Revert "Avoid redundant uniform uploads (#2090)"

This reverts commit d9d9865ad9389ac17177b4c8aaded302be21ce38.

src/glimports.h

@@ -205,7 +205,6 @@ typedef void           (GL_APIENTRYP PFNGLUNIFORM1IVPROC) (GLint location, GLsiz
 typedef void           (GL_APIENTRYP PFNGLUNIFORM2FVPROC) (GLint location, GLsizei count, const GLfloat *value);
 typedef void           (GL_APIENTRYP PFNGLUNIFORM3FVPROC) (GLint location, GLsizei count, const GLfloat *value);
 typedef void           (GL_APIENTRYP PFNGLUNIFORM4FVPROC) (GLint location, GLsizei count, const GLfloat *value);
-typedef void           (GL_APIENTRYP PFNGLUNIFORM4FPROC) (GLint location, GLfloat x, GLfloat y, GLfloat z, GLfloat w);
 typedef void           (GL_APIENTRYP PFNGLUNIFORMMATRIX3FVPROC) (GLint location, GLsizei count, GLboolean transpose, const GLfloat *value);
 typedef void           (GL_APIENTRYP PFNGLUNIFORMMATRIX4FVPROC) (GLint location, GLsizei count, GLboolean transpose, const GLfloat *value);
 typedef void           (GL_APIENTRYP PFNGLUSEPROGRAMPROC) (GLuint program);
@@ -403,7 +402,6 @@ GL_IMPORT______(false, PFNGLUNIFORM1FVPROC,                        glUniform1fv)
 GL_IMPORT______(false, PFNGLUNIFORM2FVPROC,                        glUniform2fv);
 GL_IMPORT______(false, PFNGLUNIFORM3FVPROC,                        glUniform3fv);
 GL_IMPORT______(false, PFNGLUNIFORM4FVPROC,                        glUniform4fv);
-GL_IMPORT______(false, PFNGLUNIFORM4FPROC,                         glUniform4f);
 GL_IMPORT______(false, PFNGLUNIFORMMATRIX3FVPROC,                  glUniformMatrix3fv);
 GL_IMPORT______(false, PFNGLUNIFORMMATRIX4FVPROC,                  glUniformMatrix4fv);
 GL_IMPORT______(false, PFNGLUSEPROGRAMPROC,                        glUseProgram);

src/renderer_gl.cpp

@@ -2735,7 +2735,7 @@ namespace bgfx { namespace gl
 				}
 
 				m_vaoSupport = !BX_ENABLED(BX_PLATFORM_EMSCRIPTEN)
-				   && (!!(BGFX_CONFIG_RENDERER_OPENGLES >= 30)
+					&& (!!(BGFX_CONFIG_RENDERER_OPENGLES >= 30)
 						|| s_extension[Extension::ARB_vertex_array_object].m_supported
 						|| s_extension[Extension::OES_vertex_array_object].m_supported
 						);
@@ -3442,17 +3442,17 @@ namespace bgfx { namespace gl
 			GL_CHECK(glColorMask(GL_TRUE, GL_TRUE, GL_TRUE, GL_TRUE) );
 
 			ProgramGL& program = m_program[_blitter.m_program.idx];
-			GlUseProgram(program.m_id);
-			GlUniform1i(program.m_sampler[0].loc, 0);
+			GL_CHECK(glUseProgram(program.m_id) );
+			GL_CHECK(glUniform1i(program.m_sampler[0].loc, 0) );
 
 			float proj[16];
 			bx::mtxOrtho(proj, 0.0f, (float)width, (float)height, 0.0f, 0.0f, 1000.0f, 0.0f, g_caps.homogeneousDepth);
 
-			GlUniformMatrix4fv(program.m_predefined[0].m_loc
+			GL_CHECK(glUniformMatrix4fv(program.m_predefined[0].m_loc
 				, 1
 				, GL_FALSE
 				, proj
-				);
+				) );
 
 			GL_CHECK(glActiveTexture(GL_TEXTURE0) );
 			GL_CHECK(glBindTexture(GL_TEXTURE_2D, m_textures[_blitter.m_texture.idx].m_id) );
@@ -3553,19 +3553,19 @@ namespace bgfx { namespace gl
 
 		void setShaderUniform4f(uint8_t /*_flags*/, uint32_t _regIndex, const void* _val, uint32_t _numRegs)
 		{
-			GlUniform4fv(_regIndex
+			GL_CHECK(glUniform4fv(_regIndex
 				, _numRegs
 				, (const GLfloat*)_val
-				);
+				) );
 		}
 
 		void setShaderUniform4x4f(uint8_t /*_flags*/, uint32_t _regIndex, const void* _val, uint32_t _numRegs)
 		{
-			GlUniformMatrix4fv(_regIndex
+			GL_CHECK(glUniformMatrix4fv(_regIndex
 				, _numRegs
 				, GL_FALSE
 				, (const GLfloat*)_val
-				);
+				) );
 		}
 
 		uint32_t setFrameBuffer(FrameBufferHandle _fbh, uint32_t _height, uint16_t _discard = BGFX_CLEAR_NONE, bool _msaa = true)
@@ -4029,7 +4029,7 @@ namespace bgfx { namespace gl
 		case UniformType::_uniform: \
 				{ \
 					_type* value = (_type*)data; \
-					GlUniform##_glsuffix(loc, num, value); \
+					GL_CHECK(glUniform##_glsuffix(loc, num, value) ); \
 				} \
 				break;
 
@@ -4037,7 +4037,7 @@ namespace bgfx { namespace gl
 		case UniformType::_uniform: \
 				{ \
 					_type* value = (_type*)data; \
-					GlUniform##_glsuffix(loc, num, GL_FALSE, value); \
+					GL_CHECK(glUniform##_glsuffix(loc, num, GL_FALSE, value) ); \
 				} \
 				break;
 
@@ -4048,12 +4048,12 @@ namespace bgfx { namespace gl
 				// since they need to marshal an array over from Wasm to JS, so optimize the case when there is exactly one
 				// uniform to upload.
 				case UniformType::Sampler:
-					if (num > 1) GlUniform1iv(loc, num, (int*)data);
-					else GlUniform1i(loc, *(int*)data);
+					if (num > 1) glUniform1iv(loc, num, (int*)data);
+					else glUniform1i(loc, *(int*)data);
 					break;
 				case UniformType::Vec4:
-					if (num > 1) GlUniform4fv(loc, num, (float*)data);
-					else GlUniform4f(loc, ((float*)data)[0], ((float*)data)[1], ((float*)data)[2], ((float*)data)[3]);
+					if (num > 1) glUniform4fv(loc, num, (float*)data);
+					else glUniform4f(loc, ((float*)data)[0], ((float*)data)[1], ((float*)data)[2], ((float*)data)[3]);
 					break;
 #else
 				CASE_IMPLEMENT_UNIFORM(Sampler, 1iv, I, int);
@@ -4178,7 +4178,7 @@ namespace bgfx { namespace gl
 				GL_CHECK(glBindBuffer(GL_ARRAY_BUFFER, vb.m_id) );
 
 				ProgramGL& program = m_program[_clearQuad.m_program[numMrt-1].idx];
-				GlUseProgram(program.m_id);
+				GL_CHECK(glUseProgram(program.m_id) );
 				program.bindAttributesBegin();
 				program.bindAttributes(layout, 0);
 				program.bindAttributesEnd();
@@ -4237,95 +4237,6 @@ namespace bgfx { namespace gl
 			}
 		}
 
-		void GlUseProgram(GLuint program)
-		{
-			m_uniformStateCache.saveCurrentProgram(program);
-			GL_CHECK(glUseProgram(program) );
-		}
-
-		// Cache uniform uploads to avoid redundant uploading of state that is
-		// already set to a shader program
-		void GlUniform1i(uint32_t loc, int value)
-		{
-			if (m_uniformStateCache.updateUniformCache(loc, value))
-			{
-				GL_CHECK(glUniform1i(loc, value) );
-			}
-		}
-
-		void GlUniform1iv(uint32_t loc, int num, const int *data)
-		{
-			bool changed = false;
-			for(int i = 0; i < num; ++i)
-			{
-				if (m_uniformStateCache.updateUniformCache(loc+i, data[i]))
-				{
-					changed = true;
-				}
-			}
-			if (changed)
-			{
-				GL_CHECK(glUniform1iv(loc, num, data) );
-			}
-		}
-
-		void GlUniform4f(uint32_t loc, float x, float y, float z, float w)
-		{
-			UniformStateCache::f4 f; f.val[0] = x; f.val[1] = y; f.val[2] = z; f.val[3] = w;
-			if (m_uniformStateCache.updateUniformCache(loc, f))
-			{
-				GL_CHECK(glUniform4f(loc, x, y, z, w));
-			}
-		}
-
-		void GlUniform4fv(uint32_t loc, int num, const float *data)
-		{
-			bool changed = false;
-			for(int i = 0; i < num; ++i)
-			{
-				if (m_uniformStateCache.updateUniformCache(loc+i, *(const UniformStateCache::f4*)&data[4*i]))
-				{
-					changed = true;
-				}
-			}
-			if (changed)
-			{
-				GL_CHECK(glUniform4fv(loc, num, data) );
-			}
-		}
-
-		void GlUniformMatrix3fv(uint32_t loc, int num, GLboolean transpose, const float *data)
-		{
-			bool changed = false;
-			for(int i = 0; i < num; ++i)
-			{
-				if (m_uniformStateCache.updateUniformCache(loc+i, *(const UniformStateCache::f3x3*)&data[9*i]))
-				{
-					changed = true;
-				}
-			}
-			if (changed)
-			{
-				GL_CHECK(glUniformMatrix3fv(loc, num, transpose, data) );
-			}
-		}
-
-		void GlUniformMatrix4fv(uint32_t loc, int num, GLboolean transpose, const float *data)
-		{
-			bool changed = false;
-			for(int i = 0; i < num; ++i)
-			{
-				if (m_uniformStateCache.updateUniformCache(loc+i, *(const UniformStateCache::f4x4*)&data[16*i]))
-				{
-					changed = true;
-				}
-			}
-			if (changed)
-			{
-				GL_CHECK(glUniformMatrix4fv(loc, num, transpose, data) );
-			}
-		}
-
 		void* m_renderdocdll;
 
 		uint16_t m_numWindows;
@@ -4345,7 +4256,6 @@ namespace bgfx { namespace gl
 		OcclusionQueryGL m_occlusionQuery;
 
 		SamplerStateCache m_samplerStateCache;
-		UniformStateCache m_uniformStateCache;
 
 		TextVideoMem m_textVideoMem;
 		bool m_rtMsaa;
@@ -4754,7 +4664,7 @@ namespace bgfx { namespace gl
 
 		if (0 != m_id)
 		{
-			s_renderGL->GlUseProgram(0);
+			GL_CHECK(glUseProgram(0) );
 			GL_CHECK(glDeleteProgram(m_id) );
 			m_id = 0;
 		}
@@ -7245,7 +7155,7 @@ namespace bgfx { namespace gl
 						const RenderCompute& compute = renderItem.compute;
 
 						ProgramGL& program = m_program[key.m_program.idx];
-						GlUseProgram(program.m_id);
+						GL_CHECK(glUseProgram(program.m_id) );
 
 						GLbitfield barrier = 0;
 						for (uint32_t ii = 0; ii < maxComputeBindings; ++ii)
@@ -7743,7 +7653,7 @@ namespace bgfx { namespace gl
 					// Skip rendering if program index is valid, but program is invalid.
 					currentProgram = 0 == id ? ProgramHandle{kInvalidHandle} : currentProgram;
 
-					GlUseProgram(id);
+					GL_CHECK(glUseProgram(id) );
 					programChanged =
 						constantsChanged =
 						bindAttribs = true;

src/renderer_gl.h

@@ -32,10 +32,6 @@
 			|| BX_PLATFORM_WINDOWS \
 			)
 
-// Keep a state cache of GL uniform values to avoid redundant uploads
-// on the following platforms.
-#define BGFX_GL_CONFIG_UNIFORM_CACHE BX_PLATFORM_EMSCRIPTEN
-
 #define BGFX_GL_PROFILER_BEGIN(_view, _abgr)                                               \
 	BX_MACRO_BLOCK_BEGIN                                                                   \
 		GL_CHECK(glPushDebugGroup(GL_DEBUG_SOURCE_APPLICATION, 0, -1, s_viewName[view]) ); \
@@ -1226,69 +1222,6 @@ namespace bgfx { namespace gl
 #define GL_IMPORT(_optional, _proto, _func, _import) extern _proto _func
 #include "glimports.h"
 
-	class UniformStateCache
-	{
-	public:
-		struct f4 { float val[4]; bool operator ==(const f4 &rhs) { const uint64_t *a = (const uint64_t *)this; const uint64_t *b = (const uint64_t *)&rhs; return a[0] == b[0] && a[1] == b[1]; }};
-		struct f3x3 { float val[9]; bool operator ==(const f3x3 &rhs) { const uint64_t *a = (const uint64_t *)this; const uint64_t *b = (const uint64_t *)&rhs; return a[0] == b[0] && a[1] == b[1] && a[2] == b[2] && a[3] == b[3] && ((const uint32_t*)a)[8] == ((const uint32_t*)b)[8]; }};
-		struct f4x4 { float val[16]; bool operator ==(const f4x4 &rhs) { const uint64_t *a = (const uint64_t *)this; const uint64_t *b = (const uint64_t *)&rhs; return a[0] == b[0] && a[1] == b[1] && a[2] == b[2] && a[3] == b[3] && a[4] == b[4] && a[5] == b[5] && a[6] == b[6] && a[7] == b[7]; }};
-
-		UniformStateCache()
-		:currentProgram(0)
-		{}
-
-		// Inserts the new value into the uniform cache, and returns true
-		// if the old value was different than the new one.
-		template<typename T>
-		bool updateUniformCache(uint32_t loc, const T &value)
-		{
-#if BGFX_GL_CONFIG_UNIFORM_CACHE
-			// Uniform state cache for various types.
-			stl::unordered_map<uint64_t, T> &uniformCacheMap = getUniformCache<T>();
-
-			uint64_t key = ((uint64_t)currentProgram << 32) | loc;
-
-			auto iter = uniformCacheMap.find(key);
-			// Not found in the cache? Add it.
-			if (iter == uniformCacheMap.end())
-			{
-				uniformCacheMap[key] = value;
-				return true;
-			}
-			// Value in the cache was the same as new state? Skip reuploading this state.
-			if (iter->second == value)
-			{
-				return false;
-			}
-			iter->second = value;
-#endif
-			return true;
-		}
-
-		void saveCurrentProgram(GLuint program)
-		{
-#if BGFX_GL_CONFIG_UNIFORM_CACHE
-			currentProgram = program;
-#endif
-		}
-
-	private:
-		GLuint currentProgram;
-
-		stl::unordered_map<uint64_t, int> uniformiCacheMap;
-		stl::unordered_map<uint64_t, f4> uniformf4CacheMap;
-		stl::unordered_map<uint64_t, f3x3> uniformf3x3CacheMap;
-		stl::unordered_map<uint64_t, f4x4> uniformf4x4CacheMap;
-
-		template<typename T>
-		stl::unordered_map<uint64_t, T> &getUniformCache();
-	};
-
-	template<> stl::unordered_map<uint64_t, int> &UniformStateCache::getUniformCache() { return uniformiCacheMap; }
-	template<> stl::unordered_map<uint64_t, UniformStateCache::f4> &UniformStateCache::getUniformCache() { return uniformf4CacheMap; }
-	template<> stl::unordered_map<uint64_t, UniformStateCache::f3x3> &UniformStateCache::getUniformCache() { return uniformf3x3CacheMap; }
-	template<> stl::unordered_map<uint64_t, UniformStateCache::f4x4> &UniformStateCache::getUniformCache() { return uniformf4x4CacheMap; }
-
 	class SamplerStateCache
 	{
 	public: