From 491c10ec96e4cb58724f0768a76e0cece63f2b07 Mon Sep 17 00:00:00 2001 From: Markus Wick Date: Fri, 4 Aug 2017 17:56:24 +0200 Subject: [PATCH] VideoBackends: Use VideoCommon shader generators for efb2tex copies. This will generate one shader per copy format. For now, it is the same shader with the colmat hard coded. So it should already improve the GPU performance a bit, but a rewrite of the shader generator is suggested. Half of the patch is done by linkmauve1: VideoCommon: Reorganise the shader writes. --- .../Core/VideoBackends/D3D/TextureCache.cpp | 3 +- Source/Core/VideoBackends/D3D/TextureCache.h | 3 +- Source/Core/VideoBackends/Null/TextureCache.h | 3 +- .../Core/VideoBackends/OGL/TextureCache.cpp | 177 +++++----- Source/Core/VideoBackends/OGL/TextureCache.h | 23 +- .../VideoBackends/Software/TextureCache.h | 3 +- .../VideoBackends/Vulkan/TextureCache.cpp | 111 ++----- .../Core/VideoBackends/Vulkan/TextureCache.h | 8 +- Source/Core/VideoCommon/CMakeLists.txt | 1 + Source/Core/VideoCommon/TextureCacheBase.cpp | 3 +- Source/Core/VideoCommon/TextureCacheBase.h | 3 +- .../VideoCommon/TextureConverterShaderGen.cpp | 301 ++++++++++++++++++ .../VideoCommon/TextureConverterShaderGen.h | 32 ++ Source/Core/VideoCommon/VideoCommon.vcxproj | 2 + .../VideoCommon/VideoCommon.vcxproj.filters | 6 + 15 files changed, 468 insertions(+), 211 deletions(-) create mode 100644 Source/Core/VideoCommon/TextureConverterShaderGen.cpp create mode 100644 Source/Core/VideoCommon/TextureConverterShaderGen.h diff --git a/Source/Core/VideoBackends/D3D/TextureCache.cpp b/Source/Core/VideoBackends/D3D/TextureCache.cpp index 7d9d62cf54..a7eab7c041 100644 --- a/Source/Core/VideoBackends/D3D/TextureCache.cpp +++ b/Source/Core/VideoBackends/D3D/TextureCache.cpp @@ -222,7 +222,8 @@ TextureCache::~TextureCache() void TextureCache::CopyEFBToCacheEntry(TCacheEntry* entry, bool is_depth_copy, const EFBRectangle& src_rect, bool scale_by_half, - unsigned int cbuf_id, const float* colmat) + unsigned int cbuf_id, const float* colmat, + EFBCopyFormat dst_format, bool is_intensity) { auto* destination_texture = static_cast(entry->texture.get()); diff --git a/Source/Core/VideoBackends/D3D/TextureCache.h b/Source/Core/VideoBackends/D3D/TextureCache.h index cfe2efdf65..d650c04d4c 100644 --- a/Source/Core/VideoBackends/D3D/TextureCache.h +++ b/Source/Core/VideoBackends/D3D/TextureCache.h @@ -34,7 +34,8 @@ private: bool scale_by_half) override; void CopyEFBToCacheEntry(TCacheEntry* entry, bool is_depth_copy, const EFBRectangle& src_rect, - bool scale_by_half, unsigned int cbuf_id, const float* colmat) override; + bool scale_by_half, unsigned int cbuf_id, const float* colmat, + EFBCopyFormat dst_format, bool is_intensity) override; bool CompileShaders() override { return true; } void DeleteShaders() override {} diff --git a/Source/Core/VideoBackends/Null/TextureCache.h b/Source/Core/VideoBackends/Null/TextureCache.h index 6f09108392..c10b6eaaf8 100644 --- a/Source/Core/VideoBackends/Null/TextureCache.h +++ b/Source/Core/VideoBackends/Null/TextureCache.h @@ -32,7 +32,8 @@ public: } void CopyEFBToCacheEntry(TCacheEntry* entry, bool is_depth_copy, const EFBRectangle& src_rect, - bool scale_by_half, unsigned int cbuf_id, const float* colmat) override + bool scale_by_half, unsigned int cbuf_id, const float* colmat, + EFBCopyFormat dst_format, bool is_intensity) override { } }; diff --git a/Source/Core/VideoBackends/OGL/TextureCache.cpp b/Source/Core/VideoBackends/OGL/TextureCache.cpp index ba945f8043..f370da7142 100644 --- a/Source/Core/VideoBackends/OGL/TextureCache.cpp +++ b/Source/Core/VideoBackends/OGL/TextureCache.cpp @@ -26,11 +26,44 @@ #include "VideoCommon/ImageWrite.h" #include "VideoCommon/TextureConversionShader.h" +#include "VideoCommon/TextureConverterShaderGen.h" #include "VideoCommon/TextureDecoder.h" +#include "VideoCommon/VideoCommon.h" #include "VideoCommon/VideoConfig.h" namespace OGL { +constexpr const char* vertex_program = + "out vec3 %c_uv0;\n" + "SAMPLER_BINDING(9) uniform sampler2DArray samp9;\n" + "uniform vec4 copy_position;\n" // left, top, right, bottom + "void main()\n" + "{\n" + " vec2 rawpos = vec2(gl_VertexID&1, gl_VertexID&2);\n" + " %c_uv0 = vec3(mix(copy_position.xy, copy_position.zw, rawpos) / vec2(textureSize(samp9, " + "0).xy), 0.0);\n" + " gl_Position = vec4(rawpos*2.0-1.0, 0.0, 1.0);\n" + "}\n"; + +constexpr const char* geometry_program = "layout(triangles) in;\n" + "layout(triangle_strip, max_vertices = 6) out;\n" + "in vec3 v_uv0[3];\n" + "out vec3 f_uv0;\n" + "SAMPLER_BINDING(9) uniform sampler2DArray samp9;\n" + "void main()\n" + "{\n" + " int layers = textureSize(samp9, 0).z;\n" + " for (int layer = 0; layer < layers; ++layer) {\n" + " for (int i = 0; i < 3; ++i) {\n" + " f_uv0 = vec3(v_uv0[i].xy, layer);\n" + " gl_Position = gl_in[i].gl_Position;\n" + " gl_Layer = layer;\n" + " EmitVertex();\n" + " }\n" + " EndPrimitive();\n" + " }\n" + "}\n"; + //#define TIME_TEXTURE_DECODING 1 void TextureCache::CopyEFB(u8* dst, const EFBCopyParams& params, u32 native_width, @@ -124,101 +157,22 @@ bool TextureCache::CompileShaders() " ocol0 = texcol;\n" "}\n"; - constexpr const char* color_matrix_program = - "SAMPLER_BINDING(9) uniform sampler2DArray samp9;\n" - "uniform vec4 colmat[7];\n" - "in vec3 f_uv0;\n" - "out vec4 ocol0;\n" - "\n" - "void main(){\n" - " vec4 texcol = texture(samp9, f_uv0);\n" - " texcol = floor(texcol * colmat[5]) * colmat[6];\n" - " ocol0 = texcol * mat4(colmat[0], colmat[1], colmat[2], colmat[3]) + colmat[4];\n" - "}\n"; - - constexpr const char* depth_matrix_program = - "SAMPLER_BINDING(9) uniform sampler2DArray samp9;\n" - "uniform vec4 colmat[5];\n" - "in vec3 f_uv0;\n" - "out vec4 ocol0;\n" - "\n" - "void main(){\n" - " vec4 texcol = texture(samp9, vec3(f_uv0.xy, %s));\n" - " int depth = int(texcol.x * 16777216.0);\n" - - // Convert to Z24 format - " ivec4 workspace;\n" - " workspace.r = (depth >> 16) & 255;\n" - " workspace.g = (depth >> 8) & 255;\n" - " workspace.b = depth & 255;\n" - - // Convert to Z4 format - " workspace.a = (depth >> 16) & 0xF0;\n" - - // Normalize components to [0.0..1.0] - " texcol = vec4(workspace) / 255.0;\n" - - " ocol0 = texcol * mat4(colmat[0], colmat[1], colmat[2], colmat[3]) + colmat[4];\n" - "}\n"; - - constexpr const char* vertex_program = - "out vec3 %s_uv0;\n" - "SAMPLER_BINDING(9) uniform sampler2DArray samp9;\n" - "uniform vec4 copy_position;\n" // left, top, right, bottom - "void main()\n" - "{\n" - " vec2 rawpos = vec2(gl_VertexID&1, gl_VertexID&2);\n" - " %s_uv0 = vec3(mix(copy_position.xy, copy_position.zw, rawpos) / vec2(textureSize(samp9, " - "0).xy), 0.0);\n" - " gl_Position = vec4(rawpos*2.0-1.0, 0.0, 1.0);\n" - "}\n"; - - const std::string geo_program = g_ActiveConfig.stereo_mode != StereoMode::Off ? - "layout(triangles) in;\n" - "layout(triangle_strip, max_vertices = 6) out;\n" - "in vec3 v_uv0[3];\n" - "out vec3 f_uv0;\n" - "SAMPLER_BINDING(9) uniform sampler2DArray samp9;\n" - "void main()\n" - "{\n" - " int layers = textureSize(samp9, 0).z;\n" - " for (int layer = 0; layer < layers; ++layer) {\n" - " for (int i = 0; i < 3; ++i) {\n" - " f_uv0 = vec3(v_uv0[i].xy, layer);\n" - " gl_Position = gl_in[i].gl_Position;\n" - " gl_Layer = layer;\n" - " EmitVertex();\n" - " }\n" - " EndPrimitive();\n" - " }\n" - "}\n" : - ""; - - const char* prefix = geo_program.empty() ? "f" : "v"; - const char* depth_layer = g_ActiveConfig.bStereoEFBMonoDepth ? "0.0" : "f_uv0.z"; + std::string geo_program = ""; + char prefix = 'f'; + if (g_ActiveConfig.stereo_mode != StereoMode::Off) + { + geo_program = geometry_program; + prefix = 'v'; + } if (!ProgramShaderCache::CompileShader(m_colorCopyProgram, StringFromFormat(vertex_program, prefix, prefix), - color_copy_program, geo_program) || - !ProgramShaderCache::CompileShader(m_colorMatrixProgram, - StringFromFormat(vertex_program, prefix, prefix), - color_matrix_program, geo_program) || - !ProgramShaderCache::CompileShader( - m_depthMatrixProgram, StringFromFormat(vertex_program, prefix, prefix), - StringFromFormat(depth_matrix_program, depth_layer), geo_program)) + color_copy_program, geo_program)) { return false; } - m_colorMatrixUniform = glGetUniformLocation(m_colorMatrixProgram.glprogid, "colmat"); - m_depthMatrixUniform = glGetUniformLocation(m_depthMatrixProgram.glprogid, "colmat"); - m_color_cbuf_id = UINT_MAX; - m_depth_cbuf_id = UINT_MAX; - m_colorCopyPositionUniform = glGetUniformLocation(m_colorCopyProgram.glprogid, "copy_position"); - m_colorMatrixPositionUniform = - glGetUniformLocation(m_colorMatrixProgram.glprogid, "copy_position"); - m_depthCopyPositionUniform = glGetUniformLocation(m_depthMatrixProgram.glprogid, "copy_position"); std::string palette_shader = R"GLSL( @@ -320,8 +274,9 @@ bool TextureCache::CompileShaders() void TextureCache::DeleteShaders() { - m_colorMatrixProgram.Destroy(); - m_depthMatrixProgram.Destroy(); + for (auto& it : m_efb_copy_programs) + it.second.shader.Destroy(); + m_efb_copy_programs.clear(); if (g_ActiveConfig.backend_info.bSupportsPaletteConversion) for (auto& shader : m_palette_shaders) @@ -531,7 +486,8 @@ void TextureCache::DecodeTextureOnGPU(TCacheEntry* entry, u32 dst_level, const u void TextureCache::CopyEFBToCacheEntry(TCacheEntry* entry, bool is_depth_copy, const EFBRectangle& src_rect, bool scale_by_half, - unsigned int cbuf_id, const float* colmat) + unsigned int cbuf_id, const float* colmat, + EFBCopyFormat dst_format, bool is_intensity) { auto* destination_texture = static_cast(entry->texture.get()); g_renderer->ResetAPIState(); // reset any game specific settings @@ -554,26 +510,35 @@ void TextureCache::CopyEFBToCacheEntry(TCacheEntry* entry, bool is_depth_copy, glViewport(0, 0, destination_texture->GetConfig().width, destination_texture->GetConfig().height); - GLuint uniform_location; - if (is_depth_copy) + auto uid = GetTextureConverterShaderUid(dst_format, is_depth_copy, is_intensity, scale_by_half); + + auto it = m_efb_copy_programs.emplace(uid, EFBCopyShader()); + EFBCopyShader& shader = it.first->second; + bool created = it.second; + + if (created) { - m_depthMatrixProgram.Bind(); - if (m_depth_cbuf_id != cbuf_id) - glUniform4fv(m_depthMatrixUniform, 5, colmat); - m_depth_cbuf_id = cbuf_id; - uniform_location = m_depthCopyPositionUniform; - } - else - { - m_colorMatrixProgram.Bind(); - if (m_color_cbuf_id != cbuf_id) - glUniform4fv(m_colorMatrixUniform, 7, colmat); - m_color_cbuf_id = cbuf_id; - uniform_location = m_colorMatrixPositionUniform; + ShaderCode code = GenerateTextureConverterShaderCode(APIType::OpenGL, uid.GetUidData()); + + std::string geo_program = ""; + char prefix = 'f'; + if (g_ActiveConfig.stereo_mode != StereoMode::Off) + { + geo_program = geometry_program; + prefix = 'v'; + } + + ProgramShaderCache::CompileShader(shader.shader, + StringFromFormat(vertex_program, prefix, prefix), + code.GetBuffer(), geo_program); + + shader.position_uniform = glGetUniformLocation(shader.shader.glprogid, "copy_position"); } + shader.shader.Bind(); + TargetRectangle R = g_renderer->ConvertEFBRectangle(src_rect); - glUniform4f(uniform_location, static_cast(R.left), static_cast(R.top), + glUniform4f(shader.position_uniform, static_cast(R.left), static_cast(R.top), static_cast(R.right), static_cast(R.bottom)); glDrawArrays(GL_TRIANGLE_STRIP, 0, 4); diff --git a/Source/Core/VideoBackends/OGL/TextureCache.h b/Source/Core/VideoBackends/OGL/TextureCache.h index 50366f7d1a..f0ab838c23 100644 --- a/Source/Core/VideoBackends/OGL/TextureCache.h +++ b/Source/Core/VideoBackends/OGL/TextureCache.h @@ -13,6 +13,7 @@ #include "VideoCommon/TextureCacheBase.h" #include "VideoCommon/TextureConversionShader.h" +#include "VideoCommon/TextureConverterShaderGen.h" #include "VideoCommon/VideoCommon.h" class AbstractTexture; @@ -67,7 +68,8 @@ private: bool scale_by_half) override; void CopyEFBToCacheEntry(TCacheEntry* entry, bool is_depth_copy, const EFBRectangle& src_rect, - bool scale_by_half, unsigned int cbuf_id, const float* colmat) override; + bool scale_by_half, unsigned int cbuf_id, const float* colmat, + EFBCopyFormat dst_format, bool is_intensity) override; bool CompileShaders() override; void DeleteShaders() override; @@ -78,17 +80,16 @@ private: void CreateTextureDecodingResources(); void DestroyTextureDecodingResources(); - SHADER m_colorCopyProgram; - SHADER m_colorMatrixProgram; - SHADER m_depthMatrixProgram; - GLuint m_colorMatrixUniform; - GLuint m_depthMatrixUniform; - GLuint m_colorCopyPositionUniform; - GLuint m_colorMatrixPositionUniform; - GLuint m_depthCopyPositionUniform; + struct EFBCopyShader + { + SHADER shader; + GLuint position_uniform; + }; - u32 m_color_cbuf_id; - u32 m_depth_cbuf_id; + std::map m_efb_copy_programs; + + SHADER m_colorCopyProgram; + GLuint m_colorCopyPositionUniform; std::array m_palette_shaders; std::unique_ptr m_palette_stream_buffer; diff --git a/Source/Core/VideoBackends/Software/TextureCache.h b/Source/Core/VideoBackends/Software/TextureCache.h index 49612394c2..d676d67a6e 100644 --- a/Source/Core/VideoBackends/Software/TextureCache.h +++ b/Source/Core/VideoBackends/Software/TextureCache.h @@ -26,7 +26,8 @@ public: private: void CopyEFBToCacheEntry(TCacheEntry* entry, bool is_depth_copy, const EFBRectangle& src_rect, - bool scale_by_half, unsigned int cbuf_id, const float* colmat) override + bool scale_by_half, unsigned int cbuf_id, const float* colmat, + EFBCopyFormat dst_format, bool is_intensity) override { // TODO: If we ever want to "fake" vram textures, we would need to implement this } diff --git a/Source/Core/VideoBackends/Vulkan/TextureCache.cpp b/Source/Core/VideoBackends/Vulkan/TextureCache.cpp index 5a396ee4b2..130a494bb2 100644 --- a/Source/Core/VideoBackends/Vulkan/TextureCache.cpp +++ b/Source/Core/VideoBackends/Vulkan/TextureCache.cpp @@ -237,80 +237,12 @@ bool TextureCache::CompileShaders() } )"; - static const char EFB_COLOR_TO_TEX_SOURCE[] = R"( - SAMPLER_BINDING(0) uniform sampler2DArray samp0; - - layout(std140, push_constant) uniform PSBlock - { - vec4 colmat[7]; - } C; - - layout(location = 0) in vec3 uv0; - layout(location = 1) in vec4 col0; - layout(location = 0) out vec4 ocol0; - - void main() - { - float4 texcol = texture(samp0, uv0); - texcol = floor(texcol * C.colmat[5]) * C.colmat[6]; - ocol0 = texcol * mat4(C.colmat[0], C.colmat[1], C.colmat[2], C.colmat[3]) + C.colmat[4]; - } - )"; - - static const char EFB_DEPTH_TO_TEX_SOURCE[] = R"( - SAMPLER_BINDING(0) uniform sampler2DArray samp0; - - layout(std140, push_constant) uniform PSBlock - { - vec4 colmat[5]; - } C; - - layout(location = 0) in vec3 uv0; - layout(location = 1) in vec4 col0; - layout(location = 0) out vec4 ocol0; - - void main() - { - #if MONO_DEPTH - vec4 texcol = texture(samp0, vec3(uv0.xy, 0.0f)); - #else - vec4 texcol = texture(samp0, uv0); - #endif - int depth = int((1.0 - texcol.x) * 16777216.0); - - // Convert to Z24 format - ivec4 workspace; - workspace.r = (depth >> 16) & 255; - workspace.g = (depth >> 8) & 255; - workspace.b = depth & 255; - - // Convert to Z4 format - workspace.a = (depth >> 16) & 0xF0; - - // Normalize components to [0.0..1.0] - texcol = vec4(workspace) / 255.0; - - ocol0 = texcol * mat4(C.colmat[0], C.colmat[1], C.colmat[2], C.colmat[3]) + C.colmat[4]; - } - )"; - std::string header = g_shader_cache->GetUtilityShaderHeader(); - std::string source; + std::string source = header + COPY_SHADER_SOURCE; - source = header + COPY_SHADER_SOURCE; m_copy_shader = Util::CompileAndCreateFragmentShader(source); - source = header + EFB_COLOR_TO_TEX_SOURCE; - m_efb_color_to_tex_shader = Util::CompileAndCreateFragmentShader(source); - - if (g_ActiveConfig.bStereoEFBMonoDepth) - source = header + "#define MONO_DEPTH 1\n" + EFB_DEPTH_TO_TEX_SOURCE; - else - source = header + EFB_DEPTH_TO_TEX_SOURCE; - m_efb_depth_to_tex_shader = Util::CompileAndCreateFragmentShader(source); - - return m_copy_shader != VK_NULL_HANDLE && m_efb_color_to_tex_shader != VK_NULL_HANDLE && - m_efb_depth_to_tex_shader != VK_NULL_HANDLE; + return m_copy_shader != VK_NULL_HANDLE; } void TextureCache::DeleteShaders() @@ -324,21 +256,18 @@ void TextureCache::DeleteShaders() vkDestroyShaderModule(g_vulkan_context->GetDevice(), m_copy_shader, nullptr); m_copy_shader = VK_NULL_HANDLE; } - if (m_efb_color_to_tex_shader != VK_NULL_HANDLE) + + for (auto& shader : m_efb_copy_to_tex_shaders) { - vkDestroyShaderModule(g_vulkan_context->GetDevice(), m_efb_color_to_tex_shader, nullptr); - m_efb_color_to_tex_shader = VK_NULL_HANDLE; - } - if (m_efb_depth_to_tex_shader != VK_NULL_HANDLE) - { - vkDestroyShaderModule(g_vulkan_context->GetDevice(), m_efb_depth_to_tex_shader, nullptr); - m_efb_depth_to_tex_shader = VK_NULL_HANDLE; + vkDestroyShaderModule(g_vulkan_context->GetDevice(), shader.second, nullptr); } + m_efb_copy_to_tex_shaders.clear(); } void TextureCache::CopyEFBToCacheEntry(TCacheEntry* entry, bool is_depth_copy, const EFBRectangle& src_rect, bool scale_by_half, - unsigned int cbuf_id, const float* colmat) + unsigned int cbuf_id, const float* colmat, + EFBCopyFormat dst_format, bool is_intensity) { VKTexture* texture = static_cast(entry->texture.get()); @@ -378,13 +307,25 @@ void TextureCache::CopyEFBToCacheEntry(TCacheEntry* entry, bool is_depth_copy, texture->GetRawTexIdentifier()->TransitionToLayout(command_buffer, VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL); - UtilityShaderDraw draw(command_buffer, - g_object_cache->GetPipelineLayout(PIPELINE_LAYOUT_PUSH_CONSTANT), - m_render_pass, g_shader_cache->GetPassthroughVertexShader(), - g_shader_cache->GetPassthroughGeometryShader(), - is_depth_copy ? m_efb_depth_to_tex_shader : m_efb_color_to_tex_shader); + auto uid = GetTextureConverterShaderUid(dst_format, is_depth_copy, is_intensity, scale_by_half); + + auto it = m_efb_copy_to_tex_shaders.emplace(uid, VkShaderModule(VK_NULL_HANDLE)); + VkShaderModule& shader = it.first->second; + bool created = it.second; + + if (created) + { + std::string source = g_shader_cache->GetUtilityShaderHeader(); + source += GenerateTextureConverterShaderCode(APIType::Vulkan, uid.GetUidData()).GetBuffer(); + + shader = Util::CompileAndCreateFragmentShader(source); + } + + UtilityShaderDraw draw(command_buffer, + g_object_cache->GetPipelineLayout(PIPELINE_LAYOUT_STANDARD), m_render_pass, + g_shader_cache->GetPassthroughVertexShader(), + g_shader_cache->GetPassthroughGeometryShader(), shader); - draw.SetPushConstants(colmat, (is_depth_copy ? sizeof(float) * 20 : sizeof(float) * 28)); draw.SetPSSampler(0, src_texture->GetView(), src_sampler); VkRect2D dest_region = {{0, 0}, {texture->GetConfig().width, texture->GetConfig().height}}; diff --git a/Source/Core/VideoBackends/Vulkan/TextureCache.h b/Source/Core/VideoBackends/Vulkan/TextureCache.h index 12bf088ad1..0e29229d18 100644 --- a/Source/Core/VideoBackends/Vulkan/TextureCache.h +++ b/Source/Core/VideoBackends/Vulkan/TextureCache.h @@ -4,11 +4,13 @@ #pragma once +#include #include #include "Common/CommonTypes.h" #include "VideoBackends/Vulkan/StreamBuffer.h" #include "VideoCommon/TextureCacheBase.h" +#include "VideoCommon/TextureConverterShaderGen.h" namespace Vulkan { @@ -53,7 +55,8 @@ private: bool CreateRenderPasses(); void CopyEFBToCacheEntry(TCacheEntry* entry, bool is_depth_copy, const EFBRectangle& src_rect, - bool scale_by_half, unsigned int cbuf_id, const float* colmat) override; + bool scale_by_half, unsigned int cbuf_id, const float* colmat, + EFBCopyFormat dst_format, bool is_intensity) override; VkRenderPass m_render_pass = VK_NULL_HANDLE; @@ -62,8 +65,7 @@ private: std::unique_ptr m_texture_converter; VkShaderModule m_copy_shader = VK_NULL_HANDLE; - VkShaderModule m_efb_color_to_tex_shader = VK_NULL_HANDLE; - VkShaderModule m_efb_depth_to_tex_shader = VK_NULL_HANDLE; + std::map m_efb_copy_to_tex_shaders; }; } // namespace Vulkan diff --git a/Source/Core/VideoCommon/CMakeLists.txt b/Source/Core/VideoCommon/CMakeLists.txt index eff095d628..e6dd1f769d 100644 --- a/Source/Core/VideoCommon/CMakeLists.txt +++ b/Source/Core/VideoCommon/CMakeLists.txt @@ -39,6 +39,7 @@ set(SRCS TextureCacheBase.cpp TextureConfig.cpp TextureConversionShader.cpp + TextureConverterShaderGen.cpp TextureDecoder_Common.cpp VertexLoader.cpp VertexLoaderBase.cpp diff --git a/Source/Core/VideoCommon/TextureCacheBase.cpp b/Source/Core/VideoCommon/TextureCacheBase.cpp index d6290001a9..7a8337895e 100644 --- a/Source/Core/VideoCommon/TextureCacheBase.cpp +++ b/Source/Core/VideoCommon/TextureCacheBase.cpp @@ -1997,7 +1997,8 @@ void TextureCacheBase::CopyRenderTargetToTexture(u32 dstAddr, EFBCopyFormat dstF entry->may_have_overlapping_textures = false; entry->is_custom_tex = false; - CopyEFBToCacheEntry(entry, is_depth_copy, srcRect, scaleByHalf, cbufid, colmat); + CopyEFBToCacheEntry(entry, is_depth_copy, srcRect, scaleByHalf, cbufid, colmat, dstFormat, + isIntensity); u64 hash = entry->CalculateHash(); entry->SetHashes(hash, hash); diff --git a/Source/Core/VideoCommon/TextureCacheBase.h b/Source/Core/VideoCommon/TextureCacheBase.h index 38666faf77..f2353bd0fc 100644 --- a/Source/Core/VideoCommon/TextureCacheBase.h +++ b/Source/Core/VideoCommon/TextureCacheBase.h @@ -316,7 +316,8 @@ private: virtual void CopyEFBToCacheEntry(TCacheEntry* entry, bool is_depth_copy, const EFBRectangle& src_rect, bool scale_by_half, - unsigned int cbuf_id, const float* colmat) = 0; + unsigned int cbuf_id, const float* colmat, + EFBCopyFormat dst_format, bool is_intensity) = 0; // Removes and unlinks texture from texture cache and returns it to the pool TexAddrCache::iterator InvalidateTexture(TexAddrCache::iterator t_iter); diff --git a/Source/Core/VideoCommon/TextureConverterShaderGen.cpp b/Source/Core/VideoCommon/TextureConverterShaderGen.cpp new file mode 100644 index 0000000000..fc9b971f8b --- /dev/null +++ b/Source/Core/VideoCommon/TextureConverterShaderGen.cpp @@ -0,0 +1,301 @@ +// Copyright 2017 Dolphin Emulator Project +// Licensed under GPLv2+ +// Refer to the license.txt file included. + +#include +#include + +#include "Common/Assert.h" +#include "Common/CommonTypes.h" +#include "VideoCommon/BPMemory.h" +#include "VideoCommon/TextureConverterShaderGen.h" +#include "VideoCommon/VideoCommon.h" +#include "VideoCommon/VideoConfig.h" + +TextureConverterShaderUid GetTextureConverterShaderUid(EFBCopyFormat dst_format, bool is_depth_copy, + bool is_intensity, bool scale_by_half) +{ + TextureConverterShaderUid out; + convertion_shader_uid_data* uid_data = out.GetUidData(); + memset(uid_data, 0, sizeof(*uid_data)); + + uid_data->dst_format = dst_format; + uid_data->efb_has_alpha = bpmem.zcontrol.pixel_format == PEControl::RGBA6_Z24; + uid_data->is_depth_copy = is_depth_copy; + uid_data->is_intensity = is_intensity; + uid_data->scale_by_half = scale_by_half; + + return out; +} + +ShaderCode GenerateTextureConverterShaderCode(APIType api_type, + const convertion_shader_uid_data* uid_data) +{ + ShaderCode out; + + std::array colmat = {}; + float* const const_add = &colmat[16]; + float* const color_mask = &colmat[20]; + color_mask[0] = color_mask[1] = color_mask[2] = color_mask[3] = 255.0f; + color_mask[4] = color_mask[5] = color_mask[6] = color_mask[7] = 1.0f / 255.0f; + + if (api_type == APIType::OpenGL) + out.Write("SAMPLER_BINDING(9) uniform sampler2DArray samp9;\n" + "#define samp0 samp9\n" + "#define uv0 f_uv0\n" + "in vec3 uv0;\n" + "out vec4 ocol0;\n"); + + else if (api_type == APIType::Vulkan) + out.Write("SAMPLER_BINDING(0) uniform sampler2DArray samp0;\n" + "layout(location = 0) in vec3 uv0;\n" + "layout(location = 1) in vec4 col0;\n" + "layout(location = 0) out vec4 ocol0;"); + + bool mono_depth = uid_data->is_depth_copy && g_ActiveConfig.bStereoEFBMonoDepth; + out.Write("void main(){\n" + " vec4 texcol = texture(samp0, %s);\n", + mono_depth ? "vec3(uv0.xy, 0.0)" : "uv0"); + + if (uid_data->is_depth_copy) + { + switch (uid_data->dst_format) + { + case EFBCopyFormat::R4: // Z4 + colmat[3] = colmat[7] = colmat[11] = colmat[15] = 1.0f; + break; + case EFBCopyFormat::R8_0x1: // Z8 + case EFBCopyFormat::R8: // Z8H + colmat[0] = colmat[4] = colmat[8] = colmat[12] = 1.0f; + break; + + case EFBCopyFormat::RA8: // Z16 + colmat[1] = colmat[5] = colmat[9] = colmat[12] = 1.0f; + break; + + case EFBCopyFormat::RG8: // Z16 (reverse order) + colmat[0] = colmat[4] = colmat[8] = colmat[13] = 1.0f; + break; + + case EFBCopyFormat::RGBA8: // Z24X8 + colmat[0] = colmat[5] = colmat[10] = 1.0f; + break; + + case EFBCopyFormat::G8: // Z8M + colmat[1] = colmat[5] = colmat[9] = colmat[13] = 1.0f; + break; + + case EFBCopyFormat::B8: // Z8L + colmat[2] = colmat[6] = colmat[10] = colmat[14] = 1.0f; + break; + + case EFBCopyFormat::GB8: // Z16L - copy lower 16 depth bits + // expected to be used as an IA8 texture (upper 8 bits stored as intensity, lower 8 bits + // stored as alpha) + // Used e.g. in Zelda: Skyward Sword + colmat[1] = colmat[5] = colmat[9] = colmat[14] = 1.0f; + break; + + default: + ERROR_LOG(VIDEO, "Unknown copy zbuf format: 0x%X", static_cast(uid_data->dst_format)); + colmat[2] = colmat[5] = colmat[8] = 1.0f; + break; + } + } + else if (uid_data->is_intensity) + { + const_add[0] = const_add[1] = const_add[2] = 16.0f / 255.0f; + switch (uid_data->dst_format) + { + case EFBCopyFormat::R4: // I4 + case EFBCopyFormat::R8_0x1: // I8 + case EFBCopyFormat::R8: // I8 + case EFBCopyFormat::RA4: // IA4 + case EFBCopyFormat::RA8: // IA8 + // TODO - verify these coefficients + colmat[0] = 0.257f; + colmat[1] = 0.504f; + colmat[2] = 0.098f; + colmat[4] = 0.257f; + colmat[5] = 0.504f; + colmat[6] = 0.098f; + colmat[8] = 0.257f; + colmat[9] = 0.504f; + colmat[10] = 0.098f; + + if (uid_data->dst_format == EFBCopyFormat::R4 || + uid_data->dst_format == EFBCopyFormat::R8_0x1 || + uid_data->dst_format == EFBCopyFormat::R8) + { + colmat[12] = 0.257f; + colmat[13] = 0.504f; + colmat[14] = 0.098f; + const_add[3] = 16.0f / 255.0f; + if (uid_data->dst_format == EFBCopyFormat::R4) + { + color_mask[0] = color_mask[1] = color_mask[2] = 255.0f / 16.0f; + color_mask[4] = color_mask[5] = color_mask[6] = 1.0f / 15.0f; + } + } + else // alpha + { + colmat[15] = 1; + if (uid_data->dst_format == EFBCopyFormat::RA4) + { + color_mask[0] = color_mask[1] = color_mask[2] = color_mask[3] = 255.0f / 16.0f; + color_mask[4] = color_mask[5] = color_mask[6] = color_mask[7] = 1.0f / 15.0f; + } + } + break; + + default: + ERROR_LOG(VIDEO, "Unknown copy intensity format: 0x%X", + static_cast(uid_data->dst_format)); + colmat[0] = colmat[5] = colmat[10] = colmat[15] = 1.0f; + break; + } + } + else + { + switch (uid_data->dst_format) + { + case EFBCopyFormat::R4: // R4 + colmat[0] = colmat[4] = colmat[8] = colmat[12] = 1; + color_mask[0] = 255.0f / 16.0f; + color_mask[4] = 1.0f / 15.0f; + break; + case EFBCopyFormat::R8_0x1: // R8 + case EFBCopyFormat::R8: // R8 + colmat[0] = colmat[4] = colmat[8] = colmat[12] = 1; + break; + + case EFBCopyFormat::RA4: // RA4 + colmat[0] = colmat[4] = colmat[8] = colmat[15] = 1.0f; + color_mask[0] = color_mask[3] = 255.0f / 16.0f; + color_mask[4] = color_mask[7] = 1.0f / 15.0f; + + if (!uid_data->efb_has_alpha) + { + color_mask[3] = 0.0f; + const_add[3] = 1.0f; + } + break; + case EFBCopyFormat::RA8: // RA8 + colmat[0] = colmat[4] = colmat[8] = colmat[15] = 1.0f; + + if (!uid_data->efb_has_alpha) + { + color_mask[3] = 0.0f; + const_add[3] = 1.0f; + } + break; + + case EFBCopyFormat::A8: // A8 + colmat[3] = colmat[7] = colmat[11] = colmat[15] = 1.0f; + + if (!uid_data->efb_has_alpha) + { + color_mask[3] = 0.0f; + const_add[0] = 1.0f; + const_add[1] = 1.0f; + const_add[2] = 1.0f; + const_add[3] = 1.0f; + } + break; + + case EFBCopyFormat::G8: // G8 + colmat[1] = colmat[5] = colmat[9] = colmat[13] = 1.0f; + break; + case EFBCopyFormat::B8: // B8 + colmat[2] = colmat[6] = colmat[10] = colmat[14] = 1.0f; + break; + + case EFBCopyFormat::RG8: // RG8 + colmat[0] = colmat[4] = colmat[8] = colmat[13] = 1.0f; + break; + + case EFBCopyFormat::GB8: // GB8 + colmat[1] = colmat[5] = colmat[9] = colmat[14] = 1.0f; + break; + + case EFBCopyFormat::RGB565: // RGB565 + colmat[0] = colmat[5] = colmat[10] = 1.0f; + color_mask[0] = color_mask[2] = 255.0f / 8.0f; + color_mask[4] = color_mask[6] = 1.0f / 31.0f; + color_mask[1] = 255.0f / 4.0f; + color_mask[5] = 1.0f / 63.0f; + const_add[3] = 1.0f; // set alpha to 1 + break; + + case EFBCopyFormat::RGB5A3: // RGB5A3 + colmat[0] = colmat[5] = colmat[10] = colmat[15] = 1.0f; + color_mask[0] = color_mask[1] = color_mask[2] = 255.0f / 8.0f; + color_mask[4] = color_mask[5] = color_mask[6] = 1.0f / 31.0f; + color_mask[3] = 255.0f / 32.0f; + color_mask[7] = 1.0f / 7.0f; + + if (!uid_data->efb_has_alpha) + { + color_mask[3] = 0.0f; + const_add[3] = 1.0f; + } + break; + case EFBCopyFormat::RGBA8: // RGBA8 + colmat[0] = colmat[5] = colmat[10] = colmat[15] = 1.0f; + + if (!uid_data->efb_has_alpha) + { + color_mask[3] = 0.0f; + const_add[3] = 1.0f; + } + break; + + case EFBCopyFormat::XFB: // XFB copy, we just pretend it's an RGBX copy + colmat[0] = colmat[5] = colmat[10] = colmat[15] = 1.0f; + color_mask[3] = 0.0f; + const_add[3] = 1.0f; + break; + + default: + ERROR_LOG(VIDEO, "Unknown copy color format: 0x%X", static_cast(uid_data->dst_format)); + colmat[0] = colmat[5] = colmat[10] = colmat[15] = 1.0f; + break; + } + } + + out.Write(" const vec4 colmat[7] = {\n"); + for (size_t i = 0; i < colmat.size() / 4; i++) + { + out.Write(" vec4(%f, %f, %f, %f)%s\n", colmat[i * 4 + 0], colmat[i * 4 + 1], + colmat[i * 4 + 2], colmat[i * 4 + 3], i < 7 ? "," : ""); + } + out.Write(" };\n"); + + if (uid_data->is_depth_copy) + { + if (api_type == APIType::Vulkan) + out.Write("texcol.x = 1.0 - texcol.x;\n"); + + out.Write(" int depth = int(texcol.x * 16777216.0);\n" + + // Convert to Z24 format + " ivec4 workspace;\n" + " workspace.r = (depth >> 16) & 255;\n" + " workspace.g = (depth >> 8) & 255;\n" + " workspace.b = depth & 255;\n" + + // Convert to Z4 format + " workspace.a = (depth >> 16) & 0xF0;\n" + + // Normalize components to [0.0..1.0] + " texcol = vec4(workspace) / 255.0;\n"); + } + else + { + out.Write(" texcol = floor(texcol * colmat[5]) * colmat[6];\n"); + } + out.Write(" ocol0 = texcol * mat4(colmat[0], colmat[1], colmat[2], colmat[3]) + colmat[4];\n" + "}\n"); + + return out; +} diff --git a/Source/Core/VideoCommon/TextureConverterShaderGen.h b/Source/Core/VideoCommon/TextureConverterShaderGen.h new file mode 100644 index 0000000000..410940a63e --- /dev/null +++ b/Source/Core/VideoCommon/TextureConverterShaderGen.h @@ -0,0 +1,32 @@ +// Copyright 2017 Dolphin Emulator Project +// Licensed under GPLv2+ +// Refer to the license.txt file included. + +#pragma once + +#include "Common/CommonTypes.h" +#include "VideoCommon/ShaderGenCommon.h" +#include "VideoCommon/TextureDecoder.h" + +enum class APIType; + +#pragma pack(1) +struct convertion_shader_uid_data +{ + u32 NumValues() const { return sizeof(convertion_shader_uid_data); } + EFBCopyFormat dst_format; + + u32 efb_has_alpha : 1; + u32 is_depth_copy : 1; + u32 is_intensity : 1; + u32 scale_by_half : 1; +}; +#pragma pack() + +using TextureConverterShaderUid = ShaderUid; + +ShaderCode GenerateTextureConverterShaderCode(APIType api_type, + const convertion_shader_uid_data* uid_data); + +TextureConverterShaderUid GetTextureConverterShaderUid(EFBCopyFormat dst_format, bool is_depth_copy, + bool is_intensity, bool scale_by_half); diff --git a/Source/Core/VideoCommon/VideoCommon.vcxproj b/Source/Core/VideoCommon/VideoCommon.vcxproj index 83be20714f..588701e43c 100644 --- a/Source/Core/VideoCommon/VideoCommon.vcxproj +++ b/Source/Core/VideoCommon/VideoCommon.vcxproj @@ -76,6 +76,7 @@ + @@ -139,6 +140,7 @@ + diff --git a/Source/Core/VideoCommon/VideoCommon.vcxproj.filters b/Source/Core/VideoCommon/VideoCommon.vcxproj.filters index e77cceec23..6f69333edd 100644 --- a/Source/Core/VideoCommon/VideoCommon.vcxproj.filters +++ b/Source/Core/VideoCommon/VideoCommon.vcxproj.filters @@ -83,6 +83,9 @@ Shader Generators + + Shader Generators + Shader Generators @@ -263,6 +266,9 @@ Shader Generators + + Shader Generators + Shader Generators