mirror of
				https://github.com/PabloMK7/citra.git
				synced 2025-10-30 21:30:04 +00:00 
			
		
		
		
	gl_rasterizer: only apply AMD hack when the vendor is AMD
This commit is contained in:
		
							parent
							
								
									05eb0ccb03
								
							
						
					
					
						commit
						266f2b6242
					
				
					 6 changed files with 34 additions and 18 deletions
				
			
		|  | @ -34,11 +34,17 @@ MICROPROFILE_DEFINE(OpenGL_Drawing, "OpenGL", "Drawing", MP_RGB(128, 128, 192)); | ||||||
| MICROPROFILE_DEFINE(OpenGL_Blits, "OpenGL", "Blits", MP_RGB(100, 100, 255)); | MICROPROFILE_DEFINE(OpenGL_Blits, "OpenGL", "Blits", MP_RGB(100, 100, 255)); | ||||||
| MICROPROFILE_DEFINE(OpenGL_CacheManagement, "OpenGL", "Cache Mgmt", MP_RGB(100, 255, 100)); | MICROPROFILE_DEFINE(OpenGL_CacheManagement, "OpenGL", "Cache Mgmt", MP_RGB(100, 255, 100)); | ||||||
| 
 | 
 | ||||||
|  | static bool IsVendorAmd() { | ||||||
|  |     std::string gpu_vendor{reinterpret_cast<char const*>(glGetString(GL_VENDOR))}; | ||||||
|  |     return gpu_vendor == "ATI Technologies Inc." || gpu_vendor == "Advanced Micro Devices, Inc."; | ||||||
|  | } | ||||||
|  | 
 | ||||||
| RasterizerOpenGL::RasterizerOpenGL() | RasterizerOpenGL::RasterizerOpenGL() | ||||||
|     : shader_dirty(true), vertex_buffer(GL_ARRAY_BUFFER, VERTEX_BUFFER_SIZE), |     : is_amd(IsVendorAmd()), shader_dirty(true), | ||||||
|       uniform_buffer(GL_UNIFORM_BUFFER, UNIFORM_BUFFER_SIZE), |       vertex_buffer(GL_ARRAY_BUFFER, VERTEX_BUFFER_SIZE, is_amd), | ||||||
|       index_buffer(GL_ELEMENT_ARRAY_BUFFER, INDEX_BUFFER_SIZE), |       uniform_buffer(GL_UNIFORM_BUFFER, UNIFORM_BUFFER_SIZE, false), | ||||||
|       texture_buffer(GL_TEXTURE_BUFFER, TEXTURE_BUFFER_SIZE) { |       index_buffer(GL_ELEMENT_ARRAY_BUFFER, INDEX_BUFFER_SIZE, false), | ||||||
|  |       texture_buffer(GL_TEXTURE_BUFFER, TEXTURE_BUFFER_SIZE, false) { | ||||||
| 
 | 
 | ||||||
|     allow_shadow = GLAD_GL_ARB_shader_image_load_store && GLAD_GL_ARB_shader_image_size && |     allow_shadow = GLAD_GL_ARB_shader_image_load_store && GLAD_GL_ARB_shader_image_size && | ||||||
|                    GLAD_GL_ARB_framebuffer_no_attachments; |                    GLAD_GL_ARB_framebuffer_no_attachments; | ||||||
|  | @ -145,7 +151,7 @@ RasterizerOpenGL::RasterizerOpenGL() | ||||||
|     glBindBuffer(GL_ELEMENT_ARRAY_BUFFER, index_buffer.GetHandle()); |     glBindBuffer(GL_ELEMENT_ARRAY_BUFFER, index_buffer.GetHandle()); | ||||||
| 
 | 
 | ||||||
|     shader_program_manager = |     shader_program_manager = | ||||||
|         std::make_unique<ShaderProgramManager>(GLAD_GL_ARB_separate_shader_objects); |         std::make_unique<ShaderProgramManager>(GLAD_GL_ARB_separate_shader_objects, is_amd); | ||||||
| 
 | 
 | ||||||
|     glEnable(GL_BLEND); |     glEnable(GL_BLEND); | ||||||
| 
 | 
 | ||||||
|  |  | ||||||
|  | @ -243,6 +243,8 @@ private: | ||||||
|     /// Setup geometry shader for AccelerateDrawBatch
 |     /// Setup geometry shader for AccelerateDrawBatch
 | ||||||
|     bool SetupGeometryShader(); |     bool SetupGeometryShader(); | ||||||
| 
 | 
 | ||||||
|  |     bool is_amd; | ||||||
|  | 
 | ||||||
|     OpenGLState state; |     OpenGLState state; | ||||||
| 
 | 
 | ||||||
|     RasterizerCacheOpenGL res_cache; |     RasterizerCacheOpenGL res_cache; | ||||||
|  |  | ||||||
|  | @ -216,8 +216,8 @@ using FragmentShaders = | ||||||
| 
 | 
 | ||||||
| class ShaderProgramManager::Impl { | class ShaderProgramManager::Impl { | ||||||
| public: | public: | ||||||
|     explicit Impl(bool separable) |     explicit Impl(bool separable, bool is_amd) | ||||||
|         : separable(separable), programmable_vertex_shaders(separable), |         : is_amd(is_amd), separable(separable), programmable_vertex_shaders(separable), | ||||||
|           trivial_vertex_shader(separable), programmable_geometry_shaders(separable), |           trivial_vertex_shader(separable), programmable_geometry_shaders(separable), | ||||||
|           fixed_geometry_shaders(separable), fragment_shaders(separable) { |           fixed_geometry_shaders(separable), fragment_shaders(separable) { | ||||||
|         if (separable) |         if (separable) | ||||||
|  | @ -248,6 +248,8 @@ public: | ||||||
|         }; |         }; | ||||||
|     }; |     }; | ||||||
| 
 | 
 | ||||||
|  |     bool is_amd; | ||||||
|  | 
 | ||||||
|     ShaderTuple current; |     ShaderTuple current; | ||||||
| 
 | 
 | ||||||
|     ProgrammableVertexShaders programmable_vertex_shaders; |     ProgrammableVertexShaders programmable_vertex_shaders; | ||||||
|  | @ -263,8 +265,8 @@ public: | ||||||
|     OGLPipeline pipeline; |     OGLPipeline pipeline; | ||||||
| }; | }; | ||||||
| 
 | 
 | ||||||
| ShaderProgramManager::ShaderProgramManager(bool separable) | ShaderProgramManager::ShaderProgramManager(bool separable, bool is_amd) | ||||||
|     : impl(std::make_unique<Impl>(separable)) {} |     : impl(std::make_unique<Impl>(separable, is_amd)) {} | ||||||
| 
 | 
 | ||||||
| ShaderProgramManager::~ShaderProgramManager() = default; | ShaderProgramManager::~ShaderProgramManager() = default; | ||||||
| 
 | 
 | ||||||
|  | @ -304,11 +306,15 @@ void ShaderProgramManager::UseFragmentShader(const GLShader::PicaFSConfig& confi | ||||||
| 
 | 
 | ||||||
| void ShaderProgramManager::ApplyTo(OpenGLState& state) { | void ShaderProgramManager::ApplyTo(OpenGLState& state) { | ||||||
|     if (impl->separable) { |     if (impl->separable) { | ||||||
|         // Without this reseting, AMD sometimes freezes when one stage is changed but not for the
 |         if (impl->is_amd) { | ||||||
|         // others
 |             // Without this reseting, AMD sometimes freezes when one stage is changed but not for
 | ||||||
|         glUseProgramStages(impl->pipeline.handle, |             // the others.
 | ||||||
|                            GL_VERTEX_SHADER_BIT | GL_GEOMETRY_SHADER_BIT | GL_FRAGMENT_SHADER_BIT, |             // On the other hand, including this reset seems to introduce memory leak in Intel
 | ||||||
|                            0); |             // Graphics.
 | ||||||
|  |             glUseProgramStages( | ||||||
|  |                 impl->pipeline.handle, | ||||||
|  |                 GL_VERTEX_SHADER_BIT | GL_GEOMETRY_SHADER_BIT | GL_FRAGMENT_SHADER_BIT, 0); | ||||||
|  |         } | ||||||
| 
 | 
 | ||||||
|         glUseProgramStages(impl->pipeline.handle, GL_VERTEX_SHADER_BIT, impl->current.vs); |         glUseProgramStages(impl->pipeline.handle, GL_VERTEX_SHADER_BIT, impl->current.vs); | ||||||
|         glUseProgramStages(impl->pipeline.handle, GL_GEOMETRY_SHADER_BIT, impl->current.gs); |         glUseProgramStages(impl->pipeline.handle, GL_GEOMETRY_SHADER_BIT, impl->current.gs); | ||||||
|  |  | ||||||
|  | @ -99,7 +99,7 @@ static_assert(sizeof(GSUniformData) < 16384, | ||||||
| /// A class that manage different shader stages and configures them with given config data.
 | /// A class that manage different shader stages and configures them with given config data.
 | ||||||
| class ShaderProgramManager { | class ShaderProgramManager { | ||||||
| public: | public: | ||||||
|     explicit ShaderProgramManager(bool separable); |     ShaderProgramManager(bool separable, bool is_amd); | ||||||
|     ~ShaderProgramManager(); |     ~ShaderProgramManager(); | ||||||
| 
 | 
 | ||||||
|     bool UseProgrammableVertexShader(const GLShader::PicaVSConfig& config, |     bool UseProgrammableVertexShader(const GLShader::PicaVSConfig& config, | ||||||
|  |  | ||||||
|  | @ -9,13 +9,14 @@ | ||||||
| #include "video_core/renderer_opengl/gl_state.h" | #include "video_core/renderer_opengl/gl_state.h" | ||||||
| #include "video_core/renderer_opengl/gl_stream_buffer.h" | #include "video_core/renderer_opengl/gl_stream_buffer.h" | ||||||
| 
 | 
 | ||||||
| OGLStreamBuffer::OGLStreamBuffer(GLenum target, GLsizeiptr size, bool prefer_coherent) | OGLStreamBuffer::OGLStreamBuffer(GLenum target, GLsizeiptr size, bool array_buffer_for_amd, | ||||||
|  |                                  bool prefer_coherent) | ||||||
|     : gl_target(target), buffer_size(size) { |     : gl_target(target), buffer_size(size) { | ||||||
|     gl_buffer.Create(); |     gl_buffer.Create(); | ||||||
|     glBindBuffer(gl_target, gl_buffer.handle); |     glBindBuffer(gl_target, gl_buffer.handle); | ||||||
| 
 | 
 | ||||||
|     GLsizeiptr allocate_size = size; |     GLsizeiptr allocate_size = size; | ||||||
|     if (target == GL_ARRAY_BUFFER) { |     if (array_buffer_for_amd) { | ||||||
|         // On AMD GPU there is a strange crash in indexed drawing. The crash happens when the buffer
 |         // On AMD GPU there is a strange crash in indexed drawing. The crash happens when the buffer
 | ||||||
|         // read position is near the end and is an out-of-bound access to the vertex buffer. This is
 |         // read position is near the end and is an out-of-bound access to the vertex buffer. This is
 | ||||||
|         // probably a bug in the driver and is related to the usage of vec3<byte> attributes in the
 |         // probably a bug in the driver and is related to the usage of vec3<byte> attributes in the
 | ||||||
|  |  | ||||||
|  | @ -9,7 +9,8 @@ | ||||||
| 
 | 
 | ||||||
| class OGLStreamBuffer : private NonCopyable { | class OGLStreamBuffer : private NonCopyable { | ||||||
| public: | public: | ||||||
|     explicit OGLStreamBuffer(GLenum target, GLsizeiptr size, bool prefer_coherent = false); |     explicit OGLStreamBuffer(GLenum target, GLsizeiptr size, bool array_buffer_for_amd, | ||||||
|  |                              bool prefer_coherent = false); | ||||||
|     ~OGLStreamBuffer(); |     ~OGLStreamBuffer(); | ||||||
| 
 | 
 | ||||||
|     GLuint GetHandle() const; |     GLuint GetHandle() const; | ||||||
|  |  | ||||||
		Loading…
	
	Add table
		Add a link
		
	
		Reference in a new issue