diff --git a/indra/llrender/llglslshader.h b/indra/llrender/llglslshader.h index 2a6c050eac..5c68cb46eb 100644 --- a/indra/llrender/llglslshader.h +++ b/indra/llrender/llglslshader.h @@ -33,9 +33,11 @@ class LLShaderFeatures { public: + bool atmosphericHelpers; bool calculatesLighting; bool calculatesAtmospherics; bool hasLighting; // implies no transport (it's possible to have neither though) + bool isAlphaLighting; // indicates lighting shaders need not be linked in (lighting performed directly in alpha shader to match deferred lighting functions) bool isShiny; bool isFullbright; // implies no lighting bool isSpecular; @@ -74,7 +76,9 @@ public: void unload(); BOOL createShader(std::vector<std::string> * attributes, - std::vector<std::string> * uniforms); + std::vector<std::string> * uniforms, + U32 varying_count = 0, + const char** varyings = NULL); BOOL attachObject(std::string object); void attachObject(GLhandleARB object); void attachObjects(GLhandleARB* objects = NULL, S32 count = 0); diff --git a/indra/llrender/llglstates.h b/indra/llrender/llglstates.h index e26aead676..0e2c3bcb44 100644 --- a/indra/llrender/llglstates.h +++ b/indra/llrender/llglstates.h @@ -59,7 +59,6 @@ protected: LLGLEnable mColorMaterial; LLGLDisable mAlphaTest, mBlend, mCullFace, mDither, mFog, mLineSmooth, mLineStipple, mNormalize, mPolygonSmooth, - mTextureGenQ, mTextureGenR, mTextureGenS, mTextureGenT, mGLMultisample; public: LLGLSDefault() @@ -76,10 +75,6 @@ public: mLineStipple(GL_LINE_STIPPLE), mNormalize(GL_NORMALIZE), mPolygonSmooth(GL_POLYGON_SMOOTH), - mTextureGenQ(GL_TEXTURE_GEN_Q), - mTextureGenR(GL_TEXTURE_GEN_R), - mTextureGenS(GL_TEXTURE_GEN_S), - mTextureGenT(GL_TEXTURE_GEN_T), mGLMultisample(GL_MULTISAMPLE_ARB) { } }; diff --git a/indra/llrender/llimagegl.cpp b/indra/llrender/llimagegl.cpp index 78591ddd38..a4d7872ec2 100644 --- a/indra/llrender/llimagegl.cpp +++ b/indra/llrender/llimagegl.cpp @@ -42,8 +42,11 @@ //---------------------------------------------------------------------------- const F32 MIN_TEXTURE_LIFETIME = 10.f; +//which power of 2 is i? +//assumes i is a power of 2 > 0 +U32 wpo2(U32 i); + //statics -LLGLuint LLImageGL::sCurrentBoundTextures[MAX_GL_TEXTURE_UNITS] = { 0 }; U32 LLImageGL::sUniqueCount = 0; U32 LLImageGL::sBindCount = 0; @@ -51,12 +54,14 @@ S32 LLImageGL::sGlobalTextureMemoryInBytes = 0; S32 LLImageGL::sBoundTextureMemoryInBytes = 0; S32 LLImageGL::sCurBoundTextureMemory = 0; S32 LLImageGL::sCount = 0; -std::list<U32> LLImageGL::sDeadTextureList; +LLImageGL::dead_texturelist_t LLImageGL::sDeadTextureList[LLTexUnit::TT_NONE]; +U32 LLImageGL::sCurTexName = 1; BOOL LLImageGL::sGlobalUseAnisotropic = FALSE; F32 LLImageGL::sLastFrameTime = 0.f; BOOL LLImageGL::sAllowReadBackRaw = FALSE ; LLImageGL* LLImageGL::sDefaultGLTexture = NULL ; +bool LLImageGL::sCompressTextures = false; std::set<LLImageGL*> LLImageGL::sImageList; @@ -65,19 +70,10 @@ std::set<LLImageGL*> LLImageGL::sImageList; //**************************************************************************************************** //----------------------- //debug use -BOOL gAuditTexture = FALSE ; -#define MAX_TEXTURE_LOG_SIZE 22 //2048 * 2048 -std::vector<S32> LLImageGL::sTextureLoadedCounter(MAX_TEXTURE_LOG_SIZE + 1) ; -std::vector<S32> LLImageGL::sTextureBoundCounter(MAX_TEXTURE_LOG_SIZE + 1) ; -std::vector<S32> LLImageGL::sTextureCurBoundCounter(MAX_TEXTURE_LOG_SIZE + 1) ; S32 LLImageGL::sCurTexSizeBar = -1 ; S32 LLImageGL::sCurTexPickSize = -1 ; -LLPointer<LLImageGL> LLImageGL::sHighlightTexturep = NULL; -S32 LLImageGL::sMaxCatagories = 1 ; +S32 LLImageGL::sMaxCategories = 1 ; -std::vector<S32> LLImageGL::sTextureMemByCategory; -std::vector<S32> LLImageGL::sTextureMemByCategoryBound ; -std::vector<S32> LLImageGL::sTextureCurMemByCategoryBound ; //------------------------ //**************************************************************************************************** //End for texture auditing use only @@ -175,49 +171,11 @@ BOOL is_little_endian() //static void LLImageGL::initClass(S32 num_catagories) { - sMaxCatagories = num_catagories ; - - sTextureMemByCategory.resize(sMaxCatagories); - sTextureMemByCategoryBound.resize(sMaxCatagories) ; - sTextureCurMemByCategoryBound.resize(sMaxCatagories) ; } //static void LLImageGL::cleanupClass() { - sTextureMemByCategory.clear() ; - sTextureMemByCategoryBound.clear() ; - sTextureCurMemByCategoryBound.clear() ; -} - -//static -void LLImageGL::setHighlightTexture(S32 category) -{ - const S32 dim = 128; - sHighlightTexturep = new LLImageGL() ; - LLPointer<LLImageRaw> image_raw = new LLImageRaw(dim,dim,3); - U8* data = image_raw->getData(); - for (S32 i = 0; i<dim; i++) - { - for (S32 j = 0; j<dim; j++) - { - const S32 border = 2; - if (i<border || j<border || i>=(dim-border) || j>=(dim-border)) - { - *data++ = 0xff; - *data++ = 0xff; - *data++ = 0xff; - } - else - { - *data++ = 0xff; - *data++ = 0xff; - *data++ = 0x00; - } - } - } - sHighlightTexturep->createGLTexture(0, image_raw, 0, TRUE, category); - image_raw = NULL; } //static @@ -279,37 +237,19 @@ S32 LLImageGL::dataFormatComponents(S32 dataformat) //---------------------------------------------------------------------------- +static LLFastTimer::DeclareTimer FTM_IMAGE_UPDATE_STATS("Image Stats"); // static void LLImageGL::updateStats(F32 current_time) { + LLFastTimer t(FTM_IMAGE_UPDATE_STATS); sLastFrameTime = current_time; sBoundTextureMemoryInBytes = sCurBoundTextureMemory; sCurBoundTextureMemory = 0; - - if(gAuditTexture) - { - for(U32 i = 0 ; i < sTextureCurBoundCounter.size() ; i++) - { - sTextureBoundCounter[i] = sTextureCurBoundCounter[i] ; - sTextureCurBoundCounter[i] = 0 ; - } - for(U32 i = 0 ; i < sTextureCurMemByCategoryBound.size() ; i++) - { - sTextureMemByCategoryBound[i] = sTextureCurMemByCategoryBound[i] ; - sTextureCurMemByCategoryBound[i] = 0 ; - } - } } //static S32 LLImageGL::updateBoundTexMem(const S32 mem, const S32 ncomponents, S32 category) { - if(gAuditTexture && ncomponents > 0 && category > -1) - { - sTextureCurBoundCounter[getTextureCounterIndex(mem / ncomponents)]++ ; - sTextureCurMemByCategoryBound[category] += mem ; - } - LLImageGL::sCurBoundTextureMemory += mem ; return LLImageGL::sCurBoundTextureMemory; } @@ -477,10 +417,13 @@ void LLImageGL::init(BOOL usemipmaps) mDiscardLevelInAtlas = -1 ; mTexelsInAtlas = 0 ; mTexelsInGLTexture = 0 ; + + mAllowCompression = true; mTarget = GL_TEXTURE_2D; mBindTarget = LLTexUnit::TT_TEXTURE; mHasMipMaps = false; + mMipLevels = -1; mIsResident = 0; @@ -535,7 +478,7 @@ bool LLImageGL::checkSize(S32 width, S32 height) return check_power_of_two(width) && check_power_of_two(height); } -void LLImageGL::setSize(S32 width, S32 height, S32 ncomponents) +void LLImageGL::setSize(S32 width, S32 height, S32 ncomponents, S32 discard_level) { if (width != mWidth || height != mHeight || ncomponents != mComponents) { @@ -568,6 +511,11 @@ void LLImageGL::setSize(S32 width, S32 height, S32 ncomponents) width >>= 1; height >>= 1; } + + if(discard_level > 0) + { + mMaxDiscardLevel = llmax(mMaxDiscardLevel, (S8)discard_level); + } } else { @@ -671,8 +619,24 @@ void LLImageGL::setImage(const U8* data_in, BOOL data_hasmips) is_compressed = true; } + + + if (mUseMipMaps) + { + //set has mip maps to true before binding image so tex parameters get set properly + gGL.getTexUnit(0)->unbind(mBindTarget); + mHasMipMaps = true; + mTexOptionsDirty = true; + setFilteringOption(LLTexUnit::TFO_ANISOTROPIC); + } + else + { + mHasMipMaps = false; + } + llverify(gGL.getTexUnit(0)->bind(this)); + if (mUseMipMaps) { if (data_hasmips) @@ -685,6 +649,9 @@ void LLImageGL::setImage(const U8* data_in, BOOL data_hasmips) S32 w = getWidth(d); S32 h = getHeight(d); S32 gl_level = d-mCurrentDiscardLevel; + + mMipLevels = llmax(mMipLevels, gl_level); + if (d > mCurrentDiscardLevel) { data_in -= dataFormatBytes(mFormatPrimary, w, h); // see above comment @@ -705,7 +672,7 @@ void LLImageGL::setImage(const U8* data_in, BOOL data_hasmips) stop_glerror(); } - LLImageGL::setManualImage(mTarget, gl_level, mFormatInternal, w, h, mFormatPrimary, GL_UNSIGNED_BYTE, (GLvoid*)data_in); + LLImageGL::setManualImage(mTarget, gl_level, mFormatInternal, w, h, mFormatPrimary, GL_UNSIGNED_BYTE, (GLvoid*)data_in, mAllowCompression); if (gl_level == 0) { analyzeAlpha(data_in, w, h); @@ -727,10 +694,6 @@ void LLImageGL::setImage(const U8* data_in, BOOL data_hasmips) { if (mAutoGenMips) { - if (!gGLManager.mHasFramebufferObject) - { - glTexParameteri(LLTexUnit::getInternalType(mBindTarget), GL_GENERATE_MIPMAP_SGIS, TRUE); - } stop_glerror(); { // LLFastTimer t2(FTM_TEMP4); @@ -744,10 +707,15 @@ void LLImageGL::setImage(const U8* data_in, BOOL data_hasmips) S32 w = getWidth(mCurrentDiscardLevel); S32 h = getHeight(mCurrentDiscardLevel); + mMipLevels = wpo2(llmax(w, h)); + + //use legacy mipmap generation mode + glTexParameteri(mTarget, GL_GENERATE_MIPMAP, GL_TRUE); + LLImageGL::setManualImage(mTarget, 0, mFormatInternal, w, h, mFormatPrimary, mFormatType, - data_in); + data_in, mAllowCompression); analyzeAlpha(data_in, w, h); stop_glerror(); @@ -759,16 +727,10 @@ void LLImageGL::setImage(const U8* data_in, BOOL data_hasmips) stop_glerror(); } } - - if (gGLManager.mHasFramebufferObject) - { - glGenerateMipmap(LLTexUnit::getInternalType(mBindTarget)); - } } else { // Create mips by hand - // about 30% faster than autogen on ATI 9800, 50% slower on nVidia 4800 // ~4x faster than gluBuild2DMipmaps S32 width = getWidth(mCurrentDiscardLevel); S32 height = getHeight(mCurrentDiscardLevel); @@ -778,6 +740,9 @@ void LLImageGL::setImage(const U8* data_in, BOOL data_hasmips) const U8* cur_mip_data = 0; S32 prev_mip_size = 0; S32 cur_mip_size = 0; + + mMipLevels = nummips; + for (int m=0; m<nummips; m++) { if (m==0) @@ -805,7 +770,7 @@ void LLImageGL::setImage(const U8* data_in, BOOL data_hasmips) stop_glerror(); } - LLImageGL::setManualImage(mTarget, m, mFormatInternal, w, h, mFormatPrimary, mFormatType, cur_mip_data); + LLImageGL::setManualImage(mTarget, m, mFormatInternal, w, h, mFormatPrimary, mFormatType, cur_mip_data, mAllowCompression); if (m == 0) { analyzeAlpha(data_in, w, h); @@ -842,10 +807,10 @@ void LLImageGL::setImage(const U8* data_in, BOOL data_hasmips) { llerrs << "Compressed Image has mipmaps but data does not (can not auto generate compressed mips)" << llendl; } - mHasMipMaps = true; } else { + mMipLevels = 0; S32 w = getWidth(); S32 h = getHeight(); if (is_compressed) @@ -863,7 +828,7 @@ void LLImageGL::setImage(const U8* data_in, BOOL data_hasmips) } LLImageGL::setManualImage(mTarget, 0, mFormatInternal, w, h, - mFormatPrimary, mFormatType, (GLvoid *)data_in); + mFormatPrimary, mFormatType, (GLvoid *)data_in, mAllowCompression); analyzeAlpha(data_in, w, h); updatePickMask(w, h, data_in); @@ -877,7 +842,6 @@ void LLImageGL::setImage(const U8* data_in, BOOL data_hasmips) } } - mHasMipMaps = false; } stop_glerror(); mGLTextureCreated = true; @@ -901,14 +865,13 @@ BOOL LLImageGL::preAddToAtlas(S32 discard_level, const LLImageRaw* raw_image) llassert(mCurrentDiscardLevel >= 0); discard_level = mCurrentDiscardLevel; } - discard_level = llclamp(discard_level, 0, (S32)mMaxDiscardLevel); - + // Actual image width/height = raw image width/height * 2^discard_level S32 w = raw_image->getWidth() << discard_level; S32 h = raw_image->getHeight() << discard_level; // setSize may call destroyGLTexture if the size does not match - setSize(w, h, raw_image->getComponents()); + setSize(w, h, raw_image->getComponents(), discard_level); if( !mHasExplicitFormat ) { @@ -1090,27 +1053,69 @@ BOOL LLImageGL::setSubImageFromFrameBuffer(S32 fb_x, S32 fb_y, S32 x_pos, S32 y_ } // static -void LLImageGL::generateTextures(S32 numTextures, U32 *textures) +void LLImageGL::generateTextures(LLTexUnit::eTextureType type, U32 format, S32 numTextures, U32 *textures) { - glGenTextures(numTextures, (GLuint*)textures); + bool empty = true; + + dead_texturelist_t::iterator iter = sDeadTextureList[type].find(format); + + if (iter != sDeadTextureList[type].end()) + { + empty = iter->second.empty(); + } + + for (S32 i = 0; i < numTextures; ++i) + { + if (!empty) + { + textures[i] = iter->second.front(); + iter->second.pop_front(); + empty = iter->second.empty(); + } + else + { + textures[i] = sCurTexName++; + } + } } // static -void LLImageGL::deleteTextures(S32 numTextures, U32 *textures, bool immediate) +void LLImageGL::deleteTextures(LLTexUnit::eTextureType type, U32 format, S32 mip_levels, S32 numTextures, U32 *textures, bool immediate) { - for (S32 i = 0; i < numTextures; i++) + if (gGLManager.mInited) { - sDeadTextureList.push_back(textures[i]); - } + if (format == 0 || type == LLTexUnit::TT_CUBE_MAP || mip_levels == -1) + { //unknown internal format or unknown number of mip levels, not safe to reuse + glDeleteTextures(numTextures, textures); + } + else + { + for (S32 i = 0; i < numTextures; ++i) + { //remove texture from VRAM by setting its size to zero + for (S32 j = 0; j <= mip_levels; j++) + { + gGL.getTexUnit(0)->bindManual(type, textures[i]); + + glTexImage2D(LLTexUnit::getInternalType(type), j, format, 0, 0, 0, GL_RGBA, GL_UNSIGNED_BYTE, NULL); + } - if (immediate) + llassert(std::find(sDeadTextureList[type][format].begin(), + sDeadTextureList[type][format].end(), textures[i]) == + sDeadTextureList[type][format].end()); + + sDeadTextureList[type][format].push_back(textures[i]); + } + } + } + + /*if (immediate) { LLImageGL::deleteDeadTextures(); - } + }*/ } // static -void LLImageGL::setManualImage(U32 target, S32 miplevel, S32 intformat, S32 width, S32 height, U32 pixformat, U32 pixtype, const void *pixels) +void LLImageGL::setManualImage(U32 target, S32 miplevel, S32 intformat, S32 width, S32 height, U32 pixformat, U32 pixtype, const void *pixels, bool allow_compression) { bool use_scratch = false; U32* scratch = NULL; @@ -1173,6 +1178,36 @@ void LLImageGL::setManualImage(U32 target, S32 miplevel, S32 intformat, S32 widt } } + if (LLImageGL::sCompressTextures && allow_compression) + { + switch (intformat) + { + case GL_RGB: + case GL_RGB8: + intformat = GL_COMPRESSED_RGB; + break; + case GL_RGBA: + case GL_RGBA8: + intformat = GL_COMPRESSED_RGBA; + break; + case GL_LUMINANCE: + case GL_LUMINANCE8: + intformat = GL_COMPRESSED_LUMINANCE; + break; + case GL_LUMINANCE_ALPHA: + case GL_LUMINANCE8_ALPHA8: + intformat = GL_COMPRESSED_LUMINANCE_ALPHA; + break; + case GL_ALPHA: + case GL_ALPHA8: + intformat = GL_COMPRESSED_ALPHA; + break; + default: + llwarns << "Could not compress format: " << std::hex << intformat << llendl; + break; + } + } + stop_glerror(); glTexImage2D(target, miplevel, intformat, width, height, 0, pixformat, pixtype, use_scratch ? scratch : pixels); stop_glerror(); @@ -1201,10 +1236,11 @@ BOOL LLImageGL::createGLTexture() if(mTexName) { - glDeleteTextures(1, (reinterpret_cast<GLuint*>(&mTexName))) ; + LLImageGL::deleteTextures(mBindTarget, mFormatInternal, mMipLevels, 1, (reinterpret_cast<GLuint*>(&mTexName))) ; } - glGenTextures(1, (GLuint*)&mTexName); + + LLImageGL::generateTextures(mBindTarget, mFormatInternal, 1, &mTexName); stop_glerror(); if (!mTexName) { @@ -1232,8 +1268,7 @@ BOOL LLImageGL::createGLTexture(S32 discard_level, const LLImageRaw* imageraw, S llassert(mCurrentDiscardLevel >= 0); discard_level = mCurrentDiscardLevel; } - discard_level = llclamp(discard_level, 0, (S32)mMaxDiscardLevel); - + // Actual image width/height = raw image width/height * 2^discard_level S32 raw_w = imageraw->getWidth() ; S32 raw_h = imageraw->getHeight() ; @@ -1241,7 +1276,7 @@ BOOL LLImageGL::createGLTexture(S32 discard_level, const LLImageRaw* imageraw, S S32 h = raw_h << discard_level; // setSize may call destroyGLTexture if the size does not match - setSize(w, h, imageraw->getComponents()); + setSize(w, h, imageraw->getComponents(), discard_level); if( !mHasExplicitFormat ) { @@ -1284,7 +1319,7 @@ BOOL LLImageGL::createGLTexture(S32 discard_level, const LLImageRaw* imageraw, S return TRUE ; } - setCategory(category) ; + setCategory(category); const U8* rawdata = imageraw->getData(); return createGLTexture(discard_level, rawdata, FALSE, usename); } @@ -1317,7 +1352,7 @@ BOOL LLImageGL::createGLTexture(S32 discard_level, const U8* data_in, BOOL data_ } else { - LLImageGL::generateTextures(1, &mTexName); + LLImageGL::generateTextures(mBindTarget, mFormatInternal, 1, &mTexName); stop_glerror(); { llverify(gGL.getTexUnit(0)->bind(this)); @@ -1362,12 +1397,7 @@ BOOL LLImageGL::createGLTexture(S32 discard_level, const U8* data_in, BOOL data_ { sGlobalTextureMemoryInBytes -= mTextureMemory; - if(gAuditTexture) - { - decTextureCounter(mTextureMemory, mComponents, mCategory) ; - } - - LLImageGL::deleteTextures(1, &old_name); + LLImageGL::deleteTextures(mBindTarget, mFormatInternal, mMipLevels, 1, &old_name); stop_glerror(); } @@ -1376,10 +1406,6 @@ BOOL LLImageGL::createGLTexture(S32 discard_level, const U8* data_in, BOOL data_ sGlobalTextureMemoryInBytes += mTextureMemory; mTexelsInGLTexture = getWidth() * getHeight() ; - if(gAuditTexture) - { - incTextureCounter(mTextureMemory, mComponents, mCategory) ; - } // mark this as bound at this point, so we don't throw it out immediately mLastBindTime = sLastFrameTime; return TRUE; @@ -1500,7 +1526,7 @@ void LLImageGL::deleteDeadTextures() { bool reset = false; - while (!sDeadTextureList.empty()) + /*while (!sDeadTextureList.empty()) { GLuint tex = sDeadTextureList.front(); sDeadTextureList.pop_front(); @@ -1522,7 +1548,7 @@ void LLImageGL::deleteDeadTextures() glDeleteTextures(1, &tex); stop_glerror(); - } + }*/ if (reset) { @@ -1536,22 +1562,29 @@ void LLImageGL::destroyGLTexture() { if(mTextureMemory) { - if(gAuditTexture) - { - decTextureCounter(mTextureMemory, mComponents, mCategory) ; - } sGlobalTextureMemoryInBytes -= mTextureMemory; mTextureMemory = 0; } - LLImageGL::deleteTextures(1, &mTexName); - mTexName = 0; + LLImageGL::deleteTextures(mBindTarget, mFormatInternal, mMipLevels, 1, &mTexName); mCurrentDiscardLevel = -1 ; //invalidate mCurrentDiscardLevel. + mTexName = 0; mGLTextureCreated = FALSE ; - } + } } - +//force to invalidate the gl texture, most likely a sculpty texture +void LLImageGL::forceToInvalidateGLTexture() +{ + if (mTexName != 0) + { + destroyGLTexture(); + } + else + { + mCurrentDiscardLevel = -1 ; //invalidate mCurrentDiscardLevel. + } +} //---------------------------------------------------------------------------- @@ -1969,70 +2002,6 @@ BOOL LLImageGL::getMask(const LLVector2 &tc) return res; } -void LLImageGL::setCategory(S32 category) -{ -#if 0 //turn this off temporarily because it is not in use now. - if(!gAuditTexture) - { - return ; - } - if(mCategory != category) - { - if(mCategory > -1) - { - sTextureMemByCategory[mCategory] -= mTextureMemory ; - } - if(category > -1 && category < sMaxCatagories) - { - sTextureMemByCategory[category] += mTextureMemory ; - mCategory = category; - } - else - { - mCategory = -1 ; - } - } -#endif -} - -//for debug use -//val is a "power of two" number -S32 LLImageGL::getTextureCounterIndex(U32 val) -{ - //index range is [0, MAX_TEXTURE_LOG_SIZE]. - if(val < 2) - { - return 0 ; - } - else if(val >= (1 << MAX_TEXTURE_LOG_SIZE)) - { - return MAX_TEXTURE_LOG_SIZE ; - } - else - { - S32 ret = 0 ; - while(val >>= 1) - { - ++ret; - } - return ret ; - } -} - -//static -void LLImageGL::incTextureCounter(U32 val, S32 ncomponents, S32 category) -{ - sTextureLoadedCounter[getTextureCounterIndex(val)]++ ; - sTextureMemByCategory[category] += (S32)val * ncomponents ; -} - -//static -void LLImageGL::decTextureCounter(U32 val, S32 ncomponents, S32 category) -{ - sTextureLoadedCounter[getTextureCounterIndex(val)]-- ; - sTextureMemByCategory[category] += (S32)val * ncomponents ; -} - void LLImageGL::setCurTexSizebar(S32 index, BOOL set_pick_size) { sCurTexSizeBar = index ; diff --git a/indra/llrender/llimagegl.h b/indra/llrender/llimagegl.h index 2cfb15b0d9..cf3c484c79 100644 --- a/indra/llrender/llimagegl.h +++ b/indra/llrender/llimagegl.h @@ -45,8 +45,16 @@ class LLImageGL : public LLRefCount { friend class LLTexUnit; public: - static std::list<U32> sDeadTextureList; + static U32 sCurTexName; + //previously used but now available texture names + // sDeadTextureList[<usage>][<internal format>] + typedef std::map<U32, std::list<U32> > dead_texturelist_t; + static dead_texturelist_t sDeadTextureList[LLTexUnit::TT_NONE]; + + // These 2 functions replace glGenTextures() and glDeleteTextures() + static void generateTextures(LLTexUnit::eTextureType type, U32 format, S32 numTextures, U32 *textures); + static void deleteTextures(LLTexUnit::eTextureType type, U32 format, S32 mip_levels, S32 numTextures, U32 *textures, bool immediate = false); static void deleteDeadTextures(); // Size calculation @@ -92,18 +100,15 @@ protected: public: virtual void dump(); // debugging info to llinfos - void setSize(S32 width, S32 height, S32 ncomponents); + void setSize(S32 width, S32 height, S32 ncomponents, S32 discard_level = -1); void setComponents(S32 ncomponents) { mComponents = (S8)ncomponents ;} + void setAllowCompression(bool allow) { mAllowCompression = allow; } - // These 3 functions currently wrap glGenTextures(), glDeleteTextures(), and glTexImage2D() - // for tracking purposes and will be deprecated in the future - static void generateTextures(S32 numTextures, U32 *textures); - static void deleteTextures(S32 numTextures, U32 *textures, bool immediate = false); - static void setManualImage(U32 target, S32 miplevel, S32 intformat, S32 width, S32 height, U32 pixformat, U32 pixtype, const void *pixels); + static void setManualImage(U32 target, S32 miplevel, S32 intformat, S32 width, S32 height, U32 pixformat, U32 pixtype, const void *pixels, bool allow_compression = true); BOOL createGLTexture() ; - BOOL createGLTexture(S32 discard_level, const LLImageRaw* imageraw, S32 usename = 0, BOOL to_create = TRUE, - S32 category = sMaxCatagories - 1); + BOOL createGLTexture(S32 discard_level, const LLImageRaw* imageraw, S32 usename = 0, BOOL to_create = TRUE, + S32 category = sMaxCategories-1); BOOL createGLTexture(S32 discard_level, const U8* data, BOOL data_hasmips = FALSE, S32 usename = 0); void setImage(const LLImageRaw* imageraw); void setImage(const U8* data_in, BOOL data_hasmips = FALSE); @@ -114,6 +119,7 @@ public: // Read back a raw image for this discard level, if it exists BOOL readBackRaw(S32 discard_level, LLImageRaw* imageraw, bool compressed_ok) const; void destroyGLTexture(); + void forceToInvalidateGLTexture(); void setExplicitFormat(LLGLint internal_format, LLGLenum primary_format, LLGLenum type_format = 0, BOOL swap_bytes = FALSE); void setComponents(S8 ncomponents) { mComponents = ncomponents; } @@ -209,11 +215,14 @@ private: U32 mTexelsInAtlas ; U32 mTexelsInGLTexture; + bool mAllowCompression; + protected: LLGLenum mTarget; // Normally GL_TEXTURE2D, sometimes something else (ex. cube maps) LLTexUnit::eTextureType mBindTarget; // Normally TT_TEXTURE, sometimes something else (ex. cube maps) bool mHasMipMaps; - + S32 mMipLevels; + LLGLboolean mIsResident; S8 mComponents; @@ -234,8 +243,6 @@ public: static S32 sCount; static F32 sLastFrameTime; - - static LLGLuint sCurrentBoundTextures[MAX_GL_TEXTURE_UNITS]; // Currently bound texture ID // Global memory statistics static S32 sGlobalTextureMemoryInBytes; // Tracks main memory texmem @@ -246,7 +253,7 @@ public: static BOOL sGlobalUseAnisotropic; static LLImageGL* sDefaultGLTexture ; static BOOL sAutomatedTest; - + static bool sCompressTextures; //use GL texture compression #if DEBUG_MISS BOOL mMissed; // Missed on last bind? BOOL getMissed() const { return mMissed; }; @@ -257,9 +264,10 @@ public: public: static void initClass(S32 num_catagories) ; static void cleanupClass() ; -private: - static S32 sMaxCatagories ; +private: + static S32 sMaxCategories; + //the flag to allow to call readBackRaw(...). //can be removed if we do not use that function at all. static BOOL sAllowReadBackRaw ; @@ -269,39 +277,22 @@ private: //**************************************************************************************************** private: S32 mCategory ; -public: - void setCategory(S32 category) ; - S32 getCategory()const {return mCategory ;} - +public: + void setCategory(S32 category) {mCategory = category;} + S32 getCategory()const {return mCategory;} + //for debug use: show texture size distribution //---------------------------------------- - static LLPointer<LLImageGL> sHighlightTexturep; //default texture to replace normal textures - static std::vector<S32> sTextureLoadedCounter ; - static std::vector<S32> sTextureBoundCounter ; - static std::vector<S32> sTextureCurBoundCounter ; static S32 sCurTexSizeBar ; static S32 sCurTexPickSize ; - static void setHighlightTexture(S32 category) ; - static S32 getTextureCounterIndex(U32 val) ; - static void incTextureCounter(U32 val, S32 ncomponents, S32 category) ; - static void decTextureCounter(U32 val, S32 ncomponents, S32 category) ; static void setCurTexSizebar(S32 index, BOOL set_pick_size = TRUE) ; static void resetCurTexSizebar(); - //---------------------------------------- - //for debug use: show texture category distribution - //---------------------------------------- - - static std::vector<S32> sTextureMemByCategory; - static std::vector<S32> sTextureMemByCategoryBound ; - static std::vector<S32> sTextureCurMemByCategoryBound ; - //---------------------------------------- //**************************************************************************************************** //End of definitions for texture auditing use only //**************************************************************************************************** }; -extern BOOL gAuditTexture; #endif // LL_LLIMAGEGL_H diff --git a/indra/llrender/llrender.cpp b/indra/llrender/llrender.cpp index 812fa7024b..4597d06260 100644 --- a/indra/llrender/llrender.cpp +++ b/indra/llrender/llrender.cpp @@ -246,14 +246,6 @@ bool LLTexUnit::bind(LLTexture* texture, bool for_rendering, bool forceBind) } //in audit, replace the selected texture by the default one. - if(gAuditTexture && for_rendering && LLImageGL::sCurTexPickSize > 0) - { - if(texture->getWidth() * texture->getHeight() == LLImageGL::sCurTexPickSize) - { - gl_tex->updateBindStats(gl_tex->mTextureMemory); - return bind(LLImageGL::sHighlightTexturep.get()); - } - } if ((mCurrTexture != gl_tex->getTexName()) || forceBind) { activate(); @@ -416,12 +408,14 @@ void LLTexUnit::unbind(eTextureType type) if (mIndex < 0) return; + //always flush and activate for consistency + // some code paths assume unbind always flushes and sets the active texture + gGL.flush(); + activate(); + // Disabled caching of binding state. if (mCurrTexType == type) { - gGL.flush(); - - activate(); mCurrTexture = 0; if (LLGLSLShader::sNoFixedFunction && type == LLTexUnit::TT_TEXTURE) { @@ -472,11 +466,25 @@ void LLTexUnit::setTextureFilteringOption(LLTexUnit::eTextureFilterOptions optio } else if (option >= TFO_BILINEAR) { - glTexParameteri(sGLTextureType[mCurrTexType], GL_TEXTURE_MIN_FILTER, GL_LINEAR); + if (mHasMipMaps) + { + glTexParameteri(sGLTextureType[mCurrTexType], GL_TEXTURE_MIN_FILTER, GL_LINEAR_MIPMAP_NEAREST); + } + else + { + glTexParameteri(sGLTextureType[mCurrTexType], GL_TEXTURE_MIN_FILTER, GL_LINEAR); + } } else { - glTexParameteri(sGLTextureType[mCurrTexType], GL_TEXTURE_MIN_FILTER, GL_NEAREST); + if (mHasMipMaps) + { + glTexParameteri(sGLTextureType[mCurrTexType], GL_TEXTURE_MIN_FILTER, GL_NEAREST_MIPMAP_NEAREST); + } + else + { + glTexParameteri(sGLTextureType[mCurrTexType], GL_TEXTURE_MIN_FILTER, GL_NEAREST); + } } if (gGLManager.mHasAnisotropic) @@ -640,7 +648,7 @@ void LLTexUnit::setTextureCombiner(eTextureBlendOp op, eTextureBlendSrc src1, eT gGL.flush(); glTexEnvi(GL_TEXTURE_ENV, GL_TEXTURE_ENV_MODE, GL_COMBINE_ARB); } - + // We want an early out, because this function does a LOT of stuff. if ( ( (isAlpha && (mCurrAlphaOp == op) && (mCurrAlphaSrc1 == src1) && (mCurrAlphaSrc2 == src2)) || (!isAlpha && (mCurrColorOp == op) && (mCurrColorSrc1 == src1) && (mCurrColorSrc2 == src2)) ) && !gGL.mDirty) @@ -997,7 +1005,7 @@ void LLLightState::setSpotDirection(const LLVector3& direction) const glh::matrix4f& mat = gGL.getModelviewMatrix(); mat.mult_matrix_dir(dir); - mSpotDirection.set(direction); + mSpotDirection.set(dir.v); } } @@ -1429,11 +1437,24 @@ void LLRender::matrixMode(U32 mode) mMatrixMode = mode; } +U32 LLRender::getMatrixMode() +{ + if (mMatrixMode >= MM_TEXTURE0 && mMatrixMode <= MM_TEXTURE3) + { //always return MM_TEXTURE if current matrix mode points at any texture matrix + return MM_TEXTURE; + } + + return mMatrixMode; +} + + void LLRender::loadIdentity() { flush(); { + llassert_always(mMatrixMode < NUM_MATRIX_MODES) ; + mMatrix[mMatrixMode][mMatIdx[mMatrixMode]].make_identity(); mMatHash[mMatrixMode]++; } @@ -1444,6 +1465,11 @@ const glh::matrix4f& LLRender::getModelviewMatrix() return mMatrix[MM_MODELVIEW][mMatIdx[MM_MODELVIEW]]; } +const glh::matrix4f& LLRender::getProjectionMatrix() +{ + return mMatrix[MM_PROJECTION][mMatIdx[MM_PROJECTION]]; +} + void LLRender::translateUI(F32 x, F32 y, F32 z) { if (mUIOffset.empty()) diff --git a/indra/llrender/llrender.h b/indra/llrender/llrender.h index 7581b9f908..78a310e525 100644 --- a/indra/llrender/llrender.h +++ b/indra/llrender/llrender.h @@ -346,8 +346,10 @@ public: void loadIdentity(); void multMatrix(const GLfloat* m); void matrixMode(U32 mode); + U32 getMatrixMode(); const glh::matrix4f& getModelviewMatrix(); + const glh::matrix4f& getProjectionMatrix(); void syncMatrices(); void syncLightState(); diff --git a/indra/llrender/llrendernavprim.cpp b/indra/llrender/llrendernavprim.cpp new file mode 100644 index 0000000000..ca72964832 --- /dev/null +++ b/indra/llrender/llrendernavprim.cpp @@ -0,0 +1,59 @@ +/** +* @file llrendernavprim.cpp +* @brief Implementation of llrendernavprim +* @author Prep@lindenlab.com +* +* $LicenseInfo:firstyear=2012&license=viewerlgpl$ +* Second Life Viewer Source Code +* Copyright (C) 2012, Linden Research, Inc. +* +* This library is free software; you can redistribute it and/or +* modify it under the terms of the GNU Lesser General Public +* License as published by the Free Software Foundation; +* version 2.1 of the License only. +* +* This library is distributed in the hope that it will be useful, +* but WITHOUT ANY WARRANTY; without even the implied warranty of +* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU +* Lesser General Public License for more details. +* +* You should have received a copy of the GNU Lesser General Public +* License along with this library; if not, write to the Free Software +* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA +* +* Linden Research, Inc., 945 Battery Street, San Francisco, CA 94111 USA +* $/LicenseInfo$ +*/ + + +#include "linden_common.h" + +#include "llrendernavprim.h" + +#include "llrender.h" +#include "llvertexbuffer.h" +#include "v4coloru.h" +#include "v3math.h" + +//============================================================================= +LLRenderNavPrim gRenderNav; +//============================================================================= +void LLRenderNavPrim::renderLLTri( const LLVector3& a, const LLVector3& b, const LLVector3& c, const LLColor4U& color ) const +{ + LLColor4 cV(color); + gGL.color4fv( cV.mV ); + gGL.begin(LLRender::TRIANGLES); + { + gGL.vertex3fv( a.mV ); + gGL.vertex3fv( b.mV ); + gGL.vertex3fv( c.mV ); + } + gGL.end(); +} +//============================================================================= +void LLRenderNavPrim::renderNavMeshVB( U32 mode, LLVertexBuffer* pVBO, int vertCnt ) +{ + pVBO->setBuffer( LLVertexBuffer::MAP_VERTEX | LLVertexBuffer::MAP_COLOR | LLVertexBuffer::MAP_NORMAL ); + pVBO->drawArrays( mode, 0, vertCnt ); +} +//============================================================================= diff --git a/indra/llrender/llrendernavprim.h b/indra/llrender/llrendernavprim.h new file mode 100644 index 0000000000..a3a5dfec3a --- /dev/null +++ b/indra/llrender/llrendernavprim.h @@ -0,0 +1,49 @@ +/** +* @file llrendernavprim.h +* @brief Header file for llrendernavprim +* @author Prep@lindenlab.com +* +* $LicenseInfo:firstyear=2012&license=viewerlgpl$ +* Second Life Viewer Source Code +* Copyright (C) 2012, Linden Research, Inc. +* +* This library is free software; you can redistribute it and/or +* modify it under the terms of the GNU Lesser General Public +* License as published by the Free Software Foundation; +* version 2.1 of the License only. +* +* This library is distributed in the hope that it will be useful, +* but WITHOUT ANY WARRANTY; without even the implied warranty of +* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU +* Lesser General Public License for more details. +* +* You should have received a copy of the GNU Lesser General Public +* License along with this library; if not, write to the Free Software +* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA +* +* Linden Research, Inc., 945 Battery Street, San Francisco, CA 94111 USA +* $/LicenseInfo$ +*/ +#ifndef LL_LLRENDERNAVPRIM_H +#define LL_LLRENDERNAVPRIM_H + +#include "stdtypes.h" + +class LLColor4U; +class LLVector3; +class LLVertexBuffer; + + +class LLRenderNavPrim +{ +public: + //Draw simple tri + void renderLLTri( const LLVector3& a, const LLVector3& b, const LLVector3& c, const LLColor4U& color ) const; + //Draw the contents of vertex buffer + void renderNavMeshVB( U32 mode, LLVertexBuffer* pVBO, int vertCnt ); +private: +}; + +extern LLRenderNavPrim gRenderNav; + +#endif // LL_LLRENDERNAVPRIM_H diff --git a/indra/llrender/llrendertarget.cpp b/indra/llrender/llrendertarget.cpp index 1aa12614ea..c1b96a43da 100644 --- a/indra/llrender/llrendertarget.cpp +++ b/indra/llrender/llrendertarget.cpp @@ -51,12 +51,13 @@ void check_framebuffer_status() } bool LLRenderTarget::sUseFBO = false; +U32 LLRenderTarget::sCurFBO = 0; LLRenderTarget::LLRenderTarget() : mResX(0), mResY(0), - mTex(0), mFBO(0), + mPreviousFBO(0), mDepth(0), mStencil(0), mUseDepth(false), @@ -70,8 +71,47 @@ LLRenderTarget::~LLRenderTarget() release(); } +void LLRenderTarget::resize(U32 resx, U32 resy, U32 color_fmt) +{ + //for accounting, get the number of pixels added/subtracted + S32 pix_diff = (resx*resy)-(mResX*mResY); + + mResX = resx; + mResY = resy; + + for (U32 i = 0; i < mTex.size(); ++i) + { //resize color attachments + gGL.getTexUnit(0)->bindManual(mUsage, mTex[i]); + LLImageGL::setManualImage(LLTexUnit::getInternalType(mUsage), 0, color_fmt, mResX, mResY, GL_RGBA, GL_UNSIGNED_BYTE, NULL, false); + sBytesAllocated += pix_diff*4; + } + + if (mDepth) + { //resize depth attachment + if (mStencil) + { + //use render buffers where stencil buffers are in play + glBindRenderbuffer(GL_RENDERBUFFER, mDepth); + glRenderbufferStorage(GL_RENDERBUFFER, GL_DEPTH24_STENCIL8, mResX, mResY); + glBindRenderbuffer(GL_RENDERBUFFER, 0); + } + else + { + gGL.getTexUnit(0)->bindManual(mUsage, mDepth); + U32 internal_type = LLTexUnit::getInternalType(mUsage); + LLImageGL::setManualImage(internal_type, 0, GL_DEPTH_COMPONENT24, mResX, mResY, GL_DEPTH_COMPONENT, GL_UNSIGNED_INT, NULL, false); + } + + sBytesAllocated += pix_diff*4; + } +} + + bool LLRenderTarget::allocate(U32 resx, U32 resy, U32 color_fmt, bool depth, bool stencil, LLTexUnit::eTextureType usage, bool use_fbo, S32 samples) { + resx = llmin(resx, (U32) 4096); + resy = llmin(resy, (U32) 4096); + stop_glerror(); release(); stop_glerror(); @@ -111,7 +151,7 @@ bool LLRenderTarget::allocate(U32 resx, U32 resy, U32 color_fmt, bool depth, boo glFramebufferTexture2D(GL_FRAMEBUFFER, GL_DEPTH_ATTACHMENT, LLTexUnit::getInternalType(mUsage), mDepth, 0); stop_glerror(); } - glBindFramebuffer(GL_FRAMEBUFFER, 0); + glBindFramebuffer(GL_FRAMEBUFFER, sCurFBO); } stop_glerror(); @@ -128,14 +168,23 @@ bool LLRenderTarget::addColorAttachment(U32 color_fmt) } U32 offset = mTex.size(); - if (offset >= 4 || - (offset > 0 && (mFBO == 0 || !gGLManager.mHasDrawBuffers))) + + if( offset >= 4 ) + { + llwarns << "Too many color attachments" << llendl; + llassert( offset < 4 ); + return false; + } + if( offset > 0 && (mFBO == 0 || !gGLManager.mHasDrawBuffers) ) { - llerrs << "Too many color attachments!" << llendl; + llwarns << "FBO not used or no drawbuffers available; mFBO=" << (U32)mFBO << " gGLManager.mHasDrawBuffers=" << (U32)gGLManager.mHasDrawBuffers << llendl; + llassert( mFBO != 0 ); + llassert( gGLManager.mHasDrawBuffers ); + return false; } U32 tex; - LLImageGL::generateTextures(1, &tex); + LLImageGL::generateTextures(mUsage, color_fmt, 1, &tex); gGL.getTexUnit(0)->bindManual(mUsage, tex); stop_glerror(); @@ -143,7 +192,7 @@ bool LLRenderTarget::addColorAttachment(U32 color_fmt) { clear_glerror(); - LLImageGL::setManualImage(LLTexUnit::getInternalType(mUsage), 0, color_fmt, mResX, mResY, GL_RGBA, GL_UNSIGNED_BYTE, NULL); + LLImageGL::setManualImage(LLTexUnit::getInternalType(mUsage), 0, color_fmt, mResX, mResY, GL_RGBA, GL_UNSIGNED_BYTE, NULL, false); if (glGetError() != GL_NO_ERROR) { llwarns << "Could not allocate color buffer for render target." << llendl; @@ -189,10 +238,11 @@ bool LLRenderTarget::addColorAttachment(U32 color_fmt) check_framebuffer_status(); - glBindFramebuffer(GL_FRAMEBUFFER, 0); + glBindFramebuffer(GL_FRAMEBUFFER, sCurFBO); } mTex.push_back(tex); + mInternalFormat.push_back(color_fmt); if (gDebugGL) { //bind and unbind to validate target @@ -217,13 +267,13 @@ bool LLRenderTarget::allocateDepth() } else { - LLImageGL::generateTextures(1, &mDepth); + LLImageGL::generateTextures(mUsage, GL_DEPTH_COMPONENT24, 1, &mDepth); gGL.getTexUnit(0)->bindManual(mUsage, mDepth); 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); + LLImageGL::setManualImage(internal_type, 0, GL_DEPTH_COMPONENT24, mResX, mResY, GL_DEPTH_COMPONENT, GL_UNSIGNED_INT, NULL, false); gGL.getTexUnit(0)->setTextureFilteringOption(LLTexUnit::TFO_POINT); } @@ -277,7 +327,7 @@ void LLRenderTarget::shareDepthBuffer(LLRenderTarget& target) check_framebuffer_status(); - glBindFramebuffer(GL_FRAMEBUFFER, 0); + glBindFramebuffer(GL_FRAMEBUFFER, sCurFBO); target.mUseDepth = true; } @@ -294,7 +344,7 @@ void LLRenderTarget::release() } else { - LLImageGL::deleteTextures(1, &mDepth, true); + LLImageGL::deleteTextures(mUsage, 0, 0, 1, &mDepth, true); stop_glerror(); } mDepth = 0; @@ -326,8 +376,9 @@ void LLRenderTarget::release() if (mTex.size() > 0) { sBytesAllocated -= mResX*mResY*4*mTex.size(); - LLImageGL::deleteTextures(mTex.size(), &mTex[0], true); + LLImageGL::deleteTextures(mUsage, mInternalFormat[0], 0, mTex.size(), &mTex[0], true); mTex.clear(); + mInternalFormat.clear(); } mResX = mResY = 0; @@ -339,9 +390,13 @@ void LLRenderTarget::bindTarget() { if (mFBO) { + mPreviousFBO = sCurFBO; + stop_glerror(); glBindFramebuffer(GL_FRAMEBUFFER, mFBO); + sCurFBO = mFBO; + stop_glerror(); if (gGLManager.mHasDrawBuffers) { //setup multiple render targets @@ -367,16 +422,6 @@ void LLRenderTarget::bindTarget() sBoundTarget = this; } -// static -void LLRenderTarget::unbindTarget() -{ - if (gGLManager.mHasFramebufferObject) - { - glBindFramebuffer(GL_FRAMEBUFFER, 0); - } - sBoundTarget = NULL; -} - void LLRenderTarget::clear(U32 mask_in) { U32 mask = GL_COLOR_BUFFER_BIT; @@ -442,7 +487,8 @@ void LLRenderTarget::flush(bool fetch_depth) else { stop_glerror(); - glBindFramebuffer(GL_FRAMEBUFFER, 0); + glBindFramebuffer(GL_FRAMEBUFFER, mPreviousFBO); + sCurFBO = mPreviousFBO; stop_glerror(); } } @@ -457,7 +503,8 @@ void LLRenderTarget::copyContents(LLRenderTarget& source, S32 srcX0, S32 srcY0, gGL.flush(); if (!source.mFBO || !mFBO) { - llerrs << "Cannot copy framebuffer contents for non FBO render targets." << llendl; + llwarns << "Cannot copy framebuffer contents for non FBO render targets." << llendl; + return; } @@ -471,7 +518,7 @@ void LLRenderTarget::copyContents(LLRenderTarget& source, S32 srcX0, S32 srcY0, stop_glerror(); glCopyTexSubImage2D(LLTexUnit::getInternalType(mUsage), 0, srcX0, srcY0, dstX0, dstY0, dstX1, dstY1); stop_glerror(); - glBindFramebuffer(GL_FRAMEBUFFER, 0); + glBindFramebuffer(GL_FRAMEBUFFER, sCurFBO); stop_glerror(); } else @@ -488,7 +535,7 @@ void LLRenderTarget::copyContents(LLRenderTarget& source, S32 srcX0, S32 srcY0, stop_glerror(); glBindFramebuffer(GL_DRAW_FRAMEBUFFER, 0); stop_glerror(); - glBindFramebuffer(GL_FRAMEBUFFER, 0); + glBindFramebuffer(GL_FRAMEBUFFER, sCurFBO); stop_glerror(); } } @@ -514,7 +561,7 @@ void LLRenderTarget::copyContentsToFramebuffer(LLRenderTarget& source, S32 srcX0 stop_glerror(); glBlitFramebuffer(srcX0, srcY0, srcX1, srcY1, dstX0, dstY0, dstX1, dstY1, mask, filter); stop_glerror(); - glBindFramebuffer(GL_FRAMEBUFFER, 0); + glBindFramebuffer(GL_FRAMEBUFFER, sCurFBO); stop_glerror(); } } diff --git a/indra/llrender/llrendertarget.h b/indra/llrender/llrendertarget.h index 2735ab21c5..cf15f66d31 100644 --- a/indra/llrender/llrendertarget.h +++ b/indra/llrender/llrendertarget.h @@ -57,14 +57,13 @@ */ -class LLMultisampleBuffer; - class LLRenderTarget { public: //whether or not to use FBO implementation static bool sUseFBO; static U32 sBytesAllocated; + static U32 sCurFBO; LLRenderTarget(); ~LLRenderTarget(); @@ -74,6 +73,12 @@ public: //multiple calls will release previously allocated resources bool allocate(U32 resx, U32 resy, U32 color_fmt, bool depth, bool stencil, LLTexUnit::eTextureType usage = LLTexUnit::TT_TEXTURE, bool use_fbo = false, S32 samples = 0); + //resize existing attachments to use new resolution and color format + // CAUTION: if the GL runs out of memory attempting to resize, this render target will be undefined + // DO NOT use for screen space buffers or for scratch space for an image that might be uploaded + // DO use for render targets that resize often and aren't likely to ruin someone's day if they break + void resize(U32 resx, U32 resy, U32 color_fmt); + //add color buffer attachment //limit of 4 color attachments per render target bool addColorAttachment(U32 color_fmt); @@ -92,9 +97,6 @@ public: //applies appropriate viewport void bindTarget(); - //unbind target for rendering - static void unbindTarget(); - //clear render targer, clears depth buffer if present, //uses scissor rect if in copy-to-texture mode void clear(U32 mask = 0xFFFFFFFF); @@ -142,7 +144,9 @@ protected: U32 mResX; U32 mResY; std::vector<U32> mTex; + std::vector<U32> mInternalFormat; U32 mFBO; + U32 mPreviousFBO; U32 mDepth; bool mStencil; bool mUseDepth; diff --git a/indra/llrender/llshadermgr.cpp b/indra/llrender/llshadermgr.cpp index eea768a3ea..b6a9a6b653 100644 --- a/indra/llrender/llshadermgr.cpp +++ b/indra/llrender/llshadermgr.cpp @@ -94,23 +94,29 @@ BOOL LLShaderMgr::attachShaderFeatures(LLGLSLShader * shader) } } - if (features->calculatesLighting) + if (features->calculatesLighting || features->atmosphericHelpers) { if (!shader->attachObject("windlight/atmosphericsHelpersV.glsl")) { return FALSE; } + } + if (features->calculatesLighting) + { if (features->isSpecular) { if (!shader->attachObject("lighting/lightFuncSpecularV.glsl")) { return FALSE; } - - if (!shader->attachObject("lighting/sumLightsSpecularV.glsl")) + + if (!features->isAlphaLighting) { - return FALSE; + if (!shader->attachObject("lighting/sumLightsSpecularV.glsl")) + { + return FALSE; + } } if (!shader->attachObject("lighting/lightSpecularV.glsl")) @@ -125,9 +131,12 @@ BOOL LLShaderMgr::attachShaderFeatures(LLGLSLShader * shader) return FALSE; } - if (!shader->attachObject("lighting/sumLightsV.glsl")) + if (!features->isAlphaLighting) { - return FALSE; + if (!shader->attachObject("lighting/sumLightsV.glsl")) + { + return FALSE; + } } if (!shader->attachObject("lighting/lightV.glsl")) @@ -296,7 +305,7 @@ BOOL LLShaderMgr::attachShaderFeatures(LLGLSLShader * shader) } shader->mFeatures.mIndexedTextureChannels = llmax(LLGLSLShader::sIndexedTextureChannels-1, 1); } - } + } } // NOTE order of shader object attaching is VERY IMPORTANT!!! @@ -566,34 +575,46 @@ GLhandleARB LLShaderMgr::loadShaderFile(const std::string& filename, S32 & shade GLcharARB* text[4096]; GLuint count = 0; - F32 version = gGLManager.mGLVersion; - -//hack to never use GLSL > 1.20 on OSX -#if LL_DARWIN - version = llmin(version, 2.9f); -#endif - - if (version < 2.1f) - { - text[count++] = strdup("#version 110\n"); - text[count++] = strdup("#define ATTRIBUTE attribute\n"); - text[count++] = strdup("#define VARYING varying\n"); - } - else if (version < 3.f) + S32 major_version = gGLManager.mGLSLVersionMajor; + S32 minor_version = gGLManager.mGLSLVersionMinor; + + if (major_version == 1 && minor_version < 30) { - //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"); + if (minor_version < 10) + { + //should NEVER get here -- if major version is 1 and minor version is less than 10, + // viewer should never attempt to use shaders, continuing will result in undefined behavior + llerrs << "Unsupported GLSL Version." << llendl; + } + + if (minor_version <= 19) + { + text[count++] = strdup("#version 110\n"); + text[count++] = strdup("#define ATTRIBUTE attribute\n"); + text[count++] = strdup("#define VARYING varying\n"); + text[count++] = strdup("#define VARYING_FLAT varying\n"); + } + else if (minor_version <= 29) + { + //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"); + text[count++] = strdup("#define VARYING_FLAT varying\n"); + } } else { - if (version < 4.f) + if (major_version < 4) { //set version to 1.30 text[count++] = strdup("#version 130\n"); + + //some implementations of GLSL 1.30 require integer precision be explicitly declared + text[count++] = strdup("precision mediump int;\n"); + text[count++] = strdup("precision highp float;\n"); } else { //set version to 400 @@ -609,16 +630,25 @@ GLhandleARB LLShaderMgr::loadShaderFile(const std::string& filename, S32 & shade { //"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"); + text[count++] = strdup("#define VARYING_FLAT flat out\n"); } else { text[count++] = strdup("#define VARYING in\n"); + text[count++] = strdup("#define VARYING_FLAT flat in\n"); } //backwards compatibility with legacy texture lookup syntax + text[count++] = strdup("#define texture2D texture\n"); text[count++] = strdup("#define textureCube texture\n"); text[count++] = strdup("#define texture2DLod textureLod\n"); text[count++] = strdup("#define shadow2D(a,b) vec2(texture(a,b))\n"); + + if (major_version > 1 || minor_version >= 40) + { //GLSL 1.40 replaces texture2DRect et al with texture + text[count++] = strdup("#define texture2DRect texture\n"); + text[count++] = strdup("#define shadow2DRect(a,b) vec2(texture(a,b))\n"); + } } //copy preprocessor definitions into buffer @@ -642,22 +672,24 @@ GLhandleARB LLShaderMgr::loadShaderFile(const std::string& filename, S32 & shade . uniform sampler2D texN; - VARYING float vary_texture_index; + VARYING_FLAT ivec4 vary_texture_index; + + vec4 ret = vec4(1,0,1,1); vec4 diffuseLookup(vec2 texcoord) { - switch (int(vary_texture_index+0.25)) + switch (vary_texture_index.r)) { - case 0: return texture2D(tex0, texcoord); - case 1: return texture2D(tex1, texcoord); - case 2: return texture2D(tex2, texcoord); + case 0: ret = texture2D(tex0, texcoord); break; + case 1: ret = texture2D(tex1, texcoord); break; + case 2: ret = texture2D(tex2, texcoord); break; . . . - case N: return texture2D(texN, texcoord); + case N: return texture2D(texN, texcoord); break; } - return vec4(0,0,0,0); + return ret; } */ @@ -670,7 +702,7 @@ GLhandleARB LLShaderMgr::loadShaderFile(const std::string& filename, S32 & shade if (texture_index_channels > 1) { - text[count++] = strdup("VARYING float vary_texture_index;\n"); + text[count++] = strdup("VARYING_FLAT int vary_texture_index;\n"); } text[count++] = strdup("vec4 diffuseLookup(vec2 texcoord)\n"); @@ -682,45 +714,41 @@ GLhandleARB LLShaderMgr::loadShaderFile(const std::string& filename, S32 & shade text[count++] = strdup("return texture2D(tex0, texcoord);\n"); text[count++] = strdup("}\n"); } - else if (gGLManager.mGLVersion >= 3.f) - { - text[count++] = strdup("\tswitch (int(vary_texture_index+0.25))\n"); - text[count++] = strdup("\t{\n"); - - //switch body - for (S32 i = 0; i < texture_index_channels; ++i) - { - std::string case_str = llformat("\t\tcase %d: return texture2D(tex%d, texcoord);\n", i, i); - text[count++] = strdup(case_str.c_str()); + else if (major_version > 1 || minor_version >= 30) + { //switches are supported in GLSL 1.30 and later + if (gGLManager.mIsNVIDIA) + { //switches are unreliable on some NVIDIA drivers + for (U32 i = 0; i < texture_index_channels; ++i) + { + std::string if_string = llformat("\t%sif (vary_texture_index == %d) { return texture2D(tex%d, texcoord); }\n", i > 0 ? "else " : "", i, i); + text[count++] = strdup(if_string.c_str()); + } + text[count++] = strdup("\treturn vec4(1,0,1,1);\n"); + text[count++] = strdup("}\n"); } + else + { + text[count++] = strdup("\tvec4 ret = vec4(1,0,1,1);\n"); + text[count++] = strdup("\tswitch (vary_texture_index)\n"); + text[count++] = strdup("\t{\n"); + + //switch body + for (S32 i = 0; i < texture_index_channels; ++i) + { + std::string case_str = llformat("\t\tcase %d: return texture2D(tex%d, texcoord);\n", i, i); + text[count++] = strdup(case_str.c_str()); + } - text[count++] = strdup("\t}\n"); - text[count++] = strdup("\treturn vec4(1,0,1,1);\n"); - text[count++] = strdup("}\n"); + text[count++] = strdup("\t}\n"); + text[count++] = strdup("\treturn ret;\n"); + text[count++] = strdup("}\n"); + } } else - { - //switches aren't supported, make block that looks like: - /* - int ti = int(vary_texture_index+0.25); - if (ti == 0) return texture2D(tex0, texcoord); - if (ti == 1) return texture2D(tex1, texcoord); - . - . - . - if (ti == N) return texture2D(texN, texcoord); - */ - - text[count++] = strdup("int ti = int(vary_texture_index+0.25);\n"); - for (S32 i = 0; i < texture_index_channels; ++i) - { - std::string if_str = llformat("if (ti == %d) return texture2D(tex%d, texcoord);\n", i, i); - text[count++] = strdup(if_str.c_str()); - } - - text[count++] = strdup("\treturn vec4(1,0,1,1);\n"); - text[count++] = strdup("}\n"); - } + { //should never get here. Indexed texture rendering requires GLSL 1.30 or later + // (for passing integers between vertex and fragment shaders) + llerrs << "Indexed texture rendering requires GLSL 1.30 or later." << llendl; + } } //copy file into memory @@ -845,28 +873,42 @@ BOOL LLShaderMgr::linkProgramObject(GLhandleARB obj, BOOL suppress_errors) LL_WARNS("ShaderLoading") << "GLSL Linker Error:" << LL_ENDL; } -// NOTE: Removing LL_DARWIN block as it doesn't seem to actually give the correct answer, -// but want it for reference once I move it. -#if 0 - // Force an evaluation of the gl state so the driver can tell if the shader will run in hardware or software - // per Apple's suggestion - glBegin(gGL.mMode); - glEnd(); - - // Query whether the shader can or cannot run in hardware - // http://developer.apple.com/qa/qa2007/qa1502.html - long vertexGPUProcessing; - CGLContextObj ctx = CGLGetCurrentContext(); - CGLGetParameter (ctx, kCGLCPGPUVertexProcessing, &vertexGPUProcessing); - long fragmentGPUProcessing; - CGLGetParameter (ctx, kCGLCPGPUFragmentProcessing, &fragmentGPUProcessing); - if (!fragmentGPUProcessing || !vertexGPUProcessing) +#if LL_DARWIN + + // For some reason this absolutely kills the frame rate when VBO's are enabled + if (0) { - LL_WARNS("ShaderLoading") << "GLSL Linker: Running in Software:" << LL_ENDL; - success = GL_FALSE; - suppress_errors = FALSE; + // Force an evaluation of the gl state so the driver can tell if the shader will run in hardware or software + // per Apple's suggestion + LLGLSLShader::sNoFixedFunction = false; + + glUseProgramObjectARB(obj); + + gGL.begin(LLRender::TRIANGLES); + gGL.vertex3f(0.0f, 0.0f, 0.0f); + gGL.vertex3f(0.0f, 0.0f, 0.0f); + gGL.vertex3f(0.0f, 0.0f, 0.0f); + gGL.end(); + gGL.flush(); + + glUseProgramObjectARB(0); + + LLGLSLShader::sNoFixedFunction = true; + + // Query whether the shader can or cannot run in hardware + // http://developer.apple.com/qa/qa2007/qa1502.html + GLint vertexGPUProcessing, fragmentGPUProcessing; + CGLContextObj ctx = CGLGetCurrentContext(); + CGLGetParameter(ctx, kCGLCPGPUVertexProcessing, &vertexGPUProcessing); + CGLGetParameter(ctx, kCGLCPGPUFragmentProcessing, &fragmentGPUProcessing); + if (!fragmentGPUProcessing || !vertexGPUProcessing) + { + LL_WARNS("ShaderLoading") << "GLSL Linker: Running in Software:" << LL_ENDL; + success = GL_FALSE; + suppress_errors = FALSE; + } } - + #else std::string log = get_object_log(obj); LLStringUtil::toLower(log); @@ -997,6 +1039,9 @@ void LLShaderMgr::initAttribsAndUniforms() mReservedUniforms.push_back("size"); mReservedUniforms.push_back("falloff"); + mReservedUniforms.push_back("box_center"); + mReservedUniforms.push_back("box_size"); + mReservedUniforms.push_back("minLuminance"); mReservedUniforms.push_back("maxExtractAlpha"); @@ -1033,8 +1078,9 @@ void LLShaderMgr::initAttribsAndUniforms() mReservedUniforms.push_back("proj_shadow_res"); mReservedUniforms.push_back("depth_cutoff"); mReservedUniforms.push_back("norm_cutoff"); + mReservedUniforms.push_back("shadow_target_width"); - llassert(mReservedUniforms.size() == LLShaderMgr::DEFERRED_NORM_CUTOFF+1); + llassert(mReservedUniforms.size() == LLShaderMgr::DEFERRED_SHADOW_TARGET_WIDTH+1); mReservedUniforms.push_back("tc_scale"); mReservedUniforms.push_back("rcp_screen_res"); @@ -1047,6 +1093,8 @@ void LLShaderMgr::initAttribsAndUniforms() mReservedUniforms.push_back("magnification"); mReservedUniforms.push_back("max_cof"); mReservedUniforms.push_back("res_scale"); + mReservedUniforms.push_back("dof_width"); + mReservedUniforms.push_back("dof_height"); mReservedUniforms.push_back("depthMap"); mReservedUniforms.push_back("shadowMap0"); diff --git a/indra/llrender/llshadermgr.h b/indra/llrender/llshadermgr.h index 950e6c9c2f..7a16b7c20f 100644 --- a/indra/llrender/llshadermgr.h +++ b/indra/llrender/llshadermgr.h @@ -97,6 +97,8 @@ public: LIGHT_CENTER, LIGHT_SIZE, LIGHT_FALLOFF, + BOX_CENTER, + BOX_SIZE, GLOW_MIN_LUMINANCE, GLOW_MAX_EXTRACT_ALPHA, @@ -130,6 +132,7 @@ public: DEFERRED_PROJ_SHADOW_RES, DEFERRED_DEPTH_CUTOFF, DEFERRED_NORM_CUTOFF, + DEFERRED_SHADOW_TARGET_WIDTH, FXAA_TC_SCALE, FXAA_RCP_SCREEN_RES, @@ -142,6 +145,8 @@ public: DOF_MAGNIFICATION, DOF_MAX_COF, DOF_RES_SCALE, + DOF_WIDTH, + DOF_HEIGHT, DEFERRED_DEPTH, DEFERRED_SHADOW0, diff --git a/indra/llrender/llvertexbuffer.cpp b/indra/llrender/llvertexbuffer.cpp index 3948145580..dfbd8cd4ee 100644 --- a/indra/llrender/llvertexbuffer.cpp +++ b/indra/llrender/llvertexbuffer.cpp @@ -31,7 +31,6 @@ #include "llvertexbuffer.h" // #include "llrender.h" #include "llglheaders.h" -#include "llmemtype.h" #include "llrender.h" #include "llvector4a.h" #include "llshadermgr.h" @@ -49,138 +48,198 @@ U32 nhpo2(U32 v) return r; } +//which power of 2 is i? +//assumes i is a power of 2 > 0 +U32 wpo2(U32 i) +{ + llassert(i > 0); + llassert(nhpo2(i) == i); + + U32 r = 0; + + while (i >>= 1) ++r; + + return r; +} + + +const U32 LL_VBO_BLOCK_SIZE = 2048; +const U32 LL_VBO_POOL_MAX_SEED_SIZE = 256*1024; + +U32 vbo_block_size(U32 size) +{ //what block size will fit size? + U32 mod = size % LL_VBO_BLOCK_SIZE; + return mod == 0 ? size : size + (LL_VBO_BLOCK_SIZE-mod); +} + +U32 vbo_block_index(U32 size) +{ + return vbo_block_size(size)/LL_VBO_BLOCK_SIZE; +} + +const U32 LL_VBO_POOL_SEED_COUNT = vbo_block_index(LL_VBO_POOL_MAX_SEED_SIZE); + //============================================================================ //static -LLVBOPool LLVertexBuffer::sStreamVBOPool; -LLVBOPool LLVertexBuffer::sDynamicVBOPool; -LLVBOPool LLVertexBuffer::sStreamIBOPool; -LLVBOPool LLVertexBuffer::sDynamicIBOPool; +LLVBOPool LLVertexBuffer::sStreamVBOPool(GL_STREAM_DRAW_ARB, GL_ARRAY_BUFFER_ARB); +LLVBOPool LLVertexBuffer::sDynamicVBOPool(GL_DYNAMIC_DRAW_ARB, GL_ARRAY_BUFFER_ARB); +LLVBOPool LLVertexBuffer::sStreamIBOPool(GL_STREAM_DRAW_ARB, GL_ELEMENT_ARRAY_BUFFER_ARB); +LLVBOPool LLVertexBuffer::sDynamicIBOPool(GL_DYNAMIC_DRAW_ARB, GL_ELEMENT_ARRAY_BUFFER_ARB); + U32 LLVBOPool::sBytesPooled = 0; +U32 LLVBOPool::sIndexBytesPooled = 0; +U32 LLVBOPool::sCurGLName = 1; -LLPrivateMemoryPool* LLVertexBuffer::sPrivatePoolp = NULL ; +std::list<U32> LLVertexBuffer::sAvailableVAOName; +U32 LLVertexBuffer::sCurVAOName = 1; + +U32 LLVertexBuffer::sAllocatedIndexBytes = 0; +U32 LLVertexBuffer::sIndexCount = 0; + +LLPrivateMemoryPool* LLVertexBuffer::sPrivatePoolp = NULL; U32 LLVertexBuffer::sBindCount = 0; U32 LLVertexBuffer::sSetCount = 0; S32 LLVertexBuffer::sCount = 0; S32 LLVertexBuffer::sGLCount = 0; S32 LLVertexBuffer::sMappedCount = 0; -BOOL LLVertexBuffer::sDisableVBOMapping = FALSE ; -BOOL LLVertexBuffer::sEnableVBOs = TRUE; +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; -BOOL LLVertexBuffer::sIBOActive = FALSE; +bool LLVertexBuffer::sVBOActive = false; +bool LLVertexBuffer::sIBOActive = false; U32 LLVertexBuffer::sAllocatedBytes = 0; -BOOL LLVertexBuffer::sMapped = FALSE; -BOOL LLVertexBuffer::sUseStreamDraw = TRUE; -BOOL LLVertexBuffer::sUseVAO = FALSE; -BOOL LLVertexBuffer::sPreferStreamDraw = FALSE; +U32 LLVertexBuffer::sVertexCount = 0; +bool LLVertexBuffer::sMapped = false; +bool LLVertexBuffer::sUseStreamDraw = true; +bool LLVertexBuffer::sUseVAO = false; +bool LLVertexBuffer::sPreferStreamDraw = false; -const U32 FENCE_WAIT_TIME_NANOSECONDS = 10000; //1 ms -class LLGLSyncFence : public LLGLFence +U32 LLVBOPool::genBuffer() { -public: -#ifdef GL_ARB_sync - GLsync mSync; -#endif - - LLGLSyncFence() - { -#ifdef GL_ARB_sync - mSync = 0; -#endif - } + U32 ret = 0; - ~LLGLSyncFence() + if (mGLNamePool.empty()) { -#ifdef GL_ARB_sync - if (mSync) - { - glDeleteSync(mSync); - } -#endif + ret = sCurGLName++; } - - void placeFence() + else { -#ifdef GL_ARB_sync - if (mSync) - { - glDeleteSync(mSync); - } - mSync = glFenceSync(GL_SYNC_GPU_COMMANDS_COMPLETE, 0); -#endif + ret = mGLNamePool.front(); + mGLNamePool.pop_front(); } - void wait() - { -#ifdef GL_ARB_sync - if (mSync) - { - while (glClientWaitSync(mSync, 0, FENCE_WAIT_TIME_NANOSECONDS) == GL_TIMEOUT_EXPIRED) - { //track the number of times we've waited here - static S32 waits = 0; - waits++; - } - } -#endif - } + return ret; +} +void LLVBOPool::deleteBuffer(U32 name) +{ + if (gGLManager.mInited) + { + LLVertexBuffer::unbind(); -}; + glBindBufferARB(mType, name); + glBufferDataARB(mType, 0, NULL, mUsage); + llassert(std::find(mGLNamePool.begin(), mGLNamePool.end(), name) == mGLNamePool.end()); -//which power of 2 is i? -//assumes i is a power of 2 > 0 -U32 wpo2(U32 i) -{ - llassert(i > 0); - llassert(nhpo2(i) == i); + mGLNamePool.push_back(name); - U32 r = 0; + glBindBufferARB(mType, 0); + } +} - while (i >>= 1) ++r; - return r; +LLVBOPool::LLVBOPool(U32 vboUsage, U32 vboType) +: mUsage(vboUsage), mType(vboType) +{ + mMissCount.resize(LL_VBO_POOL_SEED_COUNT); + std::fill(mMissCount.begin(), mMissCount.end(), 0); } -U8* LLVBOPool::allocate(U32& name, U32 size) +volatile U8* LLVBOPool::allocate(U32& name, U32 size, bool for_seed) { - llassert(nhpo2(size) == size); + llassert(vbo_block_size(size) == size); + + volatile U8* ret = NULL; - U32 i = wpo2(size); + U32 i = vbo_block_index(size); if (mFreeList.size() <= i) { mFreeList.resize(i+1); } - U8* ret = NULL; - - if (mFreeList[i].empty()) + if (mFreeList[i].empty() || for_seed) { //make a new buffer - glGenBuffersARB(1, &name); + name = genBuffer(); + glBindBufferARB(mType, name); - glBufferDataARB(mType, size, 0, mUsage); - LLVertexBuffer::sAllocatedBytes += size; - if (LLVertexBuffer::sDisableVBOMapping) + if (!for_seed && i < LL_VBO_POOL_SEED_COUNT) + { //record this miss + mMissCount[i]++; + } + + if (mType == GL_ARRAY_BUFFER_ARB) { + LLVertexBuffer::sAllocatedBytes += size; + } + else + { + LLVertexBuffer::sAllocatedIndexBytes += size; + } + + if (LLVertexBuffer::sDisableVBOMapping || mUsage != GL_DYNAMIC_DRAW_ARB) + { + glBufferDataARB(mType, size, 0, mUsage); ret = (U8*) ll_aligned_malloc_16(size); } + else + { //always use a true hint of static draw when allocating non-client-backed buffers + glBufferDataARB(mType, size, 0, GL_STATIC_DRAW_ARB); + } + glBindBufferARB(mType, 0); + + if (for_seed) + { //put into pool for future use + llassert(mFreeList.size() > i); + + Record rec; + rec.mGLName = name; + rec.mClientData = ret; + + if (mType == GL_ARRAY_BUFFER_ARB) + { + sBytesPooled += size; + } + else + { + sIndexBytesPooled += size; + } + mFreeList[i].push_back(rec); + } } else { name = mFreeList[i].front().mGLName; ret = mFreeList[i].front().mClientData; - sBytesPooled -= size; + if (mType == GL_ARRAY_BUFFER_ARB) + { + sBytesPooled -= size; + } + else + { + sIndexBytesPooled -= size; + } mFreeList[i].pop_front(); } @@ -188,26 +247,52 @@ U8* LLVBOPool::allocate(U32& name, U32 size) return ret; } -void LLVBOPool::release(U32 name, U8* buffer, U32 size) +void LLVBOPool::release(U32 name, volatile U8* buffer, U32 size) { - llassert(nhpo2(size) == size); + llassert(vbo_block_size(size) == size); - U32 i = wpo2(size); + deleteBuffer(name); + ll_aligned_free_16((U8*) buffer); - llassert(mFreeList.size() > i); + if (mType == GL_ARRAY_BUFFER_ARB) + { + LLVertexBuffer::sAllocatedBytes -= size; + } + else + { + LLVertexBuffer::sAllocatedIndexBytes -= size; + } +} - Record rec; - rec.mGLName = name; - rec.mClientData = buffer; +void LLVBOPool::seedPool() +{ + U32 dummy_name = 0; - sBytesPooled += size; + if (mFreeList.size() < LL_VBO_POOL_SEED_COUNT) + { + mFreeList.resize(LL_VBO_POOL_SEED_COUNT); + } - mFreeList[i].push_back(rec); + for (U32 i = 0; i < LL_VBO_POOL_SEED_COUNT; i++) + { + if (mMissCount[i] > mFreeList[i].size()) + { + U32 size = i*LL_VBO_BLOCK_SIZE; + + S32 count = mMissCount[i] - mFreeList[i].size(); + for (U32 j = 0; j < count; ++j) + { + allocate(dummy_name, size, true); + } + } + } } + + void LLVBOPool::cleanup() { - U32 size = 1; + U32 size = LL_VBO_BLOCK_SIZE; for (U32 i = 0; i < mFreeList.size(); ++i) { @@ -217,21 +302,32 @@ void LLVBOPool::cleanup() { Record& r = l.front(); - glDeleteBuffersARB(1, &r.mGLName); - + deleteBuffer(r.mGLName); + if (r.mClientData) { - ll_aligned_free_16(r.mClientData); + ll_aligned_free_16((void*) r.mClientData); } l.pop_front(); - LLVertexBuffer::sAllocatedBytes -= size; - sBytesPooled -= size; + if (mType == GL_ARRAY_BUFFER_ARB) + { + sBytesPooled -= size; + LLVertexBuffer::sAllocatedBytes -= size; + } + else + { + sIndexBytesPooled -= size; + LLVertexBuffer::sAllocatedIndexBytes -= size; + } } - size *= 2; + size += LL_VBO_BLOCK_SIZE; } + + //reset miss counts + std::fill(mMissCount.begin(), mMissCount.end(), 0); } @@ -265,13 +361,54 @@ U32 LLVertexBuffer::sGLMode[LLRender::NUM_MODES] = GL_LINE_LOOP, }; +//static +U32 LLVertexBuffer::getVAOName() +{ + U32 ret = 0; + + if (!sAvailableVAOName.empty()) + { + ret = sAvailableVAOName.front(); + sAvailableVAOName.pop_front(); + } + else + { +#ifdef GL_ARB_vertex_array_object + glGenVertexArrays(1, &ret); +#endif + } + + return ret; +} + +//static +void LLVertexBuffer::releaseVAOName(U32 name) +{ + sAvailableVAOName.push_back(name); +} + + +//static +void LLVertexBuffer::seedPools() +{ + sStreamVBOPool.seedPool(); + sDynamicVBOPool.seedPool(); + sStreamIBOPool.seedPool(); + sDynamicIBOPool.seedPool(); +} //static void LLVertexBuffer::setupClientArrays(U32 data_mask) { if (sLastMask != data_mask) { - BOOL error = FALSE; + bool error = false; + + if (gGLManager.mGLSLVersionMajor < 2 && gGLManager.mGLSLVersionMinor < 30) + { + //make sure texture index is disabled + data_mask = data_mask & ~MAP_TEXTURE_INDEX; + } if (LLGLSLShader::sNoFixedFunction) { @@ -332,7 +469,7 @@ void LLVertexBuffer::setupClientArrays(U32 data_mask) { if (gDebugSession) { - error = TRUE; + error = true; gFailLog << "Bad client state! " << array[i] << " disabled." << std::endl; } else @@ -352,7 +489,7 @@ void LLVertexBuffer::setupClientArrays(U32 data_mask) { //needs to be disabled, make sure it was (DEBUG TEMPORARY) if (gDebugSession) { - error = TRUE; + error = true; gFailLog << "Bad client state! " << array[i] << " enabled." << std::endl; } else @@ -417,8 +554,21 @@ void LLVertexBuffer::drawArrays(U32 mode, const std::vector<LLVector3>& pos, con gGL.syncMatrices(); U32 count = pos.size(); - llassert_always(norm.size() >= pos.size()); - llassert_always(count > 0) ; + + llassert(norm.size() >= pos.size()); + llassert(count > 0); + + if( count == 0 ) + { + llwarns << "Called drawArrays with 0 vertices" << llendl; + return; + } + + if( norm.size() < pos.size() ) + { + llwarns << "Called drawArrays with #" << norm.size() << " normals and #" << pos.size() << " vertices" << llendl; + return; + } unbind(); @@ -536,7 +686,7 @@ void LLVertexBuffer::validateRange(U32 start, U32 end, U32 count, U32 indices_of void LLVertexBuffer::drawRange(U32 mode, U32 start, U32 end, U32 count, U32 indices_offset) const { validateRange(start, end, count, indices_offset); - + mMappable = false; gGL.syncMatrices(); llassert(mNumVerts >= 0); @@ -591,7 +741,7 @@ 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); - + mMappable = false; gGL.syncMatrices(); llassert(mNumIndices >= 0); @@ -637,7 +787,7 @@ 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); - + mMappable = false; gGL.syncMatrices(); llassert(mNumVerts >= 0); @@ -677,23 +827,13 @@ void LLVertexBuffer::drawArrays(U32 mode, U32 first, U32 count) const //static void LLVertexBuffer::initClass(bool use_vbo, bool no_vbo_mapping) { - sEnableVBOs = use_vbo && gGLManager.mHasVertexBufferObject ; - sDisableVBOMapping = sEnableVBOs && no_vbo_mapping ; + sEnableVBOs = use_vbo && gGLManager.mHasVertexBufferObject; + sDisableVBOMapping = sEnableVBOs && no_vbo_mapping; - if(!sPrivatePoolp) + if (!sPrivatePoolp) { - sPrivatePoolp = LLPrivateMemoryPoolManager::getInstance()->newPool(LLPrivateMemoryPool::STATIC) ; + sPrivatePoolp = LLPrivateMemoryPoolManager::getInstance()->newPool(LLPrivateMemoryPool::STATIC); } - - sStreamVBOPool.mType = GL_ARRAY_BUFFER_ARB; - sStreamVBOPool.mUsage= GL_STREAM_DRAW_ARB; - sStreamIBOPool.mType = GL_ELEMENT_ARRAY_BUFFER_ARB; - sStreamIBOPool.mUsage= GL_STREAM_DRAW_ARB; - - sDynamicVBOPool.mType = GL_ARRAY_BUFFER_ARB; - sDynamicVBOPool.mUsage= GL_DYNAMIC_DRAW_ARB; - sDynamicIBOPool.mType = GL_ELEMENT_ARRAY_BUFFER_ARB; - sDynamicIBOPool.mUsage= GL_DYNAMIC_DRAW_ARB; } //static @@ -706,18 +846,18 @@ void LLVertexBuffer::unbind() #endif sGLRenderArray = 0; sGLRenderIndices = 0; - sIBOActive = FALSE; + sIBOActive = false; } if (sVBOActive) { glBindBufferARB(GL_ARRAY_BUFFER_ARB, 0); - sVBOActive = FALSE; + sVBOActive = false; } if (sIBOActive) { glBindBufferARB(GL_ELEMENT_ARRAY_BUFFER_ARB, 0); - sIBOActive = FALSE; + sIBOActive = false; } sGLRenderBuffer = 0; @@ -729,7 +869,6 @@ void LLVertexBuffer::unbind() //static void LLVertexBuffer::cleanupClass() { - LLMemType mt2(LLMemType::MTYPE_VERTEX_CLEANUP_CLASS); unbind(); sStreamIBOPool.cleanup(); @@ -739,69 +878,85 @@ void LLVertexBuffer::cleanupClass() if(sPrivatePoolp) { - LLPrivateMemoryPoolManager::getInstance()->deletePool(sPrivatePoolp) ; - sPrivatePoolp = NULL ; + LLPrivateMemoryPoolManager::getInstance()->deletePool(sPrivatePoolp); + sPrivatePoolp = NULL; } } //---------------------------------------------------------------------------- -LLVertexBuffer::LLVertexBuffer(U32 typemask, S32 usage) : - LLRefCount(), - - mNumVerts(0), - mNumIndices(0), - mUsage(usage), - mGLBuffer(0), - mGLArray(0), - mGLIndices(0), - mMappedData(NULL), - mMappedIndexData(NULL), - mVertexLocked(FALSE), - mIndexLocked(FALSE), - mFinal(FALSE), - mEmpty(TRUE), - mFence(NULL) +S32 LLVertexBuffer::determineUsage(S32 usage) { - LLMemType mt2(LLMemType::MTYPE_VERTEX_CONSTRUCTOR); - mFence = NULL; + S32 ret_usage = usage; + if (!sEnableVBOs) { - mUsage = 0 ; + ret_usage = 0; } - - if (mUsage == GL_STREAM_DRAW_ARB && !sUseStreamDraw) + + if (ret_usage == GL_STREAM_DRAW_ARB && !sUseStreamDraw) { - mUsage = 0; + ret_usage = 0; } - if (mUsage == GL_DYNAMIC_DRAW_ARB && sPreferStreamDraw) + if (ret_usage == GL_DYNAMIC_DRAW_ARB && sPreferStreamDraw) { - mUsage = GL_STREAM_DRAW_ARB; + ret_usage = GL_STREAM_DRAW_ARB; } - - if (mUsage == 0 && LLRender::sGLCoreProfile) + + if (ret_usage == 0 && LLRender::sGLCoreProfile) { //MUST use VBOs for all rendering - mUsage = GL_STREAM_DRAW_ARB; + ret_usage = GL_STREAM_DRAW_ARB; } - - if (mUsage && mUsage != GL_STREAM_DRAW_ARB) + + if (ret_usage && ret_usage != GL_STREAM_DRAW_ARB) { //only stream_draw and dynamic_draw are supported when using VBOs, dynamic draw is the default - mUsage = GL_DYNAMIC_DRAW_ARB; + if (sDisableVBOMapping) + { //always use stream draw if VBO mapping is disabled + ret_usage = GL_STREAM_DRAW_ARB; + } + else + { + ret_usage = GL_DYNAMIC_DRAW_ARB; + } } - + + return ret_usage; +} + +LLVertexBuffer::LLVertexBuffer(U32 typemask, S32 usage) : + LLRefCount(), + + mNumVerts(0), + mNumIndices(0), + mAlignedOffset(0), + mAlignedIndexOffset(0), + mSize(0), + mIndicesSize(0), + mTypeMask(typemask), + mUsage(LLVertexBuffer::determineUsage(usage)), + mGLBuffer(0), + mGLIndices(0), + mGLArray(0), + mMappedData(NULL), + mMappedIndexData(NULL), + mMappedDataUsingVBOs(false), + mMappedIndexDataUsingVBOs(false), + mVertexLocked(false), + mIndexLocked(false), + mFinal(false), + mEmpty(true), + mMappable(false), + mFence(NULL) +{ + mMappable = (mUsage == GL_DYNAMIC_DRAW_ARB && !sDisableVBOMapping); + //zero out offsets for (U32 i = 0; i < TYPE_MAX; i++) { mOffsets[i] = 0; } - mTypeMask = typemask; - mSize = 0; - mIndicesSize = 0; - mAlignedOffset = 0; - mAlignedIndexOffset = 0; - sCount++; } @@ -853,14 +1008,13 @@ S32 LLVertexBuffer::getSize() const //virtual LLVertexBuffer::~LLVertexBuffer() { - LLMemType mt2(LLMemType::MTYPE_VERTEX_DESTRUCTOR); destroyGLBuffer(); destroyGLIndices(); if (mGLArray) { #if GL_ARB_vertex_array_object - glDeleteVertexArrays(1, &mGLArray); + releaseVAOName(mGLArray); #endif } @@ -873,7 +1027,10 @@ LLVertexBuffer::~LLVertexBuffer() mFence = NULL; - llassert_always(!mMappedData && !mMappedIndexData) ; + sVertexCount -= mNumVerts; + sIndexCount -= mNumIndices; + + llassert_always(!mMappedData && !mMappedIndexData); }; void LLVertexBuffer::placeFence() const @@ -904,7 +1061,7 @@ void LLVertexBuffer::waitFence() const void LLVertexBuffer::genBuffer(U32 size) { - mSize = nhpo2(size); + mSize = vbo_block_size(size); if (mUsage == GL_STREAM_DRAW_ARB) { @@ -920,7 +1077,7 @@ void LLVertexBuffer::genBuffer(U32 size) void LLVertexBuffer::genIndices(U32 size) { - mIndicesSize = nhpo2(size); + mIndicesSize = vbo_block_size(size); if (mUsage == GL_STREAM_DRAW_ARB) { @@ -970,8 +1127,6 @@ void LLVertexBuffer::releaseIndices() void LLVertexBuffer::createGLBuffer(U32 size) { - LLMemType mt2(LLMemType::MTYPE_VERTEX_CREATE_VERTICES); - if (mGLBuffer) { destroyGLBuffer(); @@ -982,9 +1137,11 @@ void LLVertexBuffer::createGLBuffer(U32 size) return; } - mEmpty = TRUE; + mEmpty = true; - if (useVBOs()) + mMappedDataUsingVBOs = useVBOs(); + + if (mMappedDataUsingVBOs) { genBuffer(size); } @@ -999,8 +1156,6 @@ void LLVertexBuffer::createGLBuffer(U32 size) void LLVertexBuffer::createGLIndices(U32 size) { - LLMemType mt2(LLMemType::MTYPE_VERTEX_CREATE_INDICES); - if (mGLIndices) { destroyGLIndices(); @@ -1011,12 +1166,14 @@ void LLVertexBuffer::createGLIndices(U32 size) return; } - mEmpty = TRUE; + mEmpty = true; //pad by 16 bytes for aligned copies size += 16; - if (useVBOs()) + mMappedIndexDataUsingVBOs = useVBOs(); + + if (mMappedIndexDataUsingVBOs) { //pad by another 16 bytes for VBO pointer adjustment size += 16; @@ -1033,18 +1190,17 @@ void LLVertexBuffer::createGLIndices(U32 size) void LLVertexBuffer::destroyGLBuffer() { - LLMemType mt2(LLMemType::MTYPE_VERTEX_DESTROY_BUFFER); if (mGLBuffer) { - if (useVBOs()) + if (mMappedDataUsingVBOs) { releaseBuffer(); } else { - FREE_MEM(sPrivatePoolp, mMappedData) ; + FREE_MEM(sPrivatePoolp, (void*) mMappedData); mMappedData = NULL; - mEmpty = TRUE; + mEmpty = true; } } @@ -1054,18 +1210,17 @@ void LLVertexBuffer::destroyGLBuffer() void LLVertexBuffer::destroyGLIndices() { - LLMemType mt2(LLMemType::MTYPE_VERTEX_DESTROY_INDICES); if (mGLIndices) { - if (useVBOs()) + if (mMappedIndexDataUsingVBOs) { releaseIndices(); } else { - FREE_MEM(sPrivatePoolp, mMappedIndexData) ; + FREE_MEM(sPrivatePoolp, (void*) mMappedIndexData); mMappedIndexData = NULL; - mEmpty = TRUE; + mEmpty = true; } } @@ -1075,14 +1230,12 @@ void LLVertexBuffer::destroyGLIndices() void LLVertexBuffer::updateNumVerts(S32 nverts) { - LLMemType mt2(LLMemType::MTYPE_VERTEX_UPDATE_VERTS); - llassert(nverts >= 0); - if (nverts >= 65535) + if (nverts > 65536) { llwarns << "Vertex buffer overflow!" << llendl; - nverts = 65535; + nverts = 65536; } U32 needed_size = calcOffsets(mTypeMask, mOffsets, nverts); @@ -1092,13 +1245,13 @@ void LLVertexBuffer::updateNumVerts(S32 nverts) createGLBuffer(needed_size); } + sVertexCount -= mNumVerts; mNumVerts = nverts; + sVertexCount += mNumVerts; } void LLVertexBuffer::updateNumIndices(S32 nindices) { - LLMemType mt2(LLMemType::MTYPE_VERTEX_UPDATE_INDICES); - llassert(nindices >= 0); U32 needed_size = sizeof(U16) * nindices; @@ -1108,13 +1261,13 @@ void LLVertexBuffer::updateNumIndices(S32 nindices) createGLIndices(needed_size); } + sIndexCount -= mNumIndices; mNumIndices = nindices; + sIndexCount += mNumIndices; } void LLVertexBuffer::allocateBuffer(S32 nverts, S32 nindices, bool create) { - LLMemType mt2(LLMemType::MTYPE_VERTEX_ALLOCATE_BUFFER); - stop_glerror(); if (nverts < 0 || nindices < 0 || @@ -1134,7 +1287,7 @@ void LLVertexBuffer::allocateBuffer(S32 nverts, S32 nindices, bool create) if (gGLManager.mHasVertexArrayObject && useVBOs() && (LLRender::sGLCoreProfile || sUseVAO)) { #if GL_ARB_vertex_array_object - glGenVertexArrays(1, &mGLArray); + mGLArray = getVAOName(); #endif setupVertexArray(); } @@ -1187,7 +1340,24 @@ void LLVertexBuffer::setupVertexArray() GL_FLOAT, //TYPE_WEIGHT, GL_FLOAT, //TYPE_WEIGHT4, GL_FLOAT, //TYPE_CLOTHWEIGHT, - GL_FLOAT, //TYPE_TEXTURE_INDEX + GL_UNSIGNED_INT, //TYPE_TEXTURE_INDEX + }; + + bool attrib_integer[] = + { + false, //TYPE_VERTEX, + false, //TYPE_NORMAL, + false, //TYPE_TEXCOORD0, + false, //TYPE_TEXCOORD1, + false, //TYPE_TEXCOORD2, + false, //TYPE_TEXCOORD3, + false, //TYPE_COLOR, + false, //TYPE_EMISSIVE, + false, //TYPE_BINORMAL, + false, //TYPE_WEIGHT, + false, //TYPE_WEIGHT4, + false, //TYPE_CLOTHWEIGHT, + true, //TYPE_TEXTURE_INDEX }; U32 attrib_normalized[] = @@ -1215,7 +1385,21 @@ void LLVertexBuffer::setupVertexArray() if (mTypeMask & (1 << i)) { glEnableVertexAttribArrayARB(i); - glVertexAttribPointerARB(i, attrib_size[i], attrib_type[i], attrib_normalized[i], sTypeSize[i], (void*) mOffsets[i]); + + if (attrib_integer[i]) + { +#if !LL_DARWIN + //glVertexattribIPointer requires GLSL 1.30 or later + if (gGLManager.mGLSLVersionMajor > 1 || gGLManager.mGLSLVersionMinor >= 30) + { + glVertexAttribIPointer(i, attrib_size[i], attrib_type[i], sTypeSize[i], (void*) mOffsets[i]); + } +#endif + } + else + { + glVertexAttribPointerARB(i, attrib_size[i], attrib_type[i], attrib_normalized[i], sTypeSize[i], (void*) mOffsets[i]); + } } else { @@ -1234,8 +1418,6 @@ void LLVertexBuffer::resizeBuffer(S32 newnverts, S32 newnindices) llassert(newnverts >= 0); llassert(newnindices >= 0); - LLMemType mt2(LLMemType::MTYPE_VERTEX_RESIZE_BUFFER); - updateNumVerts(newnverts); updateNumIndices(newnindices); @@ -1250,16 +1432,10 @@ void LLVertexBuffer::resizeBuffer(S32 newnverts, S32 newnindices) } } -BOOL LLVertexBuffer::useVBOs() const +bool LLVertexBuffer::useVBOs() const { //it's generally ineffective to use VBO for things that are streaming on apple - - if (!mUsage) - { - return FALSE; - } - - return TRUE; + return (mUsage != 0); } //---------------------------------------------------------------------------- @@ -1282,11 +1458,13 @@ bool expand_region(LLVertexBuffer::MappedRegion& region, S32 index, S32 count) return true; } +static LLFastTimer::DeclareTimer FTM_VBO_MAP_BUFFER_RANGE("VBO Map Range"); +static LLFastTimer::DeclareTimer FTM_VBO_MAP_BUFFER("VBO Map"); + // Map for data access -U8* LLVertexBuffer::mapVertexBuffer(S32 type, S32 index, S32 count, bool map_range) +volatile U8* LLVertexBuffer::mapVertexBuffer(S32 type, S32 index, S32 count, bool map_range) { bindGLBuffer(true); - LLMemType mt2(LLMemType::MTYPE_VERTEX_MAP_BUFFER); if (mFinal) { llerrs << "LLVertexBuffer::mapVeretxBuffer() called on a finalized buffer." << llendl; @@ -1298,7 +1476,7 @@ U8* LLVertexBuffer::mapVertexBuffer(S32 type, S32 index, S32 count, bool map_ran if (useVBOs()) { - if (sDisableVBOMapping || gGLManager.mHasMapBufferRange || gGLManager.mHasFlushBufferRange) + if (!mMappable || gGLManager.mHasMapBufferRange || gGLManager.mHasFlushBufferRange) { if (count == -1) { @@ -1323,7 +1501,7 @@ U8* LLVertexBuffer::mapVertexBuffer(S32 type, S32 index, S32 count, bool map_ran if (!mapped) { //not already mapped, map new region - MappedRegion region(type, !sDisableVBOMapping && map_range ? -1 : index, count); + MappedRegion region(type, mMappable && map_range ? -1 : index, count); mMappedVertexRegions.push_back(region); } } @@ -1335,24 +1513,24 @@ U8* LLVertexBuffer::mapVertexBuffer(S32 type, S32 index, S32 count, bool map_ran if (!mVertexLocked) { - LLMemType mt_v(LLMemType::MTYPE_VERTEX_MAP_BUFFER_VERTICES); - mVertexLocked = TRUE; + mVertexLocked = true; sMappedCount++; stop_glerror(); - if(sDisableVBOMapping) + if(!mMappable) { map_range = false; } else { - U8* src = NULL; + volatile U8* src = NULL; waitFence(); if (gGLManager.mHasMapBufferRange) { if (map_range) { #ifdef GL_ARB_map_buffer_range + LLFastTimer t(FTM_VBO_MAP_BUFFER_RANGE); S32 offset = mOffsets[type] + sTypeSize[type]*index; S32 length = (sTypeSize[type]*count+0xF) & ~0xF; src = (U8*) glMapBufferRange(GL_ARRAY_BUFFER_ARB, offset, length, @@ -1376,6 +1554,7 @@ U8* LLVertexBuffer::mapVertexBuffer(S32 type, S32 index, S32 count, bool map_ran } } + LLFastTimer t(FTM_VBO_MAP_BUFFER); src = (U8*) glMapBufferRange(GL_ARRAY_BUFFER_ARB, 0, mSize, GL_MAP_WRITE_BIT | GL_MAP_FLUSH_EXPLICIT_BIT); @@ -1403,7 +1582,7 @@ U8* LLVertexBuffer::mapVertexBuffer(S32 type, S32 index, S32 count, bool map_ran llassert(src != NULL); - mMappedData = LL_NEXT_ALIGNED_ADDRESS<U8>(src); + mMappedData = LL_NEXT_ALIGNED_ADDRESS<volatile U8>(src); mAlignedOffset = mMappedData - src; stop_glerror(); @@ -1413,17 +1592,17 @@ U8* LLVertexBuffer::mapVertexBuffer(S32 type, S32 index, S32 count, bool map_ran { log_glerror(); - //check the availability of memory - LLMemory::logMemoryInfo(TRUE) ; + //check the availability of memory + LLMemory::logMemoryInfo(true); - if(!sDisableVBOMapping) + if(mMappable) { //-------------------- //print out more debug info before crash - llinfos << "vertex buffer size: (num verts : num indices) = " << getNumVerts() << " : " << getNumIndices() << llendl ; - GLint size ; - glGetBufferParameterivARB(GL_ARRAY_BUFFER_ARB, GL_BUFFER_SIZE_ARB, &size) ; - llinfos << "GL_ARRAY_BUFFER_ARB size is " << size << llendl ; + llinfos << "vertex buffer size: (num verts : num indices) = " << getNumVerts() << " : " << getNumIndices() << llendl; + GLint size; + glGetBufferParameterivARB(GL_ARRAY_BUFFER_ARB, GL_BUFFER_SIZE_ARB, &size); + llinfos << "GL_ARRAY_BUFFER_ARB size is " << size << llendl; //-------------------- GLint buff; @@ -1438,7 +1617,7 @@ U8* LLVertexBuffer::mapVertexBuffer(S32 type, S32 index, S32 count, bool map_ran } else { - llerrs << "memory allocation for vertex data failed." << llendl ; + llerrs << "memory allocation for vertex data failed." << llendl; } } } @@ -1448,7 +1627,7 @@ U8* LLVertexBuffer::mapVertexBuffer(S32 type, S32 index, S32 count, bool map_ran map_range = false; } - if (map_range && gGLManager.mHasMapBufferRange && !sDisableVBOMapping) + if (map_range && gGLManager.mHasMapBufferRange && mMappable) { return mMappedData; } @@ -1458,9 +1637,12 @@ U8* LLVertexBuffer::mapVertexBuffer(S32 type, S32 index, S32 count, bool map_ran } } -U8* LLVertexBuffer::mapIndexBuffer(S32 index, S32 count, bool map_range) + +static LLFastTimer::DeclareTimer FTM_VBO_MAP_INDEX_RANGE("IBO Map Range"); +static LLFastTimer::DeclareTimer FTM_VBO_MAP_INDEX("IBO Map"); + +volatile U8* LLVertexBuffer::mapIndexBuffer(S32 index, S32 count, bool map_range) { - LLMemType mt2(LLMemType::MTYPE_VERTEX_MAP_BUFFER); bindGLIndices(true); if (mFinal) { @@ -1473,7 +1655,7 @@ U8* LLVertexBuffer::mapIndexBuffer(S32 index, S32 count, bool map_range) if (useVBOs()) { - if (sDisableVBOMapping || gGLManager.mHasMapBufferRange || gGLManager.mHasFlushBufferRange) + if (!mMappable || gGLManager.mHasMapBufferRange || gGLManager.mHasFlushBufferRange) { if (count == -1) { @@ -1495,7 +1677,7 @@ U8* LLVertexBuffer::mapIndexBuffer(S32 index, S32 count, bool map_range) if (!mapped) { //not already mapped, map new region - MappedRegion region(TYPE_INDEX, !sDisableVBOMapping && map_range ? -1 : index, count); + MappedRegion region(TYPE_INDEX, mMappable && map_range ? -1 : index, count); mMappedIndexRegions.push_back(region); } } @@ -1507,9 +1689,7 @@ U8* LLVertexBuffer::mapIndexBuffer(S32 index, S32 count, bool map_range) if (!mIndexLocked) { - LLMemType mt_v(LLMemType::MTYPE_VERTEX_MAP_BUFFER_INDICES); - - mIndexLocked = TRUE; + mIndexLocked = true; sMappedCount++; stop_glerror(); @@ -1524,19 +1704,20 @@ U8* LLVertexBuffer::mapIndexBuffer(S32 index, S32 count, bool map_range) } } - if(sDisableVBOMapping) + if(!mMappable) { map_range = false; } else { - U8* src = NULL; + volatile U8* src = NULL; waitFence(); if (gGLManager.mHasMapBufferRange) { if (map_range) { #ifdef GL_ARB_map_buffer_range + LLFastTimer t(FTM_VBO_MAP_INDEX_RANGE); S32 offset = sizeof(U16)*index; S32 length = sizeof(U16)*count; src = (U8*) glMapBufferRange(GL_ELEMENT_ARRAY_BUFFER_ARB, offset, length, @@ -1548,6 +1729,7 @@ U8* LLVertexBuffer::mapIndexBuffer(S32 index, S32 count, bool map_range) else { #ifdef GL_ARB_map_buffer_range + LLFastTimer t(FTM_VBO_MAP_INDEX); src = (U8*) glMapBufferRange(GL_ELEMENT_ARRAY_BUFFER_ARB, 0, sizeof(U16)*mNumIndices, GL_MAP_WRITE_BIT | GL_MAP_FLUSH_EXPLICIT_BIT); @@ -1569,6 +1751,7 @@ U8* LLVertexBuffer::mapIndexBuffer(S32 index, S32 count, bool map_range) } else { + LLFastTimer t(FTM_VBO_MAP_INDEX); map_range = false; src = (U8*) glMapBufferARB(GL_ELEMENT_ARRAY_BUFFER_ARB, GL_WRITE_ONLY_ARB); } @@ -1585,9 +1768,9 @@ U8* LLVertexBuffer::mapIndexBuffer(S32 index, S32 count, bool map_range) if (!mMappedIndexData) { log_glerror(); - LLMemory::logMemoryInfo(TRUE) ; + LLMemory::logMemoryInfo(true); - if(!sDisableVBOMapping) + if(mMappable) { GLint buff; glGetIntegerv(GL_ELEMENT_ARRAY_BUFFER_BINDING_ARB, &buff); @@ -1600,7 +1783,7 @@ U8* LLVertexBuffer::mapIndexBuffer(S32 index, S32 count, bool map_range) } else { - llerrs << "memory allocation for Index data failed. " << llendl ; + llerrs << "memory allocation for Index data failed. " << llendl; } } } @@ -1609,7 +1792,7 @@ U8* LLVertexBuffer::mapIndexBuffer(S32 index, S32 count, bool map_range) map_range = false; } - if (map_range && gGLManager.mHasMapBufferRange && !sDisableVBOMapping) + if (map_range && gGLManager.mHasMapBufferRange && mMappable) { return mMappedIndexData; } @@ -1619,22 +1802,29 @@ U8* LLVertexBuffer::mapIndexBuffer(S32 index, S32 count, bool map_range) } } +static LLFastTimer::DeclareTimer FTM_VBO_UNMAP("VBO Unmap"); +static LLFastTimer::DeclareTimer FTM_VBO_FLUSH_RANGE("Flush VBO Range"); + + +static LLFastTimer::DeclareTimer FTM_IBO_UNMAP("IBO Unmap"); +static LLFastTimer::DeclareTimer FTM_IBO_FLUSH_RANGE("Flush IBO Range"); + void LLVertexBuffer::unmapBuffer() { - LLMemType mt2(LLMemType::MTYPE_VERTEX_UNMAP_BUFFER); if (!useVBOs()) { - return ; //nothing to unmap + return; //nothing to unmap } - bool updated_all = false ; + bool updated_all = false; if (mMappedData && mVertexLocked) { + LLFastTimer t(FTM_VBO_UNMAP); bindGLBuffer(true); updated_all = mIndexLocked; //both vertex and index buffers done updating - if(sDisableVBOMapping) + if(!mMappable) { if (!mMappedVertexRegions.empty()) { @@ -1644,7 +1834,7 @@ void LLVertexBuffer::unmapBuffer() const MappedRegion& region = mMappedVertexRegions[i]; S32 offset = region.mIndex >= 0 ? mOffsets[region.mType]+sTypeSize[region.mType]*region.mIndex : 0; S32 length = sTypeSize[region.mType]*region.mCount; - glBufferSubDataARB(GL_ARRAY_BUFFER_ARB, offset, length, mMappedData+offset); + glBufferSubDataARB(GL_ARRAY_BUFFER_ARB, offset, length, (U8*) mMappedData+offset); stop_glerror(); } @@ -1653,7 +1843,7 @@ void LLVertexBuffer::unmapBuffer() else { stop_glerror(); - glBufferSubDataARB(GL_ARRAY_BUFFER_ARB, 0, getSize(), mMappedData); + glBufferSubDataARB(GL_ARRAY_BUFFER_ARB, 0, getSize(), (U8*) mMappedData); stop_glerror(); } } @@ -1671,6 +1861,7 @@ void LLVertexBuffer::unmapBuffer() S32 length = sTypeSize[region.mType]*region.mCount; if (gGLManager.mHasMapBufferRange) { + LLFastTimer t(FTM_VBO_FLUSH_RANGE); #ifdef GL_ARB_map_buffer_range glFlushMappedBufferRange(GL_ARRAY_BUFFER_ARB, offset, length); #endif @@ -1692,14 +1883,15 @@ void LLVertexBuffer::unmapBuffer() mMappedData = NULL; } - mVertexLocked = FALSE ; + mVertexLocked = false; sMappedCount--; } if (mMappedIndexData && mIndexLocked) { + LLFastTimer t(FTM_IBO_UNMAP); bindGLIndices(); - if(sDisableVBOMapping) + if(!mMappable) { if (!mMappedIndexRegions.empty()) { @@ -1708,7 +1900,7 @@ void LLVertexBuffer::unmapBuffer() const MappedRegion& region = mMappedIndexRegions[i]; S32 offset = region.mIndex >= 0 ? sizeof(U16)*region.mIndex : 0; S32 length = sizeof(U16)*region.mCount; - glBufferSubDataARB(GL_ELEMENT_ARRAY_BUFFER_ARB, offset, length, mMappedIndexData+offset); + glBufferSubDataARB(GL_ELEMENT_ARRAY_BUFFER_ARB, offset, length, (U8*) mMappedIndexData+offset); stop_glerror(); } @@ -1717,7 +1909,7 @@ void LLVertexBuffer::unmapBuffer() else { stop_glerror(); - glBufferSubDataARB(GL_ELEMENT_ARRAY_BUFFER_ARB, 0, getIndicesSize(), mMappedIndexData); + glBufferSubDataARB(GL_ELEMENT_ARRAY_BUFFER_ARB, 0, getIndicesSize(), (U8*) mMappedIndexData); stop_glerror(); } } @@ -1734,6 +1926,7 @@ void LLVertexBuffer::unmapBuffer() S32 length = sizeof(U16)*region.mCount; if (gGLManager.mHasMapBufferRange) { + LLFastTimer t(FTM_IBO_FLUSH_RANGE); #ifdef GL_ARB_map_buffer_range glFlushMappedBufferRange(GL_ELEMENT_ARRAY_BUFFER_ARB, offset, length); #endif @@ -1754,16 +1947,16 @@ void LLVertexBuffer::unmapBuffer() glUnmapBufferARB(GL_ELEMENT_ARRAY_BUFFER_ARB); stop_glerror(); - mMappedIndexData = NULL ; + mMappedIndexData = NULL; } - mIndexLocked = FALSE ; + mIndexLocked = false; sMappedCount--; } if(updated_all) { - mEmpty = FALSE; + mEmpty = false; } } @@ -1778,39 +1971,39 @@ template <class T,S32 type> struct VertexBufferStrider { if (type == LLVertexBuffer::TYPE_INDEX) { - U8* ptr = vbo.mapIndexBuffer(index, count, map_range); + volatile U8* ptr = vbo.mapIndexBuffer(index, count, map_range); if (ptr == NULL) { llwarns << "mapIndexBuffer failed!" << llendl; - return FALSE; + return false; } strider = (T*)ptr; strider.setStride(0); - return TRUE; + return true; } else if (vbo.hasDataType(type)) { S32 stride = LLVertexBuffer::sTypeSize[type]; - U8* ptr = vbo.mapVertexBuffer(type, index, count, map_range); + volatile U8* ptr = vbo.mapVertexBuffer(type, index, count, map_range); if (ptr == NULL) { llwarns << "mapVertexBuffer failed!" << llendl; - return FALSE; + return false; } strider = (T*)ptr; strider.setStride(stride); - return TRUE; + return true; } else { llerrs << "VertexBufferStrider could not find valid vertex data." << llendl; } - return FALSE; + return false; } }; @@ -1909,7 +2102,7 @@ bool LLVertexBuffer::bindGLBuffer(bool force_bind) glBindBufferARB(GL_ARRAY_BUFFER_ARB, mGLBuffer); sGLRenderBuffer = mGLBuffer; sBindCount++; - sVBOActive = TRUE; + sVBOActive = true; if (mGLArray) { @@ -1941,7 +2134,7 @@ bool LLVertexBuffer::bindGLIndices(bool force_bind) sGLRenderIndices = mGLIndices; stop_glerror(); sBindCount++; - sIBOActive = TRUE; + sIBOActive = true; ret = true; } @@ -1956,14 +2149,23 @@ void LLVertexBuffer::flush() } } +// bind for transform feedback (quick 'n dirty) +void LLVertexBuffer::bindForFeedback(U32 channel, U32 type, U32 index, U32 count) +{ +#ifdef GL_TRANSFORM_FEEDBACK_BUFFER + U32 offset = mOffsets[type] + sTypeSize[type]*index; + U32 size= (sTypeSize[type]*count); + glBindBufferRange(GL_TRANSFORM_FEEDBACK_BUFFER, channel, mGLBuffer, offset, size); +#endif +} + // Set for rendering 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); + bool setup = (sLastMask != data_mask); if (gDebugGL && data_mask != 0) { //make sure data requirements are fulfilled @@ -1997,21 +2199,17 @@ void LLVertexBuffer::setBuffer(U32 data_mask) if (mGLArray) { bindGLArray(); - setup = FALSE; //do NOT perform pointer setup if using VAO + setup = false; //do NOT perform pointer setup if using VAO } else { - if (bindGLBuffer()) - { - setup = TRUE; - } - if (bindGLIndices()) - { - setup = TRUE; - } + const bool bindBuffer = bindGLBuffer(); + const bool bindIndices = bindGLIndices(); + + setup = setup || bindBuffer || bindIndices; } - BOOL error = FALSE; + bool error = false; if (gDebugGL && !mGLArray) { GLint buff; @@ -2020,7 +2218,7 @@ void LLVertexBuffer::setBuffer(U32 data_mask) { if (gDebugSession) { - error = TRUE; + error = true; gFailLog << "Invalid GL vertex buffer bound: " << buff << std::endl; } else @@ -2036,7 +2234,7 @@ void LLVertexBuffer::setBuffer(U32 data_mask) { if (gDebugSession) { - error = TRUE; + error = true; gFailLog << "Invalid GL index buffer bound: " << buff << std::endl; } else @@ -2058,7 +2256,7 @@ void LLVertexBuffer::setBuffer(U32 data_mask) #endif sGLRenderArray = 0; sGLRenderIndices = 0; - sIBOActive = FALSE; + sIBOActive = false; } if (mGLBuffer) @@ -2067,13 +2265,13 @@ void LLVertexBuffer::setBuffer(U32 data_mask) { glBindBufferARB(GL_ARRAY_BUFFER_ARB, 0); sBindCount++; - sVBOActive = FALSE; - setup = TRUE; // ... or a VBO is deactivated + sVBOActive = false; + setup = true; // ... or a VBO is deactivated } if (sGLRenderBuffer != mGLBuffer) { sGLRenderBuffer = mGLBuffer; - setup = TRUE; // ... or a client memory pointer changed + setup = true; // ... or a client memory pointer changed } } if (mGLIndices) @@ -2082,7 +2280,7 @@ void LLVertexBuffer::setBuffer(U32 data_mask) { glBindBufferARB(GL_ELEMENT_ARRAY_BUFFER_ARB, 0); sBindCount++; - sIBOActive = FALSE; + sIBOActive = false; } sGLRenderIndices = mGLIndices; @@ -2107,14 +2305,13 @@ void LLVertexBuffer::setBuffer(U32 data_mask) // virtual (default) void LLVertexBuffer::setupVertexBuffer(U32 data_mask) { - LLMemType mt2(LLMemType::MTYPE_VERTEX_SETUP_VERTEX_BUFFER); stop_glerror(); - U8* base = useVBOs() ? (U8*) mAlignedOffset : mMappedData; + volatile U8* base = useVBOs() ? (U8*) mAlignedOffset : mMappedData; - /*if ((data_mask & mTypeMask) != data_mask) + if (gDebugGL && ((data_mask & mTypeMask) != data_mask)) { llerrs << "LLVertexBuffer::setupVertexBuffer missing required components for supplied data mask." << llendl; - }*/ + } if (LLGLSLShader::sNoFixedFunction) { @@ -2170,25 +2367,28 @@ void LLVertexBuffer::setupVertexBuffer(U32 data_mask) { S32 loc = TYPE_WEIGHT; void* ptr = (void*)(base + mOffsets[TYPE_WEIGHT]); - glVertexAttribPointerARB(loc, 1, GL_FLOAT, FALSE, LLVertexBuffer::sTypeSize[TYPE_WEIGHT], ptr); + glVertexAttribPointerARB(loc, 1, GL_FLOAT, GL_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); + glVertexAttribPointerARB(loc, 4, GL_FLOAT, GL_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); + glVertexAttribPointerARB(loc, 4, GL_FLOAT, GL_TRUE, LLVertexBuffer::sTypeSize[TYPE_CLOTHWEIGHT], ptr); } - if (data_mask & MAP_TEXTURE_INDEX) + if (data_mask & MAP_TEXTURE_INDEX && + (gGLManager.mGLSLVersionMajor >= 2 || gGLManager.mGLSLVersionMinor >= 30)) //indexed texture rendering requires GLSL 1.30 or later { +#if !LL_DARWIN 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); + glVertexAttribIPointer(loc, 1, GL_UNSIGNED_INT, LLVertexBuffer::sTypeSize[TYPE_VERTEX], ptr); +#endif } if (data_mask & MAP_VERTEX) { diff --git a/indra/llrender/llvertexbuffer.h b/indra/llrender/llvertexbuffer.h index 3e6f6a959a..11fa4ab6a0 100644 --- a/indra/llrender/llvertexbuffer.h +++ b/indra/llrender/llvertexbuffer.h @@ -55,40 +55,49 @@ class LLVBOPool { public: static U32 sBytesPooled; + static U32 sIndexBytesPooled; + + static U32 sCurGLName; - U32 mUsage; - U32 mType; + LLVBOPool(U32 vboUsage, U32 vboType); + + const U32 mUsage; + const U32 mType; //size MUST be a power of 2 - U8* allocate(U32& name, U32 size); + volatile U8* allocate(U32& name, U32 size, bool for_seed = false); //size MUST be the size provided to allocate that returned the given name - void release(U32 name, U8* buffer, U32 size); + void release(U32 name, volatile U8* buffer, U32 size); + //batch allocate buffers to be provided to the application on demand + void seedPool(); + //destroy all records in mFreeList void cleanup(); + U32 genBuffer(); + void deleteBuffer(U32 name); + class Record { public: U32 mGLName; - U8* mClientData; + volatile U8* mClientData; }; + std::list<U32> mGLNamePool; + typedef std::list<Record> record_list_t; std::vector<record_list_t> mFreeList; -}; + std::vector<U32> mMissCount; -class LLGLFence -{ -public: - virtual void placeFence() = 0; - virtual void wait() = 0; }; + //============================================================================ // base class -class LLPrivateMemoryPool ; +class LLPrivateMemoryPool; class LLVertexBuffer : public LLRefCount { public: @@ -103,6 +112,7 @@ public: }; LLVertexBuffer(const LLVertexBuffer& rhs) + : mUsage(rhs.mUsage) { *this = rhs; } @@ -118,13 +128,22 @@ public: static LLVBOPool sStreamIBOPool; static LLVBOPool sDynamicIBOPool; - static BOOL sUseStreamDraw; - static BOOL sUseVAO; - static BOOL sPreferStreamDraw; + static std::list<U32> sAvailableVAOName; + static U32 sCurVAOName; + + static bool sUseStreamDraw; + static bool sUseVAO; + static bool sPreferStreamDraw; + + static void seedPools(); + + static U32 getVAOName(); + static void releaseVAOName(U32 name); static void initClass(bool use_vbo, bool no_vbo_mapping); static void cleanupClass(); static void setupClientArrays(U32 data_mask); + static void pushPositions(U32 mode, const LLVector4a* pos, U32 count); 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); @@ -201,15 +220,16 @@ protected: void destroyGLIndices(); void updateNumVerts(S32 nverts); void updateNumIndices(S32 nindices); - virtual BOOL useVBOs() const; void unmapBuffer(); public: LLVertexBuffer(U32 typemask, S32 usage); // map for data access - U8* mapVertexBuffer(S32 type, S32 index, S32 count, bool map_range); - U8* mapIndexBuffer(S32 index, S32 count, bool map_range); + volatile U8* mapVertexBuffer(S32 type, S32 index, S32 count, bool map_range); + volatile U8* mapIndexBuffer(S32 index, S32 count, bool map_range); + + void bindForFeedback(U32 channel, U32 type, U32 index, U32 count); // set for rendering virtual void setBuffer(U32 data_mask); // calls setupVertexBuffer() if data_mask is not 0 @@ -233,27 +253,30 @@ public: 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 getTextureIndexStrider(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; } + bool useVBOs() const; + bool isEmpty() const { return mEmpty; } + bool isLocked() const { return mVertexLocked || mIndexLocked; } S32 getNumVerts() const { return mNumVerts; } S32 getNumIndices() const { return mNumIndices; } - U8* getIndicesPointer() const { return useVBOs() ? (U8*) mAlignedIndexOffset : mMappedIndexData; } - U8* getVerticesPointer() const { return useVBOs() ? (U8*) mAlignedOffset : mMappedData; } + volatile U8* getIndicesPointer() const { return useVBOs() ? (U8*) mAlignedIndexOffset : mMappedIndexData; } + volatile U8* getVerticesPointer() const { return useVBOs() ? (U8*) mAlignedOffset : mMappedData; } U32 getTypeMask() const { return mTypeMask; } bool hasDataType(S32 type) const { return ((1 << type) & getTypeMask()); } S32 getSize() const; S32 getIndicesSize() const { return mIndicesSize; } - U8* getMappedData() const { return mMappedData; } - U8* getMappedIndices() const { return mMappedIndexData; } + volatile U8* getMappedData() const { return mMappedData; } + volatile U8* getMappedIndices() const { return mMappedIndexData; } S32 getOffset(S32 type) const { return mOffsets[type]; } S32 getUsage() const { return mUsage; } + bool isWriteable() const { return (mMappable || mUsage == GL_STREAM_DRAW_ARB) ? true : false; } void draw(U32 mode, U32 count, U32 indices_offset) const; void drawArrays(U32 mode, U32 offset, U32 count) const; @@ -273,17 +296,25 @@ protected: S32 mSize; S32 mIndicesSize; U32 mTypeMask; - S32 mUsage; // GL usage + + const 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 - BOOL mIndexLocked; // if TRUE, index buffer is being or has been written to in client memory - BOOL mFinal; // if TRUE, buffer can not be mapped again - BOOL mEmpty; // if TRUE, client buffer is empty (or NULL). Old values have been discarded. + volatile U8* mMappedData; // pointer to currently mapped data (NULL if unmapped) + volatile U8* mMappedIndexData; // pointer to currently mapped indices (NULL if unmapped) + + U32 mMappedDataUsingVBOs : 1; + U32 mMappedIndexDataUsingVBOs : 1; + U32 mVertexLocked : 1; // if true, vertex buffer is being or has been written to in client memory + U32 mIndexLocked : 1; // if true, index buffer is being or has been written to in client memory + U32 mFinal : 1; // if true, buffer can not be mapped again + U32 mEmpty : 1; // if true, client buffer is empty (or NULL). Old values have been discarded. + + mutable bool mMappable; // if true, use memory mapping to upload data (otherwise doublebuffer and use glBufferSubData) + S32 mOffsets[TYPE_MAX]; std::vector<MappedRegion> mMappedVertexRegions; @@ -294,28 +325,32 @@ protected: void placeFence() const; void waitFence() const; + static S32 determineUsage(S32 usage); private: - static LLPrivateMemoryPool* sPrivatePoolp ; + static LLPrivateMemoryPool* sPrivatePoolp; public: static S32 sCount; static S32 sGLCount; static S32 sMappedCount; - static BOOL sMapped; + static bool sMapped; typedef std::list<LLVertexBuffer*> buffer_list_t; - static BOOL sDisableVBOMapping; //disable glMapBufferARB - static BOOL sEnableVBOs; + static bool sDisableVBOMapping; //disable glMapBufferARB + static bool sEnableVBOs; 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; + static bool sVBOActive; + static bool sIBOActive; static U32 sLastMask; static U32 sAllocatedBytes; + static U32 sAllocatedIndexBytes; + static U32 sVertexCount; + static U32 sIndexCount; static U32 sBindCount; static U32 sSetCount; }; |