diff options
Diffstat (limited to 'indra/llrender')
40 files changed, 3209 insertions, 1805 deletions
diff --git a/indra/llrender/CMakeLists.txt b/indra/llrender/CMakeLists.txt index 7f881c8bb3..ccff7c7a8c 100644 --- a/indra/llrender/CMakeLists.txt +++ b/indra/llrender/CMakeLists.txt @@ -17,6 +17,7 @@ set(llrender_SOURCE_FILES llfontfreetype.cpp llfontfreetypesvg.cpp llfontgl.cpp + llfontvertexbuffer.cpp llfontregistry.cpp llgl.cpp llglslshader.cpp @@ -43,6 +44,7 @@ set(llrender_HEADER_FILES llcubemap.h llcubemaparray.h llfontgl.h + llfontvertexbuffer.h llfontfreetype.h llfontfreetypesvg.h llfontbitmapcache.h diff --git a/indra/llrender/llcubemap.cpp b/indra/llrender/llcubemap.cpp index f16d5b6e53..26e4aaad52 100644 --- a/indra/llrender/llcubemap.cpp +++ b/indra/llrender/llcubemap.cpp @@ -78,7 +78,7 @@ void LLCubeMap::initGL() for (int i = 0; i < 6; i++) { - mImages[i] = new LLImageGL(RESOLUTION, RESOLUTION, 4, FALSE); + mImages[i] = new LLImageGL(RESOLUTION, RESOLUTION, 4, false); #if USE_SRGB_DECODE if (mIssRGB) { mImages[i]->setExplicitFormat(GL_SRGB8_ALPHA8, GL_RGBA); @@ -111,6 +111,9 @@ void LLCubeMap::initRawData(const std::vector<LLPointer<LLImageRaw> >& rawimages // Yes, I know that this is inefficient! - djs 08/08/02 for (int i = 0; i < 6; i++) { + LLImageDataSharedLock lockIn(rawimages[i]); + LLImageDataLock lockOut(mRawImages[i]); + const U8 *sd = rawimages[i]->getData(); U8 *td = mRawImages[i]->getData(); @@ -173,7 +176,7 @@ void LLCubeMap::initReflectionMap(U32 resolution, U32 components) LLImageGL::generateTextures(1, &texname); - mImages[0] = new LLImageGL(resolution, resolution, components, TRUE); + mImages[0] = new LLImageGL(resolution, resolution, components, true); mImages[0]->setTexName(texname); mImages[0]->setTarget(mTargets[0], LLTexUnit::TT_CUBE_MAP); gGL.getTexUnit(0)->bindManual(LLTexUnit::TT_CUBE_MAP, texname); @@ -197,7 +200,7 @@ void LLCubeMap::initEnvironmentMap(const std::vector<LLPointer<LLImageRaw> >& ra llassert(rawimages[i]->getHeight() == resolution); llassert(rawimages[i]->getComponents() == components); - mImages[i] = new LLImageGL(resolution, resolution, components, TRUE); + mImages[i] = new LLImageGL(resolution, resolution, components, true); mImages[i]->setTarget(mTargets[i], LLTexUnit::TT_CUBE_MAP); mRawImages[i] = rawimages[i]; mImages[i]->createGLTexture(0, mRawImages[i], texname); @@ -221,8 +224,8 @@ void LLCubeMap::generateMipMaps() { LL_PROFILE_ZONE_SCOPED_CATEGORY_TEXTURE; - mImages[0]->setUseMipMaps(TRUE); - mImages[0]->setHasMipMaps(TRUE); + mImages[0]->setUseMipMaps(true); + mImages[0]->setHasMipMaps(true); enableTexture(0); bind(); mImages[0]->setFilteringOption(LLTexUnit::TFO_BILINEAR); diff --git a/indra/llrender/llcubemaparray.cpp b/indra/llrender/llcubemaparray.cpp index 03fbb90bf6..d0a97dc2c6 100644 --- a/indra/llrender/llcubemaparray.cpp +++ b/indra/llrender/llcubemaparray.cpp @@ -42,6 +42,8 @@ //#pragma optimize("", off) +using namespace LLImageGLMemory; + // MUST match order of OpenGL face-layers GLenum LLCubeMapArray::sTargets[6] = { @@ -107,7 +109,7 @@ LLCubeMapArray::~LLCubeMapArray() { } -void LLCubeMapArray::allocate(U32 resolution, U32 components, U32 count, BOOL use_mips) +void LLCubeMapArray::allocate(U32 resolution, U32 components, U32 count, bool use_mips, bool hdr) { U32 texname = 0; mWidth = resolution; @@ -123,24 +125,30 @@ void LLCubeMapArray::allocate(U32 resolution, U32 components, U32 count, BOOL us mImage->setHasMipMaps(use_mips); bind(0); + free_cur_tex_image(); - U32 format = components == 4 ? GL_RGBA16F : GL_RGB16F; - + U32 format = components == 4 ? GL_RGBA16F : GL_R11F_G11F_B10F; + if (!hdr) + { + format = components == 4 ? GL_RGBA8 : GL_RGB8; + } U32 mip = 0; - - while (resolution >= 1) + U32 mip_resolution = resolution; + while (mip_resolution >= 1) { - glTexImage3D(GL_TEXTURE_CUBE_MAP_ARRAY, mip, format, resolution, resolution, count * 6, 0, + glTexImage3D(GL_TEXTURE_CUBE_MAP_ARRAY, mip, format, mip_resolution, mip_resolution, count * 6, 0, GL_RGBA, GL_UNSIGNED_BYTE, nullptr); if (!use_mips) { break; } - resolution /= 2; + mip_resolution /= 2; ++mip; } + alloc_tex_image(resolution, resolution, format, count * 6); + mImage->setAddressMode(LLTexUnit::TAM_CLAMP); if (use_mips) diff --git a/indra/llrender/llcubemaparray.h b/indra/llrender/llcubemaparray.h index 9d78c5b0d4..bfc72a321d 100644 --- a/indra/llrender/llcubemaparray.h +++ b/indra/llrender/llcubemaparray.h @@ -51,8 +51,8 @@ public: // res - resolution of each cube face // components - number of components per pixel // count - number of cube maps in the array - // use_mips - if TRUE, mipmaps will be allocated for this cube map array and anisotropic filtering will be used - void allocate(U32 res, U32 components, U32 count, BOOL use_mips = TRUE); + // use_mips - if true, mipmaps will be allocated for this cube map array and anisotropic filtering will be used + void allocate(U32 res, U32 components, U32 count, bool use_mips = true, bool hdr = true); void bind(S32 stage); void unbind(); diff --git a/indra/llrender/llfontbitmapcache.cpp b/indra/llrender/llfontbitmapcache.cpp index 243041945a..6a3af1e608 100644 --- a/indra/llrender/llfontbitmapcache.cpp +++ b/indra/llrender/llfontbitmapcache.cpp @@ -78,11 +78,11 @@ LLImageGL *LLFontBitmapCache::getImageGL(EFontGlyphType bitmap_type, U32 bitmap_ } -BOOL LLFontBitmapCache::nextOpenPos(S32 width, S32& pos_x, S32& pos_y, EFontGlyphType bitmap_type, U32& bitmap_num) +bool LLFontBitmapCache::nextOpenPos(S32 width, S32& pos_x, S32& pos_y, EFontGlyphType bitmap_type, U32& bitmap_num) { if (bitmap_type >= EFontGlyphType::Count) { - return FALSE; + return false; } const U32 bitmap_idx = static_cast<U32>(bitmap_type); @@ -107,8 +107,8 @@ BOOL LLFontBitmapCache::nextOpenPos(S32 width, S32& pos_x, S32& pos_y, EFontGlyp mBitmapHeight = image_height; S32 num_components = getNumComponents(bitmap_type); - mImageRawVec[bitmap_idx].push_back(new LLImageRaw(mBitmapWidth, mBitmapHeight, num_components)); - bitmap_num = mImageRawVec[bitmap_idx].size() - 1; + mImageRawVec[bitmap_idx].emplace_back(new LLImageRaw(mBitmapWidth, mBitmapHeight, num_components)); + bitmap_num = static_cast<U32>(mImageRawVec[bitmap_idx].size()) - 1; LLImageRaw* image_raw = getImageRaw(bitmap_type, bitmap_num); if (EFontGlyphType::Grayscale == bitmap_type) @@ -117,7 +117,7 @@ BOOL LLFontBitmapCache::nextOpenPos(S32 width, S32& pos_x, S32& pos_y, EFontGlyp } // Make corresponding GL image. - mImageGLVec[bitmap_idx].push_back(new LLImageGL(image_raw, false)); + mImageGLVec[bitmap_idx].emplace_back(new LLImageGL(image_raw, false, false)); LLImageGL* image_gl = getImageGL(bitmap_type, bitmap_num); // Start at beginning of the new image. @@ -126,7 +126,7 @@ BOOL LLFontBitmapCache::nextOpenPos(S32 width, S32& pos_x, S32& pos_y, EFontGlyp // Attach corresponding GL texture. (*TODO: is this needed?) gGL.getTexUnit(0)->bind(image_gl); - image_gl->setFilteringOption(LLTexUnit::TFO_POINT); // was setMipFilterNearest(TRUE, TRUE); + image_gl->setFilteringOption(LLTexUnit::TFO_POINT); // was setMipFilterNearest(true, true); } else { @@ -141,8 +141,9 @@ BOOL LLFontBitmapCache::nextOpenPos(S32 width, S32& pos_x, S32& pos_y, EFontGlyp bitmap_num = getNumBitmaps(bitmap_type) - 1; mCurrentOffsetX[bitmap_idx] += width + 1; + mGeneration++; - return TRUE; + return true; } void LLFontBitmapCache::destroyGL() @@ -168,6 +169,7 @@ void LLFontBitmapCache::reset() mBitmapWidth = 0; mBitmapHeight = 0; + mGeneration++; } //static diff --git a/indra/llrender/llfontbitmapcache.h b/indra/llrender/llfontbitmapcache.h index 8f704df44f..0ae4e6bed0 100644 --- a/indra/llrender/llfontbitmapcache.h +++ b/indra/llrender/llfontbitmapcache.h @@ -52,7 +52,7 @@ public: void reset(); - BOOL nextOpenPos(S32 width, S32& posX, S32& posY, EFontGlyphType bitmapType, U32& bitmapNum); + bool nextOpenPos(S32 width, S32& posX, S32& posY, EFontGlyphType bitmapType, U32& bitmapNum); void destroyGL(); @@ -60,9 +60,10 @@ public: LLImageGL* getImageGL(EFontGlyphType bitmapType, U32 bitmapNum) const; S32 getMaxCharWidth() const { return mMaxCharWidth; } - U32 getNumBitmaps(EFontGlyphType bitmapType) const { return (bitmapType < EFontGlyphType::Count) ? mImageRawVec[static_cast<U32>(bitmapType)].size() : 0; } + U32 getNumBitmaps(EFontGlyphType bitmapType) const { return (bitmapType < EFontGlyphType::Count) ? static_cast<U32>(mImageRawVec[static_cast<U32>(bitmapType)].size()) : 0U; } S32 getBitmapWidth() const { return mBitmapWidth; } S32 getBitmapHeight() const { return mBitmapHeight; } + S32 getCacheGeneration() const { return mGeneration; } protected: static U32 getNumComponents(EFontGlyphType bitmap_type); @@ -74,6 +75,7 @@ private: S32 mCurrentOffsetY[static_cast<U32>(EFontGlyphType::Count)] = { 1 }; S32 mMaxCharWidth = 0; S32 mMaxCharHeight = 0; + S32 mGeneration = 0; std::vector<LLPointer<LLImageRaw>> mImageRawVec[static_cast<U32>(EFontGlyphType::Count)]; std::vector<LLPointer<LLImageGL>> mImageGLVec[static_cast<U32>(EFontGlyphType::Count)]; }; diff --git a/indra/llrender/llfontfreetype.cpp b/indra/llrender/llfontfreetype.cpp index 92b373c835..62b551f1e0 100644 --- a/indra/llrender/llfontfreetype.cpp +++ b/indra/llrender/llfontfreetype.cpp @@ -143,10 +143,9 @@ LLFontFreetype::LLFontFreetype() pFileStream(NULL), pFtStream(NULL), #endif - mIsFallback(FALSE), + mIsFallback(false), mFTFace(NULL), mRenderGlyphCount(0), - mAddGlyphCount(0), mStyle(0), mPointSize(0) { @@ -178,7 +177,7 @@ unsigned long ft_read_cb(FT_Stream stream, unsigned long offset, unsigned char * llifstream *file_stream = static_cast<llifstream *>(stream->descriptor.pointer); file_stream->seekg(offset, std::ios::beg); file_stream->read((char*)buffer, count); - return file_stream->gcount(); + return (unsigned long)file_stream->gcount(); } void ft_close_cb(FT_Stream stream) { @@ -187,7 +186,7 @@ void ft_close_cb(FT_Stream stream) { } #endif -BOOL LLFontFreetype::loadFace(const std::string& filename, F32 point_size, F32 vert_dpi, F32 horz_dpi, bool is_fallback, S32 face_n) +bool LLFontFreetype::loadFace(const std::string& filename, F32 point_size, F32 vert_dpi, F32 horz_dpi, bool is_fallback, S32 face_n) { // Don't leak face objects. This is also needed to deal with // changed font file names. @@ -212,7 +211,7 @@ BOOL LLFontFreetype::loadFace(const std::string& filename, F32 point_size, F32 v #ifdef LL_WINDOWS clearFontStreams(); #endif - return FALSE; + return false; } mIsFallback = is_fallback; @@ -232,7 +231,7 @@ BOOL LLFontFreetype::loadFace(const std::string& filename, F32 point_size, F32 v clearFontStreams(); #endif mFTFace = NULL; - return FALSE; + return false; } F32 y_max, y_min, x_max, x_min; @@ -279,7 +278,7 @@ BOOL LLFontFreetype::loadFace(const std::string& filename, F32 point_size, F32 v mStyle |= LLFontGL::ITALIC; } - return TRUE; + return true; } S32 LLFontFreetype::getNumFaces(const std::string& filename) @@ -328,7 +327,7 @@ S32 LLFontFreetype::ftOpenFace(const std::string& filename, S32 face_n) pFtStream = new LLFT_Stream(); pFtStream->base = 0; pFtStream->pos = 0; - pFtStream->size = file_size; + pFtStream->size = static_cast<unsigned long>(file_size); pFtStream->descriptor.pointer = pFileStream; pFtStream->read = ft_read_cb; pFtStream->close = ft_close_cb; @@ -441,7 +440,7 @@ F32 LLFontFreetype::getXKerning(const LLFontGlyphInfo* left_glyph_info, const LL return delta.x*(1.f/64.f); } -BOOL LLFontFreetype::hasGlyph(llwchar wch) const +bool LLFontFreetype::hasGlyph(llwchar wch) const { llassert(!mIsFallback); return(mCharGlyphInfoMap.find(wch) != mCharGlyphInfoMap.end()); @@ -552,7 +551,7 @@ LLFontGlyphInfo* LLFontFreetype::addGlyphFromFont(const LLFontFreetype *fontp, l return NULL; llassert(!mIsFallback); - fontp->renderGlyph(requested_glyph_type, glyph_index); + fontp->renderGlyph(requested_glyph_type, glyph_index, wch); EFontGlyphType bitmap_glyph_type = EFontGlyphType::Unspecified; switch (fontp->mFTFace->glyph->bitmap.pixel_mode) @@ -574,7 +573,6 @@ LLFontGlyphInfo* LLFontFreetype::addGlyphFromFont(const LLFontFreetype *fontp, l S32 pos_x, pos_y; U32 bitmap_num; mFontBitmapCachep->nextOpenPos(width, pos_x, pos_y, bitmap_glyph_type, bitmap_num); - mAddGlyphCount++; LLFontGlyphInfo* gi = new LLFontGlyphInfo(glyph_index, requested_glyph_type); gi->mXBitmapOffset = pos_x; @@ -697,7 +695,7 @@ void LLFontFreetype::insertGlyphInfo(llwchar wch, LLFontGlyphInfo* gi) const } } -void LLFontFreetype::renderGlyph(EFontGlyphType bitmap_type, U32 glyph_index) const +void LLFontFreetype::renderGlyph(EFontGlyphType bitmap_type, U32 glyph_index, llwchar wch) const { if (mFTFace == NULL) return; @@ -712,11 +710,28 @@ void LLFontFreetype::renderGlyph(EFontGlyphType bitmap_type, U32 glyph_index) co FT_Error error = FT_Load_Glyph(mFTFace, glyph_index, load_flags); if (FT_Err_Ok != error) { + if (error == FT_Err_Out_Of_Memory) + { + LLError::LLUserWarningMsg::showOutOfMemory(); + LL_ERRS() << "Out of memory loading glyph for character " << llformat("U+%xu", U32(wch)) << LL_ENDL; + } + std::string message = llformat( - "Error %d (%s) loading glyph %u: bitmap_type=%u, load_flags=%d", - error, FT_Error_String(error), glyph_index, bitmap_type, load_flags); + "Error %d (%s) loading wchar %u glyph %u/%u: bitmap_type=%u, load_flags=%d", + error, FT_Error_String(error), wch, glyph_index, mFTFace->num_glyphs, bitmap_type, load_flags); LL_WARNS_ONCE() << message << LL_ENDL; error = FT_Load_Glyph(mFTFace, glyph_index, load_flags ^ FT_LOAD_COLOR); + if (FT_Err_Invalid_Outline == error + || FT_Err_Invalid_Composite == error + || (FT_Err_Ok != error && LLStringOps::isEmoji(wch))) + { + glyph_index = FT_Get_Char_Index(mFTFace, '?'); + // if '?' is not present, potentially can use last index, that's supposed to be null glyph + if (glyph_index > 0) + { + error = FT_Load_Glyph(mFTFace, glyph_index, load_flags ^ FT_LOAD_COLOR); + } + } llassert_always_msg(FT_Err_Ok == error, message.c_str()); } @@ -851,6 +866,7 @@ bool LLFontFreetype::setSubImageBGRA(U32 x, U32 y, U32 bitmap_num, U16 width, U1 void LLFontFreetype::setSubImageLuminanceAlpha(U32 x, U32 y, U32 bitmap_num, U32 width, U32 height, U8 *data, S32 stride) const { LLImageRaw *image_raw = mFontBitmapCachep->getImageRaw(EFontGlyphType::Grayscale, bitmap_num); + LLImageDataLock lock(image_raw); llassert(!mIsFallback); llassert(image_raw && (image_raw->getComponents() == 2)); diff --git a/indra/llrender/llfontfreetype.h b/indra/llrender/llfontfreetype.h index 19112830a4..783bf4a4b3 100644 --- a/indra/llrender/llfontfreetype.h +++ b/indra/llrender/llfontfreetype.h @@ -86,7 +86,7 @@ public: // is_fallback should be true for fallback fonts that aren't used // to render directly (Unicode backup, primarily) - BOOL loadFace(const std::string& filename, F32 point_size, F32 vert_dpi, F32 horz_dpi, bool is_fallback, S32 face_n); + bool loadFace(const std::string& filename, F32 point_size, F32 vert_dpi, F32 horz_dpi, bool is_fallback, S32 face_n); S32 getNumFaces(const std::string& filename); @@ -153,10 +153,10 @@ private: void resetBitmapCache(); void setSubImageLuminanceAlpha(U32 x, U32 y, U32 bitmap_num, U32 width, U32 height, U8 *data, S32 stride = 0) const; bool setSubImageBGRA(U32 x, U32 y, U32 bitmap_num, U16 width, U16 height, const U8* data, U32 stride) const; - BOOL hasGlyph(llwchar wch) const; // Has a glyph for this character + bool hasGlyph(llwchar wch) const; // Has a glyph for this character LLFontGlyphInfo* addGlyph(llwchar wch, EFontGlyphType glyph_type) const; // Add a new character to the font if necessary LLFontGlyphInfo* addGlyphFromFont(const LLFontFreetype *fontp, llwchar wch, U32 glyph_index, EFontGlyphType bitmap_type) const; // Add a glyph from this font to the other (returns the glyph_index, 0 if not found) - void renderGlyph(EFontGlyphType bitmap_type, U32 glyph_index) const; + void renderGlyph(EFontGlyphType bitmap_type, U32 glyph_index, llwchar wch) const; void insertGlyphInfo(llwchar wch, LLFontGlyphInfo* gi) const; std::string mName; @@ -175,7 +175,7 @@ private: LLFT_Stream *pFtStream; #endif - BOOL mIsFallback; + bool mIsFallback; typedef std::pair<LLPointer<LLFontFreetype>, char_functor_t> fallback_font_t; typedef std::vector<fallback_font_t> fallback_font_vector_t; fallback_font_vector_t mFallbackFonts; // A list of fallback fonts to look for glyphs in (for Unicode chars) @@ -187,7 +187,6 @@ private: mutable LLFontBitmapCache* mFontBitmapCachep; mutable S32 mRenderGlyphCount; - mutable S32 mAddGlyphCount; }; #endif // LL_FONTFREETYPE_H diff --git a/indra/llrender/llfontfreetypesvg.cpp b/indra/llrender/llfontfreetypesvg.cpp index 355e8432aa..71f751329e 100644 --- a/indra/llrender/llfontfreetypesvg.cpp +++ b/indra/llrender/llfontfreetypesvg.cpp @@ -136,18 +136,18 @@ FT_Error LLFontFreeTypeSvgRenderer::OnPresetGlypthSlot(FT_GlyphSlot glyph_slot, float svg_scale = llmin(svg_x_scale, svg_y_scale); datap->Scale = svg_scale; - glyph_slot->bitmap.width = floorf(svg_width) * svg_scale; - glyph_slot->bitmap.rows = floorf(svg_height) * svg_scale; + glyph_slot->bitmap.width = (unsigned int)(floorf(svg_width) * svg_scale); + glyph_slot->bitmap.rows = (unsigned int)(floorf(svg_height) * svg_scale); glyph_slot->bitmap_left = (document->metrics.x_ppem - glyph_slot->bitmap.width) / 2; - glyph_slot->bitmap_top = glyph_slot->face->size->metrics.ascender / 64.f; + glyph_slot->bitmap_top = (FT_Int)(glyph_slot->face->size->metrics.ascender / 64.f); glyph_slot->bitmap.pitch = glyph_slot->bitmap.width * 4; glyph_slot->bitmap.pixel_mode = FT_PIXEL_MODE_BGRA; /* Copied as-is from fcft (MIT license) */ // Compute all the bearings and set them correctly. The outline is scaled already, we just need to use the bounding box. - float horiBearingX = 0.; - float horiBearingY = -glyph_slot->bitmap_top; + float horiBearingX = 0.f; + float horiBearingY = -(float)glyph_slot->bitmap_top; // XXX parentheses correct? float vertBearingX = glyph_slot->metrics.horiBearingX / 64.0f - glyph_slot->metrics.horiAdvance / 64.0f / 2; @@ -156,13 +156,13 @@ FT_Error LLFontFreeTypeSvgRenderer::OnPresetGlypthSlot(FT_GlyphSlot glyph_slot, // Do conversion in two steps to avoid 'bad function cast' warning glyph_slot->metrics.width = glyph_slot->bitmap.width * 64; glyph_slot->metrics.height = glyph_slot->bitmap.rows * 64; - glyph_slot->metrics.horiBearingX = horiBearingX * 64; - glyph_slot->metrics.horiBearingY = horiBearingY * 64; - glyph_slot->metrics.vertBearingX = vertBearingX * 64; - glyph_slot->metrics.vertBearingY = vertBearingY * 64; + glyph_slot->metrics.horiBearingX = (FT_Pos)(horiBearingX * 64); + glyph_slot->metrics.horiBearingY = (FT_Pos)(horiBearingY * 64); + glyph_slot->metrics.vertBearingX = (FT_Pos)(vertBearingX * 64); + glyph_slot->metrics.vertBearingY = (FT_Pos)(vertBearingY * 64); if (glyph_slot->metrics.vertAdvance == 0) { - glyph_slot->metrics.vertAdvance = glyph_slot->bitmap.rows * 1.2f * 64; + glyph_slot->metrics.vertAdvance = (FT_Pos)(glyph_slot->bitmap.rows * 1.2f * 64); } return FT_Err_Ok; diff --git a/indra/llrender/llfontfreetypesvg.h b/indra/llrender/llfontfreetypesvg.h index 72ac5293ce..4170cab273 100644 --- a/indra/llrender/llfontfreetypesvg.h +++ b/indra/llrender/llfontfreetypesvg.h @@ -46,7 +46,7 @@ public: // - right before the svg module calls the render callback hook. (with cache == true) static FT_Error OnPresetGlypthSlot(FT_GlyphSlot glyph_slot, FT_Bool cache, FT_Pointer* state); - // Called to render an OT-SVG glyph (right after the preset hook OnPresetGlypthSlot was called with cache set to TRUE) + // Called to render an OT-SVG glyph (right after the preset hook OnPresetGlypthSlot was called with cache set to true) static FT_Error OnRender(FT_GlyphSlot glyph_slot, FT_Pointer* state); // Called to deallocate our per glyph slot data diff --git a/indra/llrender/llfontgl.cpp b/indra/llrender/llfontgl.cpp index 3714bb1883..16eec1fdd2 100644 --- a/indra/llrender/llfontgl.cpp +++ b/indra/llrender/llfontgl.cpp @@ -58,7 +58,8 @@ F32 LLFontGL::sVertDPI = 96.f; F32 LLFontGL::sHorizDPI = 96.f; F32 LLFontGL::sScaleX = 1.f; F32 LLFontGL::sScaleY = 1.f; -BOOL LLFontGL::sDisplayFont = TRUE ; +S32 LLFontGL::sResolutionGeneration = 0; +bool LLFontGL::sDisplayFont = true ; std::string LLFontGL::sAppDir; LLColor4 LLFontGL::sShadowColor(0.f, 0.f, 0.f, 1.f); @@ -89,7 +90,7 @@ void LLFontGL::destroyGL() mFontFreetype->destroyGL(); } -BOOL LLFontGL::loadFace(const std::string& filename, F32 point_size, const F32 vert_dpi, const F32 horz_dpi, bool is_fallback, S32 face_n) +bool LLFontGL::loadFace(const std::string& filename, F32 point_size, const F32 vert_dpi, const F32 horz_dpi, bool is_fallback, S32 face_n) { if(mFontFreetype == reinterpret_cast<LLFontFreetype*>(NULL)) { @@ -109,15 +110,21 @@ S32 LLFontGL::getNumFaces(const std::string& filename) return mFontFreetype->getNumFaces(filename); } +S32 LLFontGL::getCacheGeneration() const +{ + const LLFontBitmapCache* font_bitmap_cache = mFontFreetype->getFontBitmapCache(); + return font_bitmap_cache->getCacheGeneration(); +} + 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, BOOL use_color) const + ShadowType shadow, S32 max_chars, F32* right_x, bool use_ellipses, bool use_color) const { - LLRectf rect_float(rect.mLeft, rect.mTop, rect.mRight, rect.mBottom); + LLRectf rect_float((F32)rect.mLeft, (F32)rect.mTop, (F32)rect.mRight, (F32)rect.mBottom); return render(wstr, begin_offset, rect_float, color, halign, valign, style, shadow, max_chars, right_x, use_ellipses, use_color); } S32 LLFontGL::render(const LLWString &wstr, S32 begin_offset, const LLRectf& rect, const LLColor4 &color, HAlign halign, VAlign valign, U8 style, - ShadowType shadow, S32 max_chars, F32* right_x, BOOL use_ellipses, BOOL use_color) const + ShadowType shadow, S32 max_chars, F32* right_x, bool use_ellipses, bool use_color) const { F32 x = rect.mLeft; F32 y = 0.f; @@ -138,18 +145,18 @@ S32 LLFontGL::render(const LLWString &wstr, S32 begin_offset, const LLRectf& rec y = rect.mBottom; break; } - return render(wstr, begin_offset, x, y, color, halign, valign, style, shadow, max_chars, rect.getWidth(), right_x, use_ellipses, use_color); + return render(wstr, begin_offset, x, y, color, halign, valign, style, shadow, max_chars, (S32)rect.getWidth(), right_x, use_ellipses, use_color); } S32 LLFontGL::render(const LLWString &wstr, S32 begin_offset, F32 x, F32 y, const LLColor4 &color, HAlign halign, VAlign valign, U8 style, - ShadowType shadow, S32 max_chars, S32 max_pixels, F32* right_x, BOOL use_ellipses, BOOL use_color) const + ShadowType shadow, S32 max_chars, S32 max_pixels, F32* right_x, bool use_ellipses, bool use_color) const { LL_PROFILE_ZONE_SCOPED_CATEGORY_UI; if(!sDisplayFont) //do not display texts { - return wstr.length() ; + return static_cast<S32>(wstr.length()); } if (wstr.empty()) @@ -249,12 +256,16 @@ S32 LLFontGL::render(const LLWString &wstr, S32 begin_offset, F32 x, F32 y, cons const LLFontBitmapCache* font_bitmap_cache = mFontFreetype->getFontBitmapCache(); + // This looks wrong, value is dynamic. + // LLFontBitmapCache::nextOpenPos can alter these values when + // new characters get added to cache, which affects whole string. + // Todo: Perhaps value should update after symbols were added? F32 inv_width = 1.f / font_bitmap_cache->getBitmapWidth(); F32 inv_height = 1.f / font_bitmap_cache->getBitmapHeight(); const S32 LAST_CHARACTER = LLFontFreetype::LAST_CHAR_FULL; - BOOL draw_ellipses = FALSE; + bool draw_ellipses = false; if (use_ellipses) { // check for too long of a string @@ -264,24 +275,29 @@ S32 LLFontGL::render(const LLWString &wstr, S32 begin_offset, F32 x, F32 y, cons // use four dots for ellipsis width to generate padding const LLWString dots(utf8str_to_wstring(std::string("...."))); scaled_max_pixels = llmax(0, scaled_max_pixels - ll_round(getWidthF32(dots.c_str()))); - draw_ellipses = TRUE; + draw_ellipses = true; } } const LLFontGlyphInfo* next_glyph = NULL; - const S32 GLYPH_BATCH_SIZE = 30; - LLVector3 vertices[GLYPH_BATCH_SIZE * 4]; - LLVector2 uvs[GLYPH_BATCH_SIZE * 4]; - LLColor4U colors[GLYPH_BATCH_SIZE * 4]; + // string can have more than one glyph per char (ex: bold or shadow), + // make sure that GLYPH_BATCH_SIZE won't end up with half a symbol. + // See drawGlyph. + // Ex: with shadows it's 6 glyps per char. 30 fits exactly 5 chars. + static constexpr S32 GLYPH_BATCH_SIZE = 30; + static thread_local LLVector4a vertices[GLYPH_BATCH_SIZE * 6]; + static thread_local LLVector2 uvs[GLYPH_BATCH_SIZE * 6]; + static thread_local LLColor4U colors[GLYPH_BATCH_SIZE * 6]; LLColor4U text_color(color); // Preserve the transparency to render fading emojis in fading text (e.g. // for the chat console)... HB - LLColor4U emoji_color(255, 255, 255, text_color.mV[VW]); + LLColor4U emoji_color(255, 255, 255, text_color.mV[VALPHA]); std::pair<EFontGlyphType, S32> bitmap_entry = std::make_pair(EFontGlyphType::Grayscale, -1); S32 glyph_count = 0; + llwchar last_char = wstr[begin_offset]; for (i = begin_offset; i < begin_offset + length; i++) { llwchar wch = wstr[i]; @@ -299,15 +315,15 @@ S32 LLFontGL::render(const LLWString &wstr, S32 begin_offset, F32 x, F32 y, cons } // Per-glyph bitmap texture. std::pair<EFontGlyphType, S32> next_bitmap_entry = fgi->mBitmapEntry; - if (next_bitmap_entry != bitmap_entry) + if (next_bitmap_entry != bitmap_entry || last_char != wch) { // 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.begin(LLRender::TRIANGLES); { - gGL.vertexBatchPreTransformed(vertices, uvs, colors, glyph_count * 4); + gGL.vertexBatchPreTransformed(vertices, uvs, colors, glyph_count * 6); } gGL.end(); glyph_count = 0; @@ -316,6 +332,11 @@ S32 LLFontGL::render(const LLWString &wstr, S32 begin_offset, F32 x, F32 y, cons bitmap_entry = next_bitmap_entry; LLImageGL* font_image = font_bitmap_cache->getImageGL(bitmap_entry.first, bitmap_entry.second); gGL.getTexUnit(0)->bind(font_image); + + // For some reason it's not enough to compare by bitmap_entry. + // Issue hits emojis, japenese and chinese glyphs, only on first run. + // Todo: figure it out, there might be a bug with raw image data. + last_char = wch; } if ((start_x + scaled_max_pixels) < (cur_x + fgi->mXBearing + fgi->mWidth)) @@ -338,9 +359,9 @@ S32 LLFontGL::render(const LLWString &wstr, S32 begin_offset, F32 x, F32 y, cons if (glyph_count >= GLYPH_BATCH_SIZE) { - gGL.begin(LLRender::QUADS); + gGL.begin(LLRender::TRIANGLES); { - gGL.vertexBatchPreTransformed(vertices, uvs, colors, glyph_count * 4); + gGL.vertexBatchPreTransformed(vertices, uvs, colors, glyph_count * 6); } gGL.end(); @@ -376,9 +397,9 @@ S32 LLFontGL::render(const LLWString &wstr, S32 begin_offset, F32 x, F32 y, cons cur_render_y = cur_y; } - gGL.begin(LLRender::QUADS); + gGL.begin(LLRender::TRIANGLES); { - gGL.vertexBatchPreTransformed(vertices, uvs, colors, glyph_count * 4); + gGL.vertexBatchPreTransformed(vertices, uvs, colors, glyph_count * 6); } gGL.end(); @@ -402,11 +423,10 @@ S32 LLFontGL::render(const LLWString &wstr, S32 begin_offset, F32 x, F32 y, cons if (draw_ellipses) { - // recursively render ellipses at end of string // we've already reserved enough room - gGL.pushUIMatrix(); - renderUTF8(std::string("..."), + static LLWString elipses_wstr(utf8string_to_wstring(std::string("..."))); + render(elipses_wstr, 0, (cur_x - origin.mV[VX]) / sScaleX, (F32)y, color, @@ -415,9 +435,8 @@ S32 LLFontGL::render(const LLWString &wstr, S32 begin_offset, F32 x, F32 y, cons shadow, S32_MAX, max_pixels, right_x, - FALSE, + false, use_color); - gGL.popUIMatrix(); } gGL.popUIMatrix(); @@ -430,7 +449,7 @@ S32 LLFontGL::render(const LLWString &text, S32 begin_offset, F32 x, F32 y, cons return render(text, begin_offset, x, y, color, LEFT, BASELINE, NORMAL, NO_SHADOW); } -S32 LLFontGL::renderUTF8(const std::string &text, S32 begin_offset, F32 x, F32 y, const LLColor4 &color, HAlign halign, VAlign valign, U8 style, ShadowType shadow, S32 max_chars, S32 max_pixels, F32* right_x, BOOL use_ellipses, BOOL use_color) const +S32 LLFontGL::renderUTF8(const std::string &text, S32 begin_offset, F32 x, F32 y, const LLColor4 &color, HAlign halign, VAlign valign, U8 style, ShadowType shadow, S32 max_chars, S32 max_pixels, F32* right_x, bool use_ellipses, bool use_color) const { return render(utf8str_to_wstring(text), begin_offset, x, y, color, halign, valign, style, shadow, max_chars, max_pixels, right_x, use_ellipses, use_color); } @@ -503,6 +522,7 @@ F32 LLFontGL::getWidthF32(const std::string& utf8text, S32 begin_offset, S32 max F32 LLFontGL::getWidthF32(const llwchar* wchars, S32 begin_offset, S32 max_chars, bool no_padding) const { + LL_PROFILE_ZONE_SCOPED_CATEGORY_UI; const S32 LAST_CHARACTER = LLFontFreetype::LAST_CHAR_FULL; F32 cur_x = 0; @@ -560,7 +580,7 @@ F32 LLFontGL::getWidthF32(const llwchar* wchars, S32 begin_offset, S32 max_chars void LLFontGL::generateASCIIglyphs() { - LL_PROFILE_ZONE_SCOPED_CATEGORY_UI + LL_PROFILE_ZONE_SCOPED_CATEGORY_UI; for (U32 i = 32; (i < 127); i++) { mFontFreetype->getGlyphInfo(i, EFontGlyphType::Grayscale); @@ -570,7 +590,7 @@ void LLFontGL::generateASCIIglyphs() // Returns the max number of complete characters from text (up to max_chars) that can be drawn in max_pixels S32 LLFontGL::maxDrawableChars(const llwchar* wchars, F32 max_pixels, S32 max_chars, EWordWrapStyle end_on_word_boundary) const { - LL_PROFILE_ZONE_SCOPED_CATEGORY_UI + LL_PROFILE_ZONE_SCOPED_CATEGORY_UI; if (!wchars || !wchars[0] || max_chars == 0) { return 0; @@ -579,11 +599,11 @@ S32 LLFontGL::maxDrawableChars(const llwchar* wchars, F32 max_pixels, S32 max_ch llassert(max_pixels >= 0.f); llassert(max_chars >= 0); - BOOL clip = FALSE; + bool clip = false; F32 cur_x = 0; S32 start_of_last_word = 0; - BOOL in_word = FALSE; + bool in_word = false; // avoid S32 overflow when max_pixels == S32_MAX by staying in floating point F32 scaled_max_pixels = max_pixels * sScaleX; @@ -608,18 +628,18 @@ S32 LLFontGL::maxDrawableChars(const llwchar* wchars, F32 max_pixels, S32 max_ch { if(wch !=(0x00A0)) { - in_word = FALSE; + in_word = false; } } if (iswindividual(wch)) { if (iswpunct(wchars[i+1])) { - in_word=TRUE; + in_word=true; } else { - in_word=FALSE; + in_word=false; start_of_last_word = i; } } @@ -629,7 +649,7 @@ S32 LLFontGL::maxDrawableChars(const llwchar* wchars, F32 max_pixels, S32 max_ch start_of_last_word = i; if (!iswspace(wch)||!iswindividual(wch)) { - in_word = TRUE; + in_word = true; } } @@ -655,7 +675,7 @@ S32 LLFontGL::maxDrawableChars(const llwchar* wchars, F32 max_pixels, S32 max_ch // clip if current character runs past scaled_max_pixels (using width_padding) if (scaled_max_pixels < cur_x + width_padding) { - clip = TRUE; + clip = true; break; } @@ -753,7 +773,7 @@ S32 LLFontGL::firstDrawableChar(const llwchar* wchars, F32 max_pixels, S32 text_ } -S32 LLFontGL::charFromPixelOffset(const llwchar* wchars, S32 begin_offset, F32 target_x, F32 max_pixels, S32 max_chars, BOOL round) const +S32 LLFontGL::charFromPixelOffset(const llwchar* wchars, S32 begin_offset, F32 target_x, F32 max_pixels, S32 max_chars, bool round) const { if (!wchars || !wchars[0] || max_chars == 0) { @@ -881,7 +901,7 @@ void LLFontGL::dumpFontTextures() // static bool LLFontGL::loadDefaultFonts() { - LL_PROFILE_ZONE_SCOPED_CATEGORY_UI + LL_PROFILE_ZONE_SCOPED_CATEGORY_UI; bool succ = true; succ &= (NULL != getFontSansSerifSmall()); succ &= (NULL != getFontSansSerif()); @@ -894,7 +914,7 @@ bool LLFontGL::loadDefaultFonts() void LLFontGL::loadCommonFonts() { - LL_PROFILE_ZONE_SCOPED_CATEGORY_UI + LL_PROFILE_ZONE_SCOPED_CATEGORY_UI; getFont(LLFontDescriptor("SansSerif", "Small", BOLD)); getFont(LLFontDescriptor("SansSerif", "Large", BOLD)); getFont(LLFontDescriptor("SansSerif", "Huge", BOLD)); @@ -1165,7 +1185,7 @@ LLFontGL* LLFontGL::getFontDefault() std::string LLFontGL::getFontPathSystem() { #if LL_DARWIN - // HACK for Mac OS X + // HACK for macOS return "/System/Library/Fonts/"; #elif LL_WINDOWS @@ -1226,31 +1246,42 @@ LLFontGL &LLFontGL::operator=(const LLFontGL &source) return *this; } -void LLFontGL::renderQuad(LLVector3* vertex_out, LLVector2* uv_out, LLColor4U* colors_out, const LLRectf& screen_rect, const LLRectf& uv_rect, const LLColor4U& color, F32 slant_amt) const +void LLFontGL::renderTriangle(LLVector4a* vertex_out, LLVector2* uv_out, LLColor4U* colors_out, const LLRectf& screen_rect, const LLRectf& uv_rect, const LLColor4U& color, F32 slant_amt) const { S32 index = 0; - vertex_out[index] = LLVector3(screen_rect.mRight, screen_rect.mTop, 0.f); - uv_out[index] = LLVector2(uv_rect.mRight, uv_rect.mTop); + vertex_out[index].set(screen_rect.mRight, screen_rect.mTop, 0.f); + uv_out[index].set(uv_rect.mRight, uv_rect.mTop); + colors_out[index] = color; + index++; + + vertex_out[index].set(screen_rect.mLeft, screen_rect.mTop, 0.f); + uv_out[index].set(uv_rect.mLeft, uv_rect.mTop); colors_out[index] = color; index++; - vertex_out[index] = LLVector3(screen_rect.mLeft, screen_rect.mTop, 0.f); - uv_out[index] = LLVector2(uv_rect.mLeft, uv_rect.mTop); + vertex_out[index].set(screen_rect.mLeft, screen_rect.mBottom, 0.f); + uv_out[index].set(uv_rect.mLeft, uv_rect.mBottom); + colors_out[index] = color; + index++; + + + vertex_out[index].set(screen_rect.mRight, screen_rect.mTop, 0.f); + uv_out[index].set(uv_rect.mRight, uv_rect.mTop); colors_out[index] = color; index++; - vertex_out[index] = LLVector3(screen_rect.mLeft, screen_rect.mBottom, 0.f); - uv_out[index] = LLVector2(uv_rect.mLeft, uv_rect.mBottom); + vertex_out[index].set(screen_rect.mLeft, screen_rect.mBottom, 0.f); + uv_out[index].set(uv_rect.mLeft, uv_rect.mBottom); colors_out[index] = color; index++; - vertex_out[index] = LLVector3(screen_rect.mRight, screen_rect.mBottom, 0.f); - uv_out[index] = LLVector2(uv_rect.mRight, uv_rect.mBottom); + vertex_out[index].set(screen_rect.mRight, screen_rect.mBottom, 0.f); + uv_out[index].set(uv_rect.mRight, uv_rect.mBottom); colors_out[index] = color; } -void LLFontGL::drawGlyph(S32& glyph_count, LLVector3* vertex_out, LLVector2* uv_out, LLColor4U* colors_out, const LLRectf& screen_rect, const LLRectf& uv_rect, const LLColor4U& color, U8 style, ShadowType shadow, F32 drop_shadow_strength) const +void LLFontGL::drawGlyph(S32& glyph_count, LLVector4a* vertex_out, LLVector2* uv_out, LLColor4U* colors_out, const LLRectf& screen_rect, const LLRectf& uv_rect, const LLColor4U& color, U8 style, ShadowType shadow, F32 drop_shadow_strength) const { F32 slant_offset; slant_offset = ((style & ITALIC) ? ( -mFontFreetype->getAscenderHeight() * 0.2f) : 0.f); @@ -1264,7 +1295,7 @@ void LLFontGL::drawGlyph(S32& glyph_count, LLVector3* vertex_out, LLVector2* uv_ LLRectf screen_rect_offset = screen_rect; screen_rect_offset.translate((F32)(pass * BOLD_OFFSET), 0.f); - renderQuad(&vertex_out[glyph_count * 4], &uv_out[glyph_count * 4], &colors_out[glyph_count * 4], screen_rect_offset, uv_rect, color, slant_offset); + renderTriangle(&vertex_out[glyph_count * 6], &uv_out[glyph_count * 6], &colors_out[glyph_count * 6], screen_rect_offset, uv_rect, color, slant_offset); glyph_count++; } } @@ -1295,10 +1326,10 @@ void LLFontGL::drawGlyph(S32& glyph_count, LLVector3* vertex_out, LLVector2* uv_ break; } - renderQuad(&vertex_out[glyph_count * 4], &uv_out[glyph_count * 4], &colors_out[glyph_count * 4], screen_rect_offset, uv_rect, shadow_color, slant_offset); + renderTriangle(&vertex_out[glyph_count * 6], &uv_out[glyph_count * 6], &colors_out[glyph_count * 6], screen_rect_offset, uv_rect, shadow_color, slant_offset); glyph_count++; } - renderQuad(&vertex_out[glyph_count * 4], &uv_out[glyph_count * 4], &colors_out[glyph_count * 4], screen_rect, uv_rect, color, slant_offset); + renderTriangle(&vertex_out[glyph_count * 6], &uv_out[glyph_count * 6], &colors_out[glyph_count * 6], screen_rect, uv_rect, color, slant_offset); glyph_count++; } else if (shadow == DROP_SHADOW) @@ -1307,14 +1338,14 @@ void LLFontGL::drawGlyph(S32& glyph_count, LLVector3* vertex_out, LLVector2* uv_ shadow_color.mV[VALPHA] = U8(color.mV[VALPHA] * drop_shadow_strength); LLRectf screen_rect_shadow = screen_rect; screen_rect_shadow.translate(1.f, -1.f); - renderQuad(&vertex_out[glyph_count * 4], &uv_out[glyph_count * 4], &colors_out[glyph_count * 4], screen_rect_shadow, uv_rect, shadow_color, slant_offset); + renderTriangle(&vertex_out[glyph_count * 6], &uv_out[glyph_count * 6], &colors_out[glyph_count * 6], screen_rect_shadow, uv_rect, shadow_color, slant_offset); glyph_count++; - renderQuad(&vertex_out[glyph_count * 4], &uv_out[glyph_count * 4], &colors_out[glyph_count * 4], screen_rect, uv_rect, color, slant_offset); + renderTriangle(&vertex_out[glyph_count * 6], &uv_out[glyph_count * 6], &colors_out[glyph_count * 6], screen_rect, uv_rect, color, slant_offset); glyph_count++; } else // normal rendering { - renderQuad(&vertex_out[glyph_count * 4], &uv_out[glyph_count * 4], &colors_out[glyph_count * 4], screen_rect, uv_rect, color, slant_offset); + renderTriangle(&vertex_out[glyph_count * 6], &uv_out[glyph_count * 6], &colors_out[glyph_count * 6], screen_rect, uv_rect, color, slant_offset); glyph_count++; } } diff --git a/indra/llrender/llfontgl.h b/indra/llrender/llfontgl.h index 65f0a8cbfd..1c8e036f58 100644 --- a/indra/llrender/llfontgl.h +++ b/indra/llrender/llfontgl.h @@ -87,9 +87,10 @@ public: void destroyGL(); - BOOL loadFace(const std::string& filename, F32 point_size, const F32 vert_dpi, const F32 horz_dpi, bool is_fallback, S32 face_n); + bool loadFace(const std::string& filename, F32 point_size, const F32 vert_dpi, const F32 horz_dpi, bool is_fallback, S32 face_n); S32 getNumFaces(const std::string& filename); + S32 getCacheGeneration() const; S32 render(const LLWString &text, S32 begin_offset, const LLRect& rect, @@ -98,8 +99,8 @@ public: U8 style = NORMAL, ShadowType shadow = NO_SHADOW, S32 max_chars = S32_MAX, F32* right_x=NULL, - BOOL use_ellipses = FALSE, - BOOL use_color = TRUE) const; + bool use_ellipses = false, + bool use_color = true) const; S32 render(const LLWString &text, S32 begin_offset, const LLRectf& rect, @@ -108,8 +109,8 @@ public: U8 style = NORMAL, ShadowType shadow = NO_SHADOW, S32 max_chars = S32_MAX, F32* right_x=NULL, - BOOL use_ellipses = FALSE, - BOOL use_color = TRUE) const; + bool use_ellipses = false, + bool use_color = true) const; S32 render(const LLWString &text, S32 begin_offset, F32 x, F32 y, @@ -118,13 +119,13 @@ public: U8 style = NORMAL, ShadowType shadow = NO_SHADOW, S32 max_chars = S32_MAX, S32 max_pixels = S32_MAX, F32* right_x=NULL, - BOOL use_ellipses = FALSE, - BOOL use_color = TRUE) const; + bool use_ellipses = false, + bool use_color = true) const; S32 render(const LLWString &text, S32 begin_offset, F32 x, F32 y, const LLColor4 &color) const; // renderUTF8 does a conversion, so is slower! - S32 renderUTF8(const std::string &text, S32 begin_offset, F32 x, F32 y, const LLColor4 &color, HAlign halign, VAlign valign, U8 style, ShadowType shadow, S32 max_chars = S32_MAX, S32 max_pixels = S32_MAX, F32* right_x = NULL, BOOL use_ellipses = FALSE, BOOL use_color = TRUE) const; + S32 renderUTF8(const std::string &text, S32 begin_offset, F32 x, F32 y, const LLColor4 &color, HAlign halign, VAlign valign, U8 style, ShadowType shadow, S32 max_chars = S32_MAX, S32 max_pixels = S32_MAX, F32* right_x = NULL, bool use_ellipses = false, bool use_color = true) const; S32 renderUTF8(const std::string &text, S32 begin_offset, S32 x, S32 y, const LLColor4 &color) const; 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; @@ -159,7 +160,7 @@ public: S32 firstDrawableChar(const llwchar* wchars, F32 max_pixels, S32 text_len, S32 start_pos=S32_MAX, S32 max_chars = S32_MAX) const; // Returns the index of the character closest to pixel position x (ignoring text to the right of max_pixels and max_chars) - S32 charFromPixelOffset(const llwchar* wchars, S32 char_offset, F32 x, F32 max_pixels=F32_MAX, S32 max_chars = S32_MAX, BOOL round = TRUE) const; + S32 charFromPixelOffset(const llwchar* wchars, S32 char_offset, F32 x, F32 max_pixels=F32_MAX, S32 max_chars = S32_MAX, bool round = true) const; const LLFontDescriptor& getFontDesc() const; @@ -192,7 +193,7 @@ public: static std::string nameFromVAlign(LLFontGL::VAlign align); static LLFontGL::VAlign vAlignFromName(const std::string& name); - static void setFontDisplay(BOOL flag) { sDisplayFont = flag; } + static void setFontDisplay(bool flag) { sDisplayFont = flag; } static LLFontGL* getFontEmojiSmall(); static LLFontGL* getFontEmojiMedium(); @@ -224,7 +225,8 @@ public: static F32 sHorizDPI; static F32 sScaleX; static F32 sScaleY; - static BOOL sDisplayFont ; + static S32 sResolutionGeneration; + static bool sDisplayFont ; static std::string sAppDir; // For loading fonts private: @@ -238,8 +240,8 @@ private: LLFontDescriptor mFontDescriptor; LLPointer<LLFontFreetype> mFontFreetype; - void renderQuad(LLVector3* vertex_out, LLVector2* uv_out, LLColor4U* colors_out, const LLRectf& screen_rect, const LLRectf& uv_rect, const LLColor4U& color, F32 slant_amt) const; - void drawGlyph(S32& glyph_count, LLVector3* vertex_out, LLVector2* uv_out, LLColor4U* colors_out, const LLRectf& screen_rect, const LLRectf& uv_rect, const LLColor4U& color, U8 style, ShadowType shadow, F32 drop_shadow_fade) const; + void renderTriangle(LLVector4a* vertex_out, LLVector2* uv_out, LLColor4U* colors_out, const LLRectf& screen_rect, const LLRectf& uv_rect, const LLColor4U& color, F32 slant_amt) const; + void drawGlyph(S32& glyph_count, LLVector4a* vertex_out, LLVector2* uv_out, LLColor4U* colors_out, const LLRectf& screen_rect, const LLRectf& uv_rect, const LLColor4U& color, U8 style, ShadowType shadow, F32 drop_shadow_fade) const; // Registry holds all instantiated fonts. static LLFontRegistry* sFontRegistry; diff --git a/indra/llrender/llfontregistry.cpp b/indra/llrender/llfontregistry.cpp index 546211aac8..c48a389f6a 100644 --- a/indra/llrender/llfontregistry.cpp +++ b/indra/llrender/llfontregistry.cpp @@ -297,7 +297,7 @@ bool font_desc_init_from_xml(LLXMLNodePtr node, LLFontDescriptor& desc) if (child->hasAttribute("load_collection")) { - BOOL col = FALSE; + bool col = false; child->getAttributeBOOL("load_collection", col); if (col) { @@ -475,7 +475,7 @@ LLFontGL *LLFontRegistry::createFont(const LLFontDescriptor& desc) // The first font will get pulled will be the "head" font, set to non-fallback. // Rest will consitute the fallback list. - BOOL is_first_found = TRUE; + bool is_first_found = true; string_vec_t font_search_paths; font_search_paths.push_back(LLFontGL::getFontPathLocal()); @@ -499,8 +499,8 @@ LLFontGL *LLFontRegistry::createFont(const LLFontDescriptor& desc) // *HACK: Fallback fonts don't render, so we can use that to suppress // creation of OpenGL textures for test apps. JC - BOOL is_fallback = !is_first_found || !mCreateGLTextures; - F32 extra_scale = (is_fallback)?fallback_scale:1.0; + bool is_fallback = !is_first_found || !mCreateGLTextures; + F32 extra_scale = (is_fallback) ? fallback_scale : 1.0f; F32 point_size_scale = extra_scale * point_size; bool is_font_loaded = false; for(string_vec_t::iterator font_search_path_it = font_search_paths.begin(); diff --git a/indra/llrender/llfontvertexbuffer.cpp b/indra/llrender/llfontvertexbuffer.cpp new file mode 100644 index 0000000000..a223509d30 --- /dev/null +++ b/indra/llrender/llfontvertexbuffer.cpp @@ -0,0 +1,239 @@ +/** + * @file llfontvertexbuffer.cpp + * @brief Buffer storage for font rendering. + * + * $LicenseInfo:firstyear=2024&license=viewerlgpl$ + * Second Life Viewer Source Code + * Copyright (C) 2024, 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 "llfontvertexbuffer.h" + +#include "llvertexbuffer.h" + + +bool LLFontVertexBuffer::sEnableBufferCollection = true; + +LLFontVertexBuffer::LLFontVertexBuffer() +{ +} + +LLFontVertexBuffer::~LLFontVertexBuffer() +{ + reset(); +} + +void LLFontVertexBuffer::reset() +{ + // Todo: some form of debug only frequecy check&assert to see if this is happening too often. + // Regenerating this list is expensive + mBufferList.clear(); +} + +S32 LLFontVertexBuffer::render( + const LLFontGL* fontp, + const LLWString& text, + S32 begin_offset, + LLRect rect, + const LLColor4& color, + LLFontGL::HAlign halign, LLFontGL::VAlign valign, + U8 style, + LLFontGL::ShadowType shadow, + S32 max_chars, S32 max_pixels, + F32* right_x, + bool use_ellipses, + bool use_color) +{ + LLRectf rect_float((F32)rect.mLeft, (F32)rect.mTop, (F32)rect.mRight, (F32)rect.mBottom); + return render(fontp, text, begin_offset, rect_float, color, halign, valign, style, shadow, max_chars, right_x, use_ellipses, use_color); +} + +S32 LLFontVertexBuffer::render( + const LLFontGL* fontp, + const LLWString& text, + S32 begin_offset, + LLRectf rect, + const LLColor4& color, + LLFontGL::HAlign halign, LLFontGL::VAlign valign, + U8 style, + LLFontGL::ShadowType shadow, + S32 max_chars, + F32* right_x, + bool use_ellipses, + bool use_color) +{ + F32 x = rect.mLeft; + F32 y = 0.f; + + switch (valign) + { + case LLFontGL::TOP: + y = rect.mTop; + break; + case LLFontGL::VCENTER: + y = rect.getCenterY(); + break; + case LLFontGL::BASELINE: + case LLFontGL::BOTTOM: + y = rect.mBottom; + break; + default: + y = rect.mBottom; + break; + } + return render(fontp, text, begin_offset, x, y, color, halign, valign, style, shadow, max_chars, (S32)rect.getWidth(), right_x, use_ellipses, use_color); +} + +S32 LLFontVertexBuffer::render( + const LLFontGL* fontp, + const LLWString& text, + S32 begin_offset, + F32 x, F32 y, + const LLColor4& color, + LLFontGL::HAlign halign, LLFontGL::VAlign valign, + U8 style, + LLFontGL::ShadowType shadow, + S32 max_chars , S32 max_pixels, + F32* right_x, + bool use_ellipses, + bool use_color ) +{ + if (!LLFontGL::sDisplayFont) //do not display texts + { + return static_cast<S32>(text.length()); + } + if (!sEnableBufferCollection) + { + // For debug purposes and performance testing + return fontp->render(text, begin_offset, x, y, color, halign, valign, style, shadow, max_chars, max_pixels, right_x, use_ellipses, use_color); + } + if (mBufferList.empty()) + { + genBuffers(fontp, text, begin_offset, x, y, color, halign, valign, + style, shadow, max_chars, max_pixels, right_x, use_ellipses, use_color); + } + else if (mLastX != x + || mLastY != y + || mLastFont != fontp + || mLastColor != color // alphas change often + || mLastHalign != halign + || mLastValign != valign + || mLastOffset != begin_offset + || mLastMaxChars != max_chars + || mLastMaxPixels != max_pixels + || mLastStyle != style + || mLastShadow != shadow // ex: buttons change shadow state + || mLastScaleX != LLFontGL::sScaleX + || mLastScaleY != LLFontGL::sScaleY + || mLastVertDPI != LLFontGL::sVertDPI + || mLastHorizDPI != LLFontGL::sHorizDPI + || mLastOrigin != LLFontGL::sCurOrigin + || mLastResGeneration != LLFontGL::sResolutionGeneration + || mLastFontCacheGen != fontp->getCacheGeneration()) + { + genBuffers(fontp, text, begin_offset, x, y, color, halign, valign, + style, shadow, max_chars, max_pixels, right_x, use_ellipses, use_color); + } + else + { + renderBuffers(); + + if (right_x) + { + *right_x = mLastRightX; + } + } + return mChars; +} + +void LLFontVertexBuffer::genBuffers( + const LLFontGL* fontp, + const LLWString& text, + S32 begin_offset, + F32 x, F32 y, + const LLColor4& color, + LLFontGL::HAlign halign, LLFontGL::VAlign valign, + U8 style, LLFontGL::ShadowType shadow, + S32 max_chars, S32 max_pixels, + F32* right_x, + bool use_ellipses, + bool use_color) +{ + // todo: add a debug build assert if this triggers too often for to long? + mBufferList.clear(); + // Save before rendreing, it can change mid-render, + // so will need to rerender previous characters + mLastFontCacheGen = fontp->getCacheGeneration(); + + gGL.beginList(&mBufferList); + mChars = fontp->render(text, begin_offset, x, y, color, halign, valign, + style, shadow, max_chars, max_pixels, right_x, use_ellipses, use_color); + gGL.endList(); + + mLastFont = fontp; + mLastOffset = begin_offset; + mLastMaxChars = max_chars; + mLastMaxPixels = max_pixels; + mLastX = x; + mLastY = y; + mLastColor = color; + mLastHalign = halign; + mLastValign = valign; + mLastStyle = style; + mLastShadow = shadow; + + mLastScaleX = LLFontGL::sScaleX; + mLastScaleY = LLFontGL::sScaleY; + mLastVertDPI = LLFontGL::sVertDPI; + mLastHorizDPI = LLFontGL::sHorizDPI; + mLastOrigin = LLFontGL::sCurOrigin; + mLastResGeneration = LLFontGL::sResolutionGeneration; + + if (right_x) + { + mLastRightX = *right_x; + } +} + +void LLFontVertexBuffer::renderBuffers() +{ + gGL.flush(); // deliberately empty pending verts + gGL.getTexUnit(0)->enable(LLTexUnit::TT_TEXTURE); + gGL.pushUIMatrix(); + + gGL.loadUIIdentity(); + + // Depth translation, so that floating text appears 'in-world' + // and is correctly occluded. + gGL.translatef(0.f, 0.f, LLFontGL::sCurDepth); + gGL.setSceneBlendType(LLRender::BT_ALPHA); + + // Note: ellipses should technically be covered by push/load/translate of their own + // but it's more complexity, values do not change, skipping doesn't appear to break + // anything, so we can skip that until it proves to cause issues. + for (LLVertexBufferData& buffer : mBufferList) + { + buffer.draw(); + } + gGL.popUIMatrix(); +} + diff --git a/indra/llrender/llfontvertexbuffer.h b/indra/llrender/llfontvertexbuffer.h new file mode 100644 index 0000000000..a9e1e2337c --- /dev/null +++ b/indra/llrender/llfontvertexbuffer.h @@ -0,0 +1,130 @@ +/** + * @file llfontgl.h + * @author Andrii Kleshchev + * @brief Buffer storage for font rendering. + * + * $LicenseInfo:firstyear=2001&license=viewerlgpl$ + * Second Life Viewer Source Code + * Copyright (C) 2010, 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_LLFONTVERTEXBUFFER_H +#define LL_LLFONTVERTEXBUFFER_H + +#include "llfontgl.h" + +class LLVertexBufferData; + +class LLFontVertexBuffer +{ +public: + LLFontVertexBuffer(); + ~LLFontVertexBuffer(); + + void reset(); + + S32 render(const LLFontGL* fontp, + const LLWString& text, + S32 begin_offset, + LLRect rect, + const LLColor4& color, + LLFontGL::HAlign halign = LLFontGL::LEFT, LLFontGL::VAlign valign = LLFontGL::BASELINE, + U8 style = LLFontGL::NORMAL, + LLFontGL::ShadowType shadow = LLFontGL::NO_SHADOW, + S32 max_chars = S32_MAX, S32 max_pixels = S32_MAX, + F32* right_x = NULL, + bool use_ellipses = false, + bool use_color = true); + + S32 render(const LLFontGL* fontp, + const LLWString& text, + S32 begin_offset, + LLRectf rect, + const LLColor4& color, + LLFontGL::HAlign halign = LLFontGL::LEFT, LLFontGL::VAlign valign = LLFontGL::BASELINE, + U8 style = LLFontGL::NORMAL, + LLFontGL::ShadowType shadow = LLFontGL::NO_SHADOW, + S32 max_chars = S32_MAX, + F32* right_x = NULL, + bool use_ellipses = false, + bool use_color = true); + + S32 render(const LLFontGL* fontp, + const LLWString& text, + S32 begin_offset, + F32 x, F32 y, + const LLColor4& color, + LLFontGL::HAlign halign = LLFontGL::LEFT, LLFontGL::VAlign valign = LLFontGL::BASELINE, + U8 style = LLFontGL::NORMAL, + LLFontGL::ShadowType shadow = LLFontGL::NO_SHADOW, + S32 max_chars = S32_MAX, S32 max_pixels = S32_MAX, + F32* right_x = NULL, + bool use_ellipses = false, + bool use_color = true); + + static void enableBufferCollection(bool enable) { sEnableBufferCollection = enable; } +private: + + void genBuffers(const LLFontGL* fontp, + const LLWString& text, + S32 begin_offset, + F32 x, F32 y, + const LLColor4& color, + LLFontGL::HAlign halign, LLFontGL::VAlign valign, + U8 style, + LLFontGL::ShadowType shadow, + S32 max_chars, S32 max_pixels, + F32* right_x, + bool use_ellipses, + bool use_color); + + void renderBuffers(); + + std::list<LLVertexBufferData> mBufferList; + S32 mChars = 0; + const LLFontGL *mLastFont = nullptr; + S32 mLastOffset = 0; + S32 mLastMaxChars = 0; + S32 mLastMaxPixels = 0; + F32 mLastX = 0.f; + F32 mLastY = 0.f; + LLColor4 mLastColor; + LLFontGL::HAlign mLastHalign = LLFontGL::LEFT; + LLFontGL::VAlign mLastValign = LLFontGL::BASELINE; + U8 mLastStyle = LLFontGL::NORMAL; + LLFontGL::ShadowType mLastShadow = LLFontGL::NO_SHADOW; + F32 mLastRightX = 0.f; + + // LLFontGL's statics + F32 mLastScaleX = 1.f; + F32 mLastScaleY = 1.f; + F32 mLastVertDPI = 0.f; + F32 mLastHorizDPI = 0.f; + S32 mLastResGeneration = 0; + LLCoordGL mLastOrigin; + + // Adding new characters to bitmap cache can alter value from getBitmapWidth(); + // which alters whole string. So rerender when new characters were added to cache. + S32 mLastFontCacheGen = 0; + + static bool sEnableBufferCollection; +}; + +#endif diff --git a/indra/llrender/llgl.cpp b/indra/llrender/llgl.cpp index 1b5566a3f7..1bc5e79835 100644 --- a/indra/llrender/llgl.cpp +++ b/indra/llrender/llgl.cpp @@ -50,6 +50,10 @@ #include "llglheaders.h" #include "llglslshader.h" +#include "glm/glm.hpp" +#include <glm/gtc/matrix_access.hpp> +#include "glm/gtc/type_ptr.hpp" + #if LL_WINDOWS #include "lldxhardware.h" #endif @@ -59,12 +63,12 @@ #endif -BOOL gDebugSession = FALSE; -BOOL gDebugGLSession = FALSE; -BOOL gClothRipple = FALSE; -BOOL gHeadlessClient = FALSE; -BOOL gNonInteractive = FALSE; -BOOL gGLActive = FALSE; +bool gDebugSession = false; +bool gDebugGLSession = false; +bool gClothRipple = false; +bool gHeadlessClient = false; +bool gNonInteractive = false; +bool gGLActive = false; static const std::string HEADLESS_VENDOR_STRING("Linden Lab"); static const std::string HEADLESS_RENDERER_STRING("Headless"); @@ -983,21 +987,21 @@ PFNGLPOLYGONOFFSETCLAMPPROC glPolygonOffsetClamp = nullptr; LLGLManager gGLManager; LLGLManager::LLGLManager() : - mInited(FALSE), - mIsDisabled(FALSE), + mInited(false), + mIsDisabled(false), mMaxSamples(0), mNumTextureImageUnits(1), mMaxSampleMaskWords(0), mMaxColorTextureSamples(0), mMaxDepthTextureSamples(0), mMaxIntegerSamples(0), - mIsAMD(FALSE), - mIsNVIDIA(FALSE), - mIsIntel(FALSE), + mIsAMD(false), + mIsNVIDIA(false), + mIsIntel(false), #if LL_DARWIN - mIsMobileGF(FALSE), + mIsMobileGF(false), #endif - mHasRequirements(TRUE), + mHasRequirements(true), mDriverVersionMajor(1), mDriverVersionMinor(0), mDriverVersionRelease(0), @@ -1038,6 +1042,7 @@ void LLGLManager::initWGL() GLH_EXT_NAME(wglGetGPUIDsAMD) = (PFNWGLGETGPUIDSAMDPROC)GLH_EXT_GET_PROC_ADDRESS("wglGetGPUIDsAMD"); GLH_EXT_NAME(wglGetGPUInfoAMD) = (PFNWGLGETGPUINFOAMDPROC)GLH_EXT_GET_PROC_ADDRESS("wglGetGPUInfoAMD"); } + mHasNVXGpuMemoryInfo = ExtensionExists("GL_NVX_gpu_memory_info", gGLHExts.mSysExts); if (ExtensionExists("WGL_EXT_swap_control", gGLHExts.mSysExts)) { @@ -1147,12 +1152,12 @@ bool LLGLManager::initGL() { mGLVendorShort = "AMD"; // *TODO: Fix this? - mIsAMD = TRUE; + mIsAMD = true; } else if (mGLVendor.find("NVIDIA ") != std::string::npos) { mGLVendorShort = "NVIDIA"; - mIsNVIDIA = TRUE; + mIsNVIDIA = true; } else if (mGLVendor.find("INTEL") != std::string::npos #if LL_LINUX @@ -1163,7 +1168,12 @@ bool LLGLManager::initGL() ) { mGLVendorShort = "INTEL"; - mIsIntel = TRUE; + mIsIntel = true; + } + else if (mGLVendor.find("APPLE") != std::string::npos) + { + mGLVendorShort = "APPLE"; + mIsApple = true; } else { @@ -1173,7 +1183,7 @@ bool LLGLManager::initGL() // This is called here because it depends on the setting of mIsGF2or4MX, and sets up mHasMultitexture. initExtensions(); - S32 old_vram = mVRAM; + U32 old_vram = mVRAM; mVRAM = 0; #if LL_WINDOWS @@ -1205,6 +1215,17 @@ bool LLGLManager::initGL() LL_WARNS("RenderInit") << "VRAM Detected (AMDAssociations):" << mVRAM << LL_ENDL; } } + else if (mHasNVXGpuMemoryInfo) + { + GLint mem_kb = 0; + glGetIntegerv(GL_GPU_MEMORY_INFO_DEDICATED_VIDMEM_NVX, &mem_kb); + mVRAM = mem_kb / 1024; + + if (mVRAM != 0) + { + LL_WARNS("RenderInit") << "VRAM Detected (NVXGpuMemoryInfo):" << mVRAM << LL_ENDL; + } + } #endif #if LL_WINDOWS @@ -1215,7 +1236,7 @@ bool LLGLManager::initGL() // Function will check all GPUs WMI knows of and will pick up the one with most // memory. We need to check all GPUs because system can switch active GPU to // weaker one, to preserve power when not under load. - S32 mem = LLDXHardware::getMBVideoMemoryViaWMI(); + U32 mem = LLDXHardware::getMBVideoMemoryViaWMI(); if (mem != 0) { mVRAM = mem; @@ -1238,6 +1259,12 @@ bool LLGLManager::initGL() glGetIntegerv(GL_MAX_INTEGER_SAMPLES, &mMaxIntegerSamples); glGetIntegerv(GL_MAX_SAMPLE_MASK_WORDS, &mMaxSampleMaskWords); glGetIntegerv(GL_MAX_SAMPLES, &mMaxSamples); + glGetIntegerv(GL_MAX_VARYING_VECTORS, &mMaxVaryingVectors); + glGetIntegerv(GL_MAX_UNIFORM_BLOCK_SIZE, &mMaxUniformBlockSize); + + // sanity clamp max uniform block size to 64k just in case + // there's some implementation that reports a crazy value + mMaxUniformBlockSize = llmin(mMaxUniformBlockSize, 65536); if (mGLVersion >= 4.59f) { @@ -1345,7 +1372,7 @@ void LLGLManager::asLLSD(LLSD& info) info["gpu_version"] = mDriverVersionVendorString; info["opengl_version"] = mGLVersionString; - info["vram"] = mVRAM; + info["vram"] = LLSD::Integer(mVRAM); // OpenGL limits info["max_samples"] = mMaxSamples; @@ -1372,12 +1399,12 @@ void LLGLManager::shutdownGL() { glFinish(); stop_glerror(); - mInited = FALSE; + mInited = false; } } // these are used to turn software blending on. They appear in the Debug/Avatar menu -// presence of vertex skinning/blending or vertex programs will set these to FALSE by default. +// presence of vertex skinning/blending or vertex programs will set these to false by default. void LLGLManager::initExtensions() { @@ -1409,7 +1436,7 @@ void LLGLManager::initExtensions() glGetIntegerv(GL_MAX_ELEMENTS_INDICES, (GLint*) &mGLMaxIndexRange); glGetIntegerv(GL_MAX_TEXTURE_SIZE, (GLint*) &mGLMaxTextureSize); - mInited = TRUE; + mInited = true; #if (LL_WINDOWS || LL_LINUX) && !LL_MESA_HEADLESS LL_DEBUGS("RenderInit") << "GL Probe: Getting symbols" << LL_ENDL; @@ -2297,10 +2324,10 @@ void do_assert_glerror() // Create or update texture to be used with this data GLenum error; error = glGetError(); - BOOL quit = FALSE; + bool quit = false; if (LL_UNLIKELY(error)) { - quit = TRUE; + quit = true; GLubyte const * gl_error_msg = gluErrorString(error); if (NULL != gl_error_msg) { @@ -2420,7 +2447,7 @@ void LLGLState::dumpStates() for (boost::unordered_map<LLGLenum, LLGLboolean>::iterator iter = sStateMap.begin(); iter != sStateMap.end(); ++iter) { - LL_INFOS("RenderState") << llformat(" 0x%04x : %s",(S32)iter->first,iter->second?"TRUE":"FALSE") << LL_ENDL; + LL_INFOS("RenderState") << llformat(" 0x%04x : %s",(S32)iter->first,iter->second?"true":"false") << LL_ENDL; } } @@ -2463,7 +2490,7 @@ void LLGLState::checkStates(GLboolean writeAlpha) /////////////////////////////////////////////////////////////////////// LLGLState::LLGLState(LLGLenum state, S32 enabled) : - mState(state), mWasEnabled(FALSE), mIsEnabled(FALSE) + mState(state), mWasEnabled(false), mIsEnabled(false) { LL_PROFILE_ZONE_SCOPED_CATEGORY_PIPELINE; @@ -2482,15 +2509,15 @@ void LLGLState::setEnabled(S32 enabled) } if (enabled == CURRENT_STATE) { - enabled = sStateMap[mState] == GL_TRUE ? TRUE : FALSE; + enabled = sStateMap[mState] == GL_TRUE ? ENABLED_STATE : DISABLED_STATE; } - else if (enabled == TRUE && sStateMap[mState] != GL_TRUE) + else if (enabled == ENABLED_STATE && sStateMap[mState] != GL_TRUE) { gGL.flush(); glEnable(mState); sStateMap[mState] = GL_TRUE; } - else if (enabled == FALSE && sStateMap[mState] != GL_FALSE) + else if (enabled == DISABLED_STATE && sStateMap[mState] != GL_FALSE) { gGL.flush(); glDisable(mState); @@ -2677,7 +2704,7 @@ void parse_glsl_version(S32& major, S32& minor) LLStringUtil::convertToS32(minor_str, minor); } -LLGLUserClipPlane::LLGLUserClipPlane(const LLPlane& p, const glh::matrix4f& modelview, const glh::matrix4f& projection, bool apply) +LLGLUserClipPlane::LLGLUserClipPlane(const LLPlane& p, const glm::mat4& modelview, const glm::mat4& projection, bool apply) { mApply = apply; @@ -2704,13 +2731,12 @@ void LLGLUserClipPlane::disable() void LLGLUserClipPlane::setPlane(F32 a, F32 b, F32 c, F32 d) { - glh::matrix4f& P = mProjection; - glh::matrix4f& M = mModelview; + const glm::mat4& P = mProjection; + const glm::mat4& M = mModelview; - glh::matrix4f invtrans_MVP = (P * M).inverse().transpose(); - glh::vec4f oplane(a,b,c,d); - glh::vec4f cplane; - invtrans_MVP.mult_matrix_vec(oplane, cplane); + glm::mat4 invtrans_MVP = glm::transpose(glm::inverse(P*M)); + glm::vec4 oplane(a,b,c,d); + glm::vec4 cplane = invtrans_MVP * oplane; cplane /= fabs(cplane[2]); // normalize such that depth is not scaled cplane[3] -= 1; @@ -2718,13 +2744,13 @@ void LLGLUserClipPlane::setPlane(F32 a, F32 b, F32 c, F32 d) if(cplane[2] < 0) cplane *= -1; - glh::matrix4f suffix; - suffix.set_row(2, cplane); - glh::matrix4f newP = suffix * P; + glm::mat4 suffix; + suffix = glm::row(suffix, 2, cplane); + glm::mat4 newP = suffix * P; gGL.matrixMode(LLRender::MM_PROJECTION); gGL.pushMatrix(); - gGL.loadMatrix(newP.m); - gGLObliqueProjectionInverse = LLMatrix4(newP.inverse().transpose().m); + gGL.loadMatrix(glm::value_ptr(newP)); + gGLObliqueProjectionInverse = LLMatrix4(glm::value_ptr(glm::transpose(glm::inverse(newP)))); gGL.matrixMode(LLRender::MM_MODELVIEW); } @@ -2737,14 +2763,14 @@ LLGLDepthTest::LLGLDepthTest(GLboolean depth_enabled, GLboolean write_enabled, G : mPrevDepthEnabled(sDepthEnabled), mPrevDepthFunc(sDepthFunc), mPrevWriteEnabled(sWriteEnabled) { stop_glerror(); - + LL_PROFILE_ZONE_SCOPED_CATEGORY_PIPELINE; checkState(); if (!depth_enabled) { // always disable depth writes if depth testing is disabled // GL spec defines this as a requirement, but some implementations allow depth writes with testing disabled // The proper way to write to depth buffer with testing disabled is to enable testing and use a depth_func of GL_ALWAYS - write_enabled = FALSE; + write_enabled = GL_FALSE; } if (depth_enabled != sDepthEnabled) @@ -2770,6 +2796,7 @@ LLGLDepthTest::LLGLDepthTest(GLboolean depth_enabled, GLboolean write_enabled, G LLGLDepthTest::~LLGLDepthTest() { + LL_PROFILE_ZONE_SCOPED_CATEGORY_PIPELINE; checkState(); if (sDepthEnabled != mPrevDepthEnabled ) { @@ -2797,7 +2824,7 @@ void LLGLDepthTest::checkState() if (gDebugGL) { GLint func = 0; - GLboolean mask = FALSE; + GLboolean mask = GL_FALSE; glGetIntegerv(GL_DEPTH_FUNC, &func); glGetBooleanv(GL_DEPTH_WRITEMASK, &mask); @@ -2820,31 +2847,27 @@ void LLGLDepthTest::checkState() LLGLSquashToFarClip::LLGLSquashToFarClip() { - glh::matrix4f proj = get_current_projection(); + glm::mat4 proj = get_current_projection(); setProjectionMatrix(proj, 0); } -LLGLSquashToFarClip::LLGLSquashToFarClip(glh::matrix4f& P, U32 layer) +LLGLSquashToFarClip::LLGLSquashToFarClip(const glm::mat4& P, U32 layer) { setProjectionMatrix(P, layer); } - -void LLGLSquashToFarClip::setProjectionMatrix(glh::matrix4f& projection, U32 layer) +void LLGLSquashToFarClip::setProjectionMatrix(glm::mat4 projection, U32 layer) { - F32 depth = 0.99999f - 0.0001f * layer; - for (U32 i = 0; i < 4; i++) - { - projection.element(2, i) = projection.element(3, i) * depth; - } + glm::vec4 P_row_3 = glm::row(projection, 3) * depth; + projection = glm::row(projection, 2, P_row_3); LLRender::eMatrixMode last_matrix_mode = gGL.getMatrixMode(); gGL.matrixMode(LLRender::MM_PROJECTION); gGL.pushMatrix(); - gGL.loadMatrix(projection.m); + gGL.loadMatrix(glm::value_ptr(projection)); gGL.matrixMode(last_matrix_mode); } diff --git a/indra/llrender/llgl.h b/indra/llrender/llgl.h index 5a7ad943df..24ba4d6deb 100644 --- a/indra/llrender/llgl.h +++ b/indra/llrender/llgl.h @@ -43,11 +43,11 @@ #include "llinstancetracker.h" #include "llglheaders.h" -#include "glh/glh_linear.h" +#include "glm/mat4x4.hpp" -extern BOOL gDebugGL; -extern BOOL gDebugSession; -extern BOOL gDebugGLSession; +extern bool gDebugGL; +extern bool gDebugSession; +extern bool gDebugGLSession; extern llofstream gFailLog; #define LL_GL_ERRS LL_ERRS("RenderState") @@ -73,8 +73,8 @@ public: std::string getRawGLString(); // For sending to simulator - BOOL mInited; - BOOL mIsDisabled; + bool mInited; + bool mIsDisabled; // OpenGL limits S32 mMaxSamples; @@ -87,6 +87,8 @@ public: S32 mGLMaxIndexRange; S32 mGLMaxTextureSize; F32 mMaxAnisotropy = 0.f; + S32 mMaxUniformBlockSize = 0; + S32 mMaxVaryingVectors = 0; // GL 4.x capabilities bool mHasCubeMapArray = false; @@ -96,18 +98,23 @@ public: // Vendor-specific extensions bool mHasAMDAssociations = false; + bool mHasNVXGpuMemoryInfo = false; - BOOL mIsAMD; - BOOL mIsNVIDIA; - BOOL mIsIntel; + bool mIsAMD; + bool mIsNVIDIA; + bool mIsIntel; + bool mIsApple = false; + + // hints to the render pipe + U32 mDownScaleMethod = 0; // see settings.xml RenderDownScaleMethod #if LL_DARWIN // Needed to distinguish problem cards on older Macs that break with Materials - BOOL mIsMobileGF; + bool mIsMobileGF; #endif // Whether this version of GL is good enough for SL to use - BOOL mHasRequirements; + bool mHasRequirements; S32 mDriverVersionMajor; S32 mDriverVersionMinor; @@ -118,9 +125,7 @@ public: std::string mDriverVersionVendorString; std::string mGLVersionString; - S32 mVRAM; // VRAM in MB - - void getPixelFormat(); // Get the best pixel format + U32 mVRAM; // VRAM in MB std::string getGLInfoString(); void printGLInfoString(); @@ -138,7 +143,6 @@ public: private: void initExtensions(); void initGLStates(); - void initGLImages(); }; extern LLGLManager gGLManager; @@ -155,13 +159,18 @@ void assert_glerror(); void clear_glerror(); -//#if LL_DEBUG + # define stop_glerror() assert_glerror() # define llglassertok() assert_glerror() -//#else -//# define stop_glerror() -//# define llglassertok() -//#endif + +// stop_glerror is still needed on OS X but has performance implications +// use macro below to conditionally add stop_glerror to non-release builds +// on OS X +#if LL_DARWIN && !LL_RELEASE_FOR_DOWNLOAD +#define STOP_GLERROR stop_glerror() +#else +#define STOP_GLERROR +#endif #define llglassertok_always() assert_glerror() @@ -241,16 +250,16 @@ protected: static boost::unordered_map<LLGLenum, LLGLboolean> sStateMap; public: - enum { CURRENT_STATE = -2 }; + enum { CURRENT_STATE = -2, DISABLED_STATE = 0, ENABLED_STATE = 1 }; LLGLState(LLGLenum state, S32 enabled = CURRENT_STATE); ~LLGLState(); void setEnabled(S32 enabled); - void enable() { setEnabled(TRUE); } - void disable() { setEnabled(FALSE); } + void enable() { setEnabled(ENABLED_STATE); } + void disable() { setEnabled(DISABLED_STATE); } protected: LLGLenum mState; - BOOL mWasEnabled; - BOOL mIsEnabled; + bool mWasEnabled; + bool mIsEnabled; }; // New LLGLState class wrappers that don't depend on actual GL flags. @@ -284,14 +293,14 @@ public: class LLGLEnable : public LLGLState { public: - LLGLEnable(LLGLenum state) : LLGLState(state, TRUE) {} + LLGLEnable(LLGLenum state) : LLGLState(state, ENABLED_STATE) {} }; /// TODO: Being deprecated. class LLGLDisable : public LLGLState { public: - LLGLDisable(LLGLenum state) : LLGLState(state, FALSE) {} + LLGLDisable(LLGLenum state) : LLGLState(state, DISABLED_STATE) {} }; /* @@ -310,7 +319,7 @@ class LLGLUserClipPlane { public: - LLGLUserClipPlane(const LLPlane& plane, const glh::matrix4f& modelview, const glh::matrix4f& projection, bool apply = true); + LLGLUserClipPlane(const LLPlane& plane, const glm::mat4& modelview, const glm::mat4& projection, bool apply = true); ~LLGLUserClipPlane(); void setPlane(F32 a, F32 b, F32 c, F32 d); @@ -319,8 +328,8 @@ public: private: bool mApply; - glh::matrix4f mProjection; - glh::matrix4f mModelview; + glm::mat4 mProjection; + glm::mat4 mModelview; }; /* @@ -334,9 +343,9 @@ class LLGLSquashToFarClip { public: LLGLSquashToFarClip(); - LLGLSquashToFarClip(glh::matrix4f& projection, U32 layer = 0); + LLGLSquashToFarClip(const glm::mat4& projection, U32 layer = 0); - void setProjectionMatrix(glh::matrix4f& projection, U32 layer); + void setProjectionMatrix(glm::mat4 projection, U32 layer); ~LLGLSquashToFarClip(); }; @@ -351,9 +360,9 @@ public: static std::list<LLGLUpdate*> sGLQ; - BOOL mInQ; + bool mInQ; LLGLUpdate() - : mInQ(FALSE) + : mInQ(false) { } virtual ~LLGLUpdate() @@ -405,10 +414,10 @@ void init_glstates(); void parse_gl_version( S32* major, S32* minor, S32* release, std::string* vendor_specific, std::string* version_string ); -extern BOOL gClothRipple; -extern BOOL gHeadlessClient; -extern BOOL gNonInteractive; -extern BOOL gGLActive; +extern bool gClothRipple; +extern bool gHeadlessClient; +extern bool gNonInteractive; +extern bool gGLActive; // Deal with changing glext.h definitions for newer SDK versions, specifically // with MAC OSX 10.5 -> 10.6 diff --git a/indra/llrender/llglheaders.h b/indra/llrender/llglheaders.h index c5e1ff3e23..3d4dc5e698 100644 --- a/indra/llrender/llglheaders.h +++ b/indra/llrender/llglheaders.h @@ -46,7 +46,7 @@ // LL_WINDOWS // windows gl headers depend on things like APIENTRY, so include windows. -#include "llwin32headerslean.h" +#include "llwin32headers.h" //---------------------------------------------------------------------------- #include <GL/gl.h> diff --git a/indra/llrender/llglslshader.cpp b/indra/llrender/llglslshader.cpp index b6fd78eefa..b3f32fdc83 100644 --- a/indra/llrender/llglslshader.cpp +++ b/indra/llrender/llglslshader.cpp @@ -54,6 +54,8 @@ using std::string; GLuint LLGLSLShader::sCurBoundShader = 0; LLGLSLShader* LLGLSLShader::sCurBoundShaderPtr = NULL; S32 LLGLSLShader::sIndexedTextureChannels = 0; +U32 LLGLSLShader::sMaxGLTFMaterials = 0; +U32 LLGLSLShader::sMaxGLTFNodes = 0; bool LLGLSLShader::sProfileEnabled = false; std::set<LLGLSLShader*> LLGLSLShader::sInstances; LLGLSLShader::defines_map_t LLGLSLShader::sGlobalDefines; @@ -61,6 +63,7 @@ U64 LLGLSLShader::sTotalTimeElapsed = 0; U32 LLGLSLShader::sTotalTrianglesDrawn = 0; U64 LLGLSLShader::sTotalSamplesDrawn = 0; U32 LLGLSLShader::sTotalBinds = 0; +boost::json::value LLGLSLShader::sDefaultStats; //UI shader -- declared here so llui_libtest will link properly LLGLSLShader gUIProgram; @@ -81,7 +84,7 @@ const std::string gShaderConstsVal[LLGLSLShader::NUM_SHADER_CONSTS] = }; -BOOL shouldChange(const LLVector4& v1, const LLVector4& v2) +bool shouldChange(const LLVector4& v1, const LLVector4& v2) { return v1 != v2; } @@ -99,9 +102,9 @@ void LLGLSLShader::initProfile() sTotalSamplesDrawn = 0; sTotalBinds = 0; - for (std::set<LLGLSLShader*>::iterator iter = sInstances.begin(); iter != sInstances.end(); ++iter) + for (auto ptr : sInstances) { - (*iter)->clearStats(); + ptr->clearStats(); } } @@ -115,45 +118,57 @@ struct LLGLSLShaderCompareTimeElapsed }; //static -void LLGLSLShader::finishProfile(bool emit_report) +void LLGLSLShader::finishProfile(boost::json::value& statsv) { sProfileEnabled = false; - if (emit_report) + if (! statsv.is_null()) { - std::vector<LLGLSLShader*> sorted; - - for (std::set<LLGLSLShader*>::iterator iter = sInstances.begin(); iter != sInstances.end(); ++iter) - { - sorted.push_back(*iter); - } - + std::vector<LLGLSLShader*> sorted(sInstances.begin(), sInstances.end()); std::sort(sorted.begin(), sorted.end(), LLGLSLShaderCompareTimeElapsed()); + auto& stats = statsv.as_object(); + auto shadersit = stats.emplace("shaders", boost::json::array_kind).first; + auto& shaders = shadersit->value().as_array(); bool unbound = false; - for (std::vector<LLGLSLShader*>::iterator iter = sorted.begin(); iter != sorted.end(); ++iter) + for (auto ptr : sorted) { - (*iter)->dumpStats(); - if ((*iter)->mBinds == 0) + if (ptr->mBinds == 0) { unbound = true; } + else + { + auto& shaderit = shaders.emplace_back(boost::json::object_kind); + ptr->dumpStats(shaderit.as_object()); + } } + constexpr float mega = 1'000'000.f; + float totalTimeMs = sTotalTimeElapsed / mega; LL_INFOS() << "-----------------------------------" << LL_ENDL; - LL_INFOS() << "Total rendering time: " << llformat("%.4f ms", sTotalTimeElapsed / 1000000.f) << LL_ENDL; - LL_INFOS() << "Total samples drawn: " << llformat("%.4f million", sTotalSamplesDrawn / 1000000.f) << LL_ENDL; - LL_INFOS() << "Total triangles drawn: " << llformat("%.3f million", sTotalTrianglesDrawn / 1000000.f) << LL_ENDL; + LL_INFOS() << "Total rendering time: " << llformat("%.4f ms", totalTimeMs) << LL_ENDL; + LL_INFOS() << "Total samples drawn: " << llformat("%.4f million", sTotalSamplesDrawn / mega) << LL_ENDL; + LL_INFOS() << "Total triangles drawn: " << llformat("%.3f million", sTotalTrianglesDrawn / mega) << LL_ENDL; LL_INFOS() << "-----------------------------------" << LL_ENDL; - + auto totalsit = stats.emplace("totals", boost::json::object_kind).first; + auto& totals = totalsit->value().as_object(); + totals.emplace("time", totalTimeMs / 1000.0); + totals.emplace("binds", sTotalBinds); + totals.emplace("samples", sTotalSamplesDrawn); + totals.emplace("triangles", sTotalTrianglesDrawn); + + auto unusedit = stats.emplace("unused", boost::json::array_kind).first; + auto& unused = unusedit->value().as_array(); if (unbound) { LL_INFOS() << "The following shaders were unused: " << LL_ENDL; - for (std::vector<LLGLSLShader*>::iterator iter = sorted.begin(); iter != sorted.end(); ++iter) + for (auto ptr : sorted) { - if ((*iter)->mBinds == 0) + if (ptr->mBinds == 0) { - LL_INFOS() << (*iter)->mName << LL_ENDL; + LL_INFOS() << ptr->mName << LL_ENDL; + unused.emplace_back(ptr->mName); } } } @@ -168,36 +183,43 @@ void LLGLSLShader::clearStats() mBinds = 0; } -void LLGLSLShader::dumpStats() +void LLGLSLShader::dumpStats(boost::json::object& stats) { - if (mBinds > 0) + stats.emplace("name", mName); + auto filesit = stats.emplace("files", boost::json::array_kind).first; + auto& files = filesit->value().as_array(); + LL_INFOS() << "=============================================" << LL_ENDL; + LL_INFOS() << mName << LL_ENDL; + for (U32 i = 0; i < mShaderFiles.size(); ++i) { - LL_INFOS() << "=============================================" << LL_ENDL; - LL_INFOS() << mName << LL_ENDL; - for (U32 i = 0; i < mShaderFiles.size(); ++i) - { - LL_INFOS() << mShaderFiles[i].first << LL_ENDL; - } - LL_INFOS() << "=============================================" << LL_ENDL; + LL_INFOS() << mShaderFiles[i].first << LL_ENDL; + files.emplace_back(mShaderFiles[i].first); + } + LL_INFOS() << "=============================================" << LL_ENDL; - F32 ms = mTimeElapsed / 1000000.f; - F32 seconds = ms / 1000.f; + constexpr float mega = 1'000'000.f; + constexpr double giga = 1'000'000'000.0; + F32 ms = mTimeElapsed / mega; + F32 seconds = ms / 1000.f; - F32 pct_tris = (F32)mTrianglesDrawn / (F32)sTotalTrianglesDrawn * 100.f; - F32 tris_sec = (F32)(mTrianglesDrawn / 1000000.0); - tris_sec /= seconds; + F32 pct_tris = (F32)mTrianglesDrawn / (F32)sTotalTrianglesDrawn * 100.f; + F32 tris_sec = (F32)(mTrianglesDrawn / mega); + tris_sec /= seconds; - F32 pct_samples = (F32)((F64)mSamplesDrawn / (F64)sTotalSamplesDrawn) * 100.f; - F32 samples_sec = (F32)mSamplesDrawn / 1000000000.0; - samples_sec /= seconds; + F32 pct_samples = (F32)((F64)mSamplesDrawn / (F64)sTotalSamplesDrawn) * 100.f; + F32 samples_sec = (F32)(mSamplesDrawn / giga); + samples_sec /= seconds; - F32 pct_binds = (F32)mBinds / (F32)sTotalBinds * 100.f; + F32 pct_binds = (F32)mBinds / (F32)sTotalBinds * 100.f; - LL_INFOS() << "Triangles Drawn: " << mTrianglesDrawn << " " << llformat("(%.2f pct of total, %.3f million/sec)", pct_tris, tris_sec) << LL_ENDL; - LL_INFOS() << "Binds: " << mBinds << " " << llformat("(%.2f pct of total)", pct_binds) << LL_ENDL; - LL_INFOS() << "SamplesDrawn: " << mSamplesDrawn << " " << llformat("(%.2f pct of total, %.3f billion/sec)", pct_samples, samples_sec) << LL_ENDL; - LL_INFOS() << "Time Elapsed: " << mTimeElapsed << " " << llformat("(%.2f pct of total, %.5f ms)\n", (F32)((F64)mTimeElapsed / (F64)sTotalTimeElapsed) * 100.f, ms) << LL_ENDL; - } + LL_INFOS() << "Triangles Drawn: " << mTrianglesDrawn << " " << llformat("(%.2f pct of total, %.3f million/sec)", pct_tris, tris_sec) << LL_ENDL; + LL_INFOS() << "Binds: " << mBinds << " " << llformat("(%.2f pct of total)", pct_binds) << LL_ENDL; + LL_INFOS() << "SamplesDrawn: " << mSamplesDrawn << " " << llformat("(%.2f pct of total, %.3f billion/sec)", pct_samples, samples_sec) << LL_ENDL; + LL_INFOS() << "Time Elapsed: " << mTimeElapsed << " " << llformat("(%.2f pct of total, %.5f ms)\n", (F32)((F64)mTimeElapsed / (F64)sTotalTimeElapsed) * 100.f, ms) << LL_ENDL; + stats.emplace("time", seconds); + stats.emplace("binds", mBinds); + stats.emplace("samples", mSamplesDrawn); + stats.emplace("triangles", mTrianglesDrawn); } //static @@ -279,7 +301,7 @@ bool LLGLSLShader::readProfileQuery(bool for_runtime, bool force_read) GLuint64 samples_passed = 0; glGetQueryObjectui64v(mSamplesQuery, GL_QUERY_RESULT, &samples_passed); - U64 primitives_generated = 0; + GLuint64 primitives_generated = 0; glGetQueryObjectui64v(mPrimitivesQuery, GL_QUERY_RESULT, &primitives_generated); sTotalTimeElapsed += time_elapsed; @@ -309,7 +331,7 @@ LLGLSLShader::LLGLSLShader() mShaderLevel(0), mShaderGroup(SG_DEFAULT), mFeatures(), - mUniformsDirty(FALSE), + mUniformsDirty(false), mTimerQuery(0), mSamplesQuery(0), mPrimitivesQuery(0) @@ -381,10 +403,7 @@ void LLGLSLShader::unloadInternal() stop_glerror(); } -BOOL LLGLSLShader::createShader(std::vector<LLStaticHashedString>* attributes, - std::vector<LLStaticHashedString>* uniforms, - U32 varying_count, - const char** varyings) +bool LLGLSLShader::createShader() { LL_PROFILE_ZONE_SCOPED_CATEGORY_SHADER; @@ -415,10 +434,10 @@ BOOL LLGLSLShader::createShader(std::vector<LLStaticHashedString>* attributes, // Shouldn't happen if shader related extensions, like ARB_vertex_shader, exist. LL_SHADER_LOADING_WARNS() << "Failed to create handle for shader: " << mName << LL_ENDL; unloadInternal(); - return FALSE; + return false; } - BOOL success = TRUE; + bool success = true; mUsingBinaryProgram = LLShaderMgr::instance()->loadCachedProgramBinary(this); @@ -440,7 +459,7 @@ BOOL LLGLSLShader::createShader(std::vector<LLStaticHashedString>* attributes, } else { - success = FALSE; + success = false; } } } @@ -449,16 +468,16 @@ BOOL LLGLSLShader::createShader(std::vector<LLStaticHashedString>* attributes, if (!LLShaderMgr::instance()->attachShaderFeatures(this)) { unloadInternal(); - return FALSE; + return false; } // Map attributes and uniforms if (success) { - success = mapAttributes(attributes); + success = mapAttributes(); } if (success) { - success = mapUniforms(uniforms); + success = mapUniforms(); } if (!success) { @@ -469,7 +488,7 @@ BOOL LLGLSLShader::createShader(std::vector<LLStaticHashedString>* attributes, { LL_SHADER_LOADING_WARNS() << "Failed to link using shader level " << mShaderLevel << " trying again using shader level " << (mShaderLevel - 1) << LL_ENDL; mShaderLevel--; - return createShader(attributes, uniforms); + return createShader(); } else { @@ -479,6 +498,7 @@ BOOL LLGLSLShader::createShader(std::vector<LLStaticHashedString>* attributes, } else if (mFeatures.mIndexedTextureChannels > 0) { //override texture channels for indexed texture rendering + llassert(mFeatures.mIndexedTextureChannels == LLGLSLShader::sIndexedTextureChannels); // these numbers must always match bind(); S32 channel_count = mFeatures.mIndexedTextureChannels; @@ -488,19 +508,41 @@ BOOL LLGLSLShader::createShader(std::vector<LLStaticHashedString>* attributes, uniform1i(uniName, i); } - S32 cur_tex = channel_count; //adjust any texture channels that might have been overwritten + //adjust any texture channels that might have been overwritten for (U32 i = 0; i < mTexture.size(); i++) { - if (mTexture[i] > -1 && mTexture[i] < channel_count) + if (mTexture[i] > -1) { - llassert(cur_tex < gGLManager.mNumTextureImageUnits); - uniform1i(i, cur_tex); - mTexture[i] = cur_tex++; + S32 new_tex = mTexture[i] + channel_count; + uniform1i(i, new_tex); + mTexture[i] = new_tex; } } + + // get the true number of active texture channels + mActiveTextureChannels = channel_count; + for (auto& tex : mTexture) + { + mActiveTextureChannels = llmax(mActiveTextureChannels, tex + 1); + } + + // when indexed texture channels are used, enforce an upper limit of 16 + // this should act as a canary in the coal mine for adding textures + // and breaking machines that are limited to 16 texture channels + llassert(mActiveTextureChannels <= 16); unbind(); } + LL_DEBUGS("GLSLTextureChannels") << mName << " has " << mActiveTextureChannels << " active texture channels" << LL_ENDL; + + for (U32 i = 0; i < mTexture.size(); i++) + { + if (mTexture[i] > -1) + { + LL_DEBUGS("GLSLTextureChannels") << "Texture " << LLShaderMgr::instance()->mReservedUniforms[i] << " assigned to channel " << mTexture[i] << LL_ENDL; + } + } + #ifdef LL_PROFILER_ENABLE_RENDER_DOC setLabel(mName.c_str()); #endif @@ -529,7 +571,7 @@ void dumpAttachObject(const char* func_name, GLuint program_object, const std::s } #endif // DEBUG_SHADER_INCLUDES -BOOL LLGLSLShader::attachVertexObject(std::string object_path) +bool LLGLSLShader::attachVertexObject(std::string object_path) { if (LLShaderMgr::instance()->mVertexShaderObjects.count(object_path) > 0) { @@ -539,19 +581,19 @@ BOOL LLGLSLShader::attachVertexObject(std::string object_path) dumpAttachObject("attachVertexObject", mProgramObject, object_path); #endif // DEBUG_SHADER_INCLUDES stop_glerror(); - return TRUE; + return true; } else { LL_SHADER_LOADING_WARNS() << "Attempting to attach shader object: '" << object_path << "' that hasn't been compiled." << LL_ENDL; - return FALSE; + return false; } } -BOOL LLGLSLShader::attachFragmentObject(std::string object_path) +bool LLGLSLShader::attachFragmentObject(std::string object_path) { if(mUsingBinaryProgram) - return TRUE; + return true; if (LLShaderMgr::instance()->mFragmentShaderObjects.count(object_path) > 0) { @@ -561,12 +603,12 @@ BOOL LLGLSLShader::attachFragmentObject(std::string object_path) dumpAttachObject("attachFragmentObject", mProgramObject, object_path); #endif // DEBUG_SHADER_INCLUDES stop_glerror(); - return TRUE; + return true; } else { LL_SHADER_LOADING_WARNS() << "Attempting to attach shader object: '" << object_path << "' that hasn't been compiled." << LL_ENDL; - return FALSE; + return false; } } @@ -602,11 +644,11 @@ void LLGLSLShader::attachObjects(GLuint* objects, S32 count) } } -BOOL LLGLSLShader::mapAttributes(const std::vector<LLStaticHashedString>* attributes) +bool LLGLSLShader::mapAttributes() { LL_PROFILE_ZONE_SCOPED_CATEGORY_SHADER; - BOOL res = TRUE; + bool res = true; if (!mUsingBinaryProgram) { //before linking, make sure reserved attributes always have consistent locations @@ -621,11 +663,10 @@ BOOL LLGLSLShader::mapAttributes(const std::vector<LLStaticHashedString>* attrib } mAttribute.clear(); - U32 numAttributes = (attributes == NULL) ? 0 : attributes->size(); #if LL_RELEASE_WITH_DEBUG_INFO - mAttribute.resize(LLShaderMgr::instance()->mReservedAttribs.size() + numAttributes, { -1, NULL }); + mAttribute.resize(LLShaderMgr::instance()->mReservedAttribs.size(), { -1, NULL }); #else - mAttribute.resize(LLShaderMgr::instance()->mReservedAttribs.size() + numAttributes, -1); + mAttribute.resize(LLShaderMgr::instance()->mReservedAttribs.size(), -1); #endif if (res) @@ -649,27 +690,14 @@ BOOL LLGLSLShader::mapAttributes(const std::vector<LLStaticHashedString>* attrib LL_DEBUGS("ShaderUniform") << "Attribute " << name << " assigned to channel " << index << LL_ENDL; } } - if (attributes != NULL) - { - for (U32 i = 0; i < numAttributes; i++) - { - const char* name = (*attributes)[i].String().c_str(); - S32 index = glGetAttribLocation(mProgramObject, name); - if (index != -1) - { - mAttribute[LLShaderMgr::instance()->mReservedAttribs.size() + i] = index; - LL_DEBUGS("ShaderUniform") << "Attribute " << name << " assigned to channel " << index << LL_ENDL; - } - } - } - return TRUE; + return true; } - return FALSE; + return false; } -void LLGLSLShader::mapUniform(GLint index, const vector<LLStaticHashedString>* uniforms) +void LLGLSLShader::mapUniform(GLint index) { LL_PROFILE_ZONE_SCOPED_CATEGORY_SHADER; @@ -753,22 +781,11 @@ void LLGLSLShader::mapUniform(GLint index, const vector<LLStaticHashedString>* u //found it mUniform[i] = location; mTexture[i] = mapUniformTextureChannel(location, type, size); - return; - } - } - - if (uniforms != NULL) - { - for (U32 i = 0; i < uniforms->size(); i++) - { - if ((mUniform[i + LLShaderMgr::instance()->mReservedUniforms.size()] == -1) - && ((*uniforms)[i].String() == name)) + if (mTexture[i] != -1) { - //found it - mUniform[i + LLShaderMgr::instance()->mReservedUniforms.size()] = location; - mTexture[i + LLShaderMgr::instance()->mReservedUniforms.size()] = mapUniformTextureChannel(location, type, size); - return; + LL_DEBUGS("GLSLTextureChannels") << name << " assigned to texture channel " << mTexture[i] << LL_ENDL; } + return; } } } @@ -806,35 +823,31 @@ GLint LLGLSLShader::mapUniformTextureChannel(GLint location, GLenum type, GLint if (size == 1) { glUniform1i(location, mActiveTextureChannels); - LL_DEBUGS("ShaderUniform") << "Assigned to texture channel " << mActiveTextureChannels << LL_ENDL; mActiveTextureChannels++; } else { //is array of textures, make sequential after this texture - GLint channel[32]; // <=== only support up to 32 texture channels - llassert(size <= 32); - size = llmin(size, 32); + GLint channel[16]; // <=== only support up to 16 texture channels + llassert(size <= 16); + size = llmin(size, 16); for (int i = 0; i < size; ++i) { channel[i] = mActiveTextureChannels++; } glUniform1iv(location, size, channel); - LL_DEBUGS("ShaderUniform") << "Assigned to texture channel " << - (mActiveTextureChannels - size) << " through " << (mActiveTextureChannels - 1) << LL_ENDL; } - llassert(mActiveTextureChannels <= 32); // too many textures (probably) return ret; } return -1; } -BOOL LLGLSLShader::mapUniforms(const vector<LLStaticHashedString>* uniforms) +bool LLGLSLShader::mapUniforms() { LL_PROFILE_ZONE_SCOPED_CATEGORY_SHADER; - BOOL res = TRUE; + bool res = true; mTotalUniformSize = 0; mActiveTextureChannels = 0; @@ -843,9 +856,8 @@ BOOL LLGLSLShader::mapUniforms(const vector<LLStaticHashedString>* uniforms) mTexture.clear(); mValue.clear(); //initialize arrays - U32 numUniforms = (uniforms == NULL) ? 0 : uniforms->size(); - mUniform.resize(numUniforms + LLShaderMgr::instance()->mReservedUniforms.size(), -1); - mTexture.resize(numUniforms + LLShaderMgr::instance()->mReservedUniforms.size(), -1); + mUniform.resize(LLShaderMgr::instance()->mReservedUniforms.size(), -1); + mTexture.resize(LLShaderMgr::instance()->mReservedUniforms.size(), -1); bind(); @@ -946,26 +958,26 @@ BOOL LLGLSLShader::mapUniforms(const vector<LLStaticHashedString>* uniforms) if (specularDiff || bumpLessDiff || envLessDiff || refLessDiff) { - mapUniform(diffuseMap, uniforms); + mapUniform(diffuseMap); skip_index.insert(diffuseMap); if (-1 != specularMap) { - mapUniform(specularMap, uniforms); + mapUniform(specularMap); skip_index.insert(specularMap); } if (-1 != bumpMap) { - mapUniform(bumpMap, uniforms); + mapUniform(bumpMap); skip_index.insert(bumpMap); } if (-1 != environmentMap) { - mapUniform(environmentMap, uniforms); + mapUniform(environmentMap); skip_index.insert(environmentMap); } if (-1 != reflectionMap) { - mapUniform(reflectionMap, uniforms); + mapUniform(reflectionMap); skip_index.insert(reflectionMap); } } @@ -979,21 +991,31 @@ BOOL LLGLSLShader::mapUniforms(const vector<LLStaticHashedString>* uniforms) if (skip_index.end() != skip_index.find(i)) continue; //........................................................................................ - mapUniform(i, uniforms); + mapUniform(i); } //........................................................................................................................................ - if (mFeatures.hasReflectionProbes) // Set up block binding, in a way supported by Apple (rather than binding = 1 in .glsl). - { // See slide 35 and more of https://docs.huihoo.com/apple/wwdc/2011/session_420__advances_in_opengl_for_mac_os_x_lion.pdf - static const GLuint BLOCKBINDING = 1; //picked by us - //Get the index, similar to a uniform location - GLuint UBOBlockIndex = glGetUniformBlockIndex(mProgramObject, "ReflectionProbes"); + // Set up block binding, in a way supported by Apple (rather than binding = 1 in .glsl). + // See slide 35 and more of https://docs.huihoo.com/apple/wwdc/2011/session_420__advances_in_opengl_for_mac_os_x_lion.pdf + const char* ubo_names[] = + { + "ReflectionProbes", // UB_REFLECTION_PROBES + "GLTFJoints", // UB_GLTF_JOINTS + "GLTFNodes", // UB_GLTF_NODES + "GLTFMaterials", // UB_GLTF_MATERIALS + }; + + llassert(LL_ARRAY_SIZE(ubo_names) == NUM_UNIFORM_BLOCKS); + + for (U32 i = 0; i < NUM_UNIFORM_BLOCKS; ++i) + { + GLuint UBOBlockIndex = glGetUniformBlockIndex(mProgramObject, ubo_names[i]); if (UBOBlockIndex != GL_INVALID_INDEX) { - //Set this index to a binding index - glUniformBlockBinding(mProgramObject, UBOBlockIndex, BLOCKBINDING); + glUniformBlockBinding(mProgramObject, UBOBlockIndex, i); } } + unbind(); LL_DEBUGS("ShaderUniform") << "Total Uniform Size: " << mTotalUniformSize << LL_ENDL; @@ -1001,11 +1023,11 @@ BOOL LLGLSLShader::mapUniforms(const vector<LLStaticHashedString>* uniforms) } -BOOL LLGLSLShader::link(BOOL suppress_errors) +bool LLGLSLShader::link(bool suppress_errors) { LL_PROFILE_ZONE_SCOPED_CATEGORY_SHADER; - BOOL success = LLShaderMgr::instance()->linkProgramObject(mProgramObject, suppress_errors); + bool success = LLShaderMgr::instance()->linkProgramObject(mProgramObject, suppress_errors); if (!success && !suppress_errors) { @@ -1024,7 +1046,7 @@ void LLGLSLShader::bind() { LL_PROFILE_ZONE_SCOPED_CATEGORY_SHADER; - llassert(mProgramObject != 0); + llassert_always(mProgramObject != 0); gGL.flush(); @@ -1045,8 +1067,18 @@ void LLGLSLShader::bind() if (mUniformsDirty) { LLShaderMgr::instance()->updateShaderUniforms(this); - mUniformsDirty = FALSE; + mUniformsDirty = false; } + + llassert_always(sCurBoundShaderPtr != nullptr); + llassert_always(sCurBoundShader == mProgramObject); +} + +void LLGLSLShader::bind(U8 variant) +{ + llassert(mGLTFVariants.size() == LLGLSLShader::NUM_GLTF_VARIANTS); + llassert(variant < LLGLSLShader::NUM_GLTF_VARIANTS); + mGLTFVariants[variant].bind(); } void LLGLSLShader::bind(bool rigged) @@ -1078,17 +1110,17 @@ void LLGLSLShader::unbind(void) sCurBoundShaderPtr = NULL; } -S32 LLGLSLShader::bindTexture(const std::string& uniform, LLTexture* texture, LLTexUnit::eTextureType mode, LLTexUnit::eTextureColorSpace colorspace) +S32 LLGLSLShader::bindTexture(const std::string& uniform, LLTexture* texture, LLTexUnit::eTextureType mode) { LL_PROFILE_ZONE_SCOPED_CATEGORY_SHADER; S32 channel = 0; channel = getUniformLocation(uniform); - return bindTexture(channel, texture, mode, colorspace); + return bindTexture(channel, texture, mode); } -S32 LLGLSLShader::bindTexture(S32 uniform, LLTexture* texture, LLTexUnit::eTextureType mode, LLTexUnit::eTextureColorSpace colorspace) +S32 LLGLSLShader::bindTexture(S32 uniform, LLTexture* texture, LLTexUnit::eTextureType mode) { LL_PROFILE_ZONE_SCOPED_CATEGORY_SHADER; @@ -1104,7 +1136,6 @@ S32 LLGLSLShader::bindTexture(S32 uniform, LLTexture* texture, LLTexUnit::eTextu if (uniform > -1) { gGL.getTexUnit(uniform)->bindFast(texture); - gGL.getTexUnit(uniform)->setTextureColorSpace(colorspace); } return uniform; @@ -1185,7 +1216,7 @@ S32 LLGLSLShader::getTextureChannel(S32 uniform) const return mTexture[uniform]; } -S32 LLGLSLShader::enableTexture(S32 uniform, LLTexUnit::eTextureType mode, LLTexUnit::eTextureColorSpace space) +S32 LLGLSLShader::enableTexture(S32 uniform, LLTexUnit::eTextureType mode) { LL_PROFILE_ZONE_SCOPED_CATEGORY_SHADER; @@ -1202,12 +1233,11 @@ S32 LLGLSLShader::enableTexture(S32 uniform, LLTexUnit::eTextureType mode, LLTex { gGL.getTexUnit(index)->activate(); gGL.getTexUnit(index)->enable(mode); - gGL.getTexUnit(index)->setTextureColorSpace(space); } return index; } -S32 LLGLSLShader::disableTexture(S32 uniform, LLTexUnit::eTextureType mode, LLTexUnit::eTextureColorSpace space) +S32 LLGLSLShader::disableTexture(S32 uniform, LLTexUnit::eTextureType mode) { LL_PROFILE_ZONE_SCOPED_CATEGORY_SHADER; @@ -1220,7 +1250,7 @@ S32 LLGLSLShader::disableTexture(S32 uniform, LLTexUnit::eTextureType mode, LLTe S32 index = mTexture[uniform]; if (index != -1 && gGL.getTexUnit(index)->getCurrType() != LLTexUnit::TT_NONE) { - if (gDebugGL && gGL.getTexUnit(index)->getCurrType() != mode && gGL.getTexUnit(index)->getCurrColorSpace() != space) + if (gDebugGL && gGL.getTexUnit(index)->getCurrType() != mode) { if (gDebugSession) { @@ -1256,7 +1286,7 @@ void LLGLSLShader::uniform1i(U32 index, GLint x) if (iter == mValue.end() || iter->second.mV[0] != x) { glUniform1i(mUniform[index], x); - mValue[mUniform[index]] = LLVector4(x, 0.f, 0.f, 0.f); + mValue[mUniform[index]] = LLVector4((F32)x, 0.f, 0.f, 0.f); } } } @@ -1396,7 +1426,7 @@ void LLGLSLShader::uniform1iv(U32 index, U32 count, const GLint* v) if (mUniform[index] >= 0) { const auto& iter = mValue.find(mUniform[index]); - LLVector4 vec(v[0], 0.f, 0.f, 0.f); + LLVector4 vec((F32)v[0], 0.f, 0.f, 0.f); if (iter == mValue.end() || shouldChange(iter->second, vec) || count != 1) { glUniform1iv(mUniform[index], count, v); @@ -1423,7 +1453,7 @@ void LLGLSLShader::uniform4iv(U32 index, U32 count, const GLint* v) if (mUniform[index] >= 0) { const auto& iter = mValue.find(mUniform[index]); - LLVector4 vec(v[0], v[1], v[2], v[3]); + LLVector4 vec((F32)v[0], (F32)v[1], (F32)v[2], (F32)v[3]); if (iter == mValue.end() || shouldChange(iter->second, vec) || count != 1) { glUniform1iv(mUniform[index], count, v); @@ -1543,6 +1573,34 @@ void LLGLSLShader::uniform4fv(U32 index, U32 count, const GLfloat* v) } } +void LLGLSLShader::uniform4uiv(U32 index, U32 count, const GLuint* v) +{ + LL_PROFILE_ZONE_SCOPED_CATEGORY_SHADER; + llassert(sCurBoundShaderPtr == this); + + if (mProgramObject) + { + if (mUniform.size() <= index) + { + LL_WARNS_ONCE("Shader") << "Uniform index out of bounds. Size: " << (S32)mUniform.size() << " index: " << index << LL_ENDL; + llassert(false); + return; + } + + if (mUniform[index] >= 0) + { + const auto& iter = mValue.find(mUniform[index]); + LLVector4 vec((F32)v[0], (F32)v[1], (F32)v[2], (F32)v[3]); + if (iter == mValue.end() || shouldChange(iter->second, vec) || count != 1) + { + LL_PROFILE_ZONE_SCOPED_CATEGORY_SHADER; + glUniform4uiv(mUniform[index], count, v); + mValue[mUniform[index]] = vec; + } + } + } +} + void LLGLSLShader::uniformMatrix2fv(U32 index, U32 count, GLboolean transpose, const GLfloat* v) { LL_PROFILE_ZONE_SCOPED_CATEGORY_SHADER; @@ -1693,7 +1751,7 @@ void LLGLSLShader::uniform1i(const LLStaticHashedString& uniform, GLint v) if (location >= 0) { const auto& iter = mValue.find(location); - LLVector4 vec(v, 0.f, 0.f, 0.f); + LLVector4 vec((F32)v, 0.f, 0.f, 0.f); if (iter == mValue.end() || shouldChange(iter->second, vec)) { glUniform1i(location, v); @@ -1709,7 +1767,7 @@ void LLGLSLShader::uniform1iv(const LLStaticHashedString& uniform, U32 count, co if (location >= 0) { - LLVector4 vec(v[0], 0, 0, 0); + LLVector4 vec((F32)v[0], 0.f, 0.f, 0.f); const auto& iter = mValue.find(location); if (iter == mValue.end() || shouldChange(iter->second, vec) || count != 1) { @@ -1727,7 +1785,7 @@ void LLGLSLShader::uniform4iv(const LLStaticHashedString& uniform, U32 count, co if (location >= 0) { - LLVector4 vec(v[0], v[1], v[2], v[3]); + LLVector4 vec((F32)v[0], (F32)v[1], (F32)v[2], (F32)v[3]); const auto& iter = mValue.find(location); if (iter == mValue.end() || shouldChange(iter->second, vec) || count != 1) { @@ -1746,7 +1804,7 @@ void LLGLSLShader::uniform2i(const LLStaticHashedString& uniform, GLint i, GLint if (location >= 0) { const auto& iter = mValue.find(location); - LLVector4 vec(i, j, 0.f, 0.f); + LLVector4 vec((F32)i, (F32)j, 0.f, 0.f); if (iter == mValue.end() || shouldChange(iter->second, vec)) { glUniform2i(location, i, j); @@ -1808,6 +1866,23 @@ void LLGLSLShader::uniform3f(const LLStaticHashedString& uniform, GLfloat x, GLf } } +void LLGLSLShader::uniform4f(const LLStaticHashedString& uniform, GLfloat x, GLfloat y, GLfloat z, GLfloat w) +{ + LL_PROFILE_ZONE_SCOPED_CATEGORY_SHADER; + GLint location = getUniformLocation(uniform); + + if (location >= 0) + { + const auto& iter = mValue.find(location); + LLVector4 vec(x, y, z, w); + if (iter == mValue.end() || shouldChange(iter->second, vec)) + { + glUniform4f(location, x, y, z, w); + mValue[location] = vec; + } + } +} + void LLGLSLShader::uniform1fv(const LLStaticHashedString& uniform, U32 count, const GLfloat* v) { LL_PROFILE_ZONE_SCOPED_CATEGORY_SHADER; @@ -1877,6 +1952,24 @@ void LLGLSLShader::uniform4fv(const LLStaticHashedString& uniform, U32 count, co } } +void LLGLSLShader::uniform4uiv(const LLStaticHashedString& uniform, U32 count, const GLuint* v) +{ + LL_PROFILE_ZONE_SCOPED_CATEGORY_SHADER; + GLint location = getUniformLocation(uniform); + + if (location >= 0) + { + LLVector4 vec((F32)v[0], (F32)v[1], (F32)v[2], (F32)v[3]); + const auto& iter = mValue.find(location); + if (iter == mValue.end() || shouldChange(iter->second, vec) || count != 1) + { + LL_PROFILE_ZONE_SCOPED_CATEGORY_SHADER; + glUniform4uiv(location, count, v); + mValue[location] = vec; + } + } +} + void LLGLSLShader::uniformMatrix4fv(const LLStaticHashedString& uniform, U32 count, GLboolean transpose, const GLfloat* v) { LL_PROFILE_ZONE_SCOPED_CATEGORY_SHADER; diff --git a/indra/llrender/llglslshader.h b/indra/llrender/llglslshader.h index d7741c51bb..873ab0cff5 100644 --- a/indra/llrender/llglslshader.h +++ b/indra/llrender/llglslshader.h @@ -30,6 +30,7 @@ #include "llgl.h" #include "llrender.h" #include "llstaticstringtable.h" +#include <boost/json.hpp> #include <unordered_map> class LLShaderFeatures @@ -44,19 +45,21 @@ public: bool hasTransport = false; // implies no lighting (it's possible to have neither though) bool hasSkinning = false; bool hasObjectSkinning = false; + bool mGLTF = false; bool hasAtmospherics = false; bool hasGamma = false; bool hasShadows = false; bool hasAmbientOcclusion = false; bool hasSrgb = false; bool isDeferred = false; + bool hasFullGBuffer = false; bool hasScreenSpaceReflections = false; - bool disableTextureIndex = false; bool hasAlphaMask = false; bool hasReflectionProbes = false; bool attachNothing = false; bool hasHeroProbes = false; bool isPBRTerrain = false; + bool hasTonemap = false; }; // ============= Structure for caching shader uniforms =============== @@ -69,8 +72,8 @@ public: template<typename T> struct UniformSetting { - S32 mUniform; - T mValue; + S32 mUniform{ 0 }; + T mValue{}; }; typedef UniformSetting<S32> IntSetting; @@ -145,6 +148,16 @@ public: SG_COUNT } eGroup; + enum UniformBlock : GLuint + { + UB_REFLECTION_PROBES, // "ReflectionProbes" + UB_GLTF_JOINTS, // "GLTFJoints" + UB_GLTF_NODES, // "GLTFNodes" + UB_GLTF_MATERIALS, // "GLTFMaterials" + NUM_UNIFORM_BLOCKS + }; + + static std::set<LLGLSLShader*> sInstances; static bool sProfileEnabled; @@ -155,15 +168,18 @@ public: static LLGLSLShader* sCurBoundShaderPtr; static S32 sIndexedTextureChannels; + static U32 sMaxGLTFMaterials; + static U32 sMaxGLTFNodes; + static void initProfile(); - static void finishProfile(bool emit_report = true); + static void finishProfile(boost::json::value& stats=sDefaultStats); static void startProfile(); static void stopProfile(); void unload(); void clearStats(); - void dumpStats(); + void dumpStats(boost::json::object& stats); // place query objects for profiling if profiling is enabled // if for_runtime is true, will place timer query only whether or not profiling is enabled @@ -175,17 +191,14 @@ public: // If force_read is true, will force an immediate readback (severe performance penalty) bool readProfileQuery(bool for_runtime = false, bool force_read = false); - BOOL createShader(std::vector<LLStaticHashedString>* attributes, - std::vector<LLStaticHashedString>* uniforms, - U32 varying_count = 0, - const char** varyings = NULL); - BOOL attachFragmentObject(std::string object); - BOOL attachVertexObject(std::string object); + bool createShader(); + bool attachFragmentObject(std::string object); + bool attachVertexObject(std::string object); void attachObject(GLuint object); void attachObjects(GLuint* objects = NULL, S32 count = 0); - BOOL mapAttributes(const std::vector<LLStaticHashedString>* attributes); - BOOL mapUniforms(const std::vector<LLStaticHashedString>*); - void mapUniform(GLint index, const std::vector<LLStaticHashedString>*); + bool mapAttributes(); + bool mapUniforms(); + void mapUniform(GLint index); void uniform1i(U32 index, GLint i); void uniform1f(U32 index, GLfloat v); void fastUniform1f(U32 index, GLfloat v); @@ -198,6 +211,7 @@ public: void uniform2fv(U32 index, U32 count, const GLfloat* v); void uniform3fv(U32 index, U32 count, const GLfloat* v); void uniform4fv(U32 index, U32 count, const GLfloat* v); + void uniform4uiv(U32 index, U32 count, const GLuint* v); void uniform2i(const LLStaticHashedString& uniform, GLint i, GLint j); void uniformMatrix2fv(U32 index, U32 count, GLboolean transpose, const GLfloat* v); void uniformMatrix3fv(U32 index, U32 count, GLboolean transpose, const GLfloat* v); @@ -209,10 +223,12 @@ public: void uniform1f(const LLStaticHashedString& uniform, GLfloat v); void uniform2f(const LLStaticHashedString& uniform, GLfloat x, GLfloat y); void uniform3f(const LLStaticHashedString& uniform, GLfloat x, GLfloat y, GLfloat z); + void uniform4f(const LLStaticHashedString& uniform, GLfloat x, GLfloat y, GLfloat z, GLfloat w); void uniform1fv(const LLStaticHashedString& uniform, U32 count, const GLfloat* v); void uniform2fv(const LLStaticHashedString& uniform, U32 count, const GLfloat* v); void uniform3fv(const LLStaticHashedString& uniform, U32 count, const GLfloat* v); void uniform4fv(const LLStaticHashedString& uniform, U32 count, const GLfloat* v); + void uniform4uiv(const LLStaticHashedString& uniform, U32 count, const GLuint* v); void uniformMatrix4fv(const LLStaticHashedString& uniform, U32 count, GLboolean transpose, const GLfloat* v); void setMinimumAlpha(F32 minimum); @@ -229,6 +245,10 @@ public: void clearPermutations(); void addPermutation(std::string name, std::string value); + void addPermutations(const std::map<std::string, std::string>& defines) + { + mDefines.insert(defines.begin(), defines.end()); + } void removePermutation(std::string name); void addConstant(const LLGLSLShader::eShaderConsts shader_const); @@ -237,22 +257,22 @@ public: //if given texture uniform is active in the shader, //the corresponding channel will be active upon return //returns channel texture is enabled in from [0-MAX) - S32 enableTexture(S32 uniform, LLTexUnit::eTextureType mode = LLTexUnit::TT_TEXTURE, LLTexUnit::eTextureColorSpace space = LLTexUnit::TCS_LINEAR); - S32 disableTexture(S32 uniform, LLTexUnit::eTextureType mode = LLTexUnit::TT_TEXTURE, LLTexUnit::eTextureColorSpace space = LLTexUnit::TCS_LINEAR); + S32 enableTexture(S32 uniform, LLTexUnit::eTextureType mode = LLTexUnit::TT_TEXTURE); + S32 disableTexture(S32 uniform, LLTexUnit::eTextureType mode = LLTexUnit::TT_TEXTURE); // get the texture channel of the given uniform, or -1 if uniform is not used as a texture S32 getTextureChannel(S32 uniform) const; // bindTexture returns the texture unit we've bound the texture to. // You can reuse the return value to unbind a texture when required. - S32 bindTexture(const std::string& uniform, LLTexture* texture, LLTexUnit::eTextureType mode = LLTexUnit::TT_TEXTURE, LLTexUnit::eTextureColorSpace space = LLTexUnit::TCS_LINEAR); - S32 bindTexture(S32 uniform, LLTexture* texture, LLTexUnit::eTextureType mode = LLTexUnit::TT_TEXTURE, LLTexUnit::eTextureColorSpace space = LLTexUnit::TCS_LINEAR); + S32 bindTexture(const std::string& uniform, LLTexture* texture, LLTexUnit::eTextureType mode = LLTexUnit::TT_TEXTURE); + S32 bindTexture(S32 uniform, LLTexture* texture, LLTexUnit::eTextureType mode = LLTexUnit::TT_TEXTURE); S32 bindTexture(const std::string& uniform, LLRenderTarget* texture, bool depth = false, LLTexUnit::eTextureFilterOptions mode = LLTexUnit::TFO_BILINEAR); S32 bindTexture(S32 uniform, LLRenderTarget* texture, bool depth = false, LLTexUnit::eTextureFilterOptions mode = LLTexUnit::TFO_BILINEAR, U32 index = 0); S32 unbindTexture(const std::string& uniform, LLTexUnit::eTextureType mode = LLTexUnit::TT_TEXTURE); S32 unbindTexture(S32 uniform, LLTexUnit::eTextureType mode = LLTexUnit::TT_TEXTURE); - BOOL link(BOOL suppress_errors = FALSE); + bool link(bool suppress_errors = false); void bind(); //helper to conditionally bind mRiggedVariant instead of this void bind(bool rigged); @@ -290,7 +310,7 @@ public: S32 mActiveTextureChannels; S32 mShaderLevel; S32 mShaderGroup; // see LLGLSLShader::eGroup - BOOL mUniformsDirty; + bool mUniformsDirty; LLShaderFeatures mFeatures; std::vector< std::pair< std::string, GLenum > > mShaderFiles; std::string mName; @@ -318,6 +338,26 @@ public: // this pointer should be set to whichever shader represents this shader's rigged variant LLGLSLShader* mRiggedVariant = nullptr; + // variants for use by GLTF renderer + // bit 0 = alpha mode blend (1) or opaque (0) + // bit 1 = rigged (1) or static (0) + // bit 2 = unlit (1) or lit (0) + // bit 3 = single (0) or multi (1) uv coordinates + struct GLTFVariant + { + constexpr static U8 ALPHA_BLEND = 1; + constexpr static U8 RIGGED = 2; + constexpr static U8 UNLIT = 4; + constexpr static U8 MULTI_UV = 8; + }; + + constexpr static U8 NUM_GLTF_VARIANTS = 16; + + std::vector<LLGLSLShader> mGLTFVariants; + + //helper to bind GLTF variant + void bind(U8 variant); + // hacky flag used for optimization in LLDrawPoolAlpha bool mCanBindFast = false; @@ -327,6 +367,11 @@ public: private: void unloadInternal(); + // This must be static because finishProfile() is called at least once + // within a __try block. If we default its stats parameter to a temporary + // json::value, that temporary must be destroyed when the stack is + // unwound, which __try forbids. + static boost::json::value sDefaultStats; }; //UI shader (declared here so llui_libtest will link properly) diff --git a/indra/llrender/llgltexture.cpp b/indra/llrender/llgltexture.cpp index 104976fcc6..4dcca5a726 100644 --- a/indra/llrender/llgltexture.cpp +++ b/indra/llrender/llgltexture.cpp @@ -27,13 +27,13 @@ #include "llgltexture.h" -LLGLTexture::LLGLTexture(BOOL usemipmaps) +LLGLTexture::LLGLTexture(bool usemipmaps) { init(); mUseMipMaps = usemipmaps; } -LLGLTexture::LLGLTexture(const U32 width, const U32 height, const U8 components, BOOL usemipmaps) +LLGLTexture::LLGLTexture(const U32 width, const U32 height, const U8 components, bool usemipmaps) { init(); mFullWidth = width ; @@ -43,14 +43,14 @@ LLGLTexture::LLGLTexture(const U32 width, const U32 height, const U8 components, setTexelsPerImage(); } -LLGLTexture::LLGLTexture(const LLImageRaw* raw, BOOL usemipmaps) +LLGLTexture::LLGLTexture(const LLImageRaw* raw, bool usemipmaps) { init(); mUseMipMaps = usemipmaps ; // Create an empty image of the specified size and width mGLTexturep = new LLImageGL(raw, usemipmaps) ; - mFullWidth = mGLTexturep->getCurrentWidth(); - mFullHeight = mGLTexturep->getCurrentHeight(); + mFullWidth = mGLTexturep->getWidth(); + mFullHeight = mGLTexturep->getHeight(); mComponents = mGLTexturep->getComponents(); setTexelsPerImage(); } @@ -67,12 +67,12 @@ void LLGLTexture::init() mFullWidth = 0; mFullHeight = 0; mTexelsPerImage = 0 ; - mUseMipMaps = FALSE ; + mUseMipMaps = false ; mComponents = 0 ; mTextureState = NO_DELETE ; - mDontDiscard = FALSE; - mNeedsGLTexture = FALSE ; + mDontDiscard = false; + mNeedsGLTexture = false ; } void LLGLTexture::cleanup() @@ -141,7 +141,7 @@ LLImageGL* LLGLTexture::getGLTexture() const return mGLTexturep ; } -BOOL LLGLTexture::createGLTexture() +bool LLGLTexture::createGLTexture() { if(mGLTexturep.isNull()) { @@ -151,11 +151,11 @@ BOOL LLGLTexture::createGLTexture() return mGLTexturep->createGLTexture() ; } -BOOL LLGLTexture::createGLTexture(S32 discard_level, const LLImageRaw* imageraw, S32 usename, BOOL to_create, S32 category, bool defer_copy, LLGLuint* tex_name) +bool LLGLTexture::createGLTexture(S32 discard_level, const LLImageRaw* imageraw, S32 usename, bool to_create, S32 category, bool defer_copy, LLGLuint* tex_name) { llassert(mGLTexturep.notNull()); - BOOL ret = mGLTexturep->createGLTexture(discard_level, imageraw, usename, to_create, category, defer_copy, tex_name) ; + bool ret = mGLTexturep->createGLTexture(discard_level, imageraw, usename, to_create, category, defer_copy, tex_name) ; if(ret) { @@ -168,7 +168,7 @@ BOOL LLGLTexture::createGLTexture(S32 discard_level, const LLImageRaw* imageraw, return ret ; } -void LLGLTexture::setExplicitFormat(LLGLint internal_format, LLGLenum primary_format, LLGLenum type_format, BOOL swap_bytes) +void LLGLTexture::setExplicitFormat(LLGLint internal_format, LLGLenum primary_format, LLGLenum type_format, bool swap_bytes) { llassert(mGLTexturep.notNull()) ; @@ -223,22 +223,22 @@ LLGLuint LLGLTexture::getTexName() const return mGLTexturep->getTexName() ; } -BOOL LLGLTexture::hasGLTexture() const +bool LLGLTexture::hasGLTexture() const { if(mGLTexturep.notNull()) { return mGLTexturep->getHasGLTexture() ; } - return FALSE ; + return false ; } -BOOL LLGLTexture::getBoundRecently() const +bool LLGLTexture::getBoundRecently() const { if(mGLTexturep.notNull()) { return mGLTexturep->getBoundRecently() ; } - return FALSE ; + return false ; } LLTexUnit::eTextureType LLGLTexture::getTarget(void) const @@ -247,7 +247,7 @@ LLTexUnit::eTextureType LLGLTexture::getTarget(void) const return mGLTexturep->getTarget() ; } -BOOL LLGLTexture::setSubImage(const LLImageRaw* imageraw, S32 x_pos, S32 y_pos, S32 width, S32 height, LLGLuint use_name) +bool LLGLTexture::setSubImage(const LLImageRaw* imageraw, S32 x_pos, S32 y_pos, S32 width, S32 height, LLGLuint use_name) { LL_PROFILE_ZONE_SCOPED_CATEGORY_TEXTURE; llassert(mGLTexturep.notNull()) ; @@ -255,7 +255,7 @@ BOOL LLGLTexture::setSubImage(const LLImageRaw* imageraw, S32 x_pos, S32 y_pos, return mGLTexturep->setSubImage(imageraw, x_pos, y_pos, width, height, 0, use_name) ; } -BOOL LLGLTexture::setSubImage(const U8* datap, S32 data_width, S32 data_height, S32 x_pos, S32 y_pos, S32 width, S32 height, LLGLuint use_name) +bool LLGLTexture::setSubImage(const U8* datap, S32 data_width, S32 data_height, S32 x_pos, S32 y_pos, S32 width, S32 height, LLGLuint use_name) { LL_PROFILE_ZONE_SCOPED_CATEGORY_TEXTURE; llassert(mGLTexturep.notNull()) ; @@ -310,14 +310,14 @@ LLGLenum LLGLTexture::getPrimaryFormat() const return mGLTexturep->getPrimaryFormat() ; } -BOOL LLGLTexture::getIsAlphaMask() const +bool LLGLTexture::getIsAlphaMask() const { llassert(mGLTexturep.notNull()) ; return mGLTexturep->getIsAlphaMask() ; } -BOOL LLGLTexture::getMask(const LLVector2 &tc) +bool LLGLTexture::getMask(const LLVector2 &tc) { llassert(mGLTexturep.notNull()) ; @@ -330,14 +330,14 @@ F32 LLGLTexture::getTimePassedSinceLastBound() return mGLTexturep->getTimePassedSinceLastBound() ; } -BOOL LLGLTexture::getMissed() const +bool LLGLTexture::getMissed() const { llassert(mGLTexturep.notNull()) ; return mGLTexturep->getMissed() ; } -BOOL LLGLTexture::isJustBound() const +bool LLGLTexture::isJustBound() const { llassert(mGLTexturep.notNull()) ; @@ -351,34 +351,13 @@ void LLGLTexture::forceUpdateBindStats(void) const return mGLTexturep->forceUpdateBindStats() ; } -U32 LLGLTexture::getTexelsInAtlas() const -{ - llassert(mGLTexturep.notNull()) ; - - return mGLTexturep->getTexelsInAtlas() ; -} - -U32 LLGLTexture::getTexelsInGLTexture() const -{ - llassert(mGLTexturep.notNull()) ; - - return mGLTexturep->getTexelsInGLTexture() ; -} - -BOOL LLGLTexture::isGLTextureCreated() const +bool LLGLTexture::isGLTextureCreated() const { llassert(mGLTexturep.notNull()) ; return mGLTexturep->isGLTextureCreated() ; } -S32 LLGLTexture::getDiscardLevelInAtlas() const -{ - llassert(mGLTexturep.notNull()) ; - - return mGLTexturep->getDiscardLevelInAtlas() ; -} - void LLGLTexture::destroyGLTexture() { if(mGLTexturep.notNull() && mGLTexturep->getHasGLTexture()) diff --git a/indra/llrender/llgltexture.h b/indra/llrender/llgltexture.h index f5bef0e291..122d2a7f9c 100644 --- a/indra/llrender/llgltexture.h +++ b/indra/llrender/llgltexture.h @@ -51,10 +51,10 @@ public: BOOST_NONE = 0, BOOST_AVATAR , BOOST_AVATAR_BAKED , - BOOST_SCULPTED , BOOST_TERRAIN , // Needed for minimap generation for now. Lower than BOOST_HIGH so the texture stats don't get forced, i.e. texture stats are manually managed by minimap/terrain instead. BOOST_HIGH = 10, + BOOST_SCULPTED , BOOST_BUMP , BOOST_UNUSED_1 , // Placeholder to avoid disrupting habits around texture debug BOOST_SELECTED , @@ -75,7 +75,6 @@ public: AVATAR_SCRATCH_TEX, DYNAMIC_TEX, MEDIA, - ATLAS, OTHER, MAX_GL_IMAGE_CATEGORY }; @@ -83,8 +82,6 @@ public: typedef enum { DELETED = 0, //removed from memory - DELETION_CANDIDATE, //ready to be removed from memory - INACTIVE, //not be used for the last certain period (i.e., 30 seconds). ACTIVE, //just being used, can become inactive if not being used for a certain time (10 seconds). NO_DELETE = 99 //stay in memory, can not be removed. } LLGLTextureState; @@ -94,9 +91,9 @@ protected: LOG_CLASS(LLGLTexture); public: - LLGLTexture(BOOL usemipmaps = TRUE); - LLGLTexture(const LLImageRaw* raw, BOOL usemipmaps) ; - LLGLTexture(const U32 width, const U32 height, const U8 components, BOOL usemipmaps) ; + LLGLTexture(bool usemipmaps = true); + LLGLTexture(const LLImageRaw* raw, bool usemipmaps) ; + LLGLTexture(const U32 width, const U32 height, const U8 components, bool usemipmaps) ; virtual void dump(); // debug info to LL_INFOS() @@ -117,25 +114,25 @@ public: /*virtual*/S32 getWidth(S32 discard_level = -1) const; /*virtual*/S32 getHeight(S32 discard_level = -1) const; - BOOL hasGLTexture() const ; + bool hasGLTexture() const ; LLGLuint getTexName() const ; - BOOL createGLTexture() ; + bool createGLTexture() ; // Create a GL Texture from an image raw // discard_level - mip level, 0 for highest resultion mip // imageraw - the image to copy from // usename - explicit GL name override - // to_create - set to FALSE to force gl texture to not be created + // to_create - set to false to force gl texture to not be created // category - LLGLTexture category for this LLGLTexture // defer_copy - set to true to allocate GL texture but NOT initialize with imageraw data // tex_name - if not null, will be set to the GL name of the texture created - BOOL createGLTexture(S32 discard_level, const LLImageRaw* imageraw, S32 usename = 0, BOOL to_create = TRUE, S32 category = LLGLTexture::OTHER, bool defer_copy = false, LLGLuint* tex_name = nullptr); + bool createGLTexture(S32 discard_level, const LLImageRaw* imageraw, S32 usename = 0, bool to_create = true, S32 category = LLGLTexture::OTHER, bool defer_copy = false, LLGLuint* tex_name = nullptr); void setFilteringOption(LLTexUnit::eTextureFilterOptions option); - void setExplicitFormat(LLGLint internal_format, LLGLenum primary_format, LLGLenum type_format = 0, BOOL swap_bytes = FALSE); + void setExplicitFormat(LLGLint internal_format, LLGLenum primary_format, LLGLenum type_format = 0, bool swap_bytes = false); void setAddressMode(LLTexUnit::eTextureAddressMode mode); - BOOL setSubImage(const LLImageRaw* imageraw, S32 x_pos, S32 y_pos, S32 width, S32 height, LLGLuint use_name = 0); - BOOL setSubImage(const U8* datap, S32 data_width, S32 data_height, S32 x_pos, S32 y_pos, S32 width, S32 height, LLGLuint use_name = 0); + bool setSubImage(const LLImageRaw* imageraw, S32 x_pos, S32 y_pos, S32 width, S32 height, LLGLuint use_name = 0); + bool setSubImage(const U8* datap, S32 data_width, S32 data_height, S32 x_pos, S32 y_pos, S32 width, S32 height, LLGLuint use_name = 0); void setGLTextureCreated (bool initialized); void setCategory(S32 category) ; void setTexName(LLGLuint); // for forcing w/ externally created textures only @@ -145,21 +142,18 @@ public: S32 getMaxDiscardLevel() const; S32 getDiscardLevel() const; S8 getComponents() const; - BOOL getBoundRecently() const; + bool getBoundRecently() const; S32Bytes getTextureMemory() const ; LLGLenum getPrimaryFormat() const; - BOOL getIsAlphaMask() const ; + bool getIsAlphaMask() const ; LLTexUnit::eTextureType getTarget(void) const ; - BOOL getMask(const LLVector2 &tc); + bool getMask(const LLVector2 &tc); F32 getTimePassedSinceLastBound(); - BOOL getMissed() const ; - BOOL isJustBound()const ; + bool getMissed() const ; + bool isJustBound()const ; void forceUpdateBindStats(void) const; - U32 getTexelsInAtlas() const ; - U32 getTexelsInGLTexture() const ; - BOOL isGLTextureCreated() const ; - S32 getDiscardLevelInAtlas() const ; + bool isGLTextureCreated() const ; LLGLTextureState getTextureState() const { return mTextureState; } //--------------------------------------------------------------------------------------------- @@ -171,7 +165,7 @@ public: void forceActive() ; void setNoDelete() ; void dontDiscard() { mDontDiscard = 1; mTextureState = NO_DELETE; } - BOOL getDontDiscard() const { return mDontDiscard; } + bool getDontDiscard() const { return mDontDiscard; } //----------------- private: @@ -188,7 +182,7 @@ protected: S32 mBoostLevel; // enum describing priority level U32 mFullWidth; U32 mFullHeight; - BOOL mUseMipMaps; + bool mUseMipMaps; S8 mComponents; U32 mTexelsPerImage; // Texels per image. mutable S8 mNeedsGLTexture; diff --git a/indra/llrender/llimagegl.cpp b/indra/llrender/llimagegl.cpp index 1609afb7ef..3f8903ca09 100644 --- a/indra/llrender/llimagegl.cpp +++ b/indra/llrender/llimagegl.cpp @@ -41,6 +41,7 @@ #include "llrender.h" #include "llwindow.h" #include "llframetimer.h" +#include <unordered_set> extern LL_COMMON_API bool on_main_thread(); @@ -50,28 +51,36 @@ extern LL_COMMON_API bool on_main_thread(); //---------------------------------------------------------------------------- const F32 MIN_TEXTURE_LIFETIME = 10.f; +const F32 CONVERSION_SCRATCH_BUFFER_GL_VERSION = 3.29f; //which power of 2 is i? //assumes i is a power of 2 > 0 U32 wpo2(U32 i); -// texture memory accounting (for OS X) +U32 LLImageGL::sFrameCount = 0; + + +// texture memory accounting (for macOS) static LLMutex sTexMemMutex; static std::unordered_map<U32, U64> sTextureAllocs; static U64 sTextureBytes = 0; // track a texture alloc on the currently bound texture. // asserts that no currently tracked alloc exists -static void alloc_tex_image(U32 width, U32 height, U32 pixformat) +void LLImageGLMemory::alloc_tex_image(U32 width, U32 height, U32 intformat, U32 count) { U32 texUnit = gGL.getCurrentTexUnitIndex(); + llassert(texUnit == 0); // allocations should always be done on tex unit 0 U32 texName = gGL.getTexUnit(texUnit)->getCurrTexture(); - U64 size = LLImageGL::dataFormatBytes(pixformat, width, height); + U64 size = LLImageGL::dataFormatBytes(intformat, width, height); + size *= count; llassert(size >= 0); sTexMemMutex.lock(); + + // it is a precondition that no existing allocation exists for this texture llassert(sTextureAllocs.find(texName) == sTextureAllocs.end()); sTextureAllocs[texName] = size; @@ -81,11 +90,11 @@ static void alloc_tex_image(U32 width, U32 height, U32 pixformat) } // track texture free on given texName -static void free_tex_image(U32 texName) +void LLImageGLMemory::free_tex_image(U32 texName) { sTexMemMutex.lock(); auto iter = sTextureAllocs.find(texName); - if (iter != sTextureAllocs.end()) + if (iter != sTextureAllocs.end()) // sometimes a texName will be "freed" before allocated (e.g. first call to setManualImage for a given texName) { llassert(iter->second <= sTextureBytes); // sTextureBytes MUST NOT go below zero @@ -98,22 +107,25 @@ static void free_tex_image(U32 texName) } // track texture free on given texNames -static void free_tex_images(U32 count, const U32* texNames) +void LLImageGLMemory::free_tex_images(U32 count, const U32* texNames) { - for (int i = 0; i < count; ++i) + for (U32 i = 0; i < count; ++i) { free_tex_image(texNames[i]); } } // track texture free on currently bound texture -static void free_cur_tex_image() +void LLImageGLMemory::free_cur_tex_image() { U32 texUnit = gGL.getCurrentTexUnitIndex(); + llassert(texUnit == 0); // frees should always be done on tex unit 0 U32 texName = gGL.getTexUnit(texUnit)->getCurrTexture(); free_tex_image(texName); } +using namespace LLImageGLMemory; + // static U64 LLImageGL::getTextureBytesAllocated() { @@ -126,12 +138,11 @@ U32 LLImageGL::sUniqueCount = 0; U32 LLImageGL::sBindCount = 0; S32 LLImageGL::sCount = 0; -BOOL LLImageGL::sGlobalUseAnisotropic = FALSE; +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; +std::unordered_set<LLImageGL*> LLImageGL::sImageList; bool LLImageGLThread::sEnabledTextures = false; @@ -147,7 +158,11 @@ S32 LLImageGL::sCurTexPickSize = -1 ; S32 LLImageGL::sMaxCategories = 1 ; //optimization for when we don't need to calculate mIsMask -BOOL LLImageGL::sSkipAnalyzeAlpha; +bool LLImageGL::sSkipAnalyzeAlpha; +U32 LLImageGL::sScratchPBO = 0; +U32 LLImageGL::sScratchPBOSize = 0; +U32* LLImageGL::sManualScratch = nullptr; + //------------------------ //**************************************************************************************************** @@ -157,20 +172,6 @@ BOOL LLImageGL::sSkipAnalyzeAlpha; //************************************************************************************** //below are functions for debug use //do not delete them even though they are not currently being used. -void check_all_images() -{ - for (std::set<LLImageGL*>::iterator iter = LLImageGL::sImageList.begin(); - iter != LLImageGL::sImageList.end(); iter++) - { - LLImageGL* glimage = *iter; - if (glimage->getTexName() && glimage->isGLTextureCreated()) - { - gGL.getTexUnit(0)->bind(glimage) ; - glimage->checkTexSize() ; - gGL.getTexUnit(0)->unbind(glimage->getTarget()) ; - } - } -} void LLImageGL::checkTexSize(bool forced) const { @@ -185,12 +186,12 @@ void LLImageGL::checkTexSize(bool forced) const GLint texname; glGetIntegerv(GL_TEXTURE_BINDING_2D, &texname); - BOOL error = FALSE; + bool error = false; if (texname != mTexName) { LL_INFOS() << "Bound: " << texname << " Should bind: " << mTexName << " Default: " << LLImageGL::sDefaultGLTexture->getTexName() << LL_ENDL; - error = TRUE; + error = true; if (gDebugSession) { gFailLog << "Invalid texture bound!" << std::endl; @@ -213,7 +214,7 @@ void LLImageGL::checkTexSize(bool forced) const } if(x != (mWidth >> mCurrentDiscardLevel) || y != (mHeight >> mCurrentDiscardLevel)) { - error = TRUE; + error = true; if (gDebugSession) { gFailLog << "wrong texture size and discard level!" << @@ -236,7 +237,7 @@ void LLImageGL::checkTexSize(bool forced) const //************************************************************************************** //---------------------------------------------------------------------------- -BOOL is_little_endian() +bool is_little_endian() { S32 a = 0x12345678; U8 *c = (U8*)(&a); @@ -245,16 +246,37 @@ BOOL is_little_endian() } //static -void LLImageGL::initClass(LLWindow* window, S32 num_catagories, BOOL skip_analyze_alpha /* = false */, bool thread_texture_loads /* = false */, bool thread_media_updates /* = false */) +void LLImageGL::initClass(LLWindow* window, S32 num_catagories, bool skip_analyze_alpha /* = false */, bool thread_texture_loads /* = false */, bool thread_media_updates /* = false */) { LL_PROFILE_ZONE_SCOPED_CATEGORY_TEXTURE; sSkipAnalyzeAlpha = skip_analyze_alpha; + if (sScratchPBO == 0) + { + glGenBuffers(1, &sScratchPBO); + } + if (thread_texture_loads || thread_media_updates) { LLImageGLThread::createInstance(window); - LLImageGLThread::sEnabledTextures = thread_texture_loads; - LLImageGLThread::sEnabledMedia = thread_media_updates; + LLImageGLThread::sEnabledTextures = gGLManager.mGLVersion > 3.95f ? thread_texture_loads : false; + LLImageGLThread::sEnabledMedia = gGLManager.mGLVersion > 3.95f ? thread_media_updates : false; + } +} + +void LLImageGL::allocateConversionBuffer() +{ + if (gGLManager.mGLVersion < CONVERSION_SCRATCH_BUFFER_GL_VERSION) + { + try + { + sManualScratch = new U32[MAX_IMAGE_AREA]; + } + catch (std::bad_alloc&) + { + LLError::LLUserWarningMsg::showOutOfMemory(); + LL_ERRS() << "Failed to allocate sManualScratch" << LL_ENDL; + } } } @@ -263,6 +285,14 @@ void LLImageGL::cleanupClass() { LL_PROFILE_ZONE_SCOPED_CATEGORY_TEXTURE; LLImageGLThread::deleteSingleton(); + if (sScratchPBO != 0) + { + glDeleteBuffers(1, &sScratchPBO); + sScratchPBO = 0; + sScratchPBOSize = 0; + } + + delete[] sManualScratch; } @@ -271,6 +301,15 @@ S32 LLImageGL::dataFormatBits(S32 dataformat) { switch (dataformat) { + case GL_COMPRESSED_RED: return 8; + case GL_COMPRESSED_RG: return 16; + case GL_COMPRESSED_RGB: return 24; + case GL_COMPRESSED_SRGB: return 32; + case GL_COMPRESSED_RGBA: return 32; + case GL_COMPRESSED_SRGB_ALPHA: return 32; + case GL_COMPRESSED_LUMINANCE: return 8; + case GL_COMPRESSED_LUMINANCE_ALPHA: return 16; + case GL_COMPRESSED_ALPHA: return 8; case GL_COMPRESSED_RGBA_S3TC_DXT1_EXT: return 4; case GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT1_EXT: return 4; case GL_COMPRESSED_RGBA_S3TC_DXT3_EXT: return 8; @@ -278,19 +317,37 @@ S32 LLImageGL::dataFormatBits(S32 dataformat) case GL_COMPRESSED_RGBA_S3TC_DXT5_EXT: return 8; case GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT5_EXT: return 8; case GL_LUMINANCE: return 8; + case GL_LUMINANCE8: return 8; case GL_ALPHA: return 8; + case GL_ALPHA8: return 8; case GL_RED: return 8; + case GL_R8: return 8; case GL_COLOR_INDEX: return 8; case GL_LUMINANCE_ALPHA: return 16; + case GL_LUMINANCE8_ALPHA8: return 16; + case GL_RG: return 16; + case GL_RG8: return 16; case GL_RGB: return 24; case GL_SRGB: return 24; case GL_RGB8: return 24; + case GL_R11F_G11F_B10F: return 32; case GL_RGBA: return 32; + case GL_RGBA8: return 32; + case GL_RGB10_A2: return 32; case GL_SRGB_ALPHA: return 32; case GL_BGRA: return 32; // Used for QuickTime media textures on the Mac case GL_DEPTH_COMPONENT: return 24; + case GL_DEPTH_COMPONENT24: return 24; + case GL_R16F: return 16; + case GL_RG16F: return 32; + case GL_RGB16F: return 48; + case GL_RGBA16F: return 64; + case GL_R32F: return 32; + case GL_RG32F: return 64; + case GL_RGB32F: return 96; + case GL_RGBA32F: return 128; default: - LL_ERRS() << "LLImageGL::Unknown format: " << dataformat << LL_ENDL; + LL_ERRS() << "LLImageGL::Unknown format: " << std::hex << dataformat << std::dec << LL_ENDL; return 0; } } @@ -333,13 +390,14 @@ S32 LLImageGL::dataFormatComponents(S32 dataformat) case GL_RED: return 1; case GL_COLOR_INDEX: return 1; case GL_LUMINANCE_ALPHA: return 2; + case GL_RG: return 2; case GL_RGB: return 3; case GL_SRGB: return 3; case GL_RGBA: return 4; case GL_SRGB_ALPHA: return 4; case GL_BGRA: return 4; // Used for QuickTime media textures on the Mac default: - LL_ERRS() << "LLImageGL::Unknown format: " << dataformat << LL_ENDL; + LL_ERRS() << "LLImageGL::Unknown format: " << std::hex << dataformat << std::dec << LL_ENDL; return 0; } } @@ -356,66 +414,19 @@ void LLImageGL::updateStats(F32 current_time) //---------------------------------------------------------------------------- //static -void LLImageGL::destroyGL(BOOL save_state) +void LLImageGL::destroyGL() { for (S32 stage = 0; stage < gGLManager.mNumTextureImageUnits; stage++) { gGL.getTexUnit(stage)->unbind(LLTexUnit::TT_TEXTURE); } - - sAllowReadBackRaw = true ; - for (std::set<LLImageGL*>::iterator iter = sImageList.begin(); - iter != sImageList.end(); iter++) - { - LLImageGL* glimage = *iter; - if (glimage->mTexName) - { - if (save_state && glimage->isGLTextureCreated() && glimage->mComponents) - { - glimage->mSaveData = new LLImageRaw; - if(!glimage->readBackRaw(glimage->mCurrentDiscardLevel, glimage->mSaveData, false)) //necessary, keep it. - { - glimage->mSaveData = NULL ; - } - } - - glimage->destroyGLTexture(); - stop_glerror(); - } - } - sAllowReadBackRaw = false ; -} - -//static -void LLImageGL::restoreGL() -{ - for (std::set<LLImageGL*>::iterator iter = sImageList.begin(); - iter != sImageList.end(); iter++) - { - LLImageGL* glimage = *iter; - if(glimage->getTexName()) - { - LL_ERRS() << "tex name is not 0." << LL_ENDL ; - } - if (glimage->mSaveData.notNull()) - { - if (glimage->getComponents() && glimage->mSaveData->getComponents()) - { - glimage->createGLTexture(glimage->mCurrentDiscardLevel, glimage->mSaveData, 0, TRUE, glimage->getCategory()); - stop_glerror(); - } - glimage->mSaveData = NULL; // deletes data - } - } } //static void LLImageGL::dirtyTexOptions() { - for (std::set<LLImageGL*>::iterator iter = sImageList.begin(); - iter != sImageList.end(); iter++) + for (auto& glimage : sImageList) { - LLImageGL* glimage = *iter; glimage->mTexOptionsDirty = true; stop_glerror(); } @@ -425,51 +436,51 @@ void LLImageGL::dirtyTexOptions() //for server side use only. //static -BOOL LLImageGL::create(LLPointer<LLImageGL>& dest, BOOL usemipmaps) +bool LLImageGL::create(LLPointer<LLImageGL>& dest, bool usemipmaps) { dest = new LLImageGL(usemipmaps); - return TRUE; + return true; } //for server side use only. -BOOL LLImageGL::create(LLPointer<LLImageGL>& dest, U32 width, U32 height, U8 components, BOOL usemipmaps) +bool LLImageGL::create(LLPointer<LLImageGL>& dest, U32 width, U32 height, U8 components, bool usemipmaps) { dest = new LLImageGL(width, height, components, usemipmaps); - return TRUE; + return true; } //for server side use only. -BOOL LLImageGL::create(LLPointer<LLImageGL>& dest, const LLImageRaw* imageraw, BOOL usemipmaps) +bool LLImageGL::create(LLPointer<LLImageGL>& dest, const LLImageRaw* imageraw, bool usemipmaps) { dest = new LLImageGL(imageraw, usemipmaps); - return TRUE; + return true; } //---------------------------------------------------------------------------- -LLImageGL::LLImageGL(BOOL usemipmaps) -: mSaveData(0), mExternalTexture(FALSE) +LLImageGL::LLImageGL(bool usemipmaps/* = true*/, bool allow_compression/* = true*/) +: mSaveData(0), mExternalTexture(false) { - init(usemipmaps); + init(usemipmaps, allow_compression); setSize(0, 0, 0); sImageList.insert(this); sCount++; } -LLImageGL::LLImageGL(U32 width, U32 height, U8 components, BOOL usemipmaps) -: mSaveData(0), mExternalTexture(FALSE) +LLImageGL::LLImageGL(U32 width, U32 height, U8 components, bool usemipmaps/* = true*/, bool allow_compression/* = true*/) +: mSaveData(0), mExternalTexture(false) { llassert( components <= 4 ); - init(usemipmaps); + init(usemipmaps, allow_compression); setSize(width, height, components); sImageList.insert(this); sCount++; } -LLImageGL::LLImageGL(const LLImageRaw* imageraw, BOOL usemipmaps) -: mSaveData(0), mExternalTexture(FALSE) +LLImageGL::LLImageGL(const LLImageRaw* imageraw, bool usemipmaps/* = true*/, bool allow_compression/* = true*/) +: mSaveData(0), mExternalTexture(false) { - init(usemipmaps); + init(usemipmaps, allow_compression); setSize(0, 0, 0); sImageList.insert(this); sCount++; @@ -486,7 +497,7 @@ LLImageGL::LLImageGL( LLGLenum formatType, LLTexUnit::eTextureAddressMode addressMode) { - init(false); + init(false, true); mTexName = texName; mTarget = target; mComponents = components; @@ -508,7 +519,7 @@ LLImageGL::~LLImageGL() } } -void LLImageGL::init(BOOL usemipmaps) +void LLImageGL::init(bool usemipmaps, bool allow_compression) { #if LL_IMAGEGL_THREAD_CHECK mActiveThread = LLThread::currentID(); @@ -525,24 +536,20 @@ void LLImageGL::init(BOOL usemipmaps) mPickMaskWidth = 0; mPickMaskHeight = 0; mUseMipMaps = usemipmaps; - mHasExplicitFormat = FALSE; + mHasExplicitFormat = false; - mIsMask = FALSE; - mNeedsAlphaAndPickMask = TRUE ; + mIsMask = false; + mNeedsAlphaAndPickMask = true ; mAlphaStride = 0 ; mAlphaOffset = 0 ; - mGLTextureCreated = FALSE ; + mGLTextureCreated = false ; mTexName = 0; mWidth = 0; mHeight = 0; mCurrentDiscardLevel = -1; - mDiscardLevelInAtlas = -1 ; - mTexelsInAtlas = 0 ; - mTexelsInGLTexture = 0 ; - - mAllowCompression = true; + mAllowCompression = allow_compression; mTarget = GL_TEXTURE_2D; mBindTarget = LLTexUnit::TT_TEXTURE; @@ -561,10 +568,10 @@ void LLImageGL::init(BOOL usemipmaps) mFormatInternal = -1; mFormatPrimary = (LLGLenum) 0; mFormatType = GL_UNSIGNED_BYTE; - mFormatSwapBytes = FALSE; + mFormatSwapBytes = false; #ifdef DEBUG_MISS - mMissed = FALSE; + mMissed = false; #endif mCategory = -1; @@ -618,9 +625,6 @@ bool LLImageGL::setSize(S32 width, S32 height, S32 ncomponents, S32 discard_leve return false; } - // pickmask validity depends on old image size, delete it - freePickMask(); - mWidth = width; mHeight = height; mComponents = ncomponents; @@ -682,12 +686,12 @@ void LLImageGL::forceUpdateBindStats(void) const mLastBindTime = sLastFrameTime; } -BOOL LLImageGL::updateBindStats() const +bool LLImageGL::updateBindStats() const { if (mTexName != 0) { #ifdef DEBUG_MISS - mMissed = ! getIsResident(TRUE); + mMissed = ! getIsResident(true); #endif sBindCount++; if (mLastBindTime != sLastFrameTime) @@ -696,10 +700,10 @@ BOOL LLImageGL::updateBindStats() const sUniqueCount++; mLastBindTime = sLastFrameTime; - return TRUE ; + return true ; } } - return FALSE ; + return false ; } F32 LLImageGL::getTimePassedSinceLastBound() @@ -707,11 +711,11 @@ F32 LLImageGL::getTimePassedSinceLastBound() return sLastFrameTime - mLastBindTime ; } -void LLImageGL::setExplicitFormat( LLGLint internal_format, LLGLenum primary_format, LLGLenum type_format, BOOL swap_bytes ) +void LLImageGL::setExplicitFormat( LLGLint internal_format, LLGLenum primary_format, LLGLenum type_format, bool swap_bytes ) { // Note: must be called before createTexture() // Note: it's up to the caller to ensure that the format matches the number of components. - mHasExplicitFormat = TRUE; + mHasExplicitFormat = true; mFormatInternal = internal_format; mFormatPrimary = primary_format; if(type_format == 0) @@ -732,10 +736,10 @@ void LLImageGL::setImage(const LLImageRaw* imageraw) (imageraw->getHeight() == getHeight(mCurrentDiscardLevel)) && (imageraw->getComponents() == getComponents())); const U8* rawdata = imageraw->getData(); - setImage(rawdata, FALSE); + setImage(rawdata, false); } -BOOL LLImageGL::setImage(const U8* data_in, BOOL data_hasmips /* = FALSE */, S32 usename /* = 0 */) +bool LLImageGL::setImage(const U8* data_in, bool data_hasmips /* = false */, S32 usename /* = 0 */) { LL_PROFILE_ZONE_SCOPED_CATEGORY_TEXTURE; @@ -785,7 +789,7 @@ BOOL LLImageGL::setImage(const U8* data_in, BOOL data_hasmips /* = FALSE */, S32 } if (is_compressed) { - S32 tex_size = dataFormatBytes(mFormatPrimary, w, h); + GLsizei tex_size = (GLsizei)dataFormatBytes(mFormatPrimary, w, h); glCompressedTexImage2D(mTarget, gl_level, mFormatPrimary, w, h, 0, tex_size, (GLvoid *)data_in); stop_glerror(); } @@ -917,7 +921,7 @@ BOOL LLImageGL::setImage(const U8* data_in, BOOL data_hasmips /* = FALSE */, S32 } mGLTextureCreated = false; - return FALSE; + return false; } else { @@ -988,7 +992,7 @@ BOOL LLImageGL::setImage(const U8* data_in, BOOL data_hasmips /* = FALSE */, S32 S32 h = getHeight(); if (is_compressed) { - S32 tex_size = dataFormatBytes(mFormatPrimary, w, h); + GLsizei tex_size = (GLsizei)dataFormatBytes(mFormatPrimary, w, h); glCompressedTexImage2D(mTarget, 0, mFormatPrimary, w, h, 0, tex_size, (GLvoid *)data_in); stop_glerror(); } @@ -1018,99 +1022,7 @@ BOOL LLImageGL::setImage(const U8* data_in, BOOL data_hasmips /* = FALSE */, S32 } stop_glerror(); mGLTextureCreated = true; - return TRUE; -} - -BOOL LLImageGL::preAddToAtlas(S32 discard_level, const LLImageRaw* raw_image) -{ - //not compatible with core GL profile - llassert(!LLRender::sGLCoreProfile); - - if (gGLManager.mIsDisabled) - { - LL_WARNS() << "Trying to create a texture while GL is disabled!" << LL_ENDL; - return FALSE; - } - llassert(gGLManager.mInited); - stop_glerror(); - - if (discard_level < 0) - { - llassert(mCurrentDiscardLevel >= 0); - discard_level = mCurrentDiscardLevel; - } - - // 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 - if (!setSize(w, h, raw_image->getComponents(), discard_level)) - { - LL_WARNS() << "Trying to create a texture with incorrect dimensions!" << LL_ENDL; - return FALSE; - } - - if (!mHasExplicitFormat) - { - switch (mComponents) - { - case 1: - // Use luminance alpha (for fonts) - mFormatInternal = GL_LUMINANCE8; - mFormatPrimary = GL_LUMINANCE; - mFormatType = GL_UNSIGNED_BYTE; - break; - case 2: - // Use luminance alpha (for fonts) - mFormatInternal = GL_LUMINANCE8_ALPHA8; - mFormatPrimary = GL_LUMINANCE_ALPHA; - mFormatType = GL_UNSIGNED_BYTE; - break; - case 3: - mFormatInternal = GL_RGB8; - mFormatPrimary = GL_RGB; - mFormatType = GL_UNSIGNED_BYTE; - break; - case 4: - mFormatInternal = GL_RGBA8; - mFormatPrimary = GL_RGBA; - mFormatType = GL_UNSIGNED_BYTE; - break; - default: - LL_ERRS() << "Bad number of components for texture: " << (U32) getComponents() << LL_ENDL; - } - } - - mCurrentDiscardLevel = discard_level; - mDiscardLevelInAtlas = discard_level; - mTexelsInAtlas = raw_image->getWidth() * raw_image->getHeight() ; - mLastBindTime = sLastFrameTime; - mGLTextureCreated = false ; - - glPixelStorei(GL_UNPACK_ROW_LENGTH, raw_image->getWidth()); - stop_glerror(); - - if(mFormatSwapBytes) - { - glPixelStorei(GL_UNPACK_SWAP_BYTES, 1); - stop_glerror(); - } - - return TRUE ; -} - -void LLImageGL::postAddToAtlas() -{ - if(mFormatSwapBytes) - { - glPixelStorei(GL_UNPACK_SWAP_BYTES, 0); - stop_glerror(); - } - - glPixelStorei(GL_UNPACK_ROW_LENGTH, 0); - gGL.getTexUnit(0)->setTextureFilteringOption(mFilterOption); - stop_glerror(); + return true; } U32 type_width_from_pixtype(U32 pixtype) @@ -1141,11 +1053,11 @@ U32 type_width_from_pixtype(U32 pixtype) bool should_stagger_image_set(bool compressed) { #if LL_DARWIN - return false; + return !compressed && on_main_thread() && gGLManager.mIsAMD; #else // glTexSubImage2D doesn't work with compressed textures on select tested Nvidia GPUs on Windows 10 -Cosmic,2023-03-08 // Setting media textures off-thread seems faster when not using sub_image_lines (Nvidia/Windows 10) -Cosmic,2023-03-31 - return !compressed && on_main_thread(); + return !compressed && on_main_thread() && !gGLManager.mIsIntel; #endif } @@ -1155,43 +1067,75 @@ void sub_image_lines(U32 target, S32 miplevel, S32 x_offset, S32 y_offset, S32 w { LL_PROFILE_ZONE_SCOPED_CATEGORY_TEXTURE; + LL_PROFILE_ZONE_NUM(width); + LL_PROFILE_ZONE_NUM(height); + U32 components = LLImageGL::dataFormatComponents(pixformat); U32 type_width = type_width_from_pixtype(pixtype); const U32 line_width = data_width * components * type_width; const U32 y_offset_end = y_offset + height; - for (U32 y_pos = y_offset; y_pos < y_offset_end; ++y_pos) + + if (width == data_width && height % 32 == 0) + { + LL_PROFILE_ZONE_NAMED_CATEGORY_TEXTURE("subimage - batched lines"); + + // full width, batch multiple lines at a time + // set batch size based on width + U32 batch_size = 32; + + if (width > 1024) + { + batch_size = 8; + } + else if (width > 512) + { + batch_size = 16; + } + + // full width texture, do 32 lines at a time + for (U32 y_pos = y_offset; y_pos < y_offset_end; y_pos += batch_size) + { + glTexSubImage2D(target, miplevel, x_offset, y_pos, width, batch_size, pixformat, pixtype, src); + src += line_width * batch_size; + } + } + else { - glTexSubImage2D(target, miplevel, x_offset, y_pos, width, 1, pixformat, pixtype, src); - src += line_width; + // partial width or strange height + for (U32 y_pos = y_offset; y_pos < y_offset_end; y_pos += 1) + { + glTexSubImage2D(target, miplevel, x_offset, y_pos, width, 1, pixformat, pixtype, src); + src += line_width; + } } } -BOOL LLImageGL::setSubImage(const U8* datap, S32 data_width, S32 data_height, S32 x_pos, S32 y_pos, S32 width, S32 height, BOOL force_fast_update /* = FALSE */, LLGLuint use_name) +bool LLImageGL::setSubImage(const U8* datap, S32 data_width, S32 data_height, S32 x_pos, S32 y_pos, S32 width, S32 height, bool force_fast_update /* = false */, LLGLuint use_name) { LL_PROFILE_ZONE_SCOPED_CATEGORY_TEXTURE; if (!width || !height) { - return TRUE; + return true; } LLGLuint tex_name = use_name != 0 ? use_name : mTexName; if (0 == tex_name) { // *TODO: Re-enable warning? Ran into thread locking issues? DK 2011-02-18 //LL_WARNS() << "Setting subimage on image without GL texture" << LL_ENDL; - return FALSE; + return false; } if (datap == NULL) { // *TODO: Re-enable warning? Ran into thread locking issues? DK 2011-02-18 //LL_WARNS() << "Setting subimage on image with NULL datap" << LL_ENDL; - return FALSE; + return false; } // HACK: allow the caller to explicitly force the fast path (i.e. using glTexSubImage2D here instead of calling setImage) even when updating the full texture. if (!force_fast_update && x_pos == 0 && y_pos == 0 && width == getWidth() && height == getHeight() && data_width == width && data_height == height) { - setImage(datap, FALSE, tex_name); + setImage(datap, false, tex_name); } else { @@ -1243,7 +1187,7 @@ BOOL LLImageGL::setSubImage(const U8* datap, S32 data_width, S32 data_height, S3 const U8* sub_datap = datap + (y_pos * data_width + x_pos) * getComponents(); // Update the GL texture - BOOL res = gGL.getTexUnit(0)->bindManual(mBindTarget, tex_name); + bool res = gGL.getTexUnit(0)->bindManual(mBindTarget, tex_name); if (!res) LL_ERRS() << "LLImageGL::setSubImage(): bindTexture failed" << LL_ENDL; stop_glerror(); @@ -1273,28 +1217,28 @@ BOOL LLImageGL::setSubImage(const U8* datap, S32 data_width, S32 data_height, S3 stop_glerror(); mGLTextureCreated = true; } - return TRUE; + return true; } -BOOL LLImageGL::setSubImage(const LLImageRaw* imageraw, S32 x_pos, S32 y_pos, S32 width, S32 height, BOOL force_fast_update /* = FALSE */, LLGLuint use_name) +bool LLImageGL::setSubImage(const LLImageRaw* imageraw, S32 x_pos, S32 y_pos, S32 width, S32 height, bool force_fast_update /* = false */, LLGLuint use_name) { LL_PROFILE_ZONE_SCOPED_CATEGORY_TEXTURE; return setSubImage(imageraw->getData(), imageraw->getWidth(), imageraw->getHeight(), x_pos, y_pos, width, height, force_fast_update, use_name); } // Copy sub image from frame buffer -BOOL LLImageGL::setSubImageFromFrameBuffer(S32 fb_x, S32 fb_y, S32 x_pos, S32 y_pos, S32 width, S32 height) +bool LLImageGL::setSubImageFromFrameBuffer(S32 fb_x, S32 fb_y, S32 x_pos, S32 y_pos, S32 width, S32 height) { if (gGL.getTexUnit(0)->bind(this, false, true)) { glCopyTexSubImage2D(GL_TEXTURE_2D, 0, fb_x, fb_y, x_pos, y_pos, width, height); mGLTextureCreated = true; stop_glerror(); - return TRUE; + return true; } else { - return FALSE; + return false; } } @@ -1314,7 +1258,7 @@ void LLImageGL::generateTextures(S32 numTextures, U32 *textures) name_count = pool_size; } - if (numTextures <= name_count) + if ((U32)numTextures <= name_count) { //copy teture names off the end of the pool memcpy(textures, name_pool + name_count - numTextures, sizeof(U32) * numTextures); @@ -1327,13 +1271,37 @@ void LLImageGL::generateTextures(S32 numTextures, U32 *textures) } } +constexpr int DELETE_DELAY = 3; // number of frames to wait before deleting textures +static std::vector<U32> sFreeList[DELETE_DELAY+1]; + +// static +void LLImageGL::updateClass() +{ + sFrameCount++; + + // wait a few frames before actually deleting the textures to avoid + // synchronization issues with the GPU + U32 idx = (sFrameCount+DELETE_DELAY) % (DELETE_DELAY+1); + + if (!sFreeList[idx].empty()) + { + free_tex_images((GLsizei) sFreeList[idx].size(), sFreeList[idx].data()); + glDeleteTextures((GLsizei)sFreeList[idx].size(), sFreeList[idx].data()); + sFreeList[idx].resize(0); + } +} + // static void LLImageGL::deleteTextures(S32 numTextures, const U32 *textures) { if (gGLManager.mInited) { - free_tex_images(numTextures, textures); - glDeleteTextures(numTextures, textures); + LL_PROFILE_ZONE_SCOPED_CATEGORY_TEXTURE; + U32 idx = sFrameCount % (DELETE_DELAY+1); + for (S32 i = 0; i < numTextures; ++i) + { + sFreeList[idx].push_back(textures[i]); + } } } @@ -1341,90 +1309,97 @@ void LLImageGL::deleteTextures(S32 numTextures, const U32 *textures) void LLImageGL::setManualImage(U32 target, S32 miplevel, S32 intformat, S32 width, S32 height, U32 pixformat, U32 pixtype, const void* pixels, bool allow_compression) { LL_PROFILE_ZONE_SCOPED_CATEGORY_TEXTURE; - bool use_scratch = false; - U32* scratch = NULL; if (LLRender::sGLCoreProfile) { - if (pixformat == GL_ALPHA && pixtype == GL_UNSIGNED_BYTE) - { //GL_ALPHA is deprecated, convert to RGBA - if (pixels != nullptr) - { - use_scratch = true; - scratch = new(std::nothrow) U32[width * height]; - if (!scratch) - { - LLError::LLUserWarningMsg::showOutOfMemory(); - LL_ERRS() << "Failed to allocate " << (U32)(width * height * sizeof(U32)) - << " bytes for a manual image W" << width << " H" << height << LL_ENDL; - } + LL_PROFILE_ZONE_SCOPED_CATEGORY_TEXTURE; + if (gGLManager.mGLVersion >= CONVERSION_SCRATCH_BUFFER_GL_VERSION) + { + if (pixformat == GL_ALPHA) + { //GL_ALPHA is deprecated, convert to RGBA + const GLint mask[] = { GL_ZERO, GL_ZERO, GL_ZERO, GL_RED }; + glTexParameteriv(GL_TEXTURE_2D, GL_TEXTURE_SWIZZLE_RGBA, mask); + pixformat = GL_RED; + intformat = GL_R8; + } - U32 pixel_count = (U32)(width * height); - for (U32 i = 0; i < pixel_count; i++) - { - U8* pix = (U8*)&scratch[i]; - pix[0] = pix[1] = pix[2] = 0; - pix[3] = ((U8*)pixels)[i]; - } + if (pixformat == GL_LUMINANCE) + { //GL_LUMINANCE is deprecated, convert to GL_RGBA + const GLint mask[] = { GL_RED, GL_RED, GL_RED, GL_ONE }; + glTexParameteriv(GL_TEXTURE_2D, GL_TEXTURE_SWIZZLE_RGBA, mask); + pixformat = GL_RED; + intformat = GL_R8; } - pixformat = GL_RGBA; - intformat = GL_RGBA8; + if (pixformat == GL_LUMINANCE_ALPHA) + { //GL_LUMINANCE_ALPHA is deprecated, convert to RGBA + const GLint mask[] = { GL_RED, GL_RED, GL_RED, GL_GREEN }; + glTexParameteriv(GL_TEXTURE_2D, GL_TEXTURE_SWIZZLE_RGBA, mask); + pixformat = GL_RG; + intformat = GL_RG8; + } } - - if (pixformat == GL_LUMINANCE_ALPHA && pixtype == GL_UNSIGNED_BYTE) - { //GL_LUMINANCE_ALPHA is deprecated, convert to RGBA - if (pixels != nullptr) - { - use_scratch = true; - scratch = new(std::nothrow) U32[width * height]; - if (!scratch) + else + { + if (pixformat == GL_ALPHA && pixtype == GL_UNSIGNED_BYTE) + { //GL_ALPHA is deprecated, convert to RGBA + if (pixels != nullptr) { - LLError::LLUserWarningMsg::showOutOfMemory(); - LL_ERRS() << "Failed to allocate " << (U32)(width * height * sizeof(U32)) - << " bytes for a manual image W" << width << " H" << height << LL_ENDL; + U32 pixel_count = (U32)(width * height); + for (U32 i = 0; i < pixel_count; i++) + { + U8* pix = (U8*)&sManualScratch[i]; + pix[0] = pix[1] = pix[2] = 0; + pix[3] = ((U8*)pixels)[i]; + } + + pixels = sManualScratch; } - U32 pixel_count = (U32)(width * height); - for (U32 i = 0; i < pixel_count; i++) + pixformat = GL_RGBA; + intformat = GL_RGBA8; + } + + if (pixformat == GL_LUMINANCE_ALPHA && pixtype == GL_UNSIGNED_BYTE) + { //GL_LUMINANCE_ALPHA is deprecated, convert to RGBA + if (pixels != nullptr) { - U8 lum = ((U8*)pixels)[i * 2 + 0]; - U8 alpha = ((U8*)pixels)[i * 2 + 1]; + U32 pixel_count = (U32)(width * height); + for (U32 i = 0; i < pixel_count; i++) + { + U8 lum = ((U8*)pixels)[i * 2 + 0]; + U8 alpha = ((U8*)pixels)[i * 2 + 1]; + + U8* pix = (U8*)&sManualScratch[i]; + pix[0] = pix[1] = pix[2] = lum; + pix[3] = alpha; + } - U8* pix = (U8*)&scratch[i]; - pix[0] = pix[1] = pix[2] = lum; - pix[3] = alpha; + pixels = sManualScratch; } - } - pixformat = GL_RGBA; - intformat = GL_RGBA8; - } + pixformat = GL_RGBA; + intformat = GL_RGBA8; + } - if (pixformat == GL_LUMINANCE && pixtype == GL_UNSIGNED_BYTE) - { //GL_LUMINANCE_ALPHA is deprecated, convert to RGB - if (pixels != nullptr) - { - use_scratch = true; - scratch = new(std::nothrow) U32[width * height]; - if (!scratch) + if (pixformat == GL_LUMINANCE && pixtype == GL_UNSIGNED_BYTE) + { //GL_LUMINANCE_ALPHA is deprecated, convert to RGB + if (pixels != nullptr) { - LLError::LLUserWarningMsg::showOutOfMemory(); - LL_ERRS() << "Failed to allocate " << (U32)(width * height * sizeof(U32)) - << " bytes for a manual image W" << width << " H" << height << LL_ENDL; - } + U32 pixel_count = (U32)(width * height); + for (U32 i = 0; i < pixel_count; i++) + { + U8 lum = ((U8*)pixels)[i]; - U32 pixel_count = (U32)(width * height); - for (U32 i = 0; i < pixel_count; i++) - { - U8 lum = ((U8*)pixels)[i]; + U8* pix = (U8*)&sManualScratch[i]; + pix[0] = pix[1] = pix[2] = lum; + pix[3] = 255; + } - U8* pix = (U8*)&scratch[i]; - pix[0] = pix[1] = pix[2] = lum; - pix[3] = 255; + pixels = sManualScratch; } + pixformat = GL_RGBA; + intformat = GL_RGB8; } - pixformat = GL_RGBA; - intformat = GL_RGB8; } } @@ -1433,6 +1408,14 @@ void LLImageGL::setManualImage(U32 target, S32 miplevel, S32 intformat, S32 widt { switch (intformat) { + case GL_RED: + case GL_R8: + intformat = GL_COMPRESSED_RED; + break; + case GL_RG: + case GL_RG8: + intformat = GL_COMPRESSED_RG; + break; case GL_RGB: case GL_RGB8: intformat = GL_COMPRESSED_RGB; @@ -1461,12 +1444,8 @@ void LLImageGL::setManualImage(U32 target, S32 miplevel, S32 intformat, S32 widt case GL_ALPHA8: intformat = GL_COMPRESSED_ALPHA; break; - case GL_RED: - case GL_R8: - intformat = GL_COMPRESSED_RED; - break; default: - LL_WARNS() << "Could not compress format: " << std::hex << intformat << LL_ENDL; + LL_WARNS() << "Could not compress format: " << std::hex << intformat << std::dec << LL_ENDL; break; } } @@ -1482,7 +1461,7 @@ void LLImageGL::setManualImage(U32 target, S32 miplevel, S32 intformat, S32 widt if (!use_sub_image) { LL_PROFILE_ZONE_NAMED("glTexImage2D alloc + copy"); - glTexImage2D(target, miplevel, intformat, width, height, 0, pixformat, pixtype, use_scratch ? scratch : pixels); + glTexImage2D(target, miplevel, intformat, width, height, 0, pixformat, pixtype, pixels); } else { @@ -1492,26 +1471,21 @@ void LLImageGL::setManualImage(U32 target, S32 miplevel, S32 intformat, S32 widt glTexImage2D(target, miplevel, intformat, width, height, 0, pixformat, pixtype, nullptr); } - U8* src = (U8*)(use_scratch ? scratch : pixels); + U8* src = (U8*)(pixels); if (src) { LL_PROFILE_ZONE_NAMED("glTexImage2D copy"); sub_image_lines(target, miplevel, 0, 0, width, height, pixformat, pixtype, src, width); } } - alloc_tex_image(width, height, pixformat); + alloc_tex_image(width, height, intformat, 1); } stop_glerror(); - - if (use_scratch) - { - delete[] scratch; - } } //create an empty GL texture: just create a texture name //the texture is assiciate with some image by calling glTexImage outside LLImageGL -BOOL LLImageGL::createGLTexture() +bool LLImageGL::createGLTexture() { LL_PROFILE_ZONE_SCOPED_CATEGORY_TEXTURE; checkActiveThread(); @@ -1519,7 +1493,7 @@ BOOL LLImageGL::createGLTexture() if (gGLManager.mIsDisabled) { LL_WARNS() << "Trying to create a texture while GL is disabled!" << LL_ENDL; - return FALSE; + return false; } mGLTextureCreated = false ; //do not save this texture when gl is destroyed. @@ -1539,13 +1513,13 @@ BOOL LLImageGL::createGLTexture() if (!mTexName) { LL_WARNS() << "LLImageGL::createGLTexture failed to make an empty texture" << LL_ENDL; - return FALSE; + return false; } - return TRUE ; + return true ; } -BOOL LLImageGL::createGLTexture(S32 discard_level, const LLImageRaw* imageraw, S32 usename/*=0*/, BOOL to_create, S32 category, bool defer_copy, LLGLuint* tex_name) +bool LLImageGL::createGLTexture(S32 discard_level, const LLImageRaw* imageraw, S32 usename/*=0*/, bool to_create, S32 category, bool defer_copy, LLGLuint* tex_name) { LL_PROFILE_ZONE_SCOPED_CATEGORY_TEXTURE; checkActiveThread(); @@ -1553,7 +1527,7 @@ BOOL LLImageGL::createGLTexture(S32 discard_level, const LLImageRaw* imageraw, S if (gGLManager.mIsDisabled) { LL_WARNS() << "Trying to create a texture while GL is disabled!" << LL_ENDL; - return FALSE; + return false; } llassert(gGLManager.mInited); @@ -1563,7 +1537,7 @@ BOOL LLImageGL::createGLTexture(S32 discard_level, const LLImageRaw* imageraw, S { LL_WARNS() << "Trying to create a texture from invalid image data" << LL_ENDL; mGLTextureCreated = false; - return FALSE; + return false; } if (discard_level < 0) @@ -1584,7 +1558,7 @@ BOOL LLImageGL::createGLTexture(S32 discard_level, const LLImageRaw* imageraw, S { LL_WARNS() << "Trying to create a texture with incorrect dimensions!" << LL_ENDL; mGLTextureCreated = false; - return FALSE; + return false; } if (mHasExplicitFormat && @@ -1593,7 +1567,7 @@ BOOL LLImageGL::createGLTexture(S32 discard_level, const LLImageRaw* imageraw, S { LL_WARNS() << "Incorrect format: " << std::hex << mFormatPrimary << " components: " << (U32)mComponents << LL_ENDL; - mHasExplicitFormat = FALSE; + mHasExplicitFormat = false; } if( !mHasExplicitFormat ) @@ -1635,15 +1609,15 @@ BOOL LLImageGL::createGLTexture(S32 discard_level, const LLImageRaw* imageraw, S mCurrentDiscardLevel = discard_level; mLastBindTime = sLastFrameTime; mGLTextureCreated = false; - return TRUE ; + return true ; } setCategory(category); const U8* rawdata = imageraw->getData(); - return createGLTexture(discard_level, rawdata, FALSE, usename, defer_copy, tex_name); + return createGLTexture(discard_level, rawdata, false, usename, defer_copy, tex_name); } -BOOL LLImageGL::createGLTexture(S32 discard_level, const U8* data_in, BOOL data_hasmips, S32 usename, bool defer_copy, LLGLuint* tex_name) +bool LLImageGL::createGLTexture(S32 discard_level, const U8* data_in, bool data_hasmips, S32 usename, bool defer_copy, LLGLuint* tex_name) // Call with void data, vmem is allocated but unitialized { LL_PROFILE_ZONE_SCOPED_CATEGORY_TEXTURE; @@ -1716,7 +1690,7 @@ BOOL LLImageGL::createGLTexture(S32 discard_level, const U8* data_in, BOOL data_ LL_PROFILE_ZONE_NAMED("cglt - late setImage"); if (!setImage(data_in, data_hasmips, new_texname)) { - return FALSE; + return false; } } @@ -1748,13 +1722,12 @@ BOOL LLImageGL::createGLTexture(S32 discard_level, const U8* data_in, BOOL data_ mTextureMemory = (S64Bytes)getMipBytes(mCurrentDiscardLevel); - mTexelsInGLTexture = getWidth() * getHeight(); // mark this as bound at this point, so we don't throw it out immediately mLastBindTime = sLastFrameTime; checkActiveThread(); - return TRUE; + return true; } void LLImageGL::syncToMainThread(LLGLuint new_tex_name) @@ -1801,7 +1774,7 @@ void LLImageGL::syncToMainThread(LLGLuint new_tex_name) ref(); LL::WorkQueue::postMaybe( mMainQueue, - [=]() + [=, this]() { LL_PROFILE_ZONE_NAMED("cglt - delete callback"); syncTexName(new_tex_name); @@ -1824,10 +1797,9 @@ void LLImageGL::syncTexName(LLGLuint texname) } } -BOOL LLImageGL::readBackRaw(S32 discard_level, LLImageRaw* imageraw, bool compressed_ok) const +bool LLImageGL::readBackRaw(S32 discard_level, LLImageRaw* imageraw, bool compressed_ok) const { - llassert_always(sAllowReadBackRaw) ; - //LL_ERRS() << "should not call this function!" << LL_ENDL ; + LL_PROFILE_ZONE_SCOPED_CATEGORY_TEXTURE; if (discard_level < 0) { @@ -1836,7 +1808,7 @@ BOOL LLImageGL::readBackRaw(S32 discard_level, LLImageRaw* imageraw, bool compre if (mTexName == 0 || discard_level < mCurrentDiscardLevel || discard_level > mMaxDiscardLevel ) { - return FALSE; + return false; } S32 gl_discard = discard_level - mCurrentDiscardLevel; @@ -1853,7 +1825,7 @@ BOOL LLImageGL::readBackRaw(S32 discard_level, LLImageRaw* imageraw, bool compre if (glwidth == 0) { // No mip data smaller than current discard level - return FALSE; + return false; } S32 width = getWidth(discard_level); @@ -1861,14 +1833,14 @@ BOOL LLImageGL::readBackRaw(S32 discard_level, LLImageRaw* imageraw, bool compre S32 ncomponents = getComponents(); if (ncomponents == 0) { - return FALSE; + return false; } if(width < glwidth) { LL_WARNS() << "texture size is smaller than it should be." << LL_ENDL ; LL_WARNS() << "width: " << width << " glwidth: " << glwidth << " mWidth: " << mWidth << " mCurrentDiscardLevel: " << (S32)mCurrentDiscardLevel << " discard_level: " << (S32)discard_level << LL_ENDL ; - return FALSE ; + return false ; } if (width <= 0 || width > 2048 || height <= 0 || height > 2048 || ncomponents < 1 || ncomponents > 4) @@ -1890,6 +1862,8 @@ BOOL LLImageGL::readBackRaw(S32 discard_level, LLImageRaw* imageraw, bool compre } //----------------------------------------------------------------------------------------------- + LLImageDataLock lock(imageraw); + if (is_compressed) { LLGLint glbytes; @@ -1898,7 +1872,7 @@ BOOL LLImageGL::readBackRaw(S32 discard_level, LLImageRaw* imageraw, bool compre { LL_WARNS() << "Memory allocation failed for reading back texture. Size is: " << glbytes << LL_ENDL ; LL_WARNS() << "width: " << width << "height: " << height << "components: " << ncomponents << LL_ENDL ; - return FALSE ; + return false ; } glGetCompressedTexImage(mTarget, gl_discard, (GLvoid*)(imageraw->getData())); @@ -1910,7 +1884,7 @@ BOOL LLImageGL::readBackRaw(S32 discard_level, LLImageRaw* imageraw, bool compre { LL_WARNS() << "Memory allocation failed for reading back texture." << LL_ENDL ; LL_WARNS() << "width: " << width << "height: " << height << "components: " << ncomponents << LL_ENDL ; - return FALSE ; + return false ; } glGetTexImage(GL_TEXTURE_2D, gl_discard, mFormatPrimary, mFormatType, (GLvoid*)(imageraw->getData())); @@ -1928,11 +1902,11 @@ BOOL LLImageGL::readBackRaw(S32 discard_level, LLImageRaw* imageraw, bool compre LL_WARNS() << "GL Error happens after reading back texture. Error code: " << error << LL_ENDL ; } - return FALSE ; + return false ; } //----------------------------------------------------------------------------------------------- - return TRUE ; + return true ; } void LLImageGL::destroyGLTexture() @@ -1949,7 +1923,7 @@ void LLImageGL::destroyGLTexture() LLImageGL::deleteTextures(1, &mTexName); mCurrentDiscardLevel = -1 ; //invalidate mCurrentDiscardLevel. mTexName = 0; - mGLTextureCreated = FALSE ; + mGLTextureCreated = false ; } } @@ -2000,7 +1974,7 @@ void LLImageGL::setFilteringOption(LLTexUnit::eTextureFilterOptions option) } } -BOOL LLImageGL::getIsResident(BOOL test_now) +bool LLImageGL::getIsResident(bool test_now) { if (test_now) { @@ -2010,7 +1984,7 @@ BOOL LLImageGL::getIsResident(BOOL test_now) } else { - mIsResident = FALSE; + mIsResident = false; } } @@ -2073,17 +2047,17 @@ S64 LLImageGL::getMipBytes(S32 discard_level) const return res; } -BOOL LLImageGL::isJustBound() const +bool LLImageGL::isJustBound() const { - return (BOOL)(sLastFrameTime - mLastBindTime < 0.5f); + return sLastFrameTime - mLastBindTime < 0.5f; } -BOOL LLImageGL::getBoundRecently() const +bool LLImageGL::getBoundRecently() const { - return (BOOL)(sLastFrameTime - mLastBindTime < MIN_TEXTURE_LIFETIME); + return (bool)(sLastFrameTime - mLastBindTime < MIN_TEXTURE_LIFETIME); } -BOOL LLImageGL::getIsAlphaMask() const +bool LLImageGL::getIsAlphaMask() const { llassert_always(!sSkipAnalyzeAlpha); return mIsMask; @@ -2096,7 +2070,7 @@ void LLImageGL::setTarget(const LLGLenum target, const LLTexUnit::eTextureType b } const S8 INVALID_OFFSET = -99 ; -void LLImageGL::setNeedsAlphaAndPickMask(BOOL need_mask) +void LLImageGL::setNeedsAlphaAndPickMask(bool need_mask) { if(mNeedsAlphaAndPickMask != need_mask) { @@ -2109,7 +2083,7 @@ void LLImageGL::setNeedsAlphaAndPickMask(BOOL need_mask) else //do not need alpha mask { mAlphaOffset = INVALID_OFFSET ; - mIsMask = FALSE; + mIsMask = false; } } } @@ -2134,8 +2108,8 @@ void LLImageGL::calcAlphaChannelOffsetAndStride() case GL_RED: case GL_RGB: case GL_SRGB: - mNeedsAlphaAndPickMask = FALSE; - mIsMask = FALSE; + mNeedsAlphaAndPickMask = false; + mIsMask = false; return; //no alpha channel. case GL_RGBA: case GL_SRGB_ALPHA: @@ -2182,8 +2156,8 @@ void LLImageGL::calcAlphaChannelOffsetAndStride() { LL_WARNS() << "Cannot analyze alpha for image with format type " << std::hex << mFormatType << std::dec << LL_ENDL; - mNeedsAlphaAndPickMask = FALSE ; - mIsMask = FALSE; + mNeedsAlphaAndPickMask = false ; + mIsMask = false; } } @@ -2194,6 +2168,8 @@ void LLImageGL::analyzeAlpha(const void* data_in, U32 w, U32 h) return ; } + LL_PROFILE_ZONE_SCOPED_CATEGORY_TEXTURE; + U32 length = w * h; U32 alphatotal = 0; @@ -2207,13 +2183,13 @@ void LLImageGL::analyzeAlpha(const void* data_in, U32 w, U32 h) // suffer the worst from aliasing when used as alpha masks. if (w >= 2 && h >= 2) { - llassert(w%2 == 0); - llassert(h%2 == 0); + llassert(w % 2 == 0); + llassert(h % 2 == 0); const GLubyte* rowstart = ((const GLubyte*) data_in) + mAlphaOffset; - for (U32 y = 0; y < h; y+=2) + for (U32 y = 0; y < h; y += 2) { const GLubyte* current = rowstart; - for (U32 x = 0; x < w; x+=2) + for (U32 x = 0; x < w; x += 2) { const U32 s1 = current[0]; alphatotal += s1; @@ -2236,7 +2212,6 @@ void LLImageGL::analyzeAlpha(const void* data_in, U32 w, U32 h) sample[asum/(16*4)] += 4; } - rowstart += 2 * w * mAlphaStride; } length *= 2; // we sampled everything twice, essentially @@ -2280,17 +2255,19 @@ void LLImageGL::analyzeAlpha(const void* data_in, U32 w, U32 h) (lowerhalftotal == length && alphatotal != 0) || // all close to transparent but not all totally transparent, or (upperhalftotal == length && alphatotal != 255*length)) // all close to opaque but not all totally opaque { - mIsMask = FALSE; // not suitable for masking + mIsMask = false; // not suitable for masking } else { - mIsMask = TRUE; + mIsMask = true; } } //---------------------------------------------------------------------------- U32 LLImageGL::createPickMask(S32 pWidth, S32 pHeight) { + LL_PROFILE_ZONE_SCOPED_CATEGORY_TEXTURE; + freePickMask(); U32 pick_width = pWidth/2 + 1; U32 pick_height = pHeight/2 + 1; @@ -2308,7 +2285,6 @@ U32 LLImageGL::createPickMask(S32 pWidth, S32 pHeight) //---------------------------------------------------------------------------- void LLImageGL::freePickMask() { - // pickmask validity depends on old image size, delete it if (mPickMask != NULL) { delete [] mPickMask; @@ -2346,16 +2322,16 @@ void LLImageGL::updatePickMask(S32 width, S32 height, const U8* data_in) return ; } - freePickMask(); - if (mFormatType != GL_UNSIGNED_BYTE || ((mFormatPrimary != GL_RGBA) && (mFormatPrimary != GL_SRGB_ALPHA))) { //cannot generate a pick mask for this texture + freePickMask(); return; } + #ifdef SHOW_ASSERT const U32 pickSize = createPickMask(width, height); #else // SHOW_ASSERT @@ -2384,9 +2360,9 @@ void LLImageGL::updatePickMask(S32 width, S32 height, const U8* data_in) } } -BOOL LLImageGL::getMask(const LLVector2 &tc) +bool LLImageGL::getMask(const LLVector2 &tc) { - BOOL res = TRUE; + bool res = true; if (mPickMask) { @@ -2430,13 +2406,13 @@ BOOL LLImageGL::getMask(const LLVector2 &tc) S32 idx = y*mPickMaskWidth+x; S32 offset = idx%8; - res = mPickMask[idx/8] & (1 << offset) ? TRUE : FALSE; + res = (mPickMask[idx/8] & (1 << offset)) != 0; } return res; } -void LLImageGL::setCurTexSizebar(S32 index, BOOL set_pick_size) +void LLImageGL::setCurTexSizebar(S32 index, bool set_pick_size) { sCurTexSizeBar = index ; @@ -2454,6 +2430,107 @@ void LLImageGL::resetCurTexSizebar() sCurTexSizeBar = -1 ; sCurTexPickSize = -1 ; } + +bool LLImageGL::scaleDown(S32 desired_discard) +{ + LL_PROFILE_ZONE_SCOPED_CATEGORY_TEXTURE; + + if (mTarget != GL_TEXTURE_2D + || mFormatInternal == -1 // not initialized + ) + { + return false; + } + + desired_discard = llmin(desired_discard, mMaxDiscardLevel); + + if (desired_discard <= mCurrentDiscardLevel) + { + return false; + } + + S32 mip = desired_discard - mCurrentDiscardLevel; + + S32 desired_width = getWidth(desired_discard); + S32 desired_height = getHeight(desired_discard); + + if (gGLManager.mDownScaleMethod == 0) + { // use an FBO to downscale the texture + glViewport(0, 0, desired_width, desired_height); + + // draw a full screen triangle + if (gGL.getTexUnit(0)->bind(this, true, true)) + { + glDrawArrays(GL_TRIANGLES, 0, 3); + + free_tex_image(mTexName); + glTexImage2D(mTarget, 0, mFormatInternal, desired_width, desired_height, 0, mFormatPrimary, mFormatType, nullptr); + glCopyTexSubImage2D(mTarget, 0, 0, 0, 0, 0, desired_width, desired_height); + alloc_tex_image(desired_width, desired_height, mFormatInternal, 1); + + mTexOptionsDirty = true; + + if (mHasMipMaps) + { // generate mipmaps if needed + LL_PROFILE_ZONE_NAMED_CATEGORY_TEXTURE("scaleDown - glGenerateMipmap"); + gGL.getTexUnit(0)->bind(this); + glGenerateMipmap(mTarget); + gGL.getTexUnit(0)->unbind(LLTexUnit::TT_TEXTURE); + } + } + else + { + LL_WARNS_ONCE("LLImageGL") << "Failed to bind texture for downscaling." << LL_ENDL; + return false; + } + } + else + { // use a PBO to downscale the texture + U64 size = getBytes(desired_discard); + llassert(size <= 2048 * 2048 * 4); // we shouldn't be using this method to downscale huge textures, but it'll work + gGL.getTexUnit(0)->bind(this, false, true); + + if (sScratchPBO == 0) + { + glGenBuffers(1, &sScratchPBO); + sScratchPBOSize = 0; + } + + glBindBuffer(GL_PIXEL_PACK_BUFFER, sScratchPBO); + + if (size > sScratchPBOSize) + { + glBufferData(GL_PIXEL_PACK_BUFFER, size, NULL, GL_STREAM_COPY); + sScratchPBOSize = (U32)size; + } + + glGetTexImage(mTarget, mip, mFormatPrimary, mFormatType, nullptr); + + free_tex_image(mTexName); + + glBindBuffer(GL_PIXEL_PACK_BUFFER, 0); + + glBindBuffer(GL_PIXEL_UNPACK_BUFFER, sScratchPBO); + glTexImage2D(mTarget, 0, mFormatInternal, desired_width, desired_height, 0, mFormatPrimary, mFormatType, nullptr); + glBindBuffer(GL_PIXEL_UNPACK_BUFFER, 0); + + alloc_tex_image(desired_width, desired_height, mFormatInternal, 1); + + if (mHasMipMaps) + { + LL_PROFILE_ZONE_NAMED_CATEGORY_TEXTURE("scaleDown - glGenerateMipmap"); + glGenerateMipmap(mTarget); + } + + gGL.getTexUnit(0)->unbind(LLTexUnit::TT_TEXTURE); + } + + mCurrentDiscardLevel = desired_discard; + + return true; +} + + //---------------------------------------------------------------------------- #if LL_IMAGEGL_THREAD_CHECK void LLImageGL::checkActiveThread() diff --git a/indra/llrender/llimagegl.h b/indra/llrender/llimagegl.h index 89e35d2226..6b4492c09e 100644 --- a/indra/llrender/llimagegl.h +++ b/indra/llrender/llimagegl.h @@ -39,6 +39,7 @@ #include "llrender.h" #include "threadpool.h" #include "workqueue.h" +#include <unordered_set> #define LL_IMAGEGL_THREAD_CHECK 0 //set to 1 to enable thread debugging for ImageGL @@ -47,12 +48,23 @@ class LLWindow; #define BYTES_TO_MEGA_BYTES(x) ((x) >> 20) #define MEGA_BYTES_TO_BYTES(x) ((x) << 20) +namespace LLImageGLMemory +{ + void alloc_tex_image(U32 width, U32 height, U32 intformat, U32 count); + void free_tex_image(U32 texName); + void free_tex_images(U32 count, const U32* texNames); + void free_cur_tex_image(); +} + //============================================================================ class LLImageGL : public LLRefCount { friend class LLTexUnit; public: + // call once per frame + static void updateClass(); + // Get an estimate of how many bytes have been allocated in vram for textures. // Does not include mipmaps. // NOTE: multiplying this number by two gives a good estimate for total @@ -68,16 +80,15 @@ public: static S64 dataFormatBytes(S32 dataformat, S32 width, S32 height); static S32 dataFormatComponents(S32 dataformat); - BOOL updateBindStats() const ; + bool updateBindStats() const ; F32 getTimePassedSinceLastBound(); void forceUpdateBindStats(void) const; // needs to be called every frame static void updateStats(F32 current_time); - // Save off / restore GL textures - static void destroyGL(BOOL save_state = TRUE); - static void restoreGL(); + // cleanup GL state + static void destroyGL(); static void dirtyTexOptions(); static bool checkSize(S32 width, S32 height); @@ -85,14 +96,14 @@ public: //for server side use only. // Not currently necessary for LLImageGL, but required in some derived classes, // so include for compatability - static BOOL create(LLPointer<LLImageGL>& dest, BOOL usemipmaps = TRUE); - static BOOL create(LLPointer<LLImageGL>& dest, U32 width, U32 height, U8 components, BOOL usemipmaps = TRUE); - static BOOL create(LLPointer<LLImageGL>& dest, const LLImageRaw* imageraw, BOOL usemipmaps = TRUE); + static bool create(LLPointer<LLImageGL>& dest, bool usemipmaps = true); + static bool create(LLPointer<LLImageGL>& dest, U32 width, U32 height, U8 components, bool usemipmaps = true); + static bool create(LLPointer<LLImageGL>& dest, const LLImageRaw* imageraw, bool usemipmaps = true); public: - LLImageGL(BOOL usemipmaps = TRUE); - LLImageGL(U32 width, U32 height, U8 components, BOOL usemipmaps = TRUE); - LLImageGL(const LLImageRaw* imageraw, BOOL usemipmaps = TRUE); + LLImageGL(bool usemipmaps = true, bool allow_compression = true); + LLImageGL(U32 width, U32 height, U8 components, bool usemipmaps = true, bool allow_compression = true); + LLImageGL(const LLImageRaw* imageraw, bool usemipmaps = true, bool allow_compression = true); // For wrapping textures created via GL elsewhere with our API only. Use with caution. LLImageGL(LLGLuint mTexName, U32 components, LLGLenum target, LLGLint formatInternal, LLGLenum formatPrimary, LLGLenum formatType, LLTexUnit::eTextureAddressMode addressMode); @@ -112,34 +123,38 @@ public: 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, + bool createGLTexture() ; + bool createGLTexture(S32 discard_level, const LLImageRaw* imageraw, S32 usename = 0, bool to_create = true, S32 category = sMaxCategories-1, bool defer_copy = false, LLGLuint* tex_name = nullptr); - BOOL createGLTexture(S32 discard_level, const U8* data, BOOL data_hasmips = FALSE, S32 usename = 0, bool defer_copy = false, LLGLuint* tex_name = nullptr); + bool createGLTexture(S32 discard_level, const U8* data, bool data_hasmips = false, S32 usename = 0, bool defer_copy = false, LLGLuint* tex_name = nullptr); void setImage(const LLImageRaw* imageraw); - BOOL setImage(const U8* data_in, BOOL data_hasmips = FALSE, S32 usename = 0); + bool setImage(const U8* data_in, bool data_hasmips = false, S32 usename = 0); // *TODO: This function may not work if the textures is compressed (i.e. // RenderCompressTextures is 0). Partial image updates do not work on // compressed textures. - BOOL setSubImage(const LLImageRaw* imageraw, S32 x_pos, S32 y_pos, S32 width, S32 height, BOOL force_fast_update = FALSE, LLGLuint use_name = 0); - BOOL setSubImage(const U8* datap, S32 data_width, S32 data_height, S32 x_pos, S32 y_pos, S32 width, S32 height, BOOL force_fast_update = FALSE, LLGLuint use_name = 0); - BOOL setSubImageFromFrameBuffer(S32 fb_x, S32 fb_y, S32 x_pos, S32 y_pos, S32 width, S32 height); + bool setSubImage(const LLImageRaw* imageraw, S32 x_pos, S32 y_pos, S32 width, S32 height, bool force_fast_update = false, LLGLuint use_name = 0); + bool setSubImage(const U8* datap, S32 data_width, S32 data_height, S32 x_pos, S32 y_pos, S32 width, S32 height, bool force_fast_update = false, LLGLuint use_name = 0); + bool setSubImageFromFrameBuffer(S32 fb_x, S32 fb_y, S32 x_pos, S32 y_pos, S32 width, S32 height); // wait for gl commands to finish on current thread and push // a lambda to main thread to swap mNewTexName and mTexName void syncToMainThread(LLGLuint new_tex_name); // Read back a raw image for this discard level, if it exists - BOOL readBackRaw(S32 discard_level, LLImageRaw* imageraw, bool compressed_ok) const; + 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 setExplicitFormat(LLGLint internal_format, LLGLenum primary_format, LLGLenum type_format = 0, bool swap_bytes = false); void setComponents(S8 ncomponents) { mComponents = ncomponents; } S32 getDiscardLevel() const { return mCurrentDiscardLevel; } S32 getMaxDiscardLevel() const { return mMaxDiscardLevel; } + // override the current discard level + // should only be used for local textures where you know exactly what you're doing + void setDiscardLevel(S32 level) { mCurrentDiscardLevel = level; } + S32 getCurrentWidth() const { return mWidth ;} S32 getCurrentHeight() const { return mHeight ;} S32 getWidth(S32 discard_level = -1) const; @@ -147,18 +162,18 @@ public: U8 getComponents() const { return mComponents; } S64 getBytes(S32 discard_level = -1) const; S64 getMipBytes(S32 discard_level = -1) const; - BOOL getBoundRecently() const; - BOOL isJustBound() const; - BOOL getHasExplicitFormat() const { return mHasExplicitFormat; } + bool getBoundRecently() const; + bool isJustBound() const; + bool getHasExplicitFormat() const { return mHasExplicitFormat; } LLGLenum getPrimaryFormat() const { return mFormatPrimary; } LLGLenum getFormatType() const { return mFormatType; } - BOOL getHasGLTexture() const { return mTexName != 0; } + bool getHasGLTexture() const { return mTexName != 0; } LLGLuint getTexName() const { return mTexName; } - BOOL getIsAlphaMask() const; + bool getIsAlphaMask() const; - BOOL getIsResident(BOOL test_now = FALSE); // not const + bool getIsResident(bool test_now = false); // not const void setTarget(const LLGLenum target, const LLTexUnit::eTextureType bind_target); @@ -166,11 +181,11 @@ public: bool isGLTextureCreated(void) const { return mGLTextureCreated ; } void setGLTextureCreated (bool initialized) { mGLTextureCreated = initialized; } - BOOL getUseMipMaps() const { return mUseMipMaps; } - void setUseMipMaps(BOOL usemips) { mUseMipMaps = usemips; } - void setHasMipMaps(BOOL hasmips) { mHasMipMaps = hasmips; } + bool getUseMipMaps() const { return mUseMipMaps; } + void setUseMipMaps(bool usemips) { mUseMipMaps = usemips; } + void setHasMipMaps(bool hasmips) { mHasMipMaps = hasmips; } void updatePickMask(S32 width, S32 height, const U8* data_in); - BOOL getMask(const LLVector2 &tc); + bool getMask(const LLVector2 &tc); void checkTexSize(bool forced = false) const ; @@ -186,19 +201,12 @@ public: void setFilteringOption(LLTexUnit::eTextureFilterOptions option); LLTexUnit::eTextureFilterOptions getFilteringOption(void) const { return mFilterOption; } - LLGLenum getTexTarget()const { return mTarget ;} - S8 getDiscardLevelInAtlas()const {return mDiscardLevelInAtlas;} - U32 getTexelsInAtlas()const { return mTexelsInAtlas ;} - U32 getTexelsInGLTexture()const {return mTexelsInGLTexture;} - + LLGLenum getTexTarget()const { return mTarget; } - void init(BOOL usemipmaps); + void init(bool usemipmaps, bool allow_compression); virtual void cleanup(); // Clean up the LLImageGL so it can be reinitialized. Be careful when using this in derived class destructors - void setNeedsAlphaAndPickMask(BOOL need_mask); - - BOOL preAddToAtlas(S32 discard_level, const LLImageRaw* raw_image); - void postAddToAtlas() ; + void setNeedsAlphaAndPickMask(bool need_mask); #if LL_IMAGEGL_THREAD_CHECK // thread debugging @@ -206,6 +214,13 @@ public: void checkActiveThread(); #endif + // scale down to the desired discard level using GPU + // returns true if texture was scaled down + // desired discard will be clamped to max discard + // if desired discard is less than or equal to current discard, no scaling will occur + // only works for GL_TEXTURE_2D target + bool scaleDown(S32 desired_discard); + public: // Various GL/Rendering options S64Bytes mTextureMemory; @@ -222,25 +237,20 @@ private: U16 mPickMaskWidth; U16 mPickMaskHeight; S8 mUseMipMaps; - BOOL mHasExplicitFormat; // If false (default), GL format is f(mComponents) + bool mHasExplicitFormat; // If false (default), GL format is f(mComponents) bool mAutoGenMips = false; - BOOL mIsMask; - BOOL mNeedsAlphaAndPickMask; + bool mIsMask; + bool mNeedsAlphaAndPickMask; S8 mAlphaStride ; S8 mAlphaOffset ; bool mGLTextureCreated ; LLGLuint mTexName; - //LLGLuint mNewTexName = 0; // tex name set by background thread to be applied in main thread U16 mWidth; U16 mHeight; S8 mCurrentDiscardLevel; - S8 mDiscardLevelInAtlas; - U32 mTexelsInAtlas ; - U32 mTexelsInGLTexture; - bool mAllowCompression; protected: @@ -261,42 +271,46 @@ protected: LLGLint mFormatInternal; // = GL internalformat LLGLenum mFormatPrimary; // = GL format (pixel data format) LLGLenum mFormatType; - BOOL mFormatSwapBytes;// if true, use glPixelStorei(GL_UNPACK_SWAP_BYTES, 1) + bool mFormatSwapBytes;// if true, use glPixelStorei(GL_UNPACK_SWAP_BYTES, 1) - BOOL mExternalTexture; + bool mExternalTexture; // STATICS public: - static std::set<LLImageGL*> sImageList; + static std::unordered_set<LLImageGL*> sImageList; static S32 sCount; - + static U32 sFrameCount; static F32 sLastFrameTime; // Global memory statistics static U32 sBindCount; // Tracks number of texture binds for current frame static U32 sUniqueCount; // Tracks number of unique texture binds for current frame - static BOOL sGlobalUseAnisotropic; + static bool sGlobalUseAnisotropic; static LLImageGL* sDefaultGLTexture ; - static BOOL sAutomatedTest; + static bool sAutomatedTest; static bool sCompressTextures; //use GL texture compression #if DEBUG_MISS - BOOL mMissed; // Missed on last bind? - BOOL getMissed() const { return mMissed; }; + bool mMissed; // Missed on last bind? + bool getMissed() const { return mMissed; }; #else - BOOL getMissed() const { return FALSE; }; + bool getMissed() const { return false; }; #endif public: - static void initClass(LLWindow* window, S32 num_catagories, BOOL skip_analyze_alpha = false, bool thread_texture_loads = false, bool thread_media_updates = false); + static void initClass(LLWindow* window, S32 num_catagories, bool skip_analyze_alpha = false, bool thread_texture_loads = false, bool thread_media_updates = false); + static void allocateConversionBuffer(); static void cleanupClass() ; private: static S32 sMaxCategories; - static BOOL sSkipAnalyzeAlpha; + static bool sSkipAnalyzeAlpha; + static U32 sScratchPBO; + static U32 sScratchPBOSize; + static U32* sManualScratch; //the flag to allow to call readBackRaw(...). //can be removed if we do not use that function at all. - static BOOL sAllowReadBackRaw ; + static bool sAllowReadBackRaw ; // //**************************************************************************************************** //The below for texture auditing use only @@ -317,7 +331,7 @@ public: static S32 sCurTexSizeBar ; static S32 sCurTexPickSize ; - static void setCurTexSizebar(S32 index, BOOL set_pick_size = TRUE) ; + static void setCurTexSizebar(S32 index, bool set_pick_size = true) ; static void resetCurTexSizebar(); //**************************************************************************************************** diff --git a/indra/llrender/llpostprocess.cpp b/indra/llrender/llpostprocess.cpp index 4242e330f4..eef7193c92 100644 --- a/indra/llrender/llpostprocess.cpp +++ b/indra/llrender/llpostprocess.cpp @@ -165,7 +165,7 @@ void LLPostProcess::invalidate() mSceneRenderTexture = NULL ; mNoiseTexture = NULL ; mTempBloomTexture = NULL ; - initialized = FALSE ; + initialized = false ; } void LLPostProcess::apply(unsigned int width, unsigned int height) @@ -343,7 +343,7 @@ void LLPostProcess::viewOrthogonal(unsigned int width, unsigned int height) gGL.matrixMode(LLRender::MM_PROJECTION); gGL.pushMatrix(); gGL.loadIdentity(); - gGL.ortho( 0.f, (GLdouble) width , (GLdouble) height , 0.f, -1.f, 1.f ); + gGL.ortho( 0.f, (GLfloat) width , (GLfloat) height , 0.f, -1.f, 1.f ); gGL.matrixMode(LLRender::MM_MODELVIEW); gGL.pushMatrix(); gGL.loadIdentity(); @@ -367,7 +367,7 @@ void LLPostProcess::createTexture(LLPointer<LLImageGL>& texture, unsigned int wi { std::vector<GLubyte> data(width * height * 4, 0) ; - texture = new LLImageGL(FALSE) ; + texture = new LLImageGL(false) ; if(texture->createGLTexture()) { gGL.getTexUnit(0)->bindManual(LLTexUnit::TT_TEXTURE, texture->getTexName()); @@ -387,7 +387,7 @@ void LLPostProcess::createNoiseTexture(LLPointer<LLImageGL>& texture) } } - texture = new LLImageGL(FALSE) ; + texture = new LLImageGL(false) ; if(texture->createGLTexture()) { gGL.getTexUnit(0)->bindManual(LLTexUnit::TT_TEXTURE, texture->getTexName()); diff --git a/indra/llrender/llrender.cpp b/indra/llrender/llrender.cpp index b92c9363fc..1dc87a66ce 100644 --- a/indra/llrender/llrender.cpp +++ b/indra/llrender/llrender.cpp @@ -36,6 +36,7 @@ #include "lltexture.h" #include "llshadermgr.h" #include "hbxxh.h" +#include "glm/gtc/type_ptr.hpp" #if LL_WINDOWS extern void APIENTRY gl_debug_callback(GLenum source, @@ -57,8 +58,8 @@ F32 gGLLastProjection[16]; F32 gGLProjection[16]; // transform from last frame's camera space to this frame's camera space (and inverse) -F32 gGLDeltaModelView[16]; -F32 gGLInverseDeltaModelView[16]; +glm::mat4 gGLDeltaModelView; +glm::mat4 gGLInverseDeltaModelView; S32 gGLViewport[4]; @@ -69,6 +70,7 @@ U32 LLTexUnit::sWhiteTexture = 0; bool LLRender::sGLCoreProfile = false; bool LLRender::sNsightDebugSupport = false; LLVector2 LLRender::sUIGLScaleFactor = LLVector2(1.f, 1.f); +bool LLRender::sClassicMode = false; struct LLVBCache { @@ -77,6 +79,7 @@ struct LLVBCache }; static std::unordered_map<U64, LLVBCache> sVBCache; +static thread_local std::list<LLVertexBufferData> *sBufferDataList = nullptr; static const GLenum sGLTextureType[] = { @@ -115,7 +118,7 @@ static const GLenum sGLBlendFactor[] = LLTexUnit::LLTexUnit(S32 index) : mCurrTexType(TT_NONE), - mCurrColorScale(1), mCurrAlphaScale(1), mCurrTexture(0), mTexColorSpace(TCS_LINEAR), + mCurrColorScale(1), mCurrAlphaScale(1), mCurrTexture(0), mHasMipMaps(false), mIndex(index) { @@ -145,8 +148,6 @@ void LLTexUnit::refreshState(void) { glBindTexture(GL_TEXTURE_2D, 0); } - - setTextureColorSpace(mTexColorSpace); } void LLTexUnit::activate(void) @@ -241,7 +242,6 @@ bool LLTexUnit::bind(LLTexture* texture, bool for_rendering, bool forceBind) setTextureAddressMode(gl_tex->mAddressMode); setTextureFilteringOption(gl_tex->mFilterOption); } - setTextureColorSpace(mTexColorSpace); } } else @@ -318,7 +318,6 @@ bool LLTexUnit::bind(LLImageGL* texture, bool for_rendering, bool forceBind, S32 setTextureFilteringOption(texture->mFilterOption); stop_glerror(); } - setTextureColorSpace(mTexColorSpace); } stop_glerror(); @@ -354,7 +353,6 @@ bool LLTexUnit::bind(LLCubeMap* cubeMap) setTextureAddressMode(cubeMap->mImages[0]->mAddressMode); setTextureFilteringOption(cubeMap->mImages[0]->mFilterOption); } - setTextureColorSpace(mTexColorSpace); return true; } else @@ -403,7 +401,6 @@ bool LLTexUnit::bindManual(eTextureType type, U32 texture, bool hasMips) mCurrTexture = texture; glBindTexture(sGLTextureType[type], texture); mHasMipMaps = hasMips; - setTextureColorSpace(mTexColorSpace); } return true; } @@ -424,8 +421,6 @@ void LLTexUnit::unbind(eTextureType type) { mCurrTexture = 0; - // Always make sure our texture color space is reset to linear. SRGB sampling should be opt-in in the vast majority of cases. Also prevents color space "popping". - mTexColorSpace = TCS_LINEAR; if (type == LLTexUnit::TT_TEXTURE) { glBindTexture(sGLTextureType[type], sWhiteTexture); @@ -447,8 +442,6 @@ void LLTexUnit::unbindFast(eTextureType type) { mCurrTexture = 0; - // Always make sure our texture color space is reset to linear. SRGB sampling should be opt-in in the vast majority of cases. Also prevents color space "popping". - mTexColorSpace = TCS_LINEAR; if (type == LLTexUnit::TT_TEXTURE) { glBindTexture(sGLTextureType[type], sWhiteTexture); @@ -642,11 +635,6 @@ void LLTexUnit::debugTextureUnit(void) } } -void LLTexUnit::setTextureColorSpace(eTextureColorSpace space) -{ - mTexColorSpace = space; -} - LLLightState::LLLightState(S32 index) : mIndex(index), mEnabled(false), @@ -749,10 +737,9 @@ void LLLightState::setPosition(const LLVector4& position) ++gGL.mLightHash; mPosition = position; //transform position by current modelview matrix - glh::vec4f pos(position.mV); - const glh::matrix4f& mat = gGL.getModelviewMatrix(); - mat.mult_matrix_vec(pos); - mPosition.set(pos.v); + glm::vec4 pos(position); + pos = gGL.getModelviewMatrix() * pos; + mPosition.set(glm::value_ptr(pos)); } void LLLightState::setConstantAttenuation(const F32& atten) @@ -804,19 +791,18 @@ void LLLightState::setSpotDirection(const LLVector3& direction) { //always set direction because modelview matrix may have changed ++gGL.mLightHash; - mSpotDirection = direction; + //transform direction by current modelview matrix - glh::vec3f dir(direction.mV); - const glh::matrix4f& mat = gGL.getModelviewMatrix(); - mat.mult_matrix_dir(dir); + glm::vec3 dir(direction); + const glm::mat3 mat(gGL.getModelviewMatrix()); + dir = mat * dir; - mSpotDirection.set(dir.v); + mSpotDirection.set(glm::value_ptr(dir)); } LLRender::LLRender() : mDirty(false), mCount(0), - mQuadCycle(0), mMode(LLRender::TRIANGLES), mCurrTextureUnitIndex(0) { @@ -844,6 +830,10 @@ LLRender::LLRender() for (U32 i = 0; i < NUM_MATRIX_MODES; ++i) { + for (U32 j = 0; j < LL_MATRIX_STACK_DEPTH; ++j) + { + mMatrix[i][j] = glm::identity<glm::mat4>(); + } mMatIdx[i] = 0; mMatHash[i] = 0; mCurMatHash[i] = 0xFFFFFFFF; @@ -982,14 +972,21 @@ void LLRender::syncLightState() shader->uniform3fv(LLShaderMgr::LIGHT_DIFFUSE, LL_NUM_LIGHT_UNITS, diffuse[0].mV); shader->uniform3fv(LLShaderMgr::LIGHT_AMBIENT, 1, mAmbientLightColor.mV); shader->uniform1i(LLShaderMgr::SUN_UP_FACTOR, sun_primary[0] ? 1 : 0); - //shader->uniform3fv(LLShaderMgr::AMBIENT, 1, mAmbientLightColor.mV); - //shader->uniform3fv(LLShaderMgr::SUNLIGHT_COLOR, 1, diffuse[0].mV); - //shader->uniform3fv(LLShaderMgr::MOONLIGHT_COLOR, 1, diffuse_b[0].mV); + + if (sClassicMode) + { + shader->uniform3fv(LLShaderMgr::AMBIENT, 1, mAmbientLightColor.mV); + shader->uniform3fv(LLShaderMgr::SUNLIGHT_COLOR, 1, diffuse[0].mV); + shader->uniform3fv(LLShaderMgr::MOONLIGHT_COLOR, 1, diffuse_b[0].mV); + } } } void LLRender::syncMatrices() { + STOP_GLERROR; + LL_PROFILE_ZONE_SCOPED_CATEGORY_DISPLAY; + static const U32 name[] = { LLShaderMgr::MODELVIEW_MATRIX, @@ -1002,32 +999,30 @@ void LLRender::syncMatrices() LLGLSLShader* shader = LLGLSLShader::sCurBoundShaderPtr; - static glh::matrix4f cached_mvp; - static glh::matrix4f cached_inv_mdv; + static glm::mat4 cached_mvp; + static glm::mat4 cached_inv_mdv; static U32 cached_mvp_mdv_hash = 0xFFFFFFFF; static U32 cached_mvp_proj_hash = 0xFFFFFFFF; - static glh::matrix4f cached_normal; + static glm::mat4 cached_normal; static U32 cached_normal_hash = 0xFFFFFFFF; if (shader) { - //llassert(shader); - bool mvp_done = false; U32 i = MM_MODELVIEW; if (mMatHash[MM_MODELVIEW] != shader->mMatHash[MM_MODELVIEW]) { //update modelview, normal, and MVP - glh::matrix4f& mat = mMatrix[MM_MODELVIEW][mMatIdx[MM_MODELVIEW]]; + const glm::mat4& mat = mMatrix[MM_MODELVIEW][mMatIdx[MM_MODELVIEW]]; // if MDV has changed, update the cached inverse as well if (cached_mvp_mdv_hash != mMatHash[MM_MODELVIEW]) { - cached_inv_mdv = mat.inverse(); + cached_inv_mdv = glm::inverse(mat); } - shader->uniformMatrix4fv(name[MM_MODELVIEW], 1, GL_FALSE, mat.m); + shader->uniformMatrix4fv(name[MM_MODELVIEW], 1, GL_FALSE, glm::value_ptr(mat)); shader->mMatHash[MM_MODELVIEW] = mMatHash[MM_MODELVIEW]; //update normal matrix @@ -1036,17 +1031,17 @@ void LLRender::syncMatrices() { if (cached_normal_hash != mMatHash[i]) { - cached_normal = cached_inv_mdv.transpose(); + cached_normal = glm::transpose(cached_inv_mdv); cached_normal_hash = mMatHash[i]; } - glh::matrix4f& norm = cached_normal; + auto norm = glm::value_ptr(cached_normal); F32 norm_mat[] = { - norm.m[0], norm.m[1], norm.m[2], - norm.m[4], norm.m[5], norm.m[6], - norm.m[8], norm.m[9], norm.m[10] + norm[0], norm[1], norm[2], + norm[4], norm[5], norm[6], + norm[8], norm[9], norm[10] }; shader->uniformMatrix3fv(LLShaderMgr::NORMAL_MATRIX, 1, GL_FALSE, norm_mat); @@ -1054,7 +1049,7 @@ void LLRender::syncMatrices() if (shader->getUniformLocation(LLShaderMgr::INVERSE_MODELVIEW_MATRIX)) { - shader->uniformMatrix4fv(LLShaderMgr::INVERSE_MODELVIEW_MATRIX, 1, GL_FALSE, cached_inv_mdv.m); + shader->uniformMatrix4fv(LLShaderMgr::INVERSE_MODELVIEW_MATRIX, 1, GL_FALSE, glm::value_ptr(cached_inv_mdv)); } //update MVP matrix @@ -1067,36 +1062,36 @@ void LLRender::syncMatrices() if (cached_mvp_mdv_hash != mMatHash[i] || cached_mvp_proj_hash != mMatHash[MM_PROJECTION]) { cached_mvp = mat; - cached_mvp.mult_left(mMatrix[proj][mMatIdx[proj]]); + cached_mvp = mMatrix[proj][mMatIdx[proj]] * cached_mvp; cached_mvp_mdv_hash = mMatHash[i]; cached_mvp_proj_hash = mMatHash[MM_PROJECTION]; } - shader->uniformMatrix4fv(LLShaderMgr::MODELVIEW_PROJECTION_MATRIX, 1, GL_FALSE, cached_mvp.m); + shader->uniformMatrix4fv(LLShaderMgr::MODELVIEW_PROJECTION_MATRIX, 1, GL_FALSE, glm::value_ptr(cached_mvp)); } } i = MM_PROJECTION; if (mMatHash[MM_PROJECTION] != shader->mMatHash[MM_PROJECTION]) { //update projection matrix, normal, and MVP - glh::matrix4f& mat = mMatrix[MM_PROJECTION][mMatIdx[MM_PROJECTION]]; + const glm::mat4& mat = mMatrix[MM_PROJECTION][mMatIdx[MM_PROJECTION]]; // GZ: This was previously disabled seemingly due to a bug involving the deferred renderer's regular pushing and popping of mats. // We're reenabling this and cleaning up the code around that - that would've been the appropriate course initially. // Anything beyond the standard proj and inv proj mats are special cases. Please setup special uniforms accordingly in the future. if (shader->getUniformLocation(LLShaderMgr::INVERSE_PROJECTION_MATRIX)) { - glh::matrix4f inv_proj = mat.inverse(); - shader->uniformMatrix4fv(LLShaderMgr::INVERSE_PROJECTION_MATRIX, 1, FALSE, inv_proj.m); + glm::mat4 inv_proj = glm::inverse(mat); + shader->uniformMatrix4fv(LLShaderMgr::INVERSE_PROJECTION_MATRIX, 1, false, glm::value_ptr(inv_proj)); } // Used by some full screen effects - such as full screen lights, glow, etc. if (shader->getUniformLocation(LLShaderMgr::IDENTITY_MATRIX)) { - shader->uniformMatrix4fv(LLShaderMgr::IDENTITY_MATRIX, 1, GL_FALSE, glh::matrix4f::identity().m); + shader->uniformMatrix4fv(LLShaderMgr::IDENTITY_MATRIX, 1, GL_FALSE, glm::value_ptr(glm::identity<glm::mat4>())); } - shader->uniformMatrix4fv(name[MM_PROJECTION], 1, GL_FALSE, mat.m); + shader->uniformMatrix4fv(name[MM_PROJECTION], 1, GL_FALSE, glm::value_ptr(mat)); shader->mMatHash[MM_PROJECTION] = mMatHash[MM_PROJECTION]; if (!mvp_done) @@ -1109,12 +1104,12 @@ void LLRender::syncMatrices() { U32 mdv = MM_MODELVIEW; cached_mvp = mat; - cached_mvp.mult_right(mMatrix[mdv][mMatIdx[mdv]]); + cached_mvp *= mMatrix[mdv][mMatIdx[mdv]]; cached_mvp_mdv_hash = mMatHash[MM_MODELVIEW]; cached_mvp_proj_hash = mMatHash[MM_PROJECTION]; } - shader->uniformMatrix4fv(LLShaderMgr::MODELVIEW_PROJECTION_MATRIX, 1, GL_FALSE, cached_mvp.m); + shader->uniformMatrix4fv(LLShaderMgr::MODELVIEW_PROJECTION_MATRIX, 1, GL_FALSE, glm::value_ptr(cached_mvp)); } } } @@ -1123,7 +1118,7 @@ void LLRender::syncMatrices() { if (mMatHash[i] != shader->mMatHash[i]) { - shader->uniformMatrix4fv(name[i], 1, GL_FALSE, mMatrix[i][mMatIdx[i]].m); + shader->uniformMatrix4fv(name[i], 1, GL_FALSE, glm::value_ptr(mMatrix[i][mMatIdx[i]])); shader->mMatHash[i] = mMatHash[i]; } } @@ -1134,6 +1129,7 @@ void LLRender::syncMatrices() syncLightState(); } } + STOP_GLERROR; } void LLRender::translatef(const GLfloat& x, const GLfloat& y, const GLfloat& z) @@ -1141,12 +1137,7 @@ void LLRender::translatef(const GLfloat& x, const GLfloat& y, const GLfloat& z) flush(); { - glh::matrix4f trans_mat(1,0,0,x, - 0,1,0,y, - 0,0,1,z, - 0,0,0,1); - - mMatrix[mMatrixMode][mMatIdx[mMatrixMode]].mult_right(trans_mat); + mMatrix[mMatrixMode][mMatIdx[mMatrixMode]] = glm::translate(mMatrix[mMatrixMode][mMatIdx[mMatrixMode]], glm::vec3(x, y, z)); mMatHash[mMatrixMode]++; } } @@ -1156,12 +1147,7 @@ void LLRender::scalef(const GLfloat& x, const GLfloat& y, const GLfloat& z) flush(); { - glh::matrix4f scale_mat(x,0,0,0, - 0,y,0,0, - 0,0,z,0, - 0,0,0,1); - - mMatrix[mMatrixMode][mMatIdx[mMatrixMode]].mult_right(scale_mat); + mMatrix[mMatrixMode][mMatIdx[mMatrixMode]] = glm::scale(mMatrix[mMatrixMode][mMatIdx[mMatrixMode]], glm::vec3(x, y, z)); mMatHash[mMatrixMode]++; } } @@ -1171,13 +1157,7 @@ void LLRender::ortho(F32 left, F32 right, F32 bottom, F32 top, F32 zNear, F32 zF flush(); { - - glh::matrix4f ortho_mat(2.f/(right-left),0,0, -(right+left)/(right-left), - 0,2.f/(top-bottom),0, -(top+bottom)/(top-bottom), - 0,0,-2.f/(zFar-zNear), -(zFar+zNear)/(zFar-zNear), - 0,0,0,1); - - mMatrix[mMatrixMode][mMatIdx[mMatrixMode]].mult_right(ortho_mat); + mMatrix[mMatrixMode][mMatIdx[mMatrixMode]] *= glm::ortho(left, right, bottom, top, zNear, zFar); mMatHash[mMatrixMode]++; } } @@ -1187,19 +1167,7 @@ void LLRender::rotatef(const GLfloat& a, const GLfloat& x, const GLfloat& y, con flush(); { - F32 r = a * DEG_TO_RAD; - - F32 c = cosf(r); - F32 s = sinf(r); - - F32 ic = 1.f-c; - - glh::matrix4f rot_mat(x*x*ic+c, x*y*ic-z*s, x*z*ic+y*s, 0, - x*y*ic+z*s, y*y*ic+c, y*z*ic-x*s, 0, - x*z*ic-y*s, y*z*ic+x*s, z*z*ic+c, 0, - 0,0,0,1); - - mMatrix[mMatrixMode][mMatIdx[mMatrixMode]].mult_right(rot_mat); + mMatrix[mMatrixMode][mMatIdx[mMatrixMode]] = glm::rotate(mMatrix[mMatrixMode][mMatIdx[mMatrixMode]], glm::radians(a), glm::vec3(x,y,z)); mMatHash[mMatrixMode]++; } } @@ -1241,7 +1209,7 @@ void LLRender::loadMatrix(const GLfloat* m) { flush(); { - mMatrix[mMatrixMode][mMatIdx[mMatrixMode]].set_value((GLfloat*) m); + mMatrix[mMatrixMode][mMatIdx[mMatrixMode]] = glm::make_mat4((GLfloat*) m); mMatHash[mMatrixMode]++; } } @@ -1250,9 +1218,7 @@ void LLRender::multMatrix(const GLfloat* m) { flush(); { - glh::matrix4f mat((GLfloat*) m); - - mMatrix[mMatrixMode][mMatIdx[mMatrixMode]].mult_right(mat); + mMatrix[mMatrixMode][mMatIdx[mMatrixMode]] *= glm::make_mat4(m); mMatHash[mMatrixMode]++; } } @@ -1295,17 +1261,17 @@ void LLRender::loadIdentity() { llassert_always(mMatrixMode < NUM_MATRIX_MODES) ; - mMatrix[mMatrixMode][mMatIdx[mMatrixMode]].make_identity(); + mMatrix[mMatrixMode][mMatIdx[mMatrixMode]] = glm::identity<glm::mat4>(); mMatHash[mMatrixMode]++; } } -const glh::matrix4f& LLRender::getModelviewMatrix() +const glm::mat4& LLRender::getModelviewMatrix() { return mMatrix[MM_MODELVIEW][mMatIdx[MM_MODELVIEW]]; } -const glh::matrix4f& LLRender::getProjectionMatrix() +const glm::mat4& LLRender::getProjectionMatrix() { return mMatrix[MM_PROJECTION][mMatIdx[MM_PROJECTION]]; } @@ -1336,7 +1302,7 @@ void LLRender::pushUIMatrix() { if (mUIOffset.empty()) { - mUIOffset.push_back(LLVector3(0,0,0)); + mUIOffset.emplace_back(0.f,0.f,0.f); } else { @@ -1345,7 +1311,7 @@ void LLRender::pushUIMatrix() if (mUIScale.empty()) { - mUIScale.push_back(LLVector3(1,1,1)); + mUIScale.emplace_back(1.f,1.f,1.f); } else { @@ -1512,7 +1478,7 @@ LLLightState* LLRender::getLight(U32 index) void LLRender::setAmbientLightColor(const LLColor4& color) { - LL_PROFILE_ZONE_SCOPED_CATEGORY_PIPELINE + LL_PROFILE_ZONE_SCOPED_CATEGORY_PIPELINE; if (color != mAmbientLightColor) { ++mLightHash; @@ -1541,17 +1507,35 @@ void LLRender::clearErrors() } } +void LLRender::beginList(std::list<LLVertexBufferData> *list) +{ + if (sBufferDataList) + { + LL_ERRS() << "beginList called while another list is open." << LL_ENDL; + } + llassert(LLGLSLShader::sCurBoundShaderPtr == &gUIProgram); + flush(); + sBufferDataList = list; +} + +void LLRender::endList() +{ + if (sBufferDataList) + { + flush(); + sBufferDataList = nullptr; + } + else + { + llassert(false); // endList called without an open list + } +} + void LLRender::begin(const GLuint& mode) { if (mode != mMode) { - if (mode == LLRender::QUADS) - { - mQuadCycle = 1; - } - - if (mMode == LLRender::QUADS || - mMode == LLRender::LINES || + if (mMode == LLRender::LINES || mMode == LLRender::TRIANGLES || mMode == LLRender::POINTS) { @@ -1574,8 +1558,7 @@ void LLRender::end() //IMM_ERRS << "GL begin and end called with no vertices specified." << LL_ENDL; } - if ((mMode != LLRender::QUADS && - mMode != LLRender::LINES && + if ((mMode != LLRender::LINES && mMode != LLRender::TRIANGLES && mMode != LLRender::POINTS) || mCount > 2048) @@ -1583,12 +1566,15 @@ void LLRender::end() flush(); } } + void LLRender::flush() { + STOP_GLERROR; if (mCount > 0) { LL_PROFILE_ZONE_SCOPED_CATEGORY_PIPELINE; - llassert(LLGLSLShader::sCurBoundShaderPtr != nullptr); + llassert_always(LLGLSLShader::sCurBoundShaderPtr != nullptr); + if (!mUIOffset.empty()) { sUICalls++; @@ -1598,28 +1584,19 @@ void LLRender::flush() //store mCount in a local variable to avoid re-entrance (drawArrays may call flush) U32 count = mCount; - if (mMode == LLRender::QUADS && !sGLCoreProfile) - { - if (mCount%4 != 0) - { - count -= (mCount % 4); - LL_WARNS() << "Incomplete quad requested." << LL_ENDL; - } - } - - if (mMode == LLRender::TRIANGLES) + if (mMode == LLRender::TRIANGLES) + { + if (mCount%3 != 0) { - if (mCount%3 != 0) - { - count -= (mCount % 3); - LL_WARNS() << "Incomplete triangle requested." << LL_ENDL; - } + count -= (mCount % 3); + LL_WARNS() << "Incomplete triangle requested." << LL_ENDL; } + } - if (mMode == LLRender::LINES) + if (mMode == LLRender::LINES) + { + if (mCount%2 != 0) { - if (mCount%2 != 0) - { count -= (mCount % 2); LL_WARNS() << "Incomplete line requested." << LL_ENDL; } @@ -1630,122 +1607,155 @@ void LLRender::flush() if (mBuffer) { - HBXXH64 hash; + LLVertexBuffer *vb; + U32 attribute_mask = LLGLSLShader::sCurBoundShaderPtr->mAttributeMask; + if (sBufferDataList) + { + vb = genBuffer(attribute_mask, count); + sBufferDataList->emplace_back( + vb, + mMode, + count, + gGL.getTexUnit(0)->mCurrTexture, + mMatrix[MM_MODELVIEW][mMatIdx[MM_MODELVIEW]], + mMatrix[MM_PROJECTION][mMatIdx[MM_PROJECTION]], + mMatrix[MM_TEXTURE0][mMatIdx[MM_TEXTURE0]] + ); + } + else { - LL_PROFILE_ZONE_NAMED_CATEGORY_VERTEX("vb cache hash"); + vb = bufferfromCache(attribute_mask, count); + } - hash.update((U8*)mVerticesp.get(), count * sizeof(LLVector4a)); - if (attribute_mask & LLVertexBuffer::MAP_TEXCOORD0) - { - hash.update((U8*)mTexcoordsp.get(), count * sizeof(LLVector2)); - } + drawBuffer(vb, mMode, count); + } + else + { + // mBuffer is present in main thread and not present in an image thread + LL_ERRS() << "A flush call from outside main rendering thread" << LL_ENDL; + } - if (attribute_mask & LLVertexBuffer::MAP_COLOR) - { - hash.update((U8*)mColorsp.get(), count * sizeof(LLColor4U)); - } + resetStriders(count); + } +} - hash.finalize(); - } +LLVertexBuffer* LLRender::bufferfromCache(U32 attribute_mask, U32 count) +{ + LLVertexBuffer *vb = nullptr; + HBXXH64 hash; + { + LL_PROFILE_ZONE_NAMED_CATEGORY_VERTEX("vb cache hash"); - U64 vhash = hash.digest(); + hash.update((U8*)mVerticesp.get(), count * sizeof(LLVector4a)); + if (attribute_mask & LLVertexBuffer::MAP_TEXCOORD0) + { + hash.update((U8*)mTexcoordsp.get(), count * sizeof(LLVector2)); + } - // check the VB cache before making a new vertex buffer - // This is a giant hack to deal with (mostly) our terrible UI rendering code - // that was built on top of OpenGL immediate mode. Huge performance wins - // can be had by not uploading geometry to VRAM unless absolutely necessary. - // Most of our usage of the "immediate mode" style draw calls is actually - // sending the same geometry over and over again. - // To leverage this, we maintain a running hash of the vertex stream being - // built up before a flush, and then check that hash against a VB - // cache just before creating a vertex buffer in VRAM - std::unordered_map<U64, LLVBCache>::iterator cache = sVBCache.find(vhash); + if (attribute_mask & LLVertexBuffer::MAP_COLOR) + { + hash.update((U8*)mColorsp.get(), count * sizeof(LLColor4U)); + } - LLPointer<LLVertexBuffer> vb; + hash.finalize(); + } - if (cache != sVBCache.end()) - { - LL_PROFILE_ZONE_NAMED_CATEGORY_VERTEX("vb cache hit"); - // cache hit, just use the cached buffer - vb = cache->second.vb; - cache->second.touched = std::chrono::steady_clock::now(); - } - else - { - LL_PROFILE_ZONE_NAMED_CATEGORY_VERTEX("vb cache miss"); - vb = new LLVertexBuffer(attribute_mask); - vb->allocateBuffer(count, 0); + U64 vhash = hash.digest(); - vb->setBuffer(); + // check the VB cache before making a new vertex buffer + // This is a giant hack to deal with (mostly) our terrible UI rendering code + // that was built on top of OpenGL immediate mode. Huge performance wins + // can be had by not uploading geometry to VRAM unless absolutely necessary. + // Most of our usage of the "immediate mode" style draw calls is actually + // sending the same geometry over and over again. + // To leverage this, we maintain a running hash of the vertex stream being + // built up before a flush, and then check that hash against a VB + // cache just before creating a vertex buffer in VRAM + std::unordered_map<U64, LLVBCache>::iterator cache = sVBCache.find(vhash); - vb->setPositionData((LLVector4a*) mVerticesp.get()); + if (cache != sVBCache.end()) + { + LL_PROFILE_ZONE_NAMED_CATEGORY_VERTEX("vb cache hit"); + // cache hit, just use the cached buffer + vb = cache->second.vb; + cache->second.touched = std::chrono::steady_clock::now(); + } + else + { + LL_PROFILE_ZONE_NAMED_CATEGORY_VERTEX("vb cache miss"); + vb = genBuffer(attribute_mask, count); + + sVBCache[vhash] = { vb , std::chrono::steady_clock::now() }; - if (attribute_mask & LLVertexBuffer::MAP_TEXCOORD0) + static U32 miss_count = 0; + miss_count++; + if (miss_count > 1024) + { + LL_PROFILE_ZONE_NAMED_CATEGORY_VERTEX("vb cache clean"); + miss_count = 0; + auto now = std::chrono::steady_clock::now(); + + using namespace std::chrono_literals; + // every 1024 misses, clean the cache of any VBs that haven't been touched in the last second + for (std::unordered_map<U64, LLVBCache>::iterator iter = sVBCache.begin(); iter != sVBCache.end(); ) + { + if (now - iter->second.touched > 1s) { - vb->setTexCoordData(mTexcoordsp.get()); + iter = sVBCache.erase(iter); } - - if (attribute_mask & LLVertexBuffer::MAP_COLOR) + else { - vb->setColorData(mColorsp.get()); + ++iter; } + } + } + } + return vb; +} - vb->unbind(); +LLVertexBuffer* LLRender::genBuffer(U32 attribute_mask, S32 count) +{ + LLVertexBuffer * vb = new LLVertexBuffer(attribute_mask); + vb->allocateBuffer(count, 0); - sVBCache[vhash] = { vb , std::chrono::steady_clock::now() }; + vb->setBuffer(); - static U32 miss_count = 0; - miss_count++; - if (miss_count > 1024) - { - LL_PROFILE_ZONE_NAMED_CATEGORY_VERTEX("vb cache clean"); - miss_count = 0; - auto now = std::chrono::steady_clock::now(); + vb->setPositionData(mVerticesp.get()); - using namespace std::chrono_literals; - // every 1024 misses, clean the cache of any VBs that haven't been touched in the last second - for (std::unordered_map<U64, LLVBCache>::iterator iter = sVBCache.begin(); iter != sVBCache.end(); ) - { - if (now - iter->second.touched > 1s) - { - iter = sVBCache.erase(iter); - } - else - { - ++iter; - } - } - } - } + if (attribute_mask & LLVertexBuffer::MAP_TEXCOORD0) + { + vb->setTexCoord0Data(mTexcoordsp.get()); + } - vb->setBuffer(); + if (attribute_mask & LLVertexBuffer::MAP_COLOR) + { + vb->setColorData(mColorsp.get()); + } - if (mMode == LLRender::QUADS && sGLCoreProfile) - { - vb->drawArrays(LLRender::TRIANGLES, 0, count); - mQuadCycle = 1; - } - else - { - vb->drawArrays(mMode, 0, count); - } - } - else - { - // mBuffer is present in main thread and not present in an image thread - LL_ERRS() << "A flush call from outside main rendering thread" << LL_ENDL; - } +#if LL_DARWIN + vb->unmapBuffer(); +#endif + vb->unbind(); + return vb; +} - mVerticesp[0] = mVerticesp[count]; - mTexcoordsp[0] = mTexcoordsp[count]; - mColorsp[0] = mColorsp[count]; +void LLRender::drawBuffer(LLVertexBuffer* vb, U32 mode, S32 count) +{ + vb->setBuffer(); + vb->drawArrays(mode, 0, count); +} - mCount = 0; - } +void LLRender::resetStriders(S32 count) +{ + mVerticesp[0] = mVerticesp[count]; + mTexcoordsp[0] = mTexcoordsp[count]; + mColorsp[0] = mColorsp[count]; + + mCount = 0; } void LLRender::vertex3f(const GLfloat& x, const GLfloat& y, const GLfloat& z) @@ -1757,7 +1767,6 @@ void LLRender::vertex3f(const GLfloat& x, const GLfloat& y, const GLfloat& z) { case LLRender::POINTS: flush(); break; case LLRender::TRIANGLES: if (mCount%3==0) flush(); break; - case LLRender::QUADS: if(mCount%4 == 0) flush(); break; case LLRender::LINES: if (mCount%2 == 0) flush(); break; } } @@ -1770,31 +1779,12 @@ void LLRender::vertex3f(const GLfloat& x, const GLfloat& y, const GLfloat& z) if (mUIOffset.empty()) { - mVerticesp[mCount] = LLVector3(x,y,z); + mVerticesp[mCount].set(x,y,z); } else { LLVector3 vert = (LLVector3(x,y,z)+mUIOffset.back()).scaledVec(mUIScale.back()); - mVerticesp[mCount] = vert; - } - - if (mMode == LLRender::QUADS && LLRender::sGLCoreProfile) - { - mQuadCycle++; - if (mQuadCycle == 4) - { //copy two vertices so fourth quad element will add a triangle - mQuadCycle = 0; - - mCount++; - mVerticesp[mCount] = mVerticesp[mCount-3]; - mColorsp[mCount] = mColorsp[mCount-3]; - mTexcoordsp[mCount] = mTexcoordsp[mCount-3]; - - mCount++; - mVerticesp[mCount] = mVerticesp[mCount-2]; - mColorsp[mCount] = mColorsp[mCount-2]; - mTexcoordsp[mCount] = mTexcoordsp[mCount-2]; - } + mVerticesp[mCount].set(vert.mV[VX], vert.mV[VY], vert.mV[VZ]); } mCount++; @@ -1803,7 +1793,7 @@ void LLRender::vertex3f(const GLfloat& x, const GLfloat& y, const GLfloat& z) mTexcoordsp[mCount] = mTexcoordsp[mCount-1]; } -void LLRender::vertexBatchPreTransformed(LLVector3* verts, S32 vert_count) +void LLRender::vertexBatchPreTransformed(LLVector4a* verts, S32 vert_count) { if (mCount + vert_count > 4094) { @@ -1811,57 +1801,20 @@ void LLRender::vertexBatchPreTransformed(LLVector3* verts, S32 vert_count) return; } - if (sGLCoreProfile && mMode == LLRender::QUADS) - { //quads are deprecated, convert to triangle list - S32 i = 0; - - while (i < vert_count) - { - //read first three - mVerticesp[mCount++] = verts[i++]; - mTexcoordsp[mCount] = mTexcoordsp[mCount-1]; - mColorsp[mCount] = mColorsp[mCount-1]; - - mVerticesp[mCount++] = verts[i++]; - mTexcoordsp[mCount] = mTexcoordsp[mCount-1]; - mColorsp[mCount] = mColorsp[mCount-1]; - - mVerticesp[mCount++] = verts[i++]; - mTexcoordsp[mCount] = mTexcoordsp[mCount-1]; - mColorsp[mCount] = mColorsp[mCount-1]; - - //copy two - mVerticesp[mCount++] = verts[i-3]; - mTexcoordsp[mCount] = mTexcoordsp[mCount-1]; - mColorsp[mCount] = mColorsp[mCount-1]; - - mVerticesp[mCount++] = verts[i-1]; - mTexcoordsp[mCount] = mTexcoordsp[mCount-1]; - mColorsp[mCount] = mColorsp[mCount-1]; - - //copy last one - mVerticesp[mCount++] = verts[i++]; - mTexcoordsp[mCount] = mTexcoordsp[mCount-1]; - mColorsp[mCount] = mColorsp[mCount-1]; - } - } - else + for (S32 i = 0; i < vert_count; i++) { - for (S32 i = 0; i < vert_count; i++) - { - mVerticesp[mCount] = verts[i]; + mVerticesp[mCount] = verts[i]; - mCount++; - mTexcoordsp[mCount] = mTexcoordsp[mCount-1]; - mColorsp[mCount] = mColorsp[mCount-1]; - } + mCount++; + mTexcoordsp[mCount] = mTexcoordsp[mCount-1]; + mColorsp[mCount] = mColorsp[mCount-1]; } if( mCount > 0 ) // ND: Guard against crashes if mCount is zero, yes it can happen mVerticesp[mCount] = mVerticesp[mCount-1]; } -void LLRender::vertexBatchPreTransformed(LLVector3* verts, LLVector2* uvs, S32 vert_count) +void LLRender::vertexBatchPreTransformed(LLVector4a* verts, LLVector2* uvs, S32 vert_count) { if (mCount + vert_count > 4094) { @@ -1869,50 +1822,13 @@ void LLRender::vertexBatchPreTransformed(LLVector3* verts, LLVector2* uvs, S32 v return; } - if (sGLCoreProfile && mMode == LLRender::QUADS) - { //quads are deprecated, convert to triangle list - S32 i = 0; - - while (i < vert_count) - { - //read first three - mVerticesp[mCount] = verts[i]; - mTexcoordsp[mCount++] = uvs[i++]; - mColorsp[mCount] = mColorsp[mCount-1]; - - mVerticesp[mCount] = verts[i]; - mTexcoordsp[mCount++] = uvs[i++]; - mColorsp[mCount] = mColorsp[mCount-1]; - - mVerticesp[mCount] = verts[i]; - mTexcoordsp[mCount++] = uvs[i++]; - mColorsp[mCount] = mColorsp[mCount-1]; - - //copy last two - mVerticesp[mCount] = verts[i-3]; - mTexcoordsp[mCount++] = uvs[i-3]; - mColorsp[mCount] = mColorsp[mCount-1]; - - mVerticesp[mCount] = verts[i-1]; - mTexcoordsp[mCount++] = uvs[i-1]; - mColorsp[mCount] = mColorsp[mCount-1]; - - //copy last one - mVerticesp[mCount] = verts[i]; - mTexcoordsp[mCount++] = uvs[i++]; - mColorsp[mCount] = mColorsp[mCount-1]; - } - } - else + for (S32 i = 0; i < vert_count; i++) { - for (S32 i = 0; i < vert_count; i++) - { - mVerticesp[mCount] = verts[i]; - mTexcoordsp[mCount] = uvs[i]; + mVerticesp[mCount] = verts[i]; + mTexcoordsp[mCount] = uvs[i]; - mCount++; - mColorsp[mCount] = mColorsp[mCount-1]; - } + mCount++; + mColorsp[mCount] = mColorsp[mCount-1]; } if (mCount > 0) @@ -1922,7 +1838,7 @@ void LLRender::vertexBatchPreTransformed(LLVector3* verts, LLVector2* uvs, S32 v } } -void LLRender::vertexBatchPreTransformed(LLVector3* verts, LLVector2* uvs, LLColor4U* colors, S32 vert_count) +void LLRender::vertexBatchPreTransformed(LLVector4a* verts, LLVector2* uvs, LLColor4U* colors, S32 vert_count) { if (mCount + vert_count > 4094) { @@ -1930,51 +1846,13 @@ void LLRender::vertexBatchPreTransformed(LLVector3* verts, LLVector2* uvs, LLCol return; } - - if (sGLCoreProfile && mMode == LLRender::QUADS) - { //quads are deprecated, convert to triangle list - S32 i = 0; - - while (i < vert_count) - { - //read first three - mVerticesp[mCount] = verts[i]; - mTexcoordsp[mCount] = uvs[i]; - mColorsp[mCount++] = colors[i++]; - - mVerticesp[mCount] = verts[i]; - mTexcoordsp[mCount] = uvs[i]; - mColorsp[mCount++] = colors[i++]; - - mVerticesp[mCount] = verts[i]; - mTexcoordsp[mCount] = uvs[i]; - mColorsp[mCount++] = colors[i++]; - - //copy last two - mVerticesp[mCount] = verts[i-3]; - mTexcoordsp[mCount] = uvs[i-3]; - mColorsp[mCount++] = colors[i-3]; - - mVerticesp[mCount] = verts[i-1]; - mTexcoordsp[mCount] = uvs[i-1]; - mColorsp[mCount++] = colors[i-1]; - - //copy last one - mVerticesp[mCount] = verts[i]; - mTexcoordsp[mCount] = uvs[i]; - mColorsp[mCount++] = colors[i++]; - } - } - else + for (S32 i = 0; i < vert_count; i++) { - for (S32 i = 0; i < vert_count; i++) - { - mVerticesp[mCount] = verts[i]; - mTexcoordsp[mCount] = uvs[i]; - mColorsp[mCount] = colors[i]; + mVerticesp[mCount] = verts[i]; + mTexcoordsp[mCount] = uvs[i]; + mColorsp[mCount] = colors[i]; - mCount++; - } + mCount++; } if (mCount > 0) @@ -2158,85 +2036,93 @@ void LLRender::debugTexUnits(void) LL_INFOS("TextureUnit") << "Active TexUnit Enabled : " << active_enabled << LL_ENDL; } - - -glh::matrix4f copy_matrix(F32* src) -{ - glh::matrix4f ret; - ret.set_value(src); - return ret; -} - -glh::matrix4f get_current_modelview() +glm::mat4 get_current_modelview() { - return copy_matrix(gGLModelView); + return glm::make_mat4(gGLModelView); } -glh::matrix4f get_current_projection() +glm::mat4 get_current_projection() { - return copy_matrix(gGLProjection); + return glm::make_mat4(gGLProjection); } -glh::matrix4f get_last_modelview() +glm::mat4 get_last_modelview() { - return copy_matrix(gGLLastModelView); + return glm::make_mat4(gGLLastModelView); } -glh::matrix4f get_last_projection() +glm::mat4 get_last_projection() { - return copy_matrix(gGLLastProjection); + return glm::make_mat4(gGLLastProjection); } -void copy_matrix(const glh::matrix4f& src, F32* dst) +void copy_matrix(const glm::mat4& src, F32* dst) { + auto matp = glm::value_ptr(src); for (U32 i = 0; i < 16; i++) { - dst[i] = src.m[i]; + dst[i] = matp[i]; } } -void set_current_modelview(const glh::matrix4f& mat) +void set_current_modelview(const glm::mat4& mat) { copy_matrix(mat, gGLModelView); } -void set_current_projection(glh::matrix4f& mat) +void set_current_projection(const glm::mat4& mat) { copy_matrix(mat, gGLProjection); } -glh::matrix4f gl_ortho(GLfloat left, GLfloat right, GLfloat bottom, GLfloat top, GLfloat znear, GLfloat zfar) +void set_last_modelview(const glm::mat4& mat) { - glh::matrix4f ret( - 2.f/(right-left), 0.f, 0.f, -(right+left)/(right-left), - 0.f, 2.f/(top-bottom), 0.f, -(top+bottom)/(top-bottom), - 0.f, 0.f, -2.f/(zfar-znear), -(zfar+znear)/(zfar-znear), - 0.f, 0.f, 0.f, 1.f); - - return ret; + copy_matrix(mat, gGLLastModelView); } -glh::matrix4f gl_perspective(GLfloat fovy, GLfloat aspect, GLfloat zNear, GLfloat zFar) +void set_last_projection(const glm::mat4& mat) { - GLfloat f = 1.f/tanf(DEG_TO_RAD*fovy/2.f); - - return glh::matrix4f(f/aspect, 0, 0, 0, - 0, f, 0, 0, - 0, 0, (zFar+zNear)/(zNear-zFar), (2.f*zFar*zNear)/(zNear-zFar), - 0, 0, -1.f, 0); + copy_matrix(mat, gGLLastProjection); } -glh::matrix4f gl_lookat(LLVector3 eye, LLVector3 center, LLVector3 up) +glm::vec3 mul_mat4_vec3(const glm::mat4& mat, const glm::vec3& vec) { - LLVector3 f = center-eye; - f.normVec(); - up.normVec(); - LLVector3 s = f % up; - LLVector3 u = s % f; +#if 1 // SIMD path results in strange crashes. Fall back to scalar for now. + const float w = vec[0] * mat[0][3] + vec[1] * mat[1][3] + vec[2] * mat[2][3] + mat[3][3]; + return glm::vec3( + (vec[0] * mat[0][0] + vec[1] * mat[1][0] + vec[2] * mat[2][0] + mat[3][0]) / w, + (vec[0] * mat[0][1] + vec[1] * mat[1][1] + vec[2] * mat[2][1] + mat[3][1]) / w, + (vec[0] * mat[0][2] + vec[1] * mat[1][2] + vec[2] * mat[2][2] + mat[3][2]) / w + ); +#else + LLVector4a x, y, z, s, t, p, q; + + x.splat(vec.x); + y.splat(vec.y); + z.splat(vec.z); + + s.splat<3>(mat[0].data); + t.splat<3>(mat[1].data); + p.splat<3>(mat[2].data); + q.splat<3>(mat[3].data); - return glh::matrix4f(s[0], s[1], s[2], 0, - u[0], u[1], u[2], 0, - -f[0], -f[1], -f[2], 0, - 0, 0, 0, 1); + s.mul(x); + t.mul(y); + p.mul(z); + q.add(s); + t.add(p); + q.add(t); + x.mul(mat[0].data); + y.mul(mat[1].data); + z.mul(mat[2].data); + + x.add(y); + z.add(mat[3].data); + LLVector4a res; + res.load3(glm::value_ptr(vec)); + res.setAdd(x, z); + res.div(q); + return glm::make_vec3(res.getF32ptr()); +#endif } diff --git a/indra/llrender/llrender.h b/indra/llrender/llrender.h index be9f3895e7..7b6bd4198b 100644 --- a/indra/llrender/llrender.h +++ b/indra/llrender/llrender.h @@ -42,15 +42,17 @@ #include "llpointer.h" #include "llglheaders.h" #include "llmatrix4a.h" -#include "glh/glh_linear.h" +#include "glm/mat4x4.hpp" #include <array> +#include <list> class LLVertexBuffer; class LLCubeMap; class LLImageGL; class LLRenderTarget; -class LLTexture ; +class LLTexture; +class LLVertexBufferData; #define LL_MATRIX_STACK_DEPTH 32 @@ -219,17 +221,12 @@ public: void setHasMipMaps(bool hasMips) { mHasMipMaps = hasMips; } - void setTextureColorSpace(eTextureColorSpace space); - - eTextureColorSpace getCurrColorSpace() { return mTexColorSpace; } - protected: friend class LLRender; S32 mIndex; U32 mCurrTexture; eTextureType mCurrTexType; - eTextureColorSpace mTexColorSpace; S32 mCurrColorScale; S32 mCurrAlphaScale; bool mHasMipMaps; @@ -293,11 +290,18 @@ public: enum eTexIndex : U8 { - DIFFUSE_MAP = 0, - ALTERNATE_DIFFUSE_MAP = 1, - NORMAL_MAP = 1, - SPECULAR_MAP = 2, - NUM_TEXTURE_CHANNELS = 3, + // Channels for material textures + DIFFUSE_MAP = 0, + ALTERNATE_DIFFUSE_MAP = 1, + NORMAL_MAP = 1, + SPECULAR_MAP = 2, + // Channels for PBR textures + BASECOLOR_MAP = 3, + METALLIC_ROUGHNESS_MAP = 4, + GLTF_NORMAL_MAP = 5, + EMISSIVE_MAP = 6, + // Total number of channels + NUM_TEXTURE_CHANNELS = 7, }; enum eVolumeTexIndex : U8 @@ -315,7 +319,6 @@ public: POINTS, LINES, LINE_STRIP, - QUADS, LINE_LOOP, NUM_MODES }; @@ -397,8 +400,8 @@ public: void matrixMode(eMatrixMode mode); eMatrixMode getMatrixMode(); - const glh::matrix4f& getModelviewMatrix(); - const glh::matrix4f& getProjectionMatrix(); + const glm::mat4& getModelviewMatrix(); + const glm::mat4& getProjectionMatrix(); void syncMatrices(); void syncLightState(); @@ -413,8 +416,15 @@ public: void flush(); + // if list is set, will store buffers in list for later use, if list isn't set, will use cache + void beginList(std::list<LLVertexBufferData> *list); + void endList(); + void begin(const GLuint& mode); void end(); + + U8 getMode() const { return mMode; } + void vertex2i(const GLint& x, const GLint& y); void vertex2f(const GLfloat& x, const GLfloat& y); void vertex3f(const GLfloat& x, const GLfloat& y, const GLfloat& z); @@ -439,9 +449,9 @@ public: void diffuseColor4ubv(const U8* c); void diffuseColor4ub(U8 r, U8 g, U8 b, U8 a); - void vertexBatchPreTransformed(LLVector3* verts, S32 vert_count); - void vertexBatchPreTransformed(LLVector3* verts, LLVector2* uvs, S32 vert_count); - void vertexBatchPreTransformed(LLVector3* verts, LLVector2* uvs, LLColor4U*, S32 vert_count); + void vertexBatchPreTransformed(LLVector4a* verts, S32 vert_count); + void vertexBatchPreTransformed(LLVector4a* verts, LLVector2* uvs, S32 vert_count); + void vertexBatchPreTransformed(LLVector4a* verts, LLVector2* uvs, LLColor4U*, S32 vert_count); void setColorMask(bool writeColor, bool writeAlpha); void setColorMask(bool writeColorR, bool writeColorG, bool writeColorB, bool writeAlpha); @@ -479,27 +489,32 @@ public: static bool sGLCoreProfile; static bool sNsightDebugSupport; static LLVector2 sUIGLScaleFactor; + static bool sClassicMode; // classic sky mode active private: friend class LLLightState; + LLVertexBuffer* bufferfromCache(U32 attribute_mask, U32 count); + LLVertexBuffer* genBuffer(U32 attribute_mask, S32 count); + void drawBuffer(LLVertexBuffer* vb, U32 mode, S32 count); + void resetStriders(S32 count); + eMatrixMode mMatrixMode; U32 mMatIdx[NUM_MATRIX_MODES]; U32 mMatHash[NUM_MATRIX_MODES]; - glh::matrix4f mMatrix[NUM_MATRIX_MODES][LL_MATRIX_STACK_DEPTH]; + glm::mat4 mMatrix[NUM_MATRIX_MODES][LL_MATRIX_STACK_DEPTH]; U32 mCurMatHash[NUM_MATRIX_MODES]; U32 mLightHash; LLColor4 mAmbientLightColor; bool mDirty; - U32 mQuadCycle; U32 mCount; U32 mMode; U32 mCurrTextureUnitIndex; bool mCurrColorMask[4]; LLPointer<LLVertexBuffer> mBuffer; - LLStrider<LLVector3> mVerticesp; + LLStrider<LLVector4a> mVerticesp; LLStrider<LLVector2> mTexcoordsp; LLStrider<LLColor4U> mColorsp; std::array<LLTexUnit, LL_NUM_TEXTURE_LAYERS> mTexUnits; @@ -513,7 +528,6 @@ private: std::vector<LLVector3> mUIOffset; std::vector<LLVector3> mUIScale; - }; extern F32 gGLModelView[16]; @@ -521,8 +535,8 @@ extern F32 gGLLastModelView[16]; extern F32 gGLLastProjection[16]; extern F32 gGLProjection[16]; extern S32 gGLViewport[4]; -extern F32 gGLDeltaModelView[16]; -extern F32 gGLInverseDeltaModelView[16]; +extern glm::mat4 gGLDeltaModelView; +extern glm::mat4 gGLInverseDeltaModelView; extern thread_local LLRender gGL; @@ -533,19 +547,20 @@ const F32 OGL_TO_CFR_ROTATION[16] = { 0.f, 0.f, -1.f, 0.f, // -Z becomes X 0.f, 1.f, 0.f, 0.f, // Y becomes Z 0.f, 0.f, 0.f, 1.f }; -glh::matrix4f copy_matrix(F32* src); -glh::matrix4f get_current_modelview(); -glh::matrix4f get_current_projection(); -glh::matrix4f get_last_modelview(); -glh::matrix4f get_last_projection(); +glm::mat4 copy_matrix(F32* src); +glm::mat4 get_current_modelview(); +glm::mat4 get_current_projection(); +glm::mat4 get_last_modelview(); +glm::mat4 get_last_projection(); -void copy_matrix(const glh::matrix4f& src, F32* dst); -void set_current_modelview(const glh::matrix4f& mat); -void set_current_projection(glh::matrix4f& mat); +void copy_matrix(const glm::mat4& src, F32* dst); +void set_current_modelview(const glm::mat4& mat); +void set_current_projection(const glm::mat4& mat); +void set_last_modelview(const glm::mat4& mat); +void set_last_projection(const glm::mat4& mat); -glh::matrix4f gl_ortho(GLfloat left, GLfloat right, GLfloat bottom, GLfloat top, GLfloat znear, GLfloat zfar); -glh::matrix4f gl_perspective(GLfloat fovy, GLfloat aspect, GLfloat zNear, GLfloat zFar); -glh::matrix4f gl_lookat(LLVector3 eye, LLVector3 center, LLVector3 up); +// glh compat +glm::vec3 mul_mat4_vec3(const glm::mat4& mat, const glm::vec3& vec); #define LL_SHADER_LOADING_WARNS(...) LL_WARNS() diff --git a/indra/llrender/llrender2dutils.cpp b/indra/llrender/llrender2dutils.cpp index e1742f84a1..5b08c29d64 100644 --- a/indra/llrender/llrender2dutils.cpp +++ b/indra/llrender/llrender2dutils.cpp @@ -51,11 +51,11 @@ const LLColor4 UI_VERTEX_COLOR(1.f, 1.f, 1.f, 1.f); // Functions // -BOOL ui_point_in_rect(S32 x, S32 y, S32 left, S32 top, S32 right, S32 bottom) +bool ui_point_in_rect(S32 x, S32 y, S32 left, S32 top, S32 right, S32 bottom) { - if (x < left || right < x) return FALSE; - if (y < bottom || top < y) return FALSE; - return TRUE; + if (x < left || right < x) return false; + if (y < bottom || top < y) return false; + return true; } @@ -91,13 +91,13 @@ void gl_draw_x(const LLRect& rect, const LLColor4& color) } -void gl_rect_2d_offset_local( S32 left, S32 top, S32 right, S32 bottom, const LLColor4 &color, S32 pixel_offset, BOOL filled) +void gl_rect_2d_offset_local( S32 left, S32 top, S32 right, S32 bottom, const LLColor4 &color, S32 pixel_offset, bool filled) { gGL.color4fv(color.mV); gl_rect_2d_offset_local(left, top, right, bottom, pixel_offset, filled); } -void gl_rect_2d_offset_local( S32 left, S32 top, S32 right, S32 bottom, S32 pixel_offset, BOOL filled) +void gl_rect_2d_offset_local( S32 left, S32 top, S32 right, S32 bottom, S32 pixel_offset, bool filled) { gGL.pushUIMatrix(); left += LLFontGL::sCurOrigin.mX; @@ -115,17 +115,20 @@ void gl_rect_2d_offset_local( S32 left, S32 top, S32 right, S32 bottom, S32 pixe } -void gl_rect_2d(S32 left, S32 top, S32 right, S32 bottom, BOOL filled ) +void gl_rect_2d(S32 left, S32 top, S32 right, S32 bottom, bool filled ) { gGL.getTexUnit(0)->unbind(LLTexUnit::TT_TEXTURE); // Counterclockwise quad will face the viewer if( filled ) { - gGL.begin( LLRender::QUADS ); + gGL.begin( LLRender::TRIANGLES ); gGL.vertex2i(left, top); gGL.vertex2i(left, bottom); gGL.vertex2i(right, bottom); + + gGL.vertex2i(left, top); + gGL.vertex2i(right, bottom); gGL.vertex2i(right, top); gGL.end(); } @@ -143,14 +146,14 @@ void gl_rect_2d(S32 left, S32 top, S32 right, S32 bottom, BOOL filled ) } } -void gl_rect_2d(S32 left, S32 top, S32 right, S32 bottom, const LLColor4 &color, BOOL filled ) +void gl_rect_2d(S32 left, S32 top, S32 right, S32 bottom, const LLColor4 &color, bool filled ) { gGL.color4fv( color.mV ); gl_rect_2d( left, top, right, bottom, filled ); } -void gl_rect_2d( const LLRect& rect, const LLColor4& color, BOOL filled ) +void gl_rect_2d( const LLRect& rect, const LLColor4& color, bool filled ) { gGL.color4fv( color.mV ); gl_rect_2d( rect.mLeft, rect.mTop, rect.mRight, rect.mBottom, filled ); @@ -172,50 +175,70 @@ void gl_drop_shadow(S32 left, S32 top, S32 right, S32 bottom, const LLColor4 &st LLColor4 end_color = start_color; end_color.mV[VALPHA] = 0.f; - gGL.begin(LLRender::QUADS); + gGL.begin(LLRender::TRIANGLES); // Right edge, CCW faces screen gGL.color4fv(start_color.mV); - gGL.vertex2i(right, top-lines); - gGL.vertex2i(right, bottom); + gGL.vertex2i(right, top - lines); + gGL.vertex2i(right, bottom); + gGL.color4fv(end_color.mV); + gGL.vertex2i(right + lines, bottom); + gGL.color4fv(start_color.mV); + gGL.vertex2i(right, top - lines); gGL.color4fv(end_color.mV); - gGL.vertex2i(right+lines, bottom); - gGL.vertex2i(right+lines, top-lines); + gGL.vertex2i(right + lines, bottom); + gGL.vertex2i(right + lines, top - lines); // Bottom edge, CCW faces screen gGL.color4fv(start_color.mV); - gGL.vertex2i(right, bottom); - gGL.vertex2i(left+lines, bottom); + gGL.vertex2i(right, bottom); + gGL.vertex2i(left + lines, bottom); + gGL.color4fv(end_color.mV); + gGL.vertex2i(left + lines, bottom - lines); + gGL.color4fv(start_color.mV); + gGL.vertex2i(right, bottom); gGL.color4fv(end_color.mV); - gGL.vertex2i(left+lines, bottom-lines); - gGL.vertex2i(right, bottom-lines); + gGL.vertex2i(left + lines, bottom - lines); + gGL.vertex2i(right, bottom - lines); // bottom left Corner gGL.color4fv(start_color.mV); - gGL.vertex2i(left+lines, bottom); + gGL.vertex2i(left + lines, bottom); gGL.color4fv(end_color.mV); - gGL.vertex2i(left, bottom); + gGL.vertex2i(left, bottom); // make the bottom left corner not sharp - gGL.vertex2i(left+1, bottom-lines+1); - gGL.vertex2i(left+lines, bottom-lines); + gGL.vertex2i(left + 1, bottom - lines + 1); + gGL.color4fv(start_color.mV); + gGL.vertex2i(left + lines, bottom); + gGL.color4fv(end_color.mV); + gGL.vertex2i(left + 1, bottom - lines + 1); + gGL.vertex2i(left + lines, bottom - lines); // bottom right corner gGL.color4fv(start_color.mV); - gGL.vertex2i(right, bottom); + gGL.vertex2i(right, bottom); gGL.color4fv(end_color.mV); - gGL.vertex2i(right, bottom-lines); + gGL.vertex2i(right, bottom - lines); // make the rightmost corner not sharp - gGL.vertex2i(right+lines-1, bottom-lines+1); - gGL.vertex2i(right+lines, bottom); + gGL.vertex2i(right + lines - 1, bottom - lines + 1); + gGL.color4fv(start_color.mV); + gGL.vertex2i(right, bottom); + gGL.color4fv(end_color.mV); + gGL.vertex2i(right + lines - 1, bottom - lines + 1); + gGL.vertex2i(right + lines, bottom); // top right corner gGL.color4fv(start_color.mV); - gGL.vertex2i( right, top-lines ); + gGL.vertex2i(right, top - lines); gGL.color4fv(end_color.mV); - gGL.vertex2i( right+lines, top-lines ); + gGL.vertex2i(right + lines, top - lines); // make the corner not sharp - gGL.vertex2i( right+lines-1, top-1 ); - gGL.vertex2i( right, top ); + gGL.vertex2i(right + lines - 1, top - 1); + gGL.color4fv(start_color.mV); + gGL.vertex2i(right, top - lines); + gGL.color4fv(end_color.mV); + gGL.vertex2i(right + lines - 1, top - 1); + gGL.vertex2i(right, top); gGL.end(); stop_glerror(); @@ -243,7 +266,7 @@ void gl_line_2d(S32 x1, S32 y1, S32 x2, S32 y2, const LLColor4 &color ) gGL.end(); } -void gl_triangle_2d(S32 x1, S32 y1, S32 x2, S32 y2, S32 x3, S32 y3, const LLColor4& color, BOOL filled) +void gl_triangle_2d(S32 x1, S32 y1, S32 x2, S32 y2, S32 x3, S32 y3, const LLColor4& color, bool filled) { gGL.getTexUnit(0)->unbind(LLTexUnit::TT_TEXTURE); @@ -322,7 +345,7 @@ void gl_draw_scaled_image(S32 x, S32 y, S32 width, S32 height, LLTexture* image, gl_draw_scaled_rotated_image( x, y, width, height, 0.f, image, color, uv_rect ); } -void gl_draw_scaled_image_with_border(S32 x, S32 y, S32 border_width, S32 border_height, S32 width, S32 height, LLTexture* image, const LLColor4& color, BOOL solid_color, const LLRectf& uv_rect, bool scale_inner) +void gl_draw_scaled_image_with_border(S32 x, S32 y, S32 border_width, S32 border_height, S32 width, S32 height, LLTexture* image, const LLColor4& color, bool solid_color, const LLRectf& uv_rect, bool scale_inner) { if (NULL == image) { @@ -338,7 +361,7 @@ void gl_draw_scaled_image_with_border(S32 x, S32 y, S32 border_width, S32 border gl_draw_scaled_image_with_border(x, y, width, height, image, color, solid_color, uv_rect, scale_rect, scale_inner); } -void gl_draw_scaled_image_with_border(S32 x, S32 y, S32 width, S32 height, LLTexture* image, const LLColor4& color, BOOL solid_color, const LLRectf& uv_outer_rect, const LLRectf& center_rect, bool scale_inner) +void gl_draw_scaled_image_with_border(S32 x, S32 y, S32 width, S32 height, LLTexture* image, const LLColor4& color, bool solid_color, const LLRectf& uv_outer_rect, const LLRectf& center_rect, bool scale_inner) { stop_glerror(); @@ -364,7 +387,7 @@ void gl_draw_scaled_image_with_border(S32 x, S32 y, S32 width, S32 height, LLTex { // add in offset of current image to current UI translation const LLVector3 ui_scale = gGL.getUIScale(); - const LLVector3 ui_translation = (gGL.getUITranslation() + LLVector3(x, y, 0.f)).scaledVec(ui_scale); + const LLVector3 ui_translation = (gGL.getUITranslation() + LLVector3((F32)x, (F32)y, 0.f)).scaledVec(ui_scale); F32 uv_width = uv_outer_rect.getWidth(); F32 uv_height = uv_outer_rect.getHeight(); @@ -375,8 +398,8 @@ void gl_draw_scaled_image_with_border(S32 x, S32 y, S32 width, S32 height, LLTex uv_outer_rect.mLeft + (center_rect.mRight * uv_width), uv_outer_rect.mBottom + (center_rect.mBottom * uv_height)); - F32 image_width = image->getWidth(0); - F32 image_height = image->getHeight(0); + F32 image_width = (F32)image->getWidth(0); + F32 image_height = (F32)image->getHeight(0); S32 image_natural_width = ll_round(image_width * uv_width); S32 image_natural_height = ll_round(image_height * uv_height); @@ -413,181 +436,261 @@ void gl_draw_scaled_image_with_border(S32 x, S32 y, S32 width, S32 height, LLTex draw_center_rect.setCenterAndSize(uv_center_rect.getCenterX() * width, uv_center_rect.getCenterY() * height, scaled_width, scaled_height); } - draw_center_rect.mLeft = ll_round(ui_translation.mV[VX] + (F32)draw_center_rect.mLeft * ui_scale.mV[VX]); - draw_center_rect.mTop = ll_round(ui_translation.mV[VY] + (F32)draw_center_rect.mTop * ui_scale.mV[VY]); - draw_center_rect.mRight = ll_round(ui_translation.mV[VX] + (F32)draw_center_rect.mRight * ui_scale.mV[VX]); - draw_center_rect.mBottom = ll_round(ui_translation.mV[VY] + (F32)draw_center_rect.mBottom * ui_scale.mV[VY]); + draw_center_rect.mLeft = (F32)ll_round(ui_translation.mV[VX] + (F32)draw_center_rect.mLeft * ui_scale.mV[VX]); + draw_center_rect.mTop = (F32)ll_round(ui_translation.mV[VY] + (F32)draw_center_rect.mTop * ui_scale.mV[VY]); + draw_center_rect.mRight = (F32)ll_round(ui_translation.mV[VX] + (F32)draw_center_rect.mRight * ui_scale.mV[VX]); + draw_center_rect.mBottom = (F32)ll_round(ui_translation.mV[VY] + (F32)draw_center_rect.mBottom * ui_scale.mV[VY]); LLRectf draw_outer_rect(ui_translation.mV[VX], ui_translation.mV[VY] + height * ui_scale.mV[VY], ui_translation.mV[VX] + width * ui_scale.mV[VX], ui_translation.mV[VY]); - LLGLSUIDefault gls_ui; - gGL.getTexUnit(0)->bind(image, true); gGL.color4fv(color.mV); - const S32 NUM_VERTICES = 9 * 4; // 9 quads - LLVector2 uv[NUM_VERTICES]; - LLVector3 pos[NUM_VERTICES]; + constexpr S32 NUM_VERTICES = 9 * 2 * 3; // 9 quads, 2 triangles per quad, 3 vertices per triangle + static thread_local LLVector2 uv[NUM_VERTICES]; + static thread_local LLVector4a pos[NUM_VERTICES]; S32 index = 0; - gGL.begin(LLRender::QUADS); + gGL.begin(LLRender::TRIANGLES); { - // draw bottom left - uv[index] = LLVector2(uv_outer_rect.mLeft, uv_outer_rect.mBottom); - pos[index] = LLVector3(draw_outer_rect.mLeft, draw_outer_rect.mBottom, 0.f); + // draw bottom left triangles + // 1 + uv[index].set(uv_outer_rect.mLeft, uv_outer_rect.mBottom); + pos[index].set(draw_outer_rect.mLeft, draw_outer_rect.mBottom, 0.f); index++; - uv[index] = LLVector2(uv_center_rect.mLeft, uv_outer_rect.mBottom); - pos[index] = LLVector3(draw_center_rect.mLeft, draw_outer_rect.mBottom, 0.f); + uv[index].set(uv_center_rect.mLeft, uv_outer_rect.mBottom); + pos[index].set(draw_center_rect.mLeft, draw_outer_rect.mBottom, 0.f); index++; - uv[index] = LLVector2(uv_center_rect.mLeft, uv_center_rect.mBottom); - pos[index] = LLVector3(draw_center_rect.mLeft, draw_center_rect.mBottom, 0.f); + uv[index].set(uv_center_rect.mLeft, uv_center_rect.mBottom); + pos[index].set(draw_center_rect.mLeft, draw_center_rect.mBottom, 0.f); index++; - uv[index] = LLVector2(uv_outer_rect.mLeft, uv_center_rect.mBottom); - pos[index] = LLVector3(draw_outer_rect.mLeft, draw_center_rect.mBottom, 0.f); + // 2 + uv[index].set(uv_outer_rect.mLeft, uv_outer_rect.mBottom); + pos[index].set(draw_outer_rect.mLeft, draw_outer_rect.mBottom, 0.f); index++; - // draw bottom middle - uv[index] = LLVector2(uv_center_rect.mLeft, uv_outer_rect.mBottom); - pos[index] = LLVector3(draw_center_rect.mLeft, draw_outer_rect.mBottom, 0.f); + uv[index].set(uv_center_rect.mLeft, uv_center_rect.mBottom); + pos[index].set(draw_center_rect.mLeft, draw_center_rect.mBottom, 0.f); index++; - uv[index] = LLVector2(uv_center_rect.mRight, uv_outer_rect.mBottom); - pos[index] = LLVector3(draw_center_rect.mRight, draw_outer_rect.mBottom, 0.f); + uv[index].set(uv_outer_rect.mLeft, uv_center_rect.mBottom); + pos[index].set(draw_outer_rect.mLeft, draw_center_rect.mBottom, 0.f); index++; - uv[index] = LLVector2(uv_center_rect.mRight, uv_center_rect.mBottom); - pos[index] = LLVector3(draw_center_rect.mRight, draw_center_rect.mBottom, 0.f); + // draw bottom middle triangles + uv[index].set(uv_center_rect.mLeft, uv_outer_rect.mBottom); + pos[index].set(draw_center_rect.mLeft, draw_outer_rect.mBottom, 0.f); index++; - uv[index] = LLVector2(uv_center_rect.mLeft, uv_center_rect.mBottom); - pos[index] = LLVector3(draw_center_rect.mLeft, draw_center_rect.mBottom, 0.f); + uv[index].set(uv_center_rect.mRight, uv_outer_rect.mBottom); + pos[index].set(draw_center_rect.mRight, draw_outer_rect.mBottom, 0.f); index++; - // draw bottom right - uv[index] = LLVector2(uv_center_rect.mRight, uv_outer_rect.mBottom); - pos[index] = LLVector3(draw_center_rect.mRight, draw_outer_rect.mBottom, 0.f); + uv[index].set(uv_center_rect.mRight, uv_center_rect.mBottom); + pos[index].set(draw_center_rect.mRight, draw_center_rect.mBottom, 0.f); index++; - uv[index] = LLVector2(uv_outer_rect.mRight, uv_outer_rect.mBottom); - pos[index] = LLVector3(draw_outer_rect.mRight, draw_outer_rect.mBottom, 0.f); + // 2 + uv[index].set(uv_center_rect.mLeft, uv_outer_rect.mBottom); + pos[index].set(draw_center_rect.mLeft, draw_outer_rect.mBottom, 0.f); index++; - uv[index] = LLVector2(uv_outer_rect.mRight, uv_center_rect.mBottom); - pos[index] = LLVector3(draw_outer_rect.mRight, draw_center_rect.mBottom, 0.f); + uv[index].set(uv_center_rect.mRight, uv_center_rect.mBottom); + pos[index].set(draw_center_rect.mRight, draw_center_rect.mBottom, 0.f); index++; - uv[index] = LLVector2(uv_center_rect.mRight, uv_center_rect.mBottom); - pos[index] = LLVector3(draw_center_rect.mRight, draw_center_rect.mBottom, 0.f); + uv[index].set(uv_center_rect.mLeft, uv_center_rect.mBottom); + pos[index].set(draw_center_rect.mLeft, draw_center_rect.mBottom, 0.f); index++; - // draw left - uv[index] = LLVector2(uv_outer_rect.mLeft, uv_center_rect.mBottom); - pos[index] = LLVector3(draw_outer_rect.mLeft, draw_center_rect.mBottom, 0.f); + // draw bottom right triangles + uv[index].set(uv_center_rect.mRight, uv_outer_rect.mBottom); + pos[index].set(draw_center_rect.mRight, draw_outer_rect.mBottom, 0.f); index++; - uv[index] = LLVector2(uv_center_rect.mLeft, uv_center_rect.mBottom); - pos[index] = LLVector3(draw_center_rect.mLeft, draw_center_rect.mBottom, 0.f); + uv[index].set(uv_outer_rect.mRight, uv_outer_rect.mBottom); + pos[index].set(draw_outer_rect.mRight, draw_outer_rect.mBottom, 0.f); index++; - uv[index] = LLVector2(uv_center_rect.mLeft, uv_center_rect.mTop); - pos[index] = LLVector3(draw_center_rect.mLeft, draw_center_rect.mTop, 0.f); + uv[index].set(uv_outer_rect.mRight, uv_center_rect.mBottom); + pos[index].set(draw_outer_rect.mRight, draw_center_rect.mBottom, 0.f); index++; - uv[index] = LLVector2(uv_outer_rect.mLeft, uv_center_rect.mTop); - pos[index] = LLVector3(draw_outer_rect.mLeft, draw_center_rect.mTop, 0.f); + // 2 + uv[index].set(uv_center_rect.mRight, uv_outer_rect.mBottom); + pos[index].set(draw_center_rect.mRight, draw_outer_rect.mBottom, 0.f); index++; - // draw middle - uv[index] = LLVector2(uv_center_rect.mLeft, uv_center_rect.mBottom); - pos[index] = LLVector3(draw_center_rect.mLeft, draw_center_rect.mBottom, 0.f); + uv[index].set(uv_outer_rect.mRight, uv_center_rect.mBottom); + pos[index].set(draw_outer_rect.mRight, draw_center_rect.mBottom, 0.f); index++; - uv[index] = LLVector2(uv_center_rect.mRight, uv_center_rect.mBottom); - pos[index] = LLVector3(draw_center_rect.mRight, draw_center_rect.mBottom, 0.f); + uv[index].set(uv_center_rect.mRight, uv_center_rect.mBottom); + pos[index].set(draw_center_rect.mRight, draw_center_rect.mBottom, 0.f); index++; - uv[index] = LLVector2(uv_center_rect.mRight, uv_center_rect.mTop); - pos[index] = LLVector3(draw_center_rect.mRight, draw_center_rect.mTop, 0.f); + // draw left triangles + uv[index].set(uv_outer_rect.mLeft, uv_center_rect.mBottom); + pos[index].set(draw_outer_rect.mLeft, draw_center_rect.mBottom, 0.f); index++; - uv[index] = LLVector2(uv_center_rect.mLeft, uv_center_rect.mTop); - pos[index] = LLVector3(draw_center_rect.mLeft, draw_center_rect.mTop, 0.f); + uv[index].set(uv_center_rect.mLeft, uv_center_rect.mBottom); + pos[index].set(draw_center_rect.mLeft, draw_center_rect.mBottom, 0.f); index++; - // draw right - uv[index] = LLVector2(uv_center_rect.mRight, uv_center_rect.mBottom); - pos[index] = LLVector3(draw_center_rect.mRight, draw_center_rect.mBottom, 0.f); + uv[index].set(uv_center_rect.mLeft, uv_center_rect.mTop); + pos[index].set(draw_center_rect.mLeft, draw_center_rect.mTop, 0.f); index++; - uv[index] = LLVector2(uv_outer_rect.mRight, uv_center_rect.mBottom); - pos[index] = LLVector3(draw_outer_rect.mRight, draw_center_rect.mBottom, 0.f); + // 2 + uv[index].set(uv_outer_rect.mLeft, uv_center_rect.mBottom); + pos[index].set(draw_outer_rect.mLeft, draw_center_rect.mBottom, 0.f); index++; - uv[index] = LLVector2(uv_outer_rect.mRight, uv_center_rect.mTop); - pos[index] = LLVector3(draw_outer_rect.mRight, draw_center_rect.mTop, 0.f); + uv[index].set(uv_center_rect.mLeft, uv_center_rect.mTop); + pos[index].set(draw_center_rect.mLeft, draw_center_rect.mTop, 0.f); index++; - uv[index] = LLVector2(uv_center_rect.mRight, uv_center_rect.mTop); - pos[index] = LLVector3(draw_center_rect.mRight, draw_center_rect.mTop, 0.f); + uv[index].set(uv_outer_rect.mLeft, uv_center_rect.mTop); + pos[index].set(draw_outer_rect.mLeft, draw_center_rect.mTop, 0.f); index++; - // draw top left - uv[index] = LLVector2(uv_outer_rect.mLeft, uv_center_rect.mTop); - pos[index] = LLVector3(draw_outer_rect.mLeft, draw_center_rect.mTop, 0.f); + // draw middle triangles + uv[index].set(uv_center_rect.mLeft, uv_center_rect.mBottom); + pos[index].set(draw_center_rect.mLeft, draw_center_rect.mBottom, 0.f); index++; - uv[index] = LLVector2(uv_center_rect.mLeft, uv_center_rect.mTop); - pos[index] = LLVector3(draw_center_rect.mLeft, draw_center_rect.mTop, 0.f); + uv[index].set(uv_center_rect.mRight, uv_center_rect.mBottom); + pos[index].set(draw_center_rect.mRight, draw_center_rect.mBottom, 0.f); index++; - uv[index] = LLVector2(uv_center_rect.mLeft, uv_outer_rect.mTop); - pos[index] = LLVector3(draw_center_rect.mLeft, draw_outer_rect.mTop, 0.f); + uv[index].set(uv_center_rect.mRight, uv_center_rect.mTop); + pos[index].set(draw_center_rect.mRight, draw_center_rect.mTop, 0.f); index++; - uv[index] = LLVector2(uv_outer_rect.mLeft, uv_outer_rect.mTop); - pos[index] = LLVector3(draw_outer_rect.mLeft, draw_outer_rect.mTop, 0.f); + // 2 + uv[index].set(uv_center_rect.mLeft, uv_center_rect.mBottom); + pos[index].set(draw_center_rect.mLeft, draw_center_rect.mBottom, 0.f); index++; - // draw top middle - uv[index] = LLVector2(uv_center_rect.mLeft, uv_center_rect.mTop); - pos[index] = LLVector3(draw_center_rect.mLeft, draw_center_rect.mTop, 0.f); + uv[index].set(uv_center_rect.mRight, uv_center_rect.mTop); + pos[index].set(draw_center_rect.mRight, draw_center_rect.mTop, 0.f); index++; - uv[index] = LLVector2(uv_center_rect.mRight, uv_center_rect.mTop); - pos[index] = LLVector3(draw_center_rect.mRight, draw_center_rect.mTop, 0.f); + uv[index].set(uv_center_rect.mLeft, uv_center_rect.mTop); + pos[index].set(draw_center_rect.mLeft, draw_center_rect.mTop, 0.f); index++; - uv[index] = LLVector2(uv_center_rect.mRight, uv_outer_rect.mTop); - pos[index] = LLVector3(draw_center_rect.mRight, draw_outer_rect.mTop, 0.f); + // draw right triangles + uv[index].set(uv_center_rect.mRight, uv_center_rect.mBottom); + pos[index].set(draw_center_rect.mRight, draw_center_rect.mBottom, 0.f); index++; - uv[index] = LLVector2(uv_center_rect.mLeft, uv_outer_rect.mTop); - pos[index] = LLVector3(draw_center_rect.mLeft, draw_outer_rect.mTop, 0.f); + uv[index].set(uv_outer_rect.mRight, uv_center_rect.mBottom); + pos[index].set(draw_outer_rect.mRight, draw_center_rect.mBottom, 0.f); index++; - // draw top right - uv[index] = LLVector2(uv_center_rect.mRight, uv_center_rect.mTop); - pos[index] = LLVector3(draw_center_rect.mRight, draw_center_rect.mTop, 0.f); + uv[index].set(uv_outer_rect.mRight, uv_center_rect.mTop); + pos[index].set(draw_outer_rect.mRight, draw_center_rect.mTop, 0.f); + index++; + + // 2 + uv[index].set(uv_center_rect.mRight, uv_center_rect.mBottom); + pos[index].set(draw_center_rect.mRight, draw_center_rect.mBottom, 0.f); + index++; + + uv[index].set(uv_outer_rect.mRight, uv_center_rect.mTop); + pos[index].set(draw_outer_rect.mRight, draw_center_rect.mTop, 0.f); + index++; + + uv[index].set(uv_center_rect.mRight, uv_center_rect.mTop); + pos[index].set(draw_center_rect.mRight, draw_center_rect.mTop, 0.f); + index++; + + // draw top left triangles + uv[index].set(uv_outer_rect.mLeft, uv_center_rect.mTop); + pos[index].set(draw_outer_rect.mLeft, draw_center_rect.mTop, 0.f); + index++; + + uv[index].set(uv_center_rect.mLeft, uv_center_rect.mTop); + pos[index].set(draw_center_rect.mLeft, draw_center_rect.mTop, 0.f); + index++; + + uv[index].set(uv_center_rect.mLeft, uv_outer_rect.mTop); + pos[index].set(draw_center_rect.mLeft, draw_outer_rect.mTop, 0.f); + index++; + + // 2 + uv[index].set(uv_outer_rect.mLeft, uv_center_rect.mTop); + pos[index].set(draw_outer_rect.mLeft, draw_center_rect.mTop, 0.f); + index++; + + uv[index].set(uv_center_rect.mLeft, uv_outer_rect.mTop); + pos[index].set(draw_center_rect.mLeft, draw_outer_rect.mTop, 0.f); + index++; + + uv[index].set(uv_outer_rect.mLeft, uv_outer_rect.mTop); + pos[index].set(draw_outer_rect.mLeft, draw_outer_rect.mTop, 0.f); + index++; + + // draw top middle triangles + uv[index].set(uv_center_rect.mLeft, uv_center_rect.mTop); + pos[index].set(draw_center_rect.mLeft, draw_center_rect.mTop, 0.f); + index++; + + uv[index].set(uv_center_rect.mRight, uv_center_rect.mTop); + pos[index].set(draw_center_rect.mRight, draw_center_rect.mTop, 0.f); + index++; + + uv[index].set(uv_center_rect.mRight, uv_outer_rect.mTop); + pos[index].set(draw_center_rect.mRight, draw_outer_rect.mTop, 0.f); + index++; + + // 2 + uv[index].set(uv_center_rect.mLeft, uv_center_rect.mTop); + pos[index].set(draw_center_rect.mLeft, draw_center_rect.mTop, 0.f); + index++; + + uv[index].set(uv_center_rect.mRight, uv_outer_rect.mTop); + pos[index].set(draw_center_rect.mRight, draw_outer_rect.mTop, 0.f); + index++; + + uv[index].set(uv_center_rect.mLeft, uv_outer_rect.mTop); + pos[index].set(draw_center_rect.mLeft, draw_outer_rect.mTop, 0.f); + index++; + + // draw top right triangles + uv[index].set(uv_center_rect.mRight, uv_center_rect.mTop); + pos[index].set(draw_center_rect.mRight, draw_center_rect.mTop, 0.f); + index++; + + uv[index].set(uv_outer_rect.mRight, uv_center_rect.mTop); + pos[index].set(draw_outer_rect.mRight, draw_center_rect.mTop, 0.f); + index++; + + uv[index].set(uv_outer_rect.mRight, uv_outer_rect.mTop); + pos[index].set(draw_outer_rect.mRight, draw_outer_rect.mTop, 0.f); index++; - uv[index] = LLVector2(uv_outer_rect.mRight, uv_center_rect.mTop); - pos[index] = LLVector3(draw_outer_rect.mRight, draw_center_rect.mTop, 0.f); + // 2 + uv[index].set(uv_center_rect.mRight, uv_center_rect.mTop); + pos[index].set(draw_center_rect.mRight, draw_center_rect.mTop, 0.f); index++; - uv[index] = LLVector2(uv_outer_rect.mRight, uv_outer_rect.mTop); - pos[index] = LLVector3(draw_outer_rect.mRight, draw_outer_rect.mTop, 0.f); + uv[index].set(uv_outer_rect.mRight, uv_outer_rect.mTop); + pos[index].set(draw_outer_rect.mRight, draw_outer_rect.mTop, 0.f); index++; - uv[index] = LLVector2(uv_center_rect.mRight, uv_outer_rect.mTop); - pos[index] = LLVector3(draw_center_rect.mRight, draw_outer_rect.mTop, 0.f); + uv[index].set(uv_center_rect.mRight, uv_outer_rect.mTop); + pos[index].set(draw_center_rect.mRight, draw_outer_rect.mTop, 0.f); index++; gGL.vertexBatchPreTransformed(pos, uv, NUM_VERTICES); @@ -614,8 +717,6 @@ void gl_draw_scaled_rotated_image(S32 x, S32 y, S32 width, S32 height, F32 degre return; } - LLGLSUIDefault gls_ui; - if(image != NULL) { gGL.getTexUnit(0)->bind(image, true); @@ -629,11 +730,11 @@ void gl_draw_scaled_rotated_image(S32 x, S32 y, S32 width, S32 height, F32 degre if (degrees == 0.f) { - const S32 NUM_VERTICES = 4; // 9 quads - LLVector2 uv[NUM_VERTICES]; - LLVector3 pos[NUM_VERTICES]; + constexpr S32 NUM_VERTICES = 2 * 3; + static thread_local LLVector2 uv[NUM_VERTICES +1]; + static thread_local LLVector4a pos[NUM_VERTICES +1]; - gGL.begin(LLRender::QUADS); + gGL.begin(LLRender::TRIANGLES); { LLVector3 ui_scale = gGL.getUIScale(); LLVector3 ui_translation = gGL.getUITranslation(); @@ -644,20 +745,28 @@ void gl_draw_scaled_rotated_image(S32 x, S32 y, S32 width, S32 height, F32 degre S32 scaled_width = ll_round(width * ui_scale.mV[VX]); S32 scaled_height = ll_round(height * ui_scale.mV[VY]); - uv[index] = LLVector2(uv_rect.mRight, uv_rect.mTop); - pos[index] = LLVector3(ui_translation.mV[VX] + scaled_width, ui_translation.mV[VY] + scaled_height, 0.f); + uv[index].set(uv_rect.mRight, uv_rect.mTop); + pos[index].set(ui_translation.mV[VX] + scaled_width, ui_translation.mV[VY] + scaled_height, 0.f); + index++; + + uv[index].set(uv_rect.mLeft, uv_rect.mTop); + pos[index].set(ui_translation.mV[VX], ui_translation.mV[VY] + scaled_height, 0.f); + index++; + + uv[index].set(uv_rect.mLeft, uv_rect.mBottom); + pos[index].set(ui_translation.mV[VX], ui_translation.mV[VY], 0.f); index++; - uv[index] = LLVector2(uv_rect.mLeft, uv_rect.mTop); - pos[index] = LLVector3(ui_translation.mV[VX], ui_translation.mV[VY] + scaled_height, 0.f); + uv[index].set(uv_rect.mRight, uv_rect.mTop); + pos[index].set(ui_translation.mV[VX] + scaled_width, ui_translation.mV[VY] + scaled_height, 0.f); index++; - uv[index] = LLVector2(uv_rect.mLeft, uv_rect.mBottom); - pos[index] = LLVector3(ui_translation.mV[VX], ui_translation.mV[VY], 0.f); + uv[index].set(uv_rect.mLeft, uv_rect.mBottom); + pos[index].set(ui_translation.mV[VX], ui_translation.mV[VY], 0.f); index++; - uv[index] = LLVector2(uv_rect.mRight, uv_rect.mBottom); - pos[index] = LLVector3(ui_translation.mV[VX] + scaled_width, ui_translation.mV[VY], 0.f); + uv[index].set(uv_rect.mRight, uv_rect.mBottom); + pos[index].set(ui_translation.mV[VX] + scaled_width, ui_translation.mV[VY], 0.f); index++; gGL.vertexBatchPreTransformed(pos, uv, NUM_VERTICES); @@ -687,7 +796,7 @@ void gl_draw_scaled_rotated_image(S32 x, S32 y, S32 width, S32 height, F32 degre gGL.color4fv(color.mV); - gGL.begin(LLRender::QUADS); + gGL.begin(LLRender::TRIANGLES); { LLVector3 v; @@ -703,6 +812,14 @@ void gl_draw_scaled_rotated_image(S32 x, S32 y, S32 width, S32 height, F32 degre gGL.texCoord2f(uv_rect.mLeft, uv_rect.mBottom); gGL.vertex2f(v.mV[0], v.mV[1] ); + v = LLVector3(offset_x, offset_y, 0.f) * quat; + gGL.texCoord2f(uv_rect.mRight, uv_rect.mTop); + gGL.vertex2f(v.mV[0], v.mV[1]); + + v = LLVector3(-offset_x, -offset_y, 0.f) * quat; + gGL.texCoord2f(uv_rect.mLeft, uv_rect.mBottom); + gGL.vertex2f(v.mV[0], v.mV[1]); + v = LLVector3(offset_x, -offset_y, 0.f) * quat; gGL.texCoord2f(uv_rect.mRight, uv_rect.mBottom); gGL.vertex2f(v.mV[0], v.mV[1] ); @@ -726,10 +843,10 @@ void gl_line_3d( const LLVector3& start, const LLVector3& end, const LLColor4& c } gGL.end(); - LLRender2D::getInstance()->setLineWidth(1.f); + LLRender2D::setLineWidth(1.f); } -void gl_arc_2d(F32 center_x, F32 center_y, F32 radius, S32 steps, BOOL filled, F32 start_angle, F32 end_angle) +void gl_arc_2d(F32 center_x, F32 center_y, F32 radius, S32 steps, bool filled, F32 start_angle, F32 end_angle) { if (end_angle < start_angle) { @@ -772,7 +889,7 @@ void gl_arc_2d(F32 center_x, F32 center_y, F32 radius, S32 steps, BOOL filled, F gGL.popUIMatrix(); } -void gl_circle_2d(F32 center_x, F32 center_y, F32 radius, S32 steps, BOOL filled) +void gl_circle_2d(F32 center_x, F32 center_y, F32 radius, S32 steps, bool filled) { gGL.pushUIMatrix(); { @@ -833,7 +950,7 @@ void gl_deep_circle( F32 radius, F32 depth, S32 steps ) gGL.end(); } -void gl_ring( F32 radius, F32 width, const LLColor4& center_color, const LLColor4& side_color, S32 steps, BOOL render_center ) +void gl_ring( F32 radius, F32 width, const LLColor4& center_color, const LLColor4& side_color, S32 steps, bool render_center ) { gGL.pushUIMatrix(); { @@ -948,7 +1065,7 @@ void gl_washer_segment_2d(F32 outer_radius, F32 inner_radius, F32 start_radians, void gl_rect_2d_simple_tex( S32 width, S32 height ) { - gGL.begin( LLRender::QUADS ); + gGL.begin( LLRender::TRIANGLES ); gGL.texCoord2f(1.f, 1.f); gGL.vertex2i(width, height); @@ -959,6 +1076,12 @@ void gl_rect_2d_simple_tex( S32 width, S32 height ) gGL.texCoord2f(0.f, 0.f); gGL.vertex2i(0, 0); + gGL.texCoord2f(1.f, 1.f); + gGL.vertex2i(width, height); + + gGL.texCoord2f(0.f, 0.f); + gGL.vertex2i(0, 0); + gGL.texCoord2f(1.f, 0.f); gGL.vertex2i(width, 0); @@ -967,10 +1090,13 @@ void gl_rect_2d_simple_tex( S32 width, S32 height ) void gl_rect_2d_simple( S32 width, S32 height ) { - gGL.begin( LLRender::QUADS ); + gGL.begin( LLRender::TRIANGLES ); gGL.vertex2i(width, height); gGL.vertex2i(0, height); gGL.vertex2i(0, 0); + + gGL.vertex2i(width, height); + gGL.vertex2i(0, 0); gGL.vertex2i(width, 0); gGL.end(); } @@ -1011,7 +1137,7 @@ void gl_segmented_rect_2d_tex(const S32 left, LLVector2 width_vec((F32)width, 0.f); LLVector2 height_vec(0.f, (F32)height); - gGL.begin(LLRender::QUADS); + gGL.begin(LLRender::TRIANGLES); { // draw bottom left gGL.texCoord2f(0.f, 0.f); @@ -1023,6 +1149,12 @@ void gl_segmented_rect_2d_tex(const S32 left, gGL.texCoord2f(border_uv_scale.mV[VX], border_uv_scale.mV[VY]); gGL.vertex2fv((border_width_left + border_height_bottom).mV); + gGL.texCoord2f(0.f, 0.f); + gGL.vertex2f(0.f, 0.f); + + gGL.texCoord2f(border_uv_scale.mV[VX], border_uv_scale.mV[VY]); + gGL.vertex2fv((border_width_left + border_height_bottom).mV); + gGL.texCoord2f(0.f, border_uv_scale.mV[VY]); gGL.vertex2fv(border_height_bottom.mV); @@ -1036,6 +1168,12 @@ void gl_segmented_rect_2d_tex(const S32 left, gGL.texCoord2f(1.f - border_uv_scale.mV[VX], border_uv_scale.mV[VY]); gGL.vertex2fv((width_vec - border_width_right + border_height_bottom).mV); + gGL.texCoord2f(border_uv_scale.mV[VX], 0.f); + gGL.vertex2fv(border_width_left.mV); + + gGL.texCoord2f(1.f - border_uv_scale.mV[VX], border_uv_scale.mV[VY]); + gGL.vertex2fv((width_vec - border_width_right + border_height_bottom).mV); + gGL.texCoord2f(border_uv_scale.mV[VX], border_uv_scale.mV[VY]); gGL.vertex2fv((border_width_left + border_height_bottom).mV); @@ -1049,6 +1187,12 @@ void gl_segmented_rect_2d_tex(const S32 left, gGL.texCoord2f(1.f, border_uv_scale.mV[VY]); gGL.vertex2fv((width_vec + border_height_bottom).mV); + gGL.texCoord2f(1.f - border_uv_scale.mV[VX], 0.f); + gGL.vertex2fv((width_vec - border_width_right).mV); + + gGL.texCoord2f(1.f, border_uv_scale.mV[VY]); + gGL.vertex2fv((width_vec + border_height_bottom).mV); + gGL.texCoord2f(1.f - border_uv_scale.mV[VX], border_uv_scale.mV[VY]); gGL.vertex2fv((width_vec - border_width_right + border_height_bottom).mV); @@ -1062,6 +1206,12 @@ void gl_segmented_rect_2d_tex(const S32 left, gGL.texCoord2f(border_uv_scale.mV[VX], 1.f - border_uv_scale.mV[VY]); gGL.vertex2fv((border_width_left + height_vec - border_height_top).mV); + gGL.texCoord2f(0.f, border_uv_scale.mV[VY]); + gGL.vertex2fv(border_height_bottom.mV); + + gGL.texCoord2f(border_uv_scale.mV[VX], 1.f - border_uv_scale.mV[VY]); + gGL.vertex2fv((border_width_left + height_vec - border_height_top).mV); + gGL.texCoord2f(0.f, 1.f - border_uv_scale.mV[VY]); gGL.vertex2fv((height_vec - border_height_top).mV); @@ -1075,6 +1225,12 @@ void gl_segmented_rect_2d_tex(const S32 left, gGL.texCoord2f(1.f - border_uv_scale.mV[VX], 1.f - border_uv_scale.mV[VY]); gGL.vertex2fv((width_vec - border_width_right + height_vec - border_height_top).mV); + gGL.texCoord2f(border_uv_scale.mV[VX], border_uv_scale.mV[VY]); + gGL.vertex2fv((border_width_left + border_height_bottom).mV); + + gGL.texCoord2f(1.f - border_uv_scale.mV[VX], 1.f - border_uv_scale.mV[VY]); + gGL.vertex2fv((width_vec - border_width_right + height_vec - border_height_top).mV); + gGL.texCoord2f(border_uv_scale.mV[VX], 1.f - border_uv_scale.mV[VY]); gGL.vertex2fv((border_width_left + height_vec - border_height_top).mV); @@ -1088,6 +1244,12 @@ void gl_segmented_rect_2d_tex(const S32 left, gGL.texCoord2f(1.f, 1.f - border_uv_scale.mV[VY]); gGL.vertex2fv((width_vec + height_vec - border_height_top).mV); + gGL.texCoord2f(1.f - border_uv_scale.mV[VX], border_uv_scale.mV[VY]); + gGL.vertex2fv((width_vec - border_width_right + border_height_bottom).mV); + + gGL.texCoord2f(1.f, 1.f - border_uv_scale.mV[VY]); + gGL.vertex2fv((width_vec + height_vec - border_height_top).mV); + gGL.texCoord2f(1.f - border_uv_scale.mV[VX], 1.f - border_uv_scale.mV[VY]); gGL.vertex2fv((width_vec - border_width_right + height_vec - border_height_top).mV); @@ -1101,6 +1263,12 @@ void gl_segmented_rect_2d_tex(const S32 left, gGL.texCoord2f(border_uv_scale.mV[VX], 1.f); gGL.vertex2fv((border_width_left + height_vec).mV); + gGL.texCoord2f(0.f, 1.f - border_uv_scale.mV[VY]); + gGL.vertex2fv((height_vec - border_height_top).mV); + + gGL.texCoord2f(border_uv_scale.mV[VX], 1.f); + gGL.vertex2fv((border_width_left + height_vec).mV); + gGL.texCoord2f(0.f, 1.f); gGL.vertex2fv((height_vec).mV); @@ -1114,6 +1282,12 @@ void gl_segmented_rect_2d_tex(const S32 left, gGL.texCoord2f(1.f - border_uv_scale.mV[VX], 1.f); gGL.vertex2fv((width_vec - border_width_right + height_vec).mV); + gGL.texCoord2f(border_uv_scale.mV[VX], 1.f - border_uv_scale.mV[VY]); + gGL.vertex2fv((border_width_left + height_vec - border_height_top).mV); + + gGL.texCoord2f(1.f - border_uv_scale.mV[VX], 1.f); + gGL.vertex2fv((width_vec - border_width_right + height_vec).mV); + gGL.texCoord2f(border_uv_scale.mV[VX], 1.f); gGL.vertex2fv((border_width_left + height_vec).mV); @@ -1127,6 +1301,12 @@ void gl_segmented_rect_2d_tex(const S32 left, gGL.texCoord2f(1.f, 1.f); gGL.vertex2fv((width_vec + height_vec).mV); + gGL.texCoord2f(1.f - border_uv_scale.mV[VX], 1.f - border_uv_scale.mV[VY]); + gGL.vertex2fv((width_vec - border_width_right + height_vec - border_height_top).mV); + + gGL.texCoord2f(1.f, 1.f); + gGL.vertex2fv((width_vec + height_vec).mV); + gGL.texCoord2f(1.f - border_uv_scale.mV[VX], 1.f); gGL.vertex2fv((width_vec - border_width_right + height_vec).mV); } @@ -1181,7 +1361,7 @@ void gl_segmented_rect_2d_fragment_tex(const LLRect& rect, LLVector2 x_min; LLVector2 x_max; - gGL.begin(LLRender::QUADS); + gGL.begin(LLRender::TRIANGLES); { if (start_fragment < middle_start) { @@ -1200,6 +1380,12 @@ void gl_segmented_rect_2d_fragment_tex(const LLRect& rect, gGL.texCoord2f(u_max, border_uv_scale.mV[VY]); gGL.vertex2fv((x_max + border_height_bottom).mV); + gGL.texCoord2f(u_min, 0.f); + gGL.vertex2fv(x_min.mV); + + gGL.texCoord2f(u_max, border_uv_scale.mV[VY]); + gGL.vertex2fv((x_max + border_height_bottom).mV); + gGL.texCoord2f(u_min, border_uv_scale.mV[VY]); gGL.vertex2fv((x_min + border_height_bottom).mV); @@ -1213,6 +1399,12 @@ void gl_segmented_rect_2d_fragment_tex(const LLRect& rect, gGL.texCoord2f(u_max, 1.f - border_uv_scale.mV[VY]); gGL.vertex2fv((x_max + height_vec - border_height_top).mV); + gGL.texCoord2f(u_min, border_uv_scale.mV[VY]); + gGL.vertex2fv((x_min + border_height_bottom).mV); + + gGL.texCoord2f(u_max, 1.f - border_uv_scale.mV[VY]); + gGL.vertex2fv((x_max + height_vec - border_height_top).mV); + gGL.texCoord2f(u_min, 1.f - border_uv_scale.mV[VY]); gGL.vertex2fv((x_min + height_vec - border_height_top).mV); @@ -1226,6 +1418,12 @@ void gl_segmented_rect_2d_fragment_tex(const LLRect& rect, gGL.texCoord2f(u_max, 1.f); gGL.vertex2fv((x_max + height_vec).mV); + gGL.texCoord2f(u_min, 1.f - border_uv_scale.mV[VY]); + gGL.vertex2fv((x_min + height_vec - border_height_top).mV); + + gGL.texCoord2f(u_max, 1.f); + gGL.vertex2fv((x_max + height_vec).mV); + gGL.texCoord2f(u_min, 1.f); gGL.vertex2fv((x_min + height_vec).mV); } @@ -1245,6 +1443,12 @@ void gl_segmented_rect_2d_fragment_tex(const LLRect& rect, gGL.texCoord2f(1.f - border_uv_scale.mV[VX], border_uv_scale.mV[VY]); gGL.vertex2fv((x_max + border_height_bottom).mV); + gGL.texCoord2f(border_uv_scale.mV[VX], 0.f); + gGL.vertex2fv(x_min.mV); + + gGL.texCoord2f(1.f - border_uv_scale.mV[VX], border_uv_scale.mV[VY]); + gGL.vertex2fv((x_max + border_height_bottom).mV); + gGL.texCoord2f(border_uv_scale.mV[VX], border_uv_scale.mV[VY]); gGL.vertex2fv((x_min + border_height_bottom).mV); @@ -1258,6 +1462,12 @@ void gl_segmented_rect_2d_fragment_tex(const LLRect& rect, gGL.texCoord2f(1.f - border_uv_scale.mV[VX], 1.f - border_uv_scale.mV[VY]); gGL.vertex2fv((x_max + height_vec - border_height_top).mV); + gGL.texCoord2f(border_uv_scale.mV[VX], border_uv_scale.mV[VY]); + gGL.vertex2fv((x_min + border_height_bottom).mV); + + gGL.texCoord2f(1.f - border_uv_scale.mV[VX], 1.f - border_uv_scale.mV[VY]); + gGL.vertex2fv((x_max + height_vec - border_height_top).mV); + gGL.texCoord2f(border_uv_scale.mV[VX], 1.f - border_uv_scale.mV[VY]); gGL.vertex2fv((x_min + height_vec - border_height_top).mV); @@ -1271,6 +1481,12 @@ void gl_segmented_rect_2d_fragment_tex(const LLRect& rect, gGL.texCoord2f(1.f - border_uv_scale.mV[VX], 1.f); gGL.vertex2fv((x_max + height_vec).mV); + gGL.texCoord2f(border_uv_scale.mV[VX], 1.f - border_uv_scale.mV[VY]); + gGL.vertex2fv((x_min + height_vec - border_height_top).mV); + + gGL.texCoord2f(1.f - border_uv_scale.mV[VX], 1.f); + gGL.vertex2fv((x_max + height_vec).mV); + gGL.texCoord2f(border_uv_scale.mV[VX], 1.f); gGL.vertex2fv((x_min + height_vec).mV); } @@ -1292,6 +1508,12 @@ void gl_segmented_rect_2d_fragment_tex(const LLRect& rect, gGL.texCoord2f(u_max, border_uv_scale.mV[VY]); gGL.vertex2fv((x_max + border_height_bottom).mV); + gGL.texCoord2f(u_min, 0.f); + gGL.vertex2fv((x_min).mV); + + gGL.texCoord2f(u_max, border_uv_scale.mV[VY]); + gGL.vertex2fv((x_max + border_height_bottom).mV); + gGL.texCoord2f(u_min, border_uv_scale.mV[VY]); gGL.vertex2fv((x_min + border_height_bottom).mV); @@ -1305,6 +1527,12 @@ void gl_segmented_rect_2d_fragment_tex(const LLRect& rect, gGL.texCoord2f(u_max, 1.f - border_uv_scale.mV[VY]); gGL.vertex2fv((x_max + height_vec - border_height_top).mV); + gGL.texCoord2f(u_min, border_uv_scale.mV[VY]); + gGL.vertex2fv((x_min + border_height_bottom).mV); + + gGL.texCoord2f(u_max, 1.f - border_uv_scale.mV[VY]); + gGL.vertex2fv((x_max + height_vec - border_height_top).mV); + gGL.texCoord2f(u_min, 1.f - border_uv_scale.mV[VY]); gGL.vertex2fv((x_min + height_vec - border_height_top).mV); @@ -1318,6 +1546,12 @@ void gl_segmented_rect_2d_fragment_tex(const LLRect& rect, gGL.texCoord2f(u_max, 1.f); gGL.vertex2fv((x_max + height_vec).mV); + gGL.texCoord2f(u_min, 1.f - border_uv_scale.mV[VY]); + gGL.vertex2fv((x_min + height_vec - border_height_top).mV); + + gGL.texCoord2f(u_max, 1.f); + gGL.vertex2fv((x_max + height_vec).mV); + gGL.texCoord2f(u_min, 1.f); gGL.vertex2fv((x_min + height_vec).mV); } @@ -1332,7 +1566,7 @@ void gl_segmented_rect_3d_tex(const LLRectf& clip_rect, const LLRectf& center_uv { LL_PROFILE_ZONE_SCOPED_CATEGORY_UI; - gGL.begin(LLRender::QUADS); + gGL.begin(LLRender::TRIANGLES); { // draw bottom left gGL.texCoord2f(clip_rect.mLeft, clip_rect.mBottom); @@ -1344,6 +1578,12 @@ void gl_segmented_rect_3d_tex(const LLRectf& clip_rect, const LLRectf& center_uv gGL.texCoord2f(center_uv_rect.mLeft, center_uv_rect.mBottom); gGL.vertex3fv((center_draw_rect.mLeft * width_vec + center_draw_rect.mBottom * height_vec).mV); + gGL.texCoord2f(clip_rect.mLeft, clip_rect.mBottom); + gGL.vertex3f(0.f, 0.f, 0.f); + + gGL.texCoord2f(center_uv_rect.mLeft, center_uv_rect.mBottom); + gGL.vertex3fv((center_draw_rect.mLeft * width_vec + center_draw_rect.mBottom * height_vec).mV); + gGL.texCoord2f(clip_rect.mLeft, center_uv_rect.mBottom); gGL.vertex3fv((center_draw_rect.mBottom * height_vec).mV); @@ -1357,6 +1597,12 @@ void gl_segmented_rect_3d_tex(const LLRectf& clip_rect, const LLRectf& center_uv gGL.texCoord2f(center_uv_rect.mRight, center_uv_rect.mBottom); gGL.vertex3fv((center_draw_rect.mRight * width_vec + center_draw_rect.mBottom * height_vec).mV); + gGL.texCoord2f(center_uv_rect.mLeft, clip_rect.mBottom); + gGL.vertex3fv((center_draw_rect.mLeft * width_vec).mV); + + gGL.texCoord2f(center_uv_rect.mRight, center_uv_rect.mBottom); + gGL.vertex3fv((center_draw_rect.mRight * width_vec + center_draw_rect.mBottom * height_vec).mV); + gGL.texCoord2f(center_uv_rect.mLeft, center_uv_rect.mBottom); gGL.vertex3fv((center_draw_rect.mLeft * width_vec + center_draw_rect.mBottom * height_vec).mV); @@ -1370,6 +1616,12 @@ void gl_segmented_rect_3d_tex(const LLRectf& clip_rect, const LLRectf& center_uv gGL.texCoord2f(clip_rect.mRight, center_uv_rect.mBottom); gGL.vertex3fv((width_vec + center_draw_rect.mBottom * height_vec).mV); + gGL.texCoord2f(center_uv_rect.mRight, clip_rect.mBottom); + gGL.vertex3fv((center_draw_rect.mRight * width_vec).mV); + + gGL.texCoord2f(clip_rect.mRight, center_uv_rect.mBottom); + gGL.vertex3fv((width_vec + center_draw_rect.mBottom * height_vec).mV); + gGL.texCoord2f(center_uv_rect.mRight, center_uv_rect.mBottom); gGL.vertex3fv((center_draw_rect.mRight * width_vec + center_draw_rect.mBottom * height_vec).mV); @@ -1383,6 +1635,12 @@ void gl_segmented_rect_3d_tex(const LLRectf& clip_rect, const LLRectf& center_uv gGL.texCoord2f(center_uv_rect.mLeft, center_uv_rect.mTop); gGL.vertex3fv((center_draw_rect.mLeft * width_vec + center_draw_rect.mTop * height_vec).mV); + gGL.texCoord2f(clip_rect.mLeft, center_uv_rect.mBottom); + gGL.vertex3fv((center_draw_rect.mBottom * height_vec).mV); + + gGL.texCoord2f(center_uv_rect.mLeft, center_uv_rect.mTop); + gGL.vertex3fv((center_draw_rect.mLeft * width_vec + center_draw_rect.mTop * height_vec).mV); + gGL.texCoord2f(clip_rect.mLeft, center_uv_rect.mTop); gGL.vertex3fv((center_draw_rect.mTop * height_vec).mV); @@ -1396,6 +1654,12 @@ void gl_segmented_rect_3d_tex(const LLRectf& clip_rect, const LLRectf& center_uv gGL.texCoord2f(center_uv_rect.mRight, center_uv_rect.mTop); gGL.vertex3fv((center_draw_rect.mRight * width_vec + center_draw_rect.mTop * height_vec).mV); + gGL.texCoord2f(center_uv_rect.mLeft, center_uv_rect.mBottom); + gGL.vertex3fv((center_draw_rect.mLeft * width_vec + center_draw_rect.mBottom * height_vec).mV); + + gGL.texCoord2f(center_uv_rect.mRight, center_uv_rect.mTop); + gGL.vertex3fv((center_draw_rect.mRight * width_vec + center_draw_rect.mTop * height_vec).mV); + gGL.texCoord2f(center_uv_rect.mLeft, center_uv_rect.mTop); gGL.vertex3fv((center_draw_rect.mLeft * width_vec + center_draw_rect.mTop * height_vec).mV); @@ -1409,6 +1673,12 @@ void gl_segmented_rect_3d_tex(const LLRectf& clip_rect, const LLRectf& center_uv gGL.texCoord2f(clip_rect.mRight, center_uv_rect.mTop); gGL.vertex3fv((width_vec + center_draw_rect.mTop * height_vec).mV); + gGL.texCoord2f(center_uv_rect.mRight, center_uv_rect.mBottom); + gGL.vertex3fv((center_draw_rect.mRight* width_vec + center_draw_rect.mBottom * height_vec).mV); + + gGL.texCoord2f(clip_rect.mRight, center_uv_rect.mTop); + gGL.vertex3fv((width_vec + center_draw_rect.mTop * height_vec).mV); + gGL.texCoord2f(center_uv_rect.mRight, center_uv_rect.mTop); gGL.vertex3fv((center_draw_rect.mRight * width_vec + center_draw_rect.mTop * height_vec).mV); @@ -1422,6 +1692,12 @@ void gl_segmented_rect_3d_tex(const LLRectf& clip_rect, const LLRectf& center_uv gGL.texCoord2f(center_uv_rect.mLeft, clip_rect.mTop); gGL.vertex3fv((center_draw_rect.mLeft * width_vec + height_vec).mV); + gGL.texCoord2f(clip_rect.mLeft, center_uv_rect.mTop); + gGL.vertex3fv((center_draw_rect.mTop * height_vec).mV); + + gGL.texCoord2f(center_uv_rect.mLeft, clip_rect.mTop); + gGL.vertex3fv((center_draw_rect.mLeft * width_vec + height_vec).mV); + gGL.texCoord2f(clip_rect.mLeft, clip_rect.mTop); gGL.vertex3fv((height_vec).mV); @@ -1435,6 +1711,12 @@ void gl_segmented_rect_3d_tex(const LLRectf& clip_rect, const LLRectf& center_uv gGL.texCoord2f(center_uv_rect.mRight, clip_rect.mTop); gGL.vertex3fv((center_draw_rect.mRight * width_vec + height_vec).mV); + gGL.texCoord2f(center_uv_rect.mLeft, center_uv_rect.mTop); + gGL.vertex3fv((center_draw_rect.mLeft * width_vec + center_draw_rect.mTop * height_vec).mV); + + gGL.texCoord2f(center_uv_rect.mRight, clip_rect.mTop); + gGL.vertex3fv((center_draw_rect.mRight* width_vec + height_vec).mV); + gGL.texCoord2f(center_uv_rect.mLeft, clip_rect.mTop); gGL.vertex3fv((center_draw_rect.mLeft * width_vec + height_vec).mV); @@ -1448,6 +1730,12 @@ void gl_segmented_rect_3d_tex(const LLRectf& clip_rect, const LLRectf& center_uv gGL.texCoord2f(clip_rect.mRight, clip_rect.mTop); gGL.vertex3fv((width_vec + height_vec).mV); + gGL.texCoord2f(center_uv_rect.mRight, center_uv_rect.mTop); + gGL.vertex3fv((center_draw_rect.mRight* width_vec + center_draw_rect.mTop * height_vec).mV); + + gGL.texCoord2f(clip_rect.mRight, clip_rect.mTop); + gGL.vertex3fv((width_vec + height_vec).mV); + gGL.texCoord2f(center_uv_rect.mRight, clip_rect.mTop); gGL.vertex3fv((center_draw_rect.mRight * width_vec + height_vec).mV); } diff --git a/indra/llrender/llrender2dutils.h b/indra/llrender/llrender2dutils.h index c3c6d66b8e..096e7584f1 100644 --- a/indra/llrender/llrender2dutils.h +++ b/indra/llrender/llrender2dutils.h @@ -43,30 +43,30 @@ class LLUUID; extern const LLColor4 UI_VERTEX_COLOR; -BOOL ui_point_in_rect(S32 x, S32 y, S32 left, S32 top, S32 right, S32 bottom); +bool ui_point_in_rect(S32 x, S32 y, S32 left, S32 top, S32 right, S32 bottom); void gl_state_for_2d(S32 width, S32 height); void gl_line_2d(S32 x1, S32 y1, S32 x2, S32 y2); void gl_line_2d(S32 x1, S32 y1, S32 x2, S32 y2, const LLColor4 &color ); -void gl_triangle_2d(S32 x1, S32 y1, S32 x2, S32 y2, S32 x3, S32 y3, const LLColor4& color, BOOL filled); +void gl_triangle_2d(S32 x1, S32 y1, S32 x2, S32 y2, S32 x3, S32 y3, const LLColor4& color, bool filled); void gl_rect_2d_simple( S32 width, S32 height ); void gl_draw_x(const LLRect& rect, const LLColor4& color); -void gl_rect_2d(S32 left, S32 top, S32 right, S32 bottom, BOOL filled = TRUE ); -void gl_rect_2d(S32 left, S32 top, S32 right, S32 bottom, const LLColor4 &color, BOOL filled = TRUE ); -void gl_rect_2d_offset_local( S32 left, S32 top, S32 right, S32 bottom, const LLColor4 &color, S32 pixel_offset = 0, BOOL filled = TRUE ); -void gl_rect_2d_offset_local( S32 left, S32 top, S32 right, S32 bottom, S32 pixel_offset = 0, BOOL filled = TRUE ); -void gl_rect_2d(const LLRect& rect, BOOL filled = TRUE ); -void gl_rect_2d(const LLRect& rect, const LLColor4& color, BOOL filled = TRUE ); +void gl_rect_2d(S32 left, S32 top, S32 right, S32 bottom, bool filled = true ); +void gl_rect_2d(S32 left, S32 top, S32 right, S32 bottom, const LLColor4 &color, bool filled = true ); +void gl_rect_2d_offset_local( S32 left, S32 top, S32 right, S32 bottom, const LLColor4 &color, S32 pixel_offset = 0, bool filled = true ); +void gl_rect_2d_offset_local( S32 left, S32 top, S32 right, S32 bottom, S32 pixel_offset = 0, bool filled = true ); +void gl_rect_2d(const LLRect& rect, bool filled = true ); +void gl_rect_2d(const LLRect& rect, const LLColor4& color, bool filled = true ); void gl_rect_2d_checkerboard(const LLRect& rect, GLfloat alpha = 1.0f); void gl_drop_shadow(S32 left, S32 top, S32 right, S32 bottom, const LLColor4 &start_color, S32 lines); -void gl_circle_2d(F32 x, F32 y, F32 radius, S32 steps, BOOL filled); -void gl_arc_2d(F32 center_x, F32 center_y, F32 radius, S32 steps, BOOL filled, F32 start_angle, F32 end_angle); +void gl_circle_2d(F32 x, F32 y, F32 radius, S32 steps, bool filled); +void gl_arc_2d(F32 center_x, F32 center_y, F32 radius, S32 steps, bool filled, F32 start_angle, F32 end_angle); void gl_deep_circle( F32 radius, F32 depth ); -void gl_ring( F32 radius, F32 width, const LLColor4& center_color, const LLColor4& side_color, S32 steps, BOOL render_center ); +void gl_ring( F32 radius, F32 width, const LLColor4& center_color, const LLColor4& side_color, S32 steps, bool render_center ); void gl_corners_2d(S32 left, S32 top, S32 right, S32 bottom, S32 length, F32 max_frac); void gl_washer_2d(F32 outer_radius, F32 inner_radius, S32 steps, const LLColor4& inner_color, const LLColor4& outer_color); void gl_washer_segment_2d(F32 outer_radius, F32 inner_radius, F32 start_radians, F32 end_radians, S32 steps, const LLColor4& inner_color, const LLColor4& outer_color); @@ -76,8 +76,8 @@ void gl_draw_scaled_target(S32 x, S32 y, S32 width, S32 height, LLRenderTarget* void gl_draw_scaled_image(S32 x, S32 y, S32 width, S32 height, LLTexture* image, const LLColor4& color = UI_VERTEX_COLOR, const LLRectf& uv_rect = LLRectf(0.f, 1.f, 1.f, 0.f)); void gl_draw_rotated_image(S32 x, S32 y, F32 degrees, LLTexture* image, const LLColor4& color = UI_VERTEX_COLOR, const LLRectf& uv_rect = LLRectf(0.f, 1.f, 1.f, 0.f)); void gl_draw_scaled_rotated_image(S32 x, S32 y, S32 width, S32 height, F32 degrees, LLTexture* image, const LLColor4& color = UI_VERTEX_COLOR, const LLRectf& uv_rect = LLRectf(0.f, 1.f, 1.f, 0.f), LLRenderTarget* target = NULL); -void gl_draw_scaled_image_with_border(S32 x, S32 y, S32 border_width, S32 border_height, S32 width, S32 height, LLTexture* image, const LLColor4 &color, BOOL solid_color = FALSE, const LLRectf& uv_rect = LLRectf(0.f, 1.f, 1.f, 0.f), bool scale_inner = true); -void gl_draw_scaled_image_with_border(S32 x, S32 y, S32 width, S32 height, LLTexture* image, const LLColor4 &color, BOOL solid_color = FALSE, const LLRectf& uv_rect = LLRectf(0.f, 1.f, 1.f, 0.f), const LLRectf& scale_rect = LLRectf(0.f, 1.f, 1.f, 0.f), bool scale_inner = true); +void gl_draw_scaled_image_with_border(S32 x, S32 y, S32 border_width, S32 border_height, S32 width, S32 height, LLTexture* image, const LLColor4 &color, bool solid_color = false, const LLRectf& uv_rect = LLRectf(0.f, 1.f, 1.f, 0.f), bool scale_inner = true); +void gl_draw_scaled_image_with_border(S32 x, S32 y, S32 width, S32 height, LLTexture* image, const LLColor4 &color, bool solid_color = false, const LLRectf& uv_rect = LLRectf(0.f, 1.f, 1.f, 0.f), const LLRectf& scale_rect = LLRectf(0.f, 1.f, 1.f, 0.f), bool scale_inner = true); void gl_line_3d( const LLVector3& start, const LLVector3& end, const LLColor4& color); @@ -110,24 +110,25 @@ void gl_segmented_rect_2d_tex(const S32 left, const S32 top, const S32 right, co void gl_segmented_rect_2d_fragment_tex(const LLRect& rect, const S32 texture_width, const S32 texture_height, const S32 border_size, const F32 start_fragment, const F32 end_fragment, const U32 edges = ROUNDED_RECT_ALL); void gl_segmented_rect_3d_tex(const LLRectf& clip_rect, const LLRectf& center_uv_rect, const LLRectf& center_draw_rect, const LLVector3& width_vec, const LLVector3& height_vec); -inline void gl_rect_2d( const LLRect& rect, BOOL filled ) +inline void gl_rect_2d( const LLRect& rect, bool filled ) { gl_rect_2d( rect.mLeft, rect.mTop, rect.mRight, rect.mBottom, filled ); } -inline void gl_rect_2d_offset_local( const LLRect& rect, S32 pixel_offset, BOOL filled) +inline void gl_rect_2d_offset_local( const LLRect& rect, S32 pixel_offset, bool filled) { gl_rect_2d_offset_local( rect.mLeft, rect.mTop, rect.mRight, rect.mBottom, pixel_offset, filled ); } class LLImageProviderInterface; -class LLRender2D : public LLParamSingleton<LLRender2D> +class LLRender2D : public LLSimpleton<LLRender2D> { - LLSINGLETON(LLRender2D, LLImageProviderInterface* image_provider); LOG_CLASS(LLRender2D); - ~LLRender2D(); public: + LLRender2D(LLImageProviderInterface* image_provider); + ~LLRender2D(); + static void pushMatrix(); static void popMatrix(); static void loadIdentity(); diff --git a/indra/llrender/llrendersphere.cpp b/indra/llrender/llrendersphere.cpp index 9570180554..cd8ef7d68e 100644 --- a/indra/llrender/llrendersphere.cpp +++ b/indra/llrender/llrendersphere.cpp @@ -34,6 +34,8 @@ #include "llerror.h" #include "llglheaders.h" +#include "llvertexbuffer.h" +#include "llglslshader.h" LLRenderSphere gSphere; @@ -53,12 +55,20 @@ inline LLVector3 polar_to_cart(F32 latitude, F32 longitude) void LLRenderSphere::renderGGL() { + LL_PROFILE_ZONE_SCOPED; S32 const LATITUDE_SLICES = 20; S32 const LONGITUDE_SLICES = 30; - if (mSpherePoints.empty()) + if (mVertexBuffer.isNull()) { mSpherePoints.resize(LATITUDE_SLICES + 1); + mVertexBuffer = new LLVertexBuffer(LLVertexBuffer::MAP_VERTEX); + + mVertexBuffer->allocateBuffer((U32)(LATITUDE_SLICES + 1) * (LONGITUDE_SLICES + 1), LATITUDE_SLICES * LONGITUDE_SLICES * 6); + + LLStrider<LLVector3> v; + mVertexBuffer->getVertexStrider(v); + for (S32 lat_i = 0; lat_i < LATITUDE_SLICES + 1; lat_i++) { mSpherePoints[lat_i].resize(LONGITUDE_SLICES + 1); @@ -68,24 +78,52 @@ void LLRenderSphere::renderGGL() F32 lon = (F32)lon_i / LONGITUDE_SLICES; mSpherePoints[lat_i][lon_i] = polar_to_cart(lat, lon); + v[lat_i * (LONGITUDE_SLICES + 1) + lon_i] = mSpherePoints[lat_i][lon_i]; } } + + LLStrider<U16> i; + mVertexBuffer->getIndexStrider(i); + + for (S32 lat_i = 0; lat_i < LATITUDE_SLICES; lat_i++) + { + for (S32 lon_i = 0; lon_i < LONGITUDE_SLICES; lon_i++) + { + i[(lat_i * LONGITUDE_SLICES + lon_i) * 6 + 0] = lat_i * (LONGITUDE_SLICES + 1) + lon_i; + i[(lat_i * LONGITUDE_SLICES + lon_i) * 6 + 1] = lat_i * (LONGITUDE_SLICES + 1) + lon_i + 1; + i[(lat_i * LONGITUDE_SLICES + lon_i) * 6 + 2] = (lat_i + 1) * (LONGITUDE_SLICES + 1) + lon_i; + + i[(lat_i * LONGITUDE_SLICES + lon_i) * 6 + 3] = (lat_i + 1) * (LONGITUDE_SLICES + 1) + lon_i; + i[(lat_i * LONGITUDE_SLICES + lon_i) * 6 + 4] = lat_i * (LONGITUDE_SLICES + 1) + lon_i + 1; + i[(lat_i * LONGITUDE_SLICES + lon_i) * 6 + 5] = (lat_i + 1) * (LONGITUDE_SLICES + 1) + lon_i + 1; + } + } + + mVertexBuffer->unmapBuffer(); } - gGL.begin(LLRender::TRIANGLES); - for (S32 lat_i = 0; lat_i < LATITUDE_SLICES; lat_i++) - { - for (S32 lon_i = 0; lon_i < LONGITUDE_SLICES; lon_i++) + if (LLGLSLShader::sCurBoundShaderPtr->mAttributeMask == LLVertexBuffer::MAP_VERTEX) + { // shader expects only vertex positions in vertex buffer, use fast path + mVertexBuffer->setBuffer(); + mVertexBuffer->drawRange(LLRender::TRIANGLES, 0, mVertexBuffer->getNumVerts(), mVertexBuffer->getNumIndices(), 0); + } + else + { //shader wants colors in the vertex stream, use slow path + gGL.begin(LLRender::TRIANGLES); + for (S32 lat_i = 0; lat_i < LATITUDE_SLICES; lat_i++) { - gGL.vertex3fv(mSpherePoints[lat_i][lon_i].mV); - gGL.vertex3fv(mSpherePoints[lat_i][lon_i+1].mV); - gGL.vertex3fv(mSpherePoints[lat_i+1][lon_i].mV); + for (S32 lon_i = 0; lon_i < LONGITUDE_SLICES; lon_i++) + { + gGL.vertex3fv(mSpherePoints[lat_i][lon_i].mV); + gGL.vertex3fv(mSpherePoints[lat_i][lon_i + 1].mV); + gGL.vertex3fv(mSpherePoints[lat_i + 1][lon_i].mV); - gGL.vertex3fv(mSpherePoints[lat_i+1][lon_i].mV); - gGL.vertex3fv(mSpherePoints[lat_i][lon_i+1].mV); - gGL.vertex3fv(mSpherePoints[lat_i+1][lon_i+1].mV); + gGL.vertex3fv(mSpherePoints[lat_i + 1][lon_i].mV); + gGL.vertex3fv(mSpherePoints[lat_i][lon_i + 1].mV); + gGL.vertex3fv(mSpherePoints[lat_i + 1][lon_i + 1].mV); + } } + gGL.end(); } - gGL.end(); } diff --git a/indra/llrender/llrendersphere.h b/indra/llrender/llrendersphere.h index e2e886fa06..5b6eabecb8 100644 --- a/indra/llrender/llrendersphere.h +++ b/indra/llrender/llrendersphere.h @@ -45,6 +45,7 @@ public: private: std::vector< std::vector<LLVector3> > mSpherePoints; + LLPointer<LLVertexBuffer> mVertexBuffer; }; extern LLRenderSphere gSphere; diff --git a/indra/llrender/llrendertarget.cpp b/indra/llrender/llrendertarget.cpp index 9cd7527d3e..38bc5ff331 100644 --- a/indra/llrender/llrendertarget.cpp +++ b/indra/llrender/llrendertarget.cpp @@ -1,25 +1,25 @@ -/** +/** * @file llrendertarget.cpp * @brief LLRenderTarget implementation * * $LicenseInfo:firstyear=2001&license=viewerlgpl$ * Second Life Viewer Source Code * Copyright (C) 2010, 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 * $/LicenseInfo$ */ @@ -44,7 +44,7 @@ void check_framebuffer_status() break; default: LL_WARNS() << "check_framebuffer_status failed -- " << std::hex << status << LL_ENDL; - ll_fail("check_framebuffer_status failed"); + ll_fail("check_framebuffer_status failed"); break; } } @@ -75,10 +75,10 @@ LLRenderTarget::~LLRenderTarget() } void LLRenderTarget::resize(U32 resx, U32 resy) -{ +{ //for accounting, get the number of pixels added/subtracted S32 pix_diff = (resx*resy)-(mResX*mResY); - + mResX = resx; mResY = resy; @@ -92,7 +92,7 @@ void LLRenderTarget::resize(U32 resx, U32 resy) } if (mDepth) - { + { 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); @@ -100,7 +100,7 @@ void LLRenderTarget::resize(U32 resx, U32 resy) sBytesAllocated += pix_diff*4; } } - + bool LLRenderTarget::allocate(U32 resx, U32 resy, U32 color_fmt, bool depth, LLTexUnit::eTextureType usage, LLTexUnit::eTextureMipGeneration generateMipMaps) { @@ -112,7 +112,7 @@ bool LLRenderTarget::allocate(U32 resx, U32 resy, U32 color_fmt, bool depth, LLT resy = llmin(resy, (U32) gGLManager.mGLMaxTextureSize); release(); - + mResX = resx; mResY = resy; @@ -123,9 +123,9 @@ bool LLRenderTarget::allocate(U32 resx, U32 resy, U32 color_fmt, bool depth, LLT if (mGenerateMipMaps != LLTexUnit::TMG_NONE) { // Calculate the number of mip levels based upon resolution that we should have. - mMipLevels = 1 + floor(log10((float)llmax(mResX, mResY))/log10(2.0)); + mMipLevels = 1 + (U32)floor(log10((float)llmax(mResX, mResY)) / log10(2.0)); } - + if (depth) { if (!allocateDepth()) @@ -140,12 +140,12 @@ bool LLRenderTarget::allocate(U32 resx, U32 resy, U32 color_fmt, bool depth, LLT if (mDepth) { glBindFramebuffer(GL_FRAMEBUFFER, mFBO); - + glFramebufferTexture2D(GL_FRAMEBUFFER, GL_DEPTH_ATTACHMENT, LLTexUnit::getInternalType(mUsage), mDepth, 0); glBindFramebuffer(GL_FRAMEBUFFER, sCurFBO); } - + return addColorAttachment(color_fmt); } @@ -190,7 +190,7 @@ void LLRenderTarget::releaseColorAttachment() llassert(!isBoundInStack()); llassert(mTex.size() == 1); //cannot use releaseColorAttachment with LLRenderTarget managed color targets llassert(mFBO != 0); // mFBO must be valid - + glBindFramebuffer(GL_FRAMEBUFFER, mFBO); glFramebufferTexture2D(GL_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, LLTexUnit::getInternalType(mUsage), 0, 0); glBindFramebuffer(GL_FRAMEBUFFER, sCurFBO); @@ -208,7 +208,7 @@ bool LLRenderTarget::addColorAttachment(U32 color_fmt) return true; } - U32 offset = mTex.size(); + U32 offset = static_cast<U32>(mTex.size()); if( offset >= 4 ) { @@ -238,12 +238,12 @@ bool LLRenderTarget::addColorAttachment(U32 color_fmt) return false; } } - + sBytesAllocated += mResX*mResY*4; stop_glerror(); - + if (offset == 0) { //use bilinear filtering on single texture render targets that aren't multisampled gGL.getTexUnit(0)->setTextureFilteringOption(LLTexUnit::TFO_BILINEAR); @@ -266,15 +266,15 @@ bool LLRenderTarget::addColorAttachment(U32 color_fmt) gGL.getTexUnit(0)->setTextureAddressMode(LLTexUnit::TAM_CLAMP); stop_glerror(); } - + if (mFBO) { glBindFramebuffer(GL_FRAMEBUFFER, mFBO); glFramebufferTexture2D(GL_FRAMEBUFFER, GL_COLOR_ATTACHMENT0+offset, LLTexUnit::getInternalType(mUsage), tex, 0); - + check_framebuffer_status(); - + glBindFramebuffer(GL_FRAMEBUFFER, sCurFBO); } @@ -286,8 +286,8 @@ bool LLRenderTarget::addColorAttachment(U32 color_fmt) bindTarget(); flush(); } - - + + return true; } @@ -296,7 +296,7 @@ bool LLRenderTarget::allocateDepth() LL_PROFILE_ZONE_SCOPED_CATEGORY_DISPLAY; LLImageGL::generateTextures(1, &mDepth); gGL.getTexUnit(0)->bindManual(mUsage, mDepth); - + U32 internal_type = LLTexUnit::getInternalType(mUsage); stop_glerror(); clear_glerror(); @@ -336,7 +336,7 @@ void LLRenderTarget::shareDepthBuffer(LLRenderTarget& target) if (mDepth) { glBindFramebuffer(GL_FRAMEBUFFER, target.mFBO); - + glFramebufferTexture2D(GL_FRAMEBUFFER, GL_DEPTH_ATTACHMENT, LLTexUnit::getInternalType(mUsage), mDepth, 0); check_framebuffer_status(); @@ -355,7 +355,7 @@ void LLRenderTarget::release() if (mDepth) { LLImageGL::deleteTextures(1, &mDepth); - + mDepth = 0; sBytesAllocated -= mResX*mResY*4; @@ -378,11 +378,11 @@ void LLRenderTarget::release() if (mFBO && (mTex.size() > 1)) { glBindFramebuffer(GL_FRAMEBUFFER, mFBO); - S32 z; + size_t z; for (z = mTex.size() - 1; z >= 1; z--) { sBytesAllocated -= mResX*mResY*4; - glFramebufferTexture2D(GL_FRAMEBUFFER, GL_COLOR_ATTACHMENT0+z, LLTexUnit::getInternalType(mUsage), 0, 0); + glFramebufferTexture2D(GL_FRAMEBUFFER, static_cast<GLenum>(GL_COLOR_ATTACHMENT0+z), LLTexUnit::getInternalType(mUsage), 0, 0); LLImageGL::deleteTextures(1, &mTex[z]); } glBindFramebuffer(GL_FRAMEBUFFER, sCurFBO); @@ -408,7 +408,7 @@ void LLRenderTarget::release() mTex.clear(); mInternalFormat.clear(); - + mResX = mResY = 0; } @@ -417,7 +417,7 @@ void LLRenderTarget::bindTarget() LL_PROFILE_GPU_ZONE("bindTarget"); llassert(mFBO); llassert(!isBoundInStack()); - + glBindFramebuffer(GL_FRAMEBUFFER, mFBO); sCurFBO = mFBO; @@ -426,14 +426,17 @@ void LLRenderTarget::bindTarget() GL_COLOR_ATTACHMENT1, GL_COLOR_ATTACHMENT2, GL_COLOR_ATTACHMENT3}; - glDrawBuffers(mTex.size(), drawbuffers); - + if (mTex.empty()) { //no color buffer to draw to glDrawBuffer(GL_NONE); glReadBuffer(GL_NONE); } - + else + { + glDrawBuffers(static_cast<GLsizei>(mTex.size()), drawbuffers); + glReadBuffer(GL_COLOR_ATTACHMENT0); + } check_framebuffer_status(); glViewport(0, 0, mResX, mResY); @@ -452,7 +455,7 @@ void LLRenderTarget::clear(U32 mask_in) if (mUseDepth) { mask |= GL_DEPTH_BUFFER_BIT; - + } if (mFBO) { @@ -472,12 +475,10 @@ void LLRenderTarget::clear(U32 mask_in) U32 LLRenderTarget::getTexture(U32 attachment) const { - if (attachment > mTex.size()-1) - { - LL_ERRS() << "Invalid attachment index." << LL_ENDL; - } - if (mTex.empty()) + if (attachment >= mTex.size()) { + LL_WARNS() << "Invalid attachment index " << attachment << " for size " << mTex.size() << LL_ENDL; + llassert(false); return 0; } return mTex[attachment]; @@ -485,7 +486,7 @@ U32 LLRenderTarget::getTexture(U32 attachment) const U32 LLRenderTarget::getNumTextures() const { - return mTex.size(); + return static_cast<U32>(mTex.size()); } void LLRenderTarget::bindTexture(U32 index, S32 channel, LLTexUnit::eTextureFilterOptions filter_options) @@ -508,7 +509,6 @@ void LLRenderTarget::bindTexture(U32 index, S32 channel, LLTexUnit::eTextureFilt } gGL.getTexUnit(channel)->setTextureFilteringOption(filter_options); - gGL.getTexUnit(channel)->setTextureColorSpace(isSRGB ? LLTexUnit::TCS_SRGB : LLTexUnit::TCS_LINEAR); } void LLRenderTarget::flush() @@ -519,7 +519,8 @@ void LLRenderTarget::flush() llassert(sCurFBO == mFBO); llassert(sBoundTarget == this); - if (mGenerateMipMaps == LLTexUnit::TMG_AUTO) { + if (mGenerateMipMaps == LLTexUnit::TMG_AUTO) + { LL_PROFILE_GPU_ZONE("rt generate mipmaps"); bindTexture(0, 0, LLTexUnit::TFO_TRILINEAR); glGenerateMipmap(GL_TEXTURE_2D); @@ -540,12 +541,14 @@ void LLRenderTarget::flush() glViewport(gGLViewport[0], gGLViewport[1], gGLViewport[2], gGLViewport[3]); sCurResX = gGLViewport[2]; sCurResY = gGLViewport[3]; + glReadBuffer(GL_BACK); + glDrawBuffer(GL_BACK); } } bool LLRenderTarget::isComplete() const { - return (!mTex.empty() || mDepth) ? true : false; + return !mTex.empty() || mDepth; } void LLRenderTarget::getViewport(S32* viewport) @@ -560,7 +563,7 @@ bool LLRenderTarget::isBoundInStack() const { LLRenderTarget* cur = sBoundTarget; while (cur && cur != this) - { + { cur = cur->mPreviousRT; } diff --git a/indra/llrender/llrendertarget.h b/indra/llrender/llrendertarget.h index 340276a752..cd3290cf66 100644 --- a/indra/llrender/llrendertarget.h +++ b/indra/llrender/llrendertarget.h @@ -61,7 +61,7 @@ class LLRenderTarget { public: - //whether or not to use FBO implementation + // Whether or not to use FBO implementation static bool sUseFBO; static U32 sBytesAllocated; static U32 sCurFBO; @@ -172,6 +172,8 @@ public: // *HACK void swapFBORefs(LLRenderTarget& other); + static LLRenderTarget* sBoundTarget; + protected: U32 mResX; U32 mResY; @@ -186,8 +188,6 @@ protected: U32 mMipLevels; LLTexUnit::eTextureType mUsage; - - static LLRenderTarget* sBoundTarget; }; #endif diff --git a/indra/llrender/llshadermgr.cpp b/indra/llrender/llshadermgr.cpp index 0f3716bc18..4807c12226 100644 --- a/indra/llrender/llshadermgr.cpp +++ b/indra/llrender/llshadermgr.cpp @@ -44,7 +44,6 @@ using std::make_pair; using std::string; LLShaderMgr * LLShaderMgr::sInstance = NULL; -bool LLShaderMgr::sMirrorsEnabled = false; LLShaderMgr::LLShaderMgr() { @@ -66,25 +65,25 @@ LLShaderMgr * LLShaderMgr::instance() return sInstance; } -BOOL LLShaderMgr::attachShaderFeatures(LLGLSLShader * shader) +bool LLShaderMgr::attachShaderFeatures(LLGLSLShader * shader) { llassert_always(shader != NULL); LLShaderFeatures *features = & shader->mFeatures; if (features->attachNothing) { - return TRUE; + return true; } ////////////////////////////////////// // Attach Vertex Shader Features First ////////////////////////////////////// // NOTE order of shader object attaching is VERY IMPORTANT!!! - if (features->calculatesAtmospherics) + if (features->calculatesAtmospherics || features->hasGamma || features->isDeferred) { if (!shader->attachVertexObject("windlight/atmosphericsVarsV.glsl")) { - return FALSE; + return false; } } @@ -92,7 +91,7 @@ BOOL LLShaderMgr::attachShaderFeatures(LLGLSLShader * shader) { if (!shader->attachVertexObject("windlight/atmosphericsHelpersV.glsl")) { - return FALSE; + return false; } } @@ -102,40 +101,40 @@ BOOL LLShaderMgr::attachShaderFeatures(LLGLSLShader * shader) { if (!shader->attachVertexObject("lighting/lightFuncSpecularV.glsl")) { - return FALSE; + return false; } if (!features->isAlphaLighting) { if (!shader->attachVertexObject("lighting/sumLightsSpecularV.glsl")) { - return FALSE; + return false; } } if (!shader->attachVertexObject("lighting/lightSpecularV.glsl")) { - return FALSE; + return false; } } else { if (!shader->attachVertexObject("lighting/lightFuncV.glsl")) { - return FALSE; + return false; } if (!features->isAlphaLighting) { if (!shader->attachVertexObject("lighting/sumLightsV.glsl")) { - return FALSE; + return false; } } if (!shader->attachVertexObject("lighting/lightV.glsl")) { - return FALSE; + return false; } } } @@ -145,16 +144,16 @@ BOOL LLShaderMgr::attachShaderFeatures(LLGLSLShader * shader) { if (!shader->attachVertexObject("environment/srgbF.glsl")) // NOTE -- "F" suffix is superfluous here, there is nothing fragment specific in srgbF { - return FALSE; + return false; } if (!shader->attachVertexObject("windlight/atmosphericsFuncs.glsl")) { - return FALSE; + return false; } if (!shader->attachVertexObject("windlight/atmosphericsV.glsl")) { - return FALSE; + return false; } } @@ -162,7 +161,7 @@ BOOL LLShaderMgr::attachShaderFeatures(LLGLSLShader * shader) { if (!shader->attachVertexObject("avatar/avatarSkinV.glsl")) { - return FALSE; + return false; } } @@ -171,13 +170,13 @@ BOOL LLShaderMgr::attachShaderFeatures(LLGLSLShader * shader) shader->mRiggedVariant = shader; if (!shader->attachVertexObject("avatar/objectSkinV.glsl")) { - return FALSE; + return false; } } if (!shader->attachVertexObject("deferred/textureUtilV.glsl")) { - return FALSE; + return false; } /////////////////////////////////////// @@ -188,14 +187,14 @@ BOOL LLShaderMgr::attachShaderFeatures(LLGLSLShader * shader) if (!shader->attachFragmentObject("deferred/globalF.glsl")) { - return FALSE; + return false; } if (features->hasSrgb || features->hasAtmospherics || features->calculatesAtmospherics || features->isDeferred) { if (!shader->attachFragmentObject("environment/srgbF.glsl")) { - return FALSE; + return false; } } @@ -203,7 +202,7 @@ BOOL LLShaderMgr::attachShaderFeatures(LLGLSLShader * shader) { if (!shader->attachFragmentObject("windlight/atmosphericsVarsF.glsl")) { - return FALSE; + return false; } } @@ -211,7 +210,7 @@ BOOL LLShaderMgr::attachShaderFeatures(LLGLSLShader * shader) { if (!shader->attachFragmentObject("windlight/atmosphericsHelpersF.glsl")) { - return FALSE; + return false; } } @@ -220,7 +219,15 @@ BOOL LLShaderMgr::attachShaderFeatures(LLGLSLShader * shader) { if (!shader->attachFragmentObject("deferred/deferredUtil.glsl")) { - return FALSE; + return false; + } + } + + if (features->hasFullGBuffer) + { + if (!shader->attachFragmentObject("deferred/gbufferUtil.glsl")) + { + return false; } } @@ -228,7 +235,7 @@ BOOL LLShaderMgr::attachShaderFeatures(LLGLSLShader * shader) { if (!shader->attachFragmentObject("deferred/screenSpaceReflUtil.glsl")) { - return FALSE; + return false; } } @@ -236,7 +243,7 @@ BOOL LLShaderMgr::attachShaderFeatures(LLGLSLShader * shader) { if (!shader->attachFragmentObject("deferred/shadowUtil.glsl")) { - return FALSE; + return false; } } @@ -244,7 +251,7 @@ BOOL LLShaderMgr::attachShaderFeatures(LLGLSLShader * shader) { if (!shader->attachFragmentObject("deferred/reflectionProbeF.glsl")) { - return FALSE; + return false; } } @@ -252,7 +259,7 @@ BOOL LLShaderMgr::attachShaderFeatures(LLGLSLShader * shader) { if (!shader->attachFragmentObject("deferred/aoUtil.glsl")) { - return FALSE; + return false; } } @@ -260,19 +267,19 @@ BOOL LLShaderMgr::attachShaderFeatures(LLGLSLShader * shader) { if (!shader->attachFragmentObject("windlight/gammaF.glsl")) { - return FALSE; + return false; } } if (features->hasAtmospherics || features->isDeferred) { if (!shader->attachFragmentObject("windlight/atmosphericsFuncs.glsl")) { - return FALSE; + return false; } if (!shader->attachFragmentObject("windlight/atmosphericsF.glsl")) { - return FALSE; + return false; } } @@ -280,7 +287,15 @@ BOOL LLShaderMgr::attachShaderFeatures(LLGLSLShader * shader) { if (!shader->attachFragmentObject("deferred/pbrterrainUtilF.glsl")) { - return FALSE; + return false; + } + } + + if (features->hasTonemap) + { + if (!shader->attachFragmentObject("deferred/tonemapUtilF.glsl")) + { + return false; } } @@ -289,26 +304,26 @@ BOOL LLShaderMgr::attachShaderFeatures(LLGLSLShader * shader) { if (!shader->attachFragmentObject("environment/waterFogF.glsl")) { - return FALSE; + return false; } } if (features->hasLighting) { - if (features->disableTextureIndex) + if (features->mIndexedTextureChannels <= 1) { if (features->hasAlphaMask) { if (!shader->attachFragmentObject("lighting/lightAlphaMaskNonIndexedF.glsl")) { - return FALSE; + return false; } } else { if (!shader->attachFragmentObject("lighting/lightNonIndexedF.glsl")) { - return FALSE; + return false; } } } @@ -318,14 +333,14 @@ BOOL LLShaderMgr::attachShaderFeatures(LLGLSLShader * shader) { if (!shader->attachFragmentObject("lighting/lightAlphaMaskF.glsl")) { - return FALSE; + return false; } } else { if (!shader->attachFragmentObject("lighting/lightF.glsl")) { - return FALSE; + return false; } } shader->mFeatures.mIndexedTextureChannels = llmax(LLGLSLShader::sIndexedTextureChannels, 1); @@ -336,18 +351,18 @@ BOOL LLShaderMgr::attachShaderFeatures(LLGLSLShader * shader) { if (!shader->attachVertexObject("objects/nonindexedTextureV.glsl")) { - return FALSE; + return false; } } else { if (!shader->attachVertexObject("objects/indexedTextureV.glsl")) { - return FALSE; + return false; } } - return TRUE; + return true; } //============================================================================ @@ -421,7 +436,7 @@ void LLShaderMgr::dumpShaderSource(U32 shader_code_count, GLchar** shader_code_t LL_CONT << LL_ENDL; } -void LLShaderMgr::dumpObjectLog(GLuint ret, BOOL warns, const std::string& filename) +void LLShaderMgr::dumpObjectLog(GLuint ret, bool warns, const std::string& filename) { std::string log; log = get_object_log(ret); @@ -459,6 +474,7 @@ GLuint LLShaderMgr::loadShaderFile(const std::string& filename, S32 & shader_lev if (filename.empty()) { + LL_WARNS("ShaderLoading") << "tried loading empty filename" << LL_ENDL; return 0; } @@ -560,17 +576,11 @@ GLuint LLShaderMgr::loadShaderFile(const std::string& filename, S32 & shader_lev } else if (major_version == 3) { - if (minor_version < 10) - { - shader_code_text[shader_code_count++] = strdup("#version 300\n"); - } - else if (minor_version <= 19) - { - shader_code_text[shader_code_count++] = strdup("#version 310\n"); - } - else if (minor_version <= 29) + if (minor_version <= 29) { - shader_code_text[shader_code_count++] = strdup("#version 320\n"); + // OpenGL 3.2 had GLSL version 1.50. anything after that the version numbers match. + // https://www.khronos.org/opengl/wiki/Core_Language_(GLSL)#OpenGL_and_GLSL_versions + shader_code_text[shader_code_count++] = strdup("#version 150\n"); } else { @@ -579,7 +589,8 @@ GLuint LLShaderMgr::loadShaderFile(const std::string& filename, S32 & shader_lev } else { - if (type == GL_GEOMETRY_SHADER) + // OpenGL 3.2 had GLSL version 1.50. anything after that the version numbers match. + if (type == GL_GEOMETRY_SHADER || minor_version >= 50) { //set version to 1.50 shader_code_text[shader_code_count++] = strdup("#version 150\n"); @@ -596,13 +607,15 @@ GLuint LLShaderMgr::loadShaderFile(const std::string& filename, S32 & shader_lev extra_code_text[extra_code_count++] = strdup("precision highp float;\n"); } } - - extra_code_text[extra_code_count++] = strdup("#define FXAA_GLSL_130 1\n"); } - if (sMirrorsEnabled) + if (type == GL_FRAGMENT_SHADER) { - extra_code_text[extra_code_count++] = strdup("#define HERO_PROBES 1\n"); + extra_code_text[extra_code_count++] = strdup("#define FRAGMENT_SHADER 1\n"); + } + else + { + extra_code_text[extra_code_count++] = strdup("#define VERTEX_SHADER 1\n"); } // Use alpha float to store bit flags @@ -611,7 +624,7 @@ GLuint LLShaderMgr::loadShaderFile(const std::string& filename, S32 & shader_lev extra_code_text[extra_code_count++] = strdup("#define GBUFFER_FLAG_HAS_ATMOS 0.34\n"); // bit 0 extra_code_text[extra_code_count++] = strdup("#define GBUFFER_FLAG_HAS_PBR 0.67\n"); // bit 1 extra_code_text[extra_code_count++] = strdup("#define GBUFFER_FLAG_HAS_HDRI 1.0\n"); // bit 2 - extra_code_text[extra_code_count++] = strdup("#define GET_GBUFFER_FLAG(flag) (abs(norm.w-flag)< 0.1)\n"); + extra_code_text[extra_code_count++] = strdup("#define GET_GBUFFER_FLAG(data, flag) (abs(data-flag)< 0.1)\n"); if (defines) { @@ -689,7 +702,7 @@ GLuint LLShaderMgr::loadShaderFile(const std::string& filename, S32 & shader_lev { //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) + for (S32 i = 0; i < texture_index_channels; ++i) { std::string if_string = llformat("\t%sif (vary_texture_index == %d) { return texture(tex%d, texcoord); }\n", i > 0 ? "else " : "", i, i); extra_code_text[extra_code_count++] = strdup(if_string.c_str()); @@ -722,6 +735,9 @@ GLuint LLShaderMgr::loadShaderFile(const std::string& filename, S32 & shader_lev } } + // Master definition can be found in deferredUtil.glsl + extra_code_text[extra_code_count++] = strdup("struct GBufferInfo { vec4 albedo; vec4 specular; vec3 normal; vec4 emissive; float gbufferFlag; float envIntensity; };\n"); + //copy file into memory enum { flag_write_to_out_of_extra_block_area = 0x01 @@ -877,7 +893,7 @@ GLuint LLShaderMgr::loadShaderFile(const std::string& filename, S32 & shader_lev { //an error occured, print log LL_WARNS("ShaderLoading") << "GLSL Compilation Error:" << LL_ENDL; - dumpObjectLog(ret, TRUE, open_file_name); + dumpObjectLog(ret, true, open_file_name); dumpShaderSource(shader_code_count, shader_code_text); glDeleteShader(ret); //no longer need handle ret = 0; @@ -916,10 +932,12 @@ GLuint LLShaderMgr::loadShaderFile(const std::string& filename, S32 & shader_lev } LL_WARNS("ShaderLoading") << "Failed to load " << filename << LL_ENDL; } + + LL_DEBUGS("ShaderLoading") << "loadShaderFile() completed, ret: " << U32(ret) << LL_ENDL; return ret; } -BOOL LLShaderMgr::linkProgramObject(GLuint obj, BOOL suppress_errors) +bool LLShaderMgr::linkProgramObject(GLuint obj, bool suppress_errors) { //check for errors { @@ -936,7 +954,7 @@ BOOL LLShaderMgr::linkProgramObject(GLuint obj, BOOL suppress_errors) { //an error occured, print log LL_SHADER_LOADING_WARNS() << "GLSL Linker Error:" << LL_ENDL; - dumpObjectLog(obj, TRUE, "linker"); + dumpObjectLog(obj, true, "linker"); return success; } } @@ -947,12 +965,12 @@ BOOL LLShaderMgr::linkProgramObject(GLuint obj, BOOL suppress_errors) { LL_SHADER_LOADING_WARNS() << "GLSL Linker: Running in Software:" << LL_ENDL; success = GL_FALSE; - suppress_errors = FALSE; + suppress_errors = false; } return success; } -BOOL LLShaderMgr::validateProgramObject(GLuint obj) +bool LLShaderMgr::validateProgramObject(GLuint obj) { //check program validity against current GL glValidateProgram(obj); @@ -965,7 +983,7 @@ BOOL LLShaderMgr::validateProgramObject(GLuint obj) } else { - dumpObjectLog(obj, FALSE); + dumpObjectLog(obj, false); } return success; @@ -1003,7 +1021,7 @@ void LLShaderMgr::initShaderCache(bool enabled, const LLUUID& old_cache_version, ProgramBinaryData binary_info = ProgramBinaryData(); binary_info.mBinaryFormat = data_pair.second["binary_format"].asInteger(); binary_info.mBinaryLength = data_pair.second["binary_size"].asInteger(); - binary_info.mLastUsedTime = data_pair.second["last_used"].asReal(); + binary_info.mLastUsedTime = (F32)data_pair.second["last_used"].asReal(); mShaderBinaryCache.insert_or_assign(LLUUID(data_pair.first), binary_info); } } @@ -1034,7 +1052,7 @@ void LLShaderMgr::persistShaderCacheMetadata() LLSD out = LLSD::emptyMap(); static const F32 LRU_TIME = (60.f * 60.f) * 24.f * 7.f; // 14 days - const F32 current_time = LLTimer::getTotalSeconds(); + const F32 current_time = (F32)LLTimer::getTotalSeconds(); for (auto it = mShaderBinaryCache.begin(); it != mShaderBinaryCache.end();) { const ProgramBinaryData& shader_metadata = it->second; @@ -1093,7 +1111,7 @@ bool LLShaderMgr::loadCachedProgramBinary(LLGLSLShader* shader) glGetProgramiv(shader->mProgramObject, GL_LINK_STATUS, &success); if (error == GL_NO_ERROR && success == GL_TRUE) { - binary_iter->second.mLastUsedTime = LLTimer::getTotalSeconds(); + binary_iter->second.mLastUsedTime = (F32)LLTimer::getTotalSeconds(); LL_INFOS() << "Loaded cached binary for shader: " << shader->mName << LL_ENDL; return true; } @@ -1120,7 +1138,7 @@ bool LLShaderMgr::saveCachedProgramBinary(LLGLSLShader* shader) program_binary.resize(binary_info.mBinaryLength); GLenum error = glGetError(); // Clear current error - glGetProgramBinary(shader->mProgramObject, program_binary.size() * sizeof(U8), nullptr, &binary_info.mBinaryFormat, program_binary.data()); + glGetProgramBinary(shader->mProgramObject, static_cast<GLsizei>(program_binary.size() * sizeof(U8)), nullptr, &binary_info.mBinaryFormat, program_binary.data()); error = glGetError(); if (error == GL_NO_ERROR) { @@ -1131,7 +1149,7 @@ bool LLShaderMgr::saveCachedProgramBinary(LLGLSLShader* shader) fwrite(program_binary.data(), sizeof(U8), program_binary.size(), outfile); outfile.close(); - binary_info.mLastUsedTime = LLTimer::getTotalSeconds(); + binary_info.mLastUsedTime = (F32)LLTimer::getTotalSeconds(); mShaderBinaryCache.insert_or_assign(shader->mShaderHash, binary_info); return true; @@ -1157,6 +1175,7 @@ void LLShaderMgr::initAttribsAndUniforms() mReservedAttribs.push_back("weight"); mReservedAttribs.push_back("weight4"); mReservedAttribs.push_back("clothing"); + mReservedAttribs.push_back("joint"); mReservedAttribs.push_back("texture_index"); //matrix state @@ -1177,9 +1196,19 @@ void LLShaderMgr::initAttribsAndUniforms() mReservedUniforms.push_back("texture_base_color_transform"); // (GLTF) mReservedUniforms.push_back("texture_normal_transform"); // (GLTF) mReservedUniforms.push_back("texture_metallic_roughness_transform"); // (GLTF) + mReservedUniforms.push_back("texture_occlusion_transform"); // (GLTF) mReservedUniforms.push_back("texture_emissive_transform"); // (GLTF) + mReservedUniforms.push_back("base_color_texcoord"); // (GLTF) + mReservedUniforms.push_back("emissive_texcoord"); // (GLTF) + mReservedUniforms.push_back("normal_texcoord"); // (GLTF) + mReservedUniforms.push_back("metallic_roughness_texcoord"); // (GLTF) + mReservedUniforms.push_back("occlusion_texcoord"); // (GLTF) + mReservedUniforms.push_back("gltf_node_id"); // (GLTF) + mReservedUniforms.push_back("gltf_material_id"); // (GLTF) - llassert(mReservedUniforms.size() == LLShaderMgr::TEXTURE_EMISSIVE_TRANSFORM+1); + mReservedUniforms.push_back("terrain_texture_transforms"); // (GLTF) + + llassert(mReservedUniforms.size() == LLShaderMgr::TERRAIN_TEXTURE_TRANSFORMS +1); mReservedUniforms.push_back("viewport"); @@ -1223,6 +1252,9 @@ void LLShaderMgr::initAttribsAndUniforms() mReservedUniforms.push_back("diffuseMap"); mReservedUniforms.push_back("altDiffuseMap"); mReservedUniforms.push_back("specularMap"); + mReservedUniforms.push_back("metallicRoughnessMap"); + mReservedUniforms.push_back("normalMap"); + mReservedUniforms.push_back("occlusionMap"); mReservedUniforms.push_back("emissiveMap"); mReservedUniforms.push_back("bumpMap"); mReservedUniforms.push_back("bumpMap2"); @@ -1234,13 +1266,13 @@ void LLShaderMgr::initAttribsAndUniforms() mReservedUniforms.push_back("heroProbes"); mReservedUniforms.push_back("cloud_noise_texture"); mReservedUniforms.push_back("cloud_noise_texture_next"); - mReservedUniforms.push_back("fullbright"); mReservedUniforms.push_back("lightnorm"); mReservedUniforms.push_back("sunlight_color"); mReservedUniforms.push_back("ambient_color"); mReservedUniforms.push_back("sky_hdr_scale"); mReservedUniforms.push_back("sky_sunlight_scale"); mReservedUniforms.push_back("sky_ambient_scale"); + mReservedUniforms.push_back("classic_mode"); mReservedUniforms.push_back("blue_horizon"); mReservedUniforms.push_back("blue_density"); mReservedUniforms.push_back("haze_horizon"); @@ -1346,7 +1378,6 @@ void LLShaderMgr::initAttribsAndUniforms() llassert(mReservedUniforms.size() == LLShaderMgr::DEFERRED_SHADOW5+1); - mReservedUniforms.push_back("normalMap"); mReservedUniforms.push_back("positionMap"); mReservedUniforms.push_back("diffuseRect"); mReservedUniforms.push_back("specularRect"); @@ -1359,7 +1390,6 @@ void LLShaderMgr::initAttribsAndUniforms() mReservedUniforms.push_back("bloomMap"); mReservedUniforms.push_back("projectionMap"); mReservedUniforms.push_back("norm_mat"); - mReservedUniforms.push_back("texture_gamma"); mReservedUniforms.push_back("specular_color"); mReservedUniforms.push_back("env_intensity"); @@ -1370,6 +1400,7 @@ void LLShaderMgr::initAttribsAndUniforms() mReservedUniforms.push_back("screenTex"); mReservedUniforms.push_back("screenDepth"); mReservedUniforms.push_back("refTex"); + mReservedUniforms.push_back("exclusionTex"); mReservedUniforms.push_back("eyeVec"); mReservedUniforms.push_back("time"); mReservedUniforms.push_back("waveDir1"); @@ -1404,6 +1435,7 @@ void LLShaderMgr::initAttribsAndUniforms() mReservedUniforms.push_back("detail_3"); mReservedUniforms.push_back("alpha_ramp"); + mReservedUniforms.push_back("paint_map"); mReservedUniforms.push_back("detail_0_base_color"); mReservedUniforms.push_back("detail_1_base_color"); @@ -1428,6 +1460,8 @@ void LLShaderMgr::initAttribsAndUniforms() mReservedUniforms.push_back("emissiveColors"); mReservedUniforms.push_back("minimum_alphas"); + mReservedUniforms.push_back("region_scale"); + mReservedUniforms.push_back("origin"); mReservedUniforms.push_back("display_gamma"); @@ -1462,6 +1496,11 @@ void LLShaderMgr::initAttribsAndUniforms() mReservedUniforms.push_back("debug_normal_draw_length"); + mReservedUniforms.push_back("edgesTex"); + mReservedUniforms.push_back("areaTex"); + mReservedUniforms.push_back("searchTex"); + mReservedUniforms.push_back("blendTex"); + llassert(mReservedUniforms.size() == END_RESERVED_UNIFORMS); std::set<std::string> dupe_check; diff --git a/indra/llrender/llshadermgr.h b/indra/llrender/llshadermgr.h index 8b13e822d5..46788841a5 100644 --- a/indra/llrender/llshadermgr.h +++ b/indra/llrender/llshadermgr.h @@ -36,6 +36,8 @@ public: LLShaderMgr(); virtual ~LLShaderMgr(); + // Note: although you can use statically hashed strings to just bind a random uniform, it's generally preferably that you use this. + // Always document what the actual shader uniform is next to the shader uniform in this struct. // clang-format off typedef enum { // Shader uniform name, set in LLShaderMgr::initAttribsAndUniforms() @@ -56,7 +58,17 @@ public: TEXTURE_BASE_COLOR_TRANSFORM, // "texture_base_color_transform" (GLTF) TEXTURE_NORMAL_TRANSFORM, // "texture_normal_transform" (GLTF) TEXTURE_METALLIC_ROUGHNESS_TRANSFORM, // "texture_metallic_roughness_transform" (GLTF) + TEXTURE_OCCLUSION_TRANSFORM, // "texture_occlusion_transform" (GLTF) TEXTURE_EMISSIVE_TRANSFORM, // "texture_emissive_transform" (GLTF) + BASE_COLOR_TEXCOORD, // "base_color_texcoord" (GLTF) + EMISSIVE_TEXCOORD, // "emissive_texcoord" (GLTF) + NORMAL_TEXCOORD, // "normal_texcoord" (GLTF) + METALLIC_ROUGHNESS_TEXCOORD, // "metallic_roughness_texcoord" (GLTF) + OCCLUSION_TEXCOORD, // "occlusion_texcoord" (GLTF) + GLTF_NODE_ID, // "gltf_node_id" (GLTF) + GLTF_MATERIAL_ID, // "gltf_material_id" (GLTF) + + TERRAIN_TEXTURE_TRANSFORMS, // "terrain_texture_transforms" (GLTF) VIEWPORT, // "viewport" LIGHT_POSITION, // "light_position" @@ -91,6 +103,9 @@ public: DIFFUSE_MAP, // "diffuseMap" ALTERNATE_DIFFUSE_MAP, // "altDiffuseMap" SPECULAR_MAP, // "specularMap" + METALLIC_ROUGHNESS_MAP, // "metallicRoughnessMap" + NORMAL_MAP, // "normalMap" + OCCLUSION_MAP, // "occlusionMap" EMISSIVE_MAP, // "emissiveMap" BUMP_MAP, // "bumpMap" BUMP_MAP2, // "bumpMap2" @@ -102,13 +117,13 @@ public: HERO_PROBE, // "heroProbes" CLOUD_NOISE_MAP, // "cloud_noise_texture" CLOUD_NOISE_MAP_NEXT, // "cloud_noise_texture_next" - FULLBRIGHT, // "fullbright" LIGHTNORM, // "lightnorm" SUNLIGHT_COLOR, // "sunlight_color" AMBIENT, // "ambient_color" SKY_HDR_SCALE, // "sky_hdr_scale" SKY_SUNLIGHT_SCALE, // "sky_sunlight_scale" SKY_AMBIENT_SCALE, // "sky_ambient_scale" + CLASSIC_MODE, // "classic_mode" BLUE_HORIZON, // "blue_horizon" BLUE_DENSITY, // "blue_density" HAZE_HORIZON, // "haze_horizon" @@ -200,7 +215,6 @@ public: DEFERRED_SHADOW3, // "shadowMap3" DEFERRED_SHADOW4, // "shadowMap4" DEFERRED_SHADOW5, // "shadowMap5" - DEFERRED_NORMAL, // "normalMap" DEFERRED_POSITION, // "positionMap" DEFERRED_DIFFUSE, // "diffuseRect" DEFERRED_SPECULAR, // "specularRect" @@ -213,7 +227,6 @@ public: DEFERRED_BLOOM, // "bloomMap" DEFERRED_PROJECTION, // "projectionMap" DEFERRED_NORM_MATRIX, // "norm_mat" - TEXTURE_GAMMA, // "texture_gamma" SPECULAR_COLOR, // "specular_color" ENVIRONMENT_INTENSITY, // "env_intensity" @@ -223,6 +236,7 @@ public: WATER_SCREENTEX, // "screenTex" WATER_SCREENDEPTH, // "screenDepth" WATER_REFTEX, // "refTex" + WATER_EXCLUSIONTEX, // "exclusionTex" WATER_EYEVEC, // "eyeVec" WATER_TIME, // "time" WATER_WAVE_DIR1, // "waveDir1" @@ -257,6 +271,7 @@ public: TERRAIN_DETAIL3, // "detail_3" TERRAIN_ALPHARAMP, // "alpha_ramp" + TERRAIN_PAINTMAP, // "paint_map" TERRAIN_DETAIL0_BASE_COLOR, // "detail_0_base_color" (GLTF) TERRAIN_DETAIL1_BASE_COLOR, // "detail_1_base_color" (GLTF) @@ -281,6 +296,8 @@ public: TERRAIN_EMISSIVE_COLORS, // "emissiveColors" (GLTF) TERRAIN_MINIMUM_ALPHAS, // "minimum_alphas" (GLTF) + REGION_SCALE, // "region_scale" (GLTF) + SHINY_ORIGIN, // "origin" DISPLAY_GAMMA, // "display_gamma" @@ -319,6 +336,11 @@ public: DEBUG_NORMAL_DRAW_LENGTH, // "debug_normal_draw_length" + SMAA_EDGE_TEX, // "edgesTex" + SMAA_AREA_TEX, // "areaTex" + SMAA_SEARCH_TEX, // "searchTex" + SMAA_BLEND_TEX, // "blendTex" + END_RESERVED_UNIFORMS } eGLSLReservedUniforms; // clang-format on @@ -328,11 +350,11 @@ public: virtual void initAttribsAndUniforms(void); - BOOL attachShaderFeatures(LLGLSLShader * shader); - void dumpObjectLog(GLuint ret, BOOL warns = TRUE, const std::string& filename = ""); + bool attachShaderFeatures(LLGLSLShader * shader); + void dumpObjectLog(GLuint ret, bool warns = true, const std::string& filename = ""); void dumpShaderSource(U32 shader_code_count, GLchar** shader_code_text); - BOOL linkProgramObject(GLuint obj, BOOL suppress_errors = FALSE); - BOOL validateProgramObject(GLuint obj); + bool linkProgramObject(GLuint obj, bool suppress_errors = false); + bool validateProgramObject(GLuint obj); GLuint loadShaderFile(const std::string& filename, S32 & shader_level, GLenum type, std::map<std::string, std::string>* defines = NULL, S32 texture_index_channels = -1); // Implemented in the application to actually point to the shader directory. @@ -368,7 +390,6 @@ public: bool mShaderCacheInitialized = false; bool mShaderCacheEnabled = false; std::string mShaderCacheDir; - static bool sMirrorsEnabled; protected: diff --git a/indra/llrender/lltexturemanagerbridge.h b/indra/llrender/lltexturemanagerbridge.h index 0b76ee4de8..e742324e4a 100644 --- a/indra/llrender/lltexturemanagerbridge.h +++ b/indra/llrender/lltexturemanagerbridge.h @@ -36,8 +36,8 @@ class LLTextureManagerBridge public: virtual ~LLTextureManagerBridge() {} - virtual LLPointer<LLGLTexture> getLocalTexture(BOOL usemipmaps = TRUE, BOOL generate_gl_tex = TRUE) = 0; - virtual LLPointer<LLGLTexture> getLocalTexture(const U32 width, const U32 height, const U8 components, BOOL usemipmaps, BOOL generate_gl_tex = TRUE) = 0; + virtual LLPointer<LLGLTexture> getLocalTexture(bool usemipmaps = true, bool generate_gl_tex = true) = 0; + virtual LLPointer<LLGLTexture> getLocalTexture(const U32 width, const U32 height, const U8 components, bool usemipmaps, bool generate_gl_tex = true) = 0; virtual LLGLTexture* getFetchedTexture(const LLUUID &image_id) = 0; }; diff --git a/indra/llrender/lluiimage.cpp b/indra/llrender/lluiimage.cpp index bcf665ca18..dc18bf16bf 100644 --- a/indra/llrender/lluiimage.cpp +++ b/indra/llrender/lluiimage.cpp @@ -81,10 +81,10 @@ void LLUIImage::draw3D(const LLVector3& origin_agent, const LLVector3& x_axis, c } } - LLRender2D::getInstance()->pushMatrix(); + LLRender2D::pushMatrix(); { - LLVector3 rect_origin = origin_agent + (rect.mLeft * x_axis) + (rect.mBottom * y_axis); - LLRender2D::getInstance()->translate(rect_origin.mV[VX], + LLVector3 rect_origin = origin_agent + ((F32)rect.mLeft * x_axis) + ((F32)rect.mBottom * y_axis); + LLRender2D::translate(rect_origin.mV[VX], rect_origin.mV[VY], rect_origin.mV[VZ]); gGL.getTexUnit(0)->bind(getImage()); @@ -100,10 +100,10 @@ void LLUIImage::draw3D(const LLVector3& origin_agent, const LLVector3& x_axis, c (rect.getHeight() - (border_height * border_scale * 0.5f)) / (F32)rect.getHeight(), (rect.getWidth() - (border_width * border_scale * 0.5f)) / (F32)rect.getWidth(), (border_height * border_scale * 0.5f) / (F32)rect.getHeight()), - rect.getWidth() * x_axis, - rect.getHeight() * y_axis); + (F32)rect.getWidth() * x_axis, + (F32)rect.getHeight() * y_axis); - } LLRender2D::getInstance()->popMatrix(); + } LLRender2D::popMatrix(); } //#include "lluiimage.inl" diff --git a/indra/llrender/lluiimage.inl b/indra/llrender/lluiimage.inl index a69616c0c1..dff1fcdfcc 100644 --- a/indra/llrender/lluiimage.inl +++ b/indra/llrender/lluiimage.inl @@ -36,7 +36,7 @@ void LLUIImage::draw(S32 x, S32 y, S32 width, S32 height, const LLColor4& color) width, height, mImage, color, - FALSE, + false, mClipRegion, mScaleRegion, mScaleStyle == SCALE_INNER); @@ -49,7 +49,7 @@ void LLUIImage::drawSolid(S32 x, S32 y, S32 width, S32 height, const LLColor4& c width, height, mImage, color, - TRUE, + true, mClipRegion, mScaleRegion, mScaleStyle == SCALE_INNER); diff --git a/indra/llrender/llvertexbuffer.cpp b/indra/llrender/llvertexbuffer.cpp index 7caf20f40b..1f0c424188 100644 --- a/indra/llrender/llvertexbuffer.cpp +++ b/indra/llrender/llvertexbuffer.cpp @@ -36,6 +36,7 @@ #include "llshadermgr.h" #include "llglslshader.h" #include "llmemory.h" +#include <glm/gtc/type_ptr.hpp> //Next Highest Power Of Two //helper function, returns first number > v that is a power of 2, or v if v is already a power of 2 @@ -271,11 +272,13 @@ static GLuint gen_buffer() { LL_PROFILE_ZONE_NAMED_CATEGORY_VERTEX("gen buffer"); sIndex = pool_size; +#if !LL_DARWIN if (!gGLManager.mIsAMD) { glGenBuffers(pool_size, sNamePool); } else +#endif { // work around for AMD driver bug for (U32 i = 0; i < pool_size; ++i) { @@ -288,13 +291,103 @@ static GLuint gen_buffer() return ret; } +static void delete_buffers(S32 count, GLuint* buffers) +{ + LL_PROFILE_ZONE_SCOPED_CATEGORY_VERTEX; + // wait a few frames before actually deleting the buffers to avoid + // synchronization issues with the GPU + static std::vector<GLuint> sFreeList[4]; + + if (gGLManager.mInited) + { + U32 idx = LLImageGL::sFrameCount % 4; + + for (S32 i = 0; i < count; ++i) + { + sFreeList[idx].push_back(buffers[i]); + } + + idx = (LLImageGL::sFrameCount + 3) % 4; + + if (!sFreeList[idx].empty()) + { + glDeleteBuffers((GLsizei)sFreeList[idx].size(), sFreeList[idx].data()); + sFreeList[idx].resize(0); + } + } +} + + #define ANALYZE_VBO_POOL 0 +// VBO Pool interface class LLVBOPool { + public: + virtual ~LLVBOPool() = default; + virtual void allocate(GLenum type, U32 size, GLuint& name, U8*& data) = 0; + virtual void free(GLenum type, U32 size, GLuint name, U8* data) = 0; + virtual U64 getVramBytesUsed() = 0; +}; + +// VBO Pool for Apple GPUs (as in M1/M2 etc, not Intel macs) +// Effectively disables VBO pooling +class LLAppleVBOPool final: public LLVBOPool +{ public: - typedef std::chrono::steady_clock::time_point Time; + U64 mAllocated = 0; + + U64 getVramBytesUsed() override + { + return mAllocated; + } + + void allocate(GLenum type, U32 size, GLuint& name, U8*& data) override + { + LL_PROFILE_ZONE_SCOPED_CATEGORY_VERTEX; + STOP_GLERROR; + llassert(type == GL_ARRAY_BUFFER || type == GL_ELEMENT_ARRAY_BUFFER); + llassert(name == 0); // non zero name indicates a gl name that wasn't freed + llassert(data == nullptr); // non null data indicates a buffer that wasn't freed + llassert(size >= 2); // any buffer size smaller than a single index is nonsensical + + mAllocated += size; + { //allocate a new buffer + LL_PROFILE_GPU_ZONE("vbo alloc"); + // ON OS X, we don't allocate a VBO until the last possible moment + // in unmapBuffer + data = (U8*) ll_aligned_malloc_16(size); + STOP_GLERROR; + } + } + + void free(GLenum type, U32 size, GLuint name, U8* data) override + { + LL_PROFILE_ZONE_SCOPED_CATEGORY_VERTEX; + llassert(type == GL_ARRAY_BUFFER || type == GL_ELEMENT_ARRAY_BUFFER); + llassert(size >= 2); + + if (data) + { + ll_aligned_free_16(data); + } + + mAllocated -= size; + STOP_GLERROR; + if (name) + { + delete_buffers(1, &name); + } + STOP_GLERROR; + } +}; + +// VBO Pool for GPUs that benefit from VBO pooling +class LLDefaultVBOPool final : public LLVBOPool +{ +public: + typedef std::chrono::steady_clock::time_point Time; struct Entry { U8* mData; @@ -302,7 +395,7 @@ public: Time mAge; }; - ~LLVBOPool() + ~LLDefaultVBOPool() override { clear(); } @@ -320,7 +413,7 @@ public: U32 mMisses = 0; U32 mHits = 0; - U64 getVramBytesUsed() + U64 getVramBytesUsed() override { return mAllocated + mReserved; } @@ -336,7 +429,7 @@ public: size += block_size - (size % block_size); } - void allocate(GLenum type, U32 size, GLuint& name, U8*& data) + void allocate(GLenum type, U32 size, GLuint& name, U8*& data) override { LL_PROFILE_ZONE_SCOPED_CATEGORY_VERTEX; llassert(type == GL_ARRAY_BUFFER || type == GL_ELEMENT_ARRAY_BUFFER); @@ -392,7 +485,7 @@ public: clean(); } - void free(GLenum type, U32 size, GLuint name, U8* data) + void free(GLenum type, U32 size, GLuint name, U8* data) override { LL_PROFILE_ZONE_SCOPED_CATEGORY_VERTEX; llassert(type == GL_ARRAY_BUFFER || type == GL_ELEMENT_ARRAY_BUFFER); @@ -455,7 +548,7 @@ public: LL_PROFILE_ZONE_NAMED_CATEGORY_VERTEX("vbo cache timeout"); auto& entry = entries.back(); ll_aligned_free_16(entry.mData); - glDeleteBuffers(1, &entry.mGLName); + delete_buffers(1, &entry.mGLName); llassert(mReserved >= iter->first); mReserved -= iter->first; entries.pop_back(); @@ -491,7 +584,7 @@ public: for (auto& entry : entries.second) { ll_aligned_free_16(entry.mData); - glDeleteBuffers(1, &entry.mGLName); + delete_buffers(1, &entry.mGLName); } } @@ -500,7 +593,7 @@ public: for (auto& entry : entries.second) { ll_aligned_free_16(entry.mData); - glDeleteBuffers(1, &entry.mGLName); + delete_buffers(1, &entry.mGLName); } } @@ -509,12 +602,72 @@ public: mIBOPool.clear(); mVBOPool.clear(); } - - }; static LLVBOPool* sVBOPool = nullptr; +void LLVertexBufferData::drawWithMatrix() +{ + if (!mVB) + { + llassert(false); + // Not supposed to happen, check buffer generation + return; + } + + if (mTexName) + { + gGL.getTexUnit(0)->bindManual(LLTexUnit::TT_TEXTURE, mTexName); + } + else + { + gGL.getTexUnit(0)->unbind(LLTexUnit::TT_TEXTURE); + } + + gGL.matrixMode(LLRender::MM_MODELVIEW); + gGL.pushMatrix(); + gGL.loadMatrix(glm::value_ptr(mModelView)); + gGL.matrixMode(LLRender::MM_PROJECTION); + gGL.pushMatrix(); + gGL.loadMatrix(glm::value_ptr(mProjection)); + gGL.matrixMode(LLRender::MM_TEXTURE0); + gGL.pushMatrix(); + gGL.loadMatrix(glm::value_ptr(mTexture0)); + + mVB->setBuffer(); + mVB->drawArrays(mMode, 0, mCount); + + gGL.popMatrix(); + gGL.matrixMode(LLRender::MM_PROJECTION); + gGL.popMatrix(); + gGL.matrixMode(LLRender::MM_MODELVIEW); + gGL.popMatrix(); +} + +void LLVertexBufferData::draw() +{ + if (!mVB) + { + llassert(false); + // Not supposed to happen, check buffer generation + return; + } + + if (mTexName) + { + gGL.getTexUnit(0)->bindManual(LLTexUnit::TT_TEXTURE, mTexName); + } + else + { + gGL.getTexUnit(0)->unbind(LLTexUnit::TT_TEXTURE); + } + + mVB->setBuffer(); + mVB->drawArrays(mMode, 0, mCount); +} + +//============================================================================ + //static U64 LLVertexBuffer::getBytesAllocated() { @@ -545,6 +698,7 @@ const U32 LLVertexBuffer::sTypeSize[LLVertexBuffer::TYPE_MAX] = sizeof(F32), // TYPE_WEIGHT, sizeof(LLVector4), // TYPE_WEIGHT4, sizeof(LLVector4), // TYPE_CLOTHWEIGHT, + sizeof(U64), // TYPE_JOINT, sizeof(LLVector4), // TYPE_TEXTURE_INDEX (actually exists as position.w), no extra data, but stride is 16 bytes }; @@ -562,6 +716,7 @@ static const std::string vb_type_name[] = "TYPE_WEIGHT", "TYPE_WEIGHT4", "TYPE_CLOTHWEIGHT", + "TYPE_JOINT" "TYPE_TEXTURE_INDEX", "TYPE_MAX", "TYPE_INDEX", @@ -575,7 +730,6 @@ const U32 LLVertexBuffer::sGLMode[LLRender::NUM_MODES] = GL_POINTS, GL_LINES, GL_LINE_STRIP, - GL_QUADS, GL_LINE_LOOP, }; @@ -629,6 +783,8 @@ void LLVertexBuffer::drawElements(U32 mode, const LLVector4a* pos, const LLVecto LL_PROFILE_ZONE_SCOPED_CATEGORY_VERTEX; llassert(LLGLSLShader::sCurBoundShaderPtr != NULL); + STOP_GLERROR; + gGL.syncMatrices(); U32 mask = LLVertexBuffer::MAP_VERTEX; @@ -643,7 +799,7 @@ void LLVertexBuffer::drawElements(U32 mode, const LLVector4a* pos, const LLVecto if (tc != nullptr) { - for (int i = 0; i < num_indices; ++i) + for (U32 i = 0; i < num_indices; ++i) { U16 idx = indicesp[i]; gGL.texCoord2fv(tc[idx].mV); @@ -652,7 +808,7 @@ void LLVertexBuffer::drawElements(U32 mode, const LLVector4a* pos, const LLVecto } else { - for (int i = 0; i < num_indices; ++i) + for (U32 i = 0; i < num_indices; ++i) { U16 idx = indicesp[i]; gGL.vertex3fv(pos[idx].getF32ptr()); @@ -669,24 +825,23 @@ bool LLVertexBuffer::validateRange(U32 start, U32 end, U32 count, U32 indices_of return true; } - llassert(start < (U32)mNumVerts); - llassert(end < (U32)mNumVerts); + llassert(start < mNumVerts); + llassert(end < mNumVerts); - if (start >= (U32) mNumVerts || - end >= (U32) mNumVerts) + if (start >= mNumVerts || + end >= mNumVerts) { LL_ERRS() << "Bad vertex buffer draw range: [" << start << ", " << end << "] vs " << mNumVerts << LL_ENDL; } - llassert(mNumIndices >= 0); - - if (indices_offset >= (U32) mNumIndices || - indices_offset + count > (U32) mNumIndices) + if (indices_offset >= mNumIndices || + indices_offset + count > mNumIndices) { LL_ERRS() << "Bad index buffer draw range: [" << indices_offset << ", " << indices_offset+count << "]" << LL_ENDL; } { +#if 0 // not a reliable test for VBOs that are not backed by a CPU buffer U16* idx = (U16*) mMappedIndexData+indices_offset; for (U32 i = 0; i < count; ++i) { @@ -718,12 +873,13 @@ bool LLVertexBuffer::validateRange(U32 start, U32 end, U32 count, U32 indices_of for (U32 i = start; i < end; i++) { U32 idx = (U32) (v[i][3]+0.25f); - if (idx >= shader->mFeatures.mIndexedTextureChannels) + if (idx >= (U32)shader->mFeatures.mIndexedTextureChannels) { LL_ERRS() << "Bad texture index found in vertex data stream." << LL_ENDL; } } } +#endif } return true; @@ -735,16 +891,37 @@ void LLVertexBuffer::setLabel(const char* label) { } #endif +void LLVertexBuffer::clone(LLVertexBuffer& target) const +{ + target.mTypeMask = mTypeMask; + target.mIndicesType = mIndicesType; + target.mIndicesStride = mIndicesStride; + if (target.getNumVerts() != getNumVerts() || + target.getNumIndices() != getNumIndices()) + { + target.allocateBuffer(getNumVerts(), getNumIndices()); + } +} + void LLVertexBuffer::drawRange(U32 mode, U32 start, U32 end, U32 count, U32 indices_offset) const { llassert(validateRange(start, end, count, indices_offset)); llassert(mGLBuffer == sGLRenderBuffer); llassert(mGLIndices == sGLRenderIndices); gGL.syncMatrices(); + STOP_GLERROR; glDrawRangeElements(sGLMode[mode], start, end, count, mIndicesType, (GLvoid*) (indices_offset * (size_t) mIndicesStride)); + STOP_GLERROR; +} + +void LLVertexBuffer::drawRangeFast(U32 mode, U32 start, U32 end, U32 count, U32 indices_offset) const +{ + glDrawRangeElements(sGLMode[mode], start, end, count, mIndicesType, + (GLvoid*)(indices_offset * (size_t)mIndicesStride)); } + void LLVertexBuffer::draw(U32 mode, U32 count, U32 indices_offset) const { drawRange(mode, 0, mNumVerts-1, count, indices_offset); @@ -758,14 +935,26 @@ void LLVertexBuffer::drawArrays(U32 mode, U32 first, U32 count) const llassert(mGLIndices == sGLRenderIndices); gGL.syncMatrices(); + STOP_GLERROR; glDrawArrays(sGLMode[mode], first, count); + STOP_GLERROR; } //static void LLVertexBuffer::initClass(LLWindow* window) { llassert(sVBOPool == nullptr); - sVBOPool = new LLVBOPool(); + + if (gGLManager.mIsApple) + { + LL_INFOS() << "VBO Pooling Disabled" << LL_ENDL; + sVBOPool = new LLAppleVBOPool(); + } + else + { + LL_INFOS() << "VBO Pooling Enabled" << LL_ENDL; + sVBOPool = new LLDefaultVBOPool(); + } #if ENABLE_GL_WORK_QUEUE sQueue = new GLWorkQueue(); @@ -781,9 +970,10 @@ void LLVertexBuffer::initClass(LLWindow* window) //static void LLVertexBuffer::unbind() { + STOP_GLERROR; glBindBuffer(GL_ARRAY_BUFFER, 0); glBindBuffer(GL_ELEMENT_ARRAY_BUFFER, 0); - + STOP_GLERROR; sGLRenderBuffer = 0; sGLRenderIndices = 0; } @@ -823,6 +1013,24 @@ LLVertexBuffer::LLVertexBuffer(U32 typemask) } } +// list of mapped buffers +// NOTE: must not be LLPointer<LLVertexBuffer> to avoid breaking non-ref-counted LLVertexBuffer instances +static std::vector<LLVertexBuffer*> sMappedBuffers; + +//static +void LLVertexBuffer::flushBuffers() +{ + LL_PROFILE_ZONE_SCOPED_CATEGORY_VERTEX; + // must only be called from main thread + for (auto& buffer : sMappedBuffers) + { + buffer->_unmapBuffer(); + buffer->mMapped = false; + } + + sMappedBuffers.resize(0); +} + //static U32 LLVertexBuffer::calcOffsets(const U32& typemask, U32* offsets, U32 num_vertices) { @@ -866,6 +1074,12 @@ U32 LLVertexBuffer::calcVertexSize(const U32& typemask) //virtual LLVertexBuffer::~LLVertexBuffer() { + if (mMapped) + { // is on the mapped buffer list but doesn't need to be flushed + mMapped = false; + unmapBuffer(); + } + destroyGLBuffer(); destroyGLIndices(); @@ -998,12 +1212,6 @@ bool LLVertexBuffer::updateNumVerts(U32 nverts) bool success = true; - if (nverts > 65536) - { - LL_WARNS() << "Vertex buffer overflow!" << LL_ENDL; - nverts = 65536; - } - U32 needed_size = calcOffsets(mTypeMask, mOffsets, nverts); if (needed_size != mSize) @@ -1036,8 +1244,7 @@ bool LLVertexBuffer::updateNumIndices(U32 nindices) bool LLVertexBuffer::allocateBuffer(U32 nverts, U32 nindices) { - if (nverts < 0 || nindices < 0 || - nverts > 65536) + if (nverts < 0 || nindices < 0) { LL_ERRS() << "Bad vertex buffer allocation: " << nverts << " : " << nindices << LL_ENDL; } @@ -1074,33 +1281,36 @@ bool expand_region(LLVertexBuffer::MappedRegion& region, U32 start, U32 end) U8* LLVertexBuffer::mapVertexBuffer(LLVertexBuffer::AttributeType type, U32 index, S32 count) { LL_PROFILE_ZONE_SCOPED_CATEGORY_VERTEX; + _mapBuffer(); if (count == -1) { count = mNumVerts - index; } - U32 start = mOffsets[type] + sTypeSize[type] * index; - U32 end = start + sTypeSize[type] * count-1; - - bool flagged = false; - // flag region as mapped - for (U32 i = 0; i < mMappedVertexRegions.size(); ++i) + if (!gGLManager.mIsApple) { - MappedRegion& region = mMappedVertexRegions[i]; - if (expand_region(region, start, end)) + U32 start = mOffsets[type] + sTypeSize[type] * index; + U32 end = start + sTypeSize[type] * count-1; + + bool flagged = false; + // flag region as mapped + for (U32 i = 0; i < mMappedVertexRegions.size(); ++i) { - flagged = true; - break; + MappedRegion& region = mMappedVertexRegions[i]; + if (expand_region(region, start, end)) + { + flagged = true; + break; + } } - } - if (!flagged) - { - //didn't expand an existing region, make a new one - mMappedVertexRegions.push_back({ start, end }); + if (!flagged) + { + //didn't expand an existing region, make a new one + mMappedVertexRegions.push_back({ start, end }); + } } - return mMappedData+mOffsets[type]+sTypeSize[type]*index; } @@ -1108,31 +1318,35 @@ U8* LLVertexBuffer::mapVertexBuffer(LLVertexBuffer::AttributeType type, U32 inde U8* LLVertexBuffer::mapIndexBuffer(U32 index, S32 count) { LL_PROFILE_ZONE_SCOPED_CATEGORY_VERTEX; + _mapBuffer(); if (count == -1) { count = mNumIndices-index; } - U32 start = sizeof(U16) * index; - U32 end = start + sizeof(U16) * count-1; - - bool flagged = false; - // flag region as mapped - for (U32 i = 0; i < mMappedIndexRegions.size(); ++i) + if (!gGLManager.mIsApple) { - MappedRegion& region = mMappedIndexRegions[i]; - if (expand_region(region, start, end)) + U32 start = sizeof(U16) * index; + U32 end = start + sizeof(U16) * count-1; + + bool flagged = false; + // flag region as mapped + for (U32 i = 0; i < mMappedIndexRegions.size(); ++i) { - flagged = true; - break; + MappedRegion& region = mMappedIndexRegions[i]; + if (expand_region(region, start, end)) + { + flagged = true; + break; + } } - } - if (!flagged) - { - //didn't expand an existing region, make a new one - mMappedIndexRegions.push_back({ start, end }); + if (!flagged) + { + //didn't expand an existing region, make a new one + mMappedIndexRegions.push_back({ start, end }); + } } return mMappedIndexData + sizeof(U16)*index; @@ -1142,31 +1356,68 @@ U8* LLVertexBuffer::mapIndexBuffer(U32 index, S32 count) // target -- "target" parameter for glBufferSubData // start -- first byte to copy // end -- last byte to copy (NOT last byte + 1) -// data -- mMappedData or mMappedIndexData -static void flush_vbo(GLenum target, U32 start, U32 end, void* data) +// data -- data to be flushed +// dst -- mMappedData or mMappedIndexData +void LLVertexBuffer::flush_vbo(GLenum target, U32 start, U32 end, void* data, U8* dst) { - if (end != 0) + if (gGLManager.mIsApple) { - LL_PROFILE_ZONE_NAMED_CATEGORY_VERTEX("glBufferSubData"); - LL_PROFILE_ZONE_NUM(start); - LL_PROFILE_ZONE_NUM(end); - LL_PROFILE_ZONE_NUM(end-start); - - constexpr U32 block_size = 8192; + // on OS X, flush_vbo doesn't actually write to the GL buffer, so be sure to call + // _mapBuffer to tag the buffer for flushing to GL + _mapBuffer(); + LL_PROFILE_ZONE_NAMED_CATEGORY_VERTEX("vb memcpy"); + STOP_GLERROR; + // copy into mapped buffer + memcpy(dst+start, data, end-start+1); + } + else + { + llassert(target == GL_ARRAY_BUFFER ? sGLRenderBuffer == mGLBuffer : sGLRenderIndices == mGLIndices); - for (U32 i = start; i <= end; i += block_size) + // skip mapped data and stream to GPU via glBufferSubData + if (end != 0) { - LL_PROFILE_ZONE_NAMED_CATEGORY_VERTEX("glBufferSubData block"); - //LL_PROFILE_GPU_ZONE("glBufferSubData"); - U32 tend = llmin(i + block_size, end); - U32 size = tend - i + 1; - glBufferSubData(target, i, size, (U8*) data + (i-start)); + LL_PROFILE_ZONE_NAMED_CATEGORY_VERTEX("glBufferSubData"); + LL_PROFILE_ZONE_NUM(start); + LL_PROFILE_ZONE_NUM(end); + LL_PROFILE_ZONE_NUM(end-start); + + constexpr U32 block_size = 65536; + + for (U32 i = start; i <= end; i += block_size) + { + //LL_PROFILE_ZONE_NAMED_CATEGORY_VERTEX("glBufferSubData block"); + //LL_PROFILE_GPU_ZONE("glBufferSubData"); + U32 tend = llmin(i + block_size, end); + U32 size = tend - i + 1; + glBufferSubData(target, i, size, (U8*) data + (i-start)); + } } } } void LLVertexBuffer::unmapBuffer() { + flushBuffers(); +} + +void LLVertexBuffer::_mapBuffer() +{ + if (!mMapped) + { + mMapped = true; + sMappedBuffers.push_back(this); + } +} + +void LLVertexBuffer::_unmapBuffer() +{ + STOP_GLERROR; + if (!mMapped) + { + return; + } + struct SortMappedRegion { bool operator()(const MappedRegion& lhs, const MappedRegion& rhs) @@ -1175,72 +1426,115 @@ void LLVertexBuffer::unmapBuffer() } }; - if (!mMappedVertexRegions.empty()) + if (gGLManager.mIsApple) { - LL_PROFILE_ZONE_NAMED_CATEGORY_VERTEX("unmapBuffer - vertex"); - if (sGLRenderBuffer != mGLBuffer) + STOP_GLERROR; + if (mMappedData) { + if (mGLBuffer) + { + delete_buffers(1, &mGLBuffer); + } + mGLBuffer = gen_buffer(); glBindBuffer(GL_ARRAY_BUFFER, mGLBuffer); sGLRenderBuffer = mGLBuffer; + glBufferData(GL_ARRAY_BUFFER, mSize, mMappedData, GL_STATIC_DRAW); } + else if (mGLBuffer != sGLRenderBuffer) + { + glBindBuffer(GL_ARRAY_BUFFER, mGLBuffer); + sGLRenderBuffer = mGLBuffer; + } + STOP_GLERROR; - U32 start = 0; - U32 end = 0; - - std::sort(mMappedVertexRegions.begin(), mMappedVertexRegions.end(), SortMappedRegion()); - - for (U32 i = 0; i < mMappedVertexRegions.size(); ++i) + if (mMappedIndexData) { - const MappedRegion& region = mMappedVertexRegions[i]; - if (region.mStart == end + 1) - { - end = region.mEnd; - } - else + if (mGLIndices) { - flush_vbo(GL_ARRAY_BUFFER, start, end, (U8*)mMappedData + start); - start = region.mStart; - end = region.mEnd; + delete_buffers(1, &mGLIndices); } - } - - flush_vbo(GL_ARRAY_BUFFER, start, end, (U8*)mMappedData + start); - - mMappedVertexRegions.clear(); - } - if (!mMappedIndexRegions.empty()) - { - LL_PROFILE_ZONE_NAMED_CATEGORY_VERTEX("unmapBuffer - index"); + mGLIndices = gen_buffer(); + glBindBuffer(GL_ELEMENT_ARRAY_BUFFER, mGLIndices); + sGLRenderIndices = mGLIndices; - if (mGLIndices != sGLRenderIndices) + glBufferData(GL_ELEMENT_ARRAY_BUFFER, mIndicesSize, mMappedIndexData, GL_STATIC_DRAW); + } + else if (mGLIndices != sGLRenderIndices) { glBindBuffer(GL_ELEMENT_ARRAY_BUFFER, mGLIndices); sGLRenderIndices = mGLIndices; } - U32 start = 0; - U32 end = 0; - - std::sort(mMappedIndexRegions.begin(), mMappedIndexRegions.end(), SortMappedRegion()); - - for (U32 i = 0; i < mMappedIndexRegions.size(); ++i) + STOP_GLERROR; + } + else + { + if (!mMappedVertexRegions.empty()) { - const MappedRegion& region = mMappedIndexRegions[i]; - if (region.mStart == end + 1) + LL_PROFILE_ZONE_NAMED_CATEGORY_VERTEX("unmapBuffer - vertex"); + + if (sGLRenderBuffer != mGLBuffer) { - end = region.mEnd; + glBindBuffer(GL_ARRAY_BUFFER, mGLBuffer); + sGLRenderBuffer = mGLBuffer; } - else + + U32 start = 0; + U32 end = 0; + + std::sort(mMappedVertexRegions.begin(), mMappedVertexRegions.end(), SortMappedRegion()); + + for (U32 i = 0; i < mMappedVertexRegions.size(); ++i) { - flush_vbo(GL_ELEMENT_ARRAY_BUFFER, start, end, (U8*)mMappedIndexData + start); - start = region.mStart; - end = region.mEnd; + const MappedRegion& region = mMappedVertexRegions[i]; + if (region.mStart == end + 1) + { + end = region.mEnd; + } + else + { + flush_vbo(GL_ARRAY_BUFFER, start, end, (U8*)mMappedData + start, mMappedData); + start = region.mStart; + end = region.mEnd; + } } + + flush_vbo(GL_ARRAY_BUFFER, start, end, (U8*)mMappedData + start, mMappedData); + mMappedVertexRegions.clear(); } - flush_vbo(GL_ELEMENT_ARRAY_BUFFER, start, end, (U8*)mMappedIndexData + start); + if (!mMappedIndexRegions.empty()) + { + LL_PROFILE_ZONE_NAMED_CATEGORY_VERTEX("unmapBuffer - index"); - mMappedIndexRegions.clear(); + if (mGLIndices != sGLRenderIndices) + { + glBindBuffer(GL_ELEMENT_ARRAY_BUFFER, mGLIndices); + sGLRenderIndices = mGLIndices; + } + U32 start = 0; + U32 end = 0; + + std::sort(mMappedIndexRegions.begin(), mMappedIndexRegions.end(), SortMappedRegion()); + + for (U32 i = 0; i < mMappedIndexRegions.size(); ++i) + { + const MappedRegion& region = mMappedIndexRegions[i]; + if (region.mStart == end + 1) + { + end = region.mEnd; + } + else + { + flush_vbo(GL_ELEMENT_ARRAY_BUFFER, start, end, (U8*)mMappedIndexData + start, mMappedIndexData); + start = region.mStart; + end = region.mEnd; + } + } + + flush_vbo(GL_ELEMENT_ARRAY_BUFFER, start, end, (U8*)mMappedIndexData + start, mMappedIndexData); + mMappedIndexRegions.clear(); + } } } @@ -1362,6 +1656,14 @@ bool LLVertexBuffer::getClothWeightStrider(LLStrider<LLVector4>& strider, U32 in // Set for rendering void LLVertexBuffer::setBuffer() { + STOP_GLERROR; + + if (mMapped) + { + LL_WARNS_ONCE() << "Missing call to unmapBuffer or flushBuffers" << LL_ENDL; + _unmapBuffer(); + } + // no data may be pending llassert(mMappedVertexRegions.empty()); llassert(mMappedIndexRegions.empty()); @@ -1394,12 +1696,15 @@ void LLVertexBuffer::setBuffer() glBindBuffer(GL_ELEMENT_ARRAY_BUFFER, mGLIndices); sGLRenderIndices = mGLIndices; } + + STOP_GLERROR; } // virtual (default) void LLVertexBuffer::setupVertexBuffer() { + STOP_GLERROR; U8* base = nullptr; U32 data_mask = LLGLSLShader::sCurBoundShaderPtr->mAttributeMask; @@ -1471,6 +1776,12 @@ void LLVertexBuffer::setupVertexBuffer() void* ptr = (void*)(base + mOffsets[TYPE_WEIGHT4]); glVertexAttribPointer(loc, 4, GL_FLOAT, GL_FALSE, LLVertexBuffer::sTypeSize[TYPE_WEIGHT4], ptr); } + if (data_mask & MAP_JOINT) + { + AttributeType loc = TYPE_JOINT; + void* ptr = (void*)(base + mOffsets[TYPE_JOINT]); + glVertexAttribIPointer(loc, 4, GL_UNSIGNED_SHORT, LLVertexBuffer::sTypeSize[TYPE_JOINT], ptr); + } if (data_mask & MAP_CLOTHWEIGHT) { AttributeType loc = TYPE_CLOTHWEIGHT; @@ -1489,59 +1800,121 @@ void LLVertexBuffer::setupVertexBuffer() void* ptr = (void*)(base + mOffsets[TYPE_VERTEX]); glVertexAttribPointer(loc, 3, GL_FLOAT, GL_FALSE, LLVertexBuffer::sTypeSize[TYPE_VERTEX], ptr); } + STOP_GLERROR; } void LLVertexBuffer::setPositionData(const LLVector4a* data) { - llassert(sGLRenderBuffer == mGLBuffer); - flush_vbo(GL_ARRAY_BUFFER, 0, sizeof(LLVector4a) * getNumVerts()-1, (U8*) data); + flush_vbo(GL_ARRAY_BUFFER, 0, sizeof(LLVector4a) * getNumVerts()-1, (U8*) data, mMappedData); +} + +void LLVertexBuffer::setTexCoord0Data(const LLVector2* data) +{ + flush_vbo(GL_ARRAY_BUFFER, mOffsets[TYPE_TEXCOORD0], mOffsets[TYPE_TEXCOORD0] + sTypeSize[TYPE_TEXCOORD0] * getNumVerts() - 1, (U8*)data, mMappedData); } -void LLVertexBuffer::setTexCoordData(const LLVector2* data) +void LLVertexBuffer::setTexCoord1Data(const LLVector2* data) { - llassert(sGLRenderBuffer == mGLBuffer); - flush_vbo(GL_ARRAY_BUFFER, mOffsets[TYPE_TEXCOORD0], mOffsets[TYPE_TEXCOORD0] + sTypeSize[TYPE_TEXCOORD0] * getNumVerts() - 1, (U8*)data); + flush_vbo(GL_ARRAY_BUFFER, mOffsets[TYPE_TEXCOORD1], mOffsets[TYPE_TEXCOORD1] + sTypeSize[TYPE_TEXCOORD1] * getNumVerts() - 1, (U8*)data, mMappedData); } void LLVertexBuffer::setColorData(const LLColor4U* data) { - llassert(sGLRenderBuffer == mGLBuffer); - flush_vbo(GL_ARRAY_BUFFER, mOffsets[TYPE_COLOR], mOffsets[TYPE_COLOR] + sTypeSize[TYPE_COLOR] * getNumVerts() - 1, (U8*) data); + flush_vbo(GL_ARRAY_BUFFER, mOffsets[TYPE_COLOR], mOffsets[TYPE_COLOR] + sTypeSize[TYPE_COLOR] * getNumVerts() - 1, (U8*) data, mMappedData); } void LLVertexBuffer::setNormalData(const LLVector4a* data) { - llassert(sGLRenderBuffer == mGLBuffer); - flush_vbo(GL_ARRAY_BUFFER, mOffsets[TYPE_NORMAL], mOffsets[TYPE_NORMAL] + sTypeSize[TYPE_NORMAL] * getNumVerts() - 1, (U8*) data); + flush_vbo(GL_ARRAY_BUFFER, mOffsets[TYPE_NORMAL], mOffsets[TYPE_NORMAL] + sTypeSize[TYPE_NORMAL] * getNumVerts() - 1, (U8*) data, mMappedData); } void LLVertexBuffer::setTangentData(const LLVector4a* data) { - llassert(sGLRenderBuffer == mGLBuffer); - flush_vbo(GL_ARRAY_BUFFER, mOffsets[TYPE_TANGENT], mOffsets[TYPE_TANGENT] + sTypeSize[TYPE_TANGENT] * getNumVerts() - 1, (U8*) data); + flush_vbo(GL_ARRAY_BUFFER, mOffsets[TYPE_TANGENT], mOffsets[TYPE_TANGENT] + sTypeSize[TYPE_TANGENT] * getNumVerts() - 1, (U8*) data, mMappedData); } void LLVertexBuffer::setWeight4Data(const LLVector4a* data) { - llassert(sGLRenderBuffer == mGLBuffer); - flush_vbo(GL_ARRAY_BUFFER, mOffsets[TYPE_WEIGHT4], mOffsets[TYPE_WEIGHT4] + sTypeSize[TYPE_WEIGHT4] * getNumVerts() - 1, (U8*) data); + flush_vbo(GL_ARRAY_BUFFER, mOffsets[TYPE_WEIGHT4], mOffsets[TYPE_WEIGHT4] + sTypeSize[TYPE_WEIGHT4] * getNumVerts() - 1, (U8*) data, mMappedData); +} + +void LLVertexBuffer::setJointData(const U64* data) +{ + flush_vbo(GL_ARRAY_BUFFER, mOffsets[TYPE_JOINT], mOffsets[TYPE_JOINT] + sTypeSize[TYPE_JOINT] * getNumVerts() - 1, (U8*) data, mMappedData); } void LLVertexBuffer::setIndexData(const U16* data) { - llassert(sGLRenderIndices == mGLIndices); - flush_vbo(GL_ELEMENT_ARRAY_BUFFER, 0, sizeof(U16) * getNumIndices() - 1, (U8*) data); + flush_vbo(GL_ELEMENT_ARRAY_BUFFER, 0, sizeof(U16) * getNumIndices() - 1, (U8*) data, mMappedIndexData); } void LLVertexBuffer::setIndexData(const U32* data) { - llassert(sGLRenderIndices == mGLIndices); if (mIndicesType != GL_UNSIGNED_INT) { // HACK -- vertex buffers are initialized as 16-bit indices, but can be switched to 32-bit indices mIndicesType = GL_UNSIGNED_INT; mIndicesStride = 4; mNumIndices /= 2; } - flush_vbo(GL_ELEMENT_ARRAY_BUFFER, 0, sizeof(U32) * getNumIndices() - 1, (U8*)data); + flush_vbo(GL_ELEMENT_ARRAY_BUFFER, 0, sizeof(U32) * getNumIndices() - 1, (U8*)data, mMappedIndexData); +} + +void LLVertexBuffer::setPositionData(const LLVector4a* data, U32 offset, U32 count) +{ + flush_vbo(GL_ARRAY_BUFFER, offset * sizeof(LLVector4a), (offset + count) * sizeof(LLVector4a) - 1, (U8*)data, mMappedData); +} + +void LLVertexBuffer::setNormalData(const LLVector4a* data, U32 offset, U32 count) +{ + flush_vbo(GL_ARRAY_BUFFER, mOffsets[TYPE_NORMAL] + offset * sTypeSize[TYPE_NORMAL], mOffsets[TYPE_NORMAL] + (offset + count) * sTypeSize[TYPE_NORMAL] - 1, (U8*)data, mMappedData); +} + +void LLVertexBuffer::setTexCoord0Data(const LLVector2* data, U32 offset, U32 count) +{ + flush_vbo(GL_ARRAY_BUFFER, mOffsets[TYPE_TEXCOORD0] + offset * sTypeSize[TYPE_TEXCOORD0], mOffsets[TYPE_TEXCOORD0] + (offset + count) * sTypeSize[TYPE_TEXCOORD0] - 1, (U8*)data, mMappedData); +} + +void LLVertexBuffer::setTexCoord1Data(const LLVector2* data, U32 offset, U32 count) +{ + flush_vbo(GL_ARRAY_BUFFER, mOffsets[TYPE_TEXCOORD1] + offset * sTypeSize[TYPE_TEXCOORD1], mOffsets[TYPE_TEXCOORD1] + (offset + count) * sTypeSize[TYPE_TEXCOORD1] - 1, (U8*)data, mMappedData); +} + +void LLVertexBuffer::setColorData(const LLColor4U* data, U32 offset, U32 count) +{ + flush_vbo(GL_ARRAY_BUFFER, mOffsets[TYPE_COLOR] + offset * sTypeSize[TYPE_COLOR], mOffsets[TYPE_COLOR] + (offset + count) * sTypeSize[TYPE_COLOR] - 1, (U8*)data, mMappedData); +} + +void LLVertexBuffer::setTangentData(const LLVector4a* data, U32 offset, U32 count) +{ + flush_vbo(GL_ARRAY_BUFFER, mOffsets[TYPE_TANGENT] + offset * sTypeSize[TYPE_TANGENT], mOffsets[TYPE_TANGENT] + (offset + count) * sTypeSize[TYPE_TANGENT] - 1, (U8*)data, mMappedData); +} + +void LLVertexBuffer::setWeight4Data(const LLVector4a* data, U32 offset, U32 count) +{ + flush_vbo(GL_ARRAY_BUFFER, mOffsets[TYPE_WEIGHT4] + offset * sTypeSize[TYPE_WEIGHT4], mOffsets[TYPE_WEIGHT4] + (offset + count) * sTypeSize[TYPE_WEIGHT4] - 1, (U8*)data, mMappedData); +} + +void LLVertexBuffer::setJointData(const U64* data, U32 offset, U32 count) +{ + flush_vbo(GL_ARRAY_BUFFER, mOffsets[TYPE_JOINT] + offset * sTypeSize[TYPE_JOINT], mOffsets[TYPE_JOINT] + (offset + count) * sTypeSize[TYPE_JOINT] - 1, (U8*)data, mMappedData); +} + +void LLVertexBuffer::setIndexData(const U16* data, U32 offset, U32 count) +{ + flush_vbo(GL_ELEMENT_ARRAY_BUFFER, offset * sizeof(U16), (offset + count) * sizeof(U16) - 1, (U8*)data, mMappedIndexData); +} + +void LLVertexBuffer::setIndexData(const U32* data, U32 offset, U32 count) +{ + if (mIndicesType != GL_UNSIGNED_INT) + { // HACK -- vertex buffers are initialized as 16-bit indices, but can be switched to 32-bit indices + mIndicesType = GL_UNSIGNED_INT; + mIndicesStride = 4; + mNumIndices /= 2; + } + flush_vbo(GL_ELEMENT_ARRAY_BUFFER, offset * sizeof(U32), (offset + count) * sizeof(U32) - 1, (U8*)data, mMappedIndexData); } + + + diff --git a/indra/llrender/llvertexbuffer.h b/indra/llrender/llvertexbuffer.h index b634609929..375ad76fb8 100644 --- a/indra/llrender/llvertexbuffer.h +++ b/indra/llrender/llvertexbuffer.h @@ -38,6 +38,7 @@ #include <set> #include <vector> #include <list> +#include <glm/gtc/matrix_transform.hpp> #define LL_MAX_VERTEX_ATTRIB_LOCATION 64 @@ -53,6 +54,41 @@ //============================================================================ // base class class LLPrivateMemoryPool; +class LLVertexBuffer; + +class LLVertexBufferData +{ +public: + LLVertexBufferData() + : mVB(nullptr) + , mMode(0) + , mCount(0) + , mTexName(0) + , mProjection(glm::identity<glm::mat4>()) + , mModelView(glm::identity<glm::mat4>()) + , mTexture0(glm::identity<glm::mat4>()) + {} + LLVertexBufferData(LLVertexBuffer* buffer, U8 mode, U32 count, U32 tex_name, const glm::mat4& model_view, const glm::mat4& projection, const glm::mat4& texture0) + : mVB(buffer) + , mMode(mode) + , mCount(count) + , mTexName(tex_name) + , mProjection(model_view) + , mModelView(projection) + , mTexture0(texture0) + {} + void drawWithMatrix(); + void draw(); + LLPointer<LLVertexBuffer> mVB; + U8 mMode; + U32 mCount; + U32 mTexName; + glm::mat4 mProjection; + glm::mat4 mModelView; + glm::mat4 mTexture0; +}; +typedef std::list<LLVertexBufferData> buffer_data_list_t; + class LLVertexBuffer final : public LLRefCount { public: @@ -89,6 +125,9 @@ public: // indexed by the following enum static U32 calcOffsets(const U32& typemask, U32* offsets, U32 num_vertices); + // flush any pending mapped buffers + static void flushBuffers(); + //WARNING -- when updating these enums you MUST // 1 - update LLVertexBuffer::sTypeSize // 2 - update LLVertexBuffer::vb_type_name @@ -110,6 +149,7 @@ public: TYPE_WEIGHT, // "weight" TYPE_WEIGHT4, // "weight4" TYPE_CLOTHWEIGHT, // "clothing" + TYPE_JOINT, // "joint" TYPE_TEXTURE_INDEX, // "texture_index" TYPE_MAX, // TYPE_MAX is the size/boundary marker for attributes that go in the vertex buffer TYPE_INDEX, // TYPE_INDEX is beyond _MAX because it lives in a separate (index) buffer @@ -129,6 +169,7 @@ public: MAP_WEIGHT = (1<<TYPE_WEIGHT), MAP_WEIGHT4 = (1<<TYPE_WEIGHT4), MAP_CLOTHWEIGHT = (1<<TYPE_CLOTHWEIGHT), + MAP_JOINT = (1<<TYPE_JOINT), MAP_TEXTURE_INDEX = (1<<TYPE_TEXTURE_INDEX), }; @@ -157,17 +198,19 @@ public: // map for data access (see also getFooStrider below) U8* mapVertexBuffer(AttributeType type, U32 index, S32 count = -1); U8* mapIndexBuffer(U32 index, S32 count = -1); + + // synonym for flushBuffers void unmapBuffer(); // set for rendering // assumes (and will assert on) the following: - // - this buffer has no pending unampBuffer call + // - this buffer has no pending unmapBuffer call // - a shader is currently bound // - This buffer has sufficient attributes within it to satisfy the needs of the currently bound shader void setBuffer(); // Only call each getVertexPointer, etc, once before calling unmapBuffer() - // call unmapBuffer() after calls to getXXXStrider() before any cals to setBuffer() + // call unmapBuffer() after calls to getXXXStrider() before any calls to setBuffer() // example: // vb->getVertexBuffer(verts); // vb->getNormalStrider(norms); @@ -193,21 +236,35 @@ public: void setNormalData(const LLVector4a* data); void setTangentData(const LLVector4a* data); void setWeight4Data(const LLVector4a* data); - void setTexCoordData(const LLVector2* data); + void setJointData(const U64* data); + void setTexCoord0Data(const LLVector2* data); + void setTexCoord1Data(const LLVector2* data); void setColorData(const LLColor4U* data); void setIndexData(const U16* data); void setIndexData(const U32* data); + void setPositionData(const LLVector4a* data, U32 offset, U32 count); + void setNormalData(const LLVector4a* data, U32 offset, U32 count); + void setTangentData(const LLVector4a* data, U32 offset, U32 count); + void setWeight4Data(const LLVector4a* data, U32 offset, U32 count); + void setJointData(const U64* data, U32 offset, U32 count); + void setTexCoord0Data(const LLVector2* data, U32 offset, U32 count); + void setTexCoord1Data(const LLVector2* data, U32 offset, U32 count); + void setColorData(const LLColor4U* data, U32 offset, U32 count); + void setIndexData(const U16* data, U32 offset, U32 count); + void setIndexData(const U32* data, U32 offset, U32 count); + + U32 getNumVerts() const { return mNumVerts; } U32 getNumIndices() const { return mNumIndices; } U32 getTypeMask() const { return mTypeMask; } - bool hasDataType(AttributeType type) const { return ((1 << type) & getTypeMask()); } + bool hasDataType(AttributeType type) const { return ((1 << type) & getTypeMask()); } U32 getSize() const { return mSize; } U32 getIndicesSize() const { return mIndicesSize; } U8* getMappedData() const { return mMappedData; } U8* getMappedIndices() const { return mMappedIndexData; } - U32 getOffset(AttributeType type) const { return mOffsets[type]; } + U32 getOffset(AttributeType type) const { return mOffsets[type]; } // these functions assume (and assert on) the current VBO being bound // Detailed error checking can be enabled by setting gDebugGL to true @@ -215,6 +272,10 @@ public: void drawArrays(U32 mode, U32 offset, U32 count) const; void drawRange(U32 mode, U32 start, U32 end, U32 count, U32 indices_offset) const; + // draw without syncing matrices. If you're positive there have been no matrix + // since the last call to syncMatrices, this is much faster than drawRange + void drawRangeFast(U32 mode, U32 start, U32 end, U32 count, U32 indices_offset) const; + //for debugging, validate data in given range is valid bool validateRange(U32 start, U32 end, U32 count, U32 offset) const; @@ -222,6 +283,7 @@ public: void setLabel(const char* label); #endif + void clone(LLVertexBuffer& target) const; protected: U32 mGLBuffer = 0; // GL VBO handle @@ -252,11 +314,20 @@ private: friend class LLNavShapeVBOManager; friend class LLNavMeshVBOManager; + void flush_vbo(GLenum target, U32 start, U32 end, void* data, U8* dst); + LLVertexBuffer(U32 typemask, U32 usage) : LLVertexBuffer(typemask) {} - bool allocateBuffer(S32 nverts, S32 nindices, BOOL create) { return allocateBuffer(nverts, nindices); } + bool allocateBuffer(S32 nverts, S32 nindices, bool create) { return allocateBuffer(nverts, nindices); } + + // actually unmap buffer + void _unmapBuffer(); + + // add to set of mapped buffers + void _mapBuffer(); + bool mMapped = false; public: |