options: anisotropic filtering, oit pixel buffer size, fast gdrom load

.github/workflows/c-cpp.yml

@@ -16,6 +16,7 @@ jobs:
           - {name: x86_64-pc-windows-msvc, os: windows-latest, cmakeArgs: -DCMAKE_SYSTEM_VERSION="10.0.18362.0" -G "Visual Studio 16 2019" -A x64}
           - {name: x86_64-w64-mingw32, os: windows-latest, cmakeArgs: -G "MinGW Makefiles", destDir: win}
           - {name: libretro-x86_64-pc-linux-gnu, os: ubuntu-latest, cmakeArgs: -DLIBRETRO=ON -DCMAKE_POSITION_INDEPENDENT_CODE=TRUE -DUSE_VULKAN=OFF}
+          - {name: libretro-x86_64-w64-mingw32, os: windows-latest, cmakeArgs: -DLIBRETRO=ON -DUSE_VULKAN=OFF -G "MinGW Makefiles"}
 
     steps:
       - name: Set up build environment (macos-latest)
@@ -34,7 +35,7 @@ jobs:
 
       - name: Set up build environment (windows-latest)
         run: echo "DXSDK_DIR=${Env:HOMEDRIVE}${Env:HOMEPATH}\ccache\" | Out-File -FilePath $Env:GITHUB_ENV -Encoding utf-8 -Append
-        if: matrix.config.os == 'windows-latest' && matrix.config.name != 'x86_64-w64-mingw32'
+        if: matrix.config.os == 'windows-latest' && matrix.config.name != 'x86_64-w64-mingw32' && matrix.config.name != 'libretro-x86_64-w64-mingw32'
 
       - uses: actions/cache@v2
         with:
@@ -52,7 +53,7 @@ jobs:
         if: matrix.config.os == 'windows-latest'
 
       - name: Download DX2010
-        if: steps.cache.outputs.cache-hit != 'true' && matrix.config.os == 'windows-latest' && matrix.config.name != 'x86_64-w64-mingw32'
+        if: steps.cache.outputs.cache-hit != 'true' && matrix.config.os == 'windows-latest' && matrix.config.name != 'x86_64-w64-mingw32' && matrix.config.name != 'libretro-x86_64-w64-mingw32'
         run:   |
                curl -L https://download.microsoft.com/download/a/e/7/ae743f1f-632b-4809-87a9-aa1bb3458e31/DXSDK_Jun10.exe -o _DX2010_.exe
                7z x _DX2010_.exe DXSDK/Include -o_DX2010_

CMakeLists.txt

@@ -879,10 +879,11 @@ if(USE_VULKAN AND NOT APPLE)
 	endif()
 endif()
 
-# TODO
 if(WIN32 AND NOT LIBRETRO)
 	target_sources(${PROJECT_NAME} PRIVATE
 		core/rend/dx9/comptr.h
+		core/rend/dx9/d3d_overlay.h
+		core/rend/dx9/d3d_overlay.cpp
 		core/rend/dx9/d3d_renderer.h
 		core/rend/dx9/d3d_renderer.cpp
 		core/rend/dx9/d3d_shaders.h
@@ -890,11 +891,7 @@ if(WIN32 AND NOT LIBRETRO)
 		core/rend/dx9/d3d_texture.h
 		core/rend/dx9/d3d_texture.cpp
 		core/rend/dx9/dxcontext.h
-		core/rend/dx9/dxcontext.cpp)
+		core/rend/dx9/dxcontext.cpp
-	if(NOT LIBRETRO)
-	    target_sources(${PROJECT_NAME} PRIVATE
-			core/rend/dx9/d3d_overlay.h
-			core/rend/dx9/d3d_overlay.cpp
 		core/rend/dx9/imgui_impl_dx9.h
 		core/rend/dx9/imgui_impl_dx9.cpp)
 	if(NOT MINGW)
@@ -905,7 +902,6 @@ if(WIN32 AND NOT LIBRETRO)
 			target_link_directories(${PROJECT_NAME} PRIVATE "$ENV{DXSDK_DIR}/Lib/x86")
 		endif()
 	endif()
-	endif()
 	target_link_libraries(${PROJECT_NAME} PRIVATE d3d9 d3dx9)
 endif()
 

core/cfg/option.cpp

@@ -90,12 +90,15 @@ Option<int> MaxThreads("pvr.MaxThreads", 3);
 Option<int> AutoSkipFrame("pvr.AutoSkipFrame", 0);
 Option<int> RenderResolution("rend.Resolution", 480);
 Option<bool> VSync("rend.vsync", true);
+Option<u64> PixelBufferSize("rend.PixelBufferSize", 512 * 1024 * 1024);
+Option<int> AnisotropicFiltering("rend.AnisotropicFiltering", 1);
 
 // Misc
 
 Option<bool> SerialConsole("Debug.SerialConsoleEnabled");
 Option<bool> SerialPTY("Debug.SerialPTY");
 Option<bool> UseReios("UseReios");
+Option<bool> FastGDRomLoad("FastGDRomLoad", false);
 
 Option<bool> OpenGlChecks("OpenGlChecks", false, "validate");
 

core/cfg/option.h

@@ -389,12 +389,15 @@ extern Option<int> MaxThreads;
 extern Option<int> AutoSkipFrame;		// 0: none, 1: some, 2: more
 extern Option<int> RenderResolution;
 extern Option<bool> VSync;
+extern Option<u64> PixelBufferSize;
+extern Option<int> AnisotropicFiltering;
 
 // Misc
 
 extern Option<bool> SerialConsole;
 extern Option<bool> SerialPTY;
 extern Option<bool> UseReios;
+extern Option<bool> FastGDRomLoad;
 
 extern Option<bool> OpenGlChecks;
 

core/hw/gdrom/gdromv3.cpp

@@ -1049,6 +1049,8 @@ static int getGDROMTicks()
 {
 	if (SB_GDST & 1)
 	{
+		if (config::FastGDRomLoad)
+			return 512;
 		u32 len = SB_GDLEN == 0 ? 0x02000000 : SB_GDLEN;
 		if (len - SB_GDLEND > 10240)
 			return 1000000;										// Large transfers: GD-ROM transfer rate 1.8 MB/s

core/reios/gdrom_hle.cpp

@@ -74,7 +74,7 @@ static void read_sectors_to(u32 addr, u32 sector, u32 count)
 	gd_hle_state.cur_sector = sector + count - 1;
 	if (virtual_addr)
 		gd_hle_state.xfer_end_time = 0;
-	else if (count > 5)
+	else if (count > 5 && !config::FastGDRomLoad)
 		// Large Transfers: GD-ROM rate (approx. 1.8 MB/s)
 		gd_hle_state.xfer_end_time = sh4_sched_now64() + (u64)count * 2048 * 1000000L / 10240;
 	else

core/rend/gl4/abuffer.cpp

@@ -29,8 +29,6 @@ static GLuint g_quadBuffer;
 static GLuint g_quadIndexBuffer;
 static GLuint g_quadVertexArray;
 
-static GLuint pixel_buffer_size = 512 * 1024 * 1024;	// Initial size 512 MB
-
 #define MAX_PIXELS_PER_FRAGMENT "32"
 
 static const char *final_shader_source = SHADER_HEADER
@@ -277,7 +275,7 @@ void initABuffer()
 		// get the max buffer size
 		GLint64 size;
 		glGetInteger64v(GL_MAX_SHADER_STORAGE_BLOCK_SIZE, &size);
-		pixel_buffer_size = (GLuint)std::min((GLint64)pixel_buffer_size, size);
+		GLsizeiptr pixel_buffer_size = std::min<GLsizeiptr>((u64)config::PixelBufferSize, size);
 
 		// Create the buffer
 		glGenBuffers(1, &pixels_buffer);

core/rend/gl4/gldraw.cpp

@@ -226,6 +226,14 @@ static void SetGPState(const PolyParam* gp)
 					glSamplerParameteri(texSamplers[i], GL_TEXTURE_MAG_FILTER, GL_LINEAR);
 					if (mipmapped)
 						glSamplerParameterf(texSamplers[i], GL_TEXTURE_LOD_BIAS, D_Adjust_LoD_Bias[tsp.MipMapD]);
+					if (gl.max_anisotropy > 1.f)
+					{
+						if (config::AnisotropicFiltering > 1)
+							glTexParameterf(GL_TEXTURE_2D, GL_TEXTURE_MAX_ANISOTROPY,
+									std::min<float>(config::AnisotropicFiltering, gl.max_anisotropy));
+						else
+							glTexParameterf(GL_TEXTURE_2D, GL_TEXTURE_MAX_ANISOTROPY, 1.f);
+					}
 				}
 			}
 		}

core/rend/gles/gldraw.cpp

@@ -194,6 +194,19 @@ __forceinline
 			if (!gl.is_gles && gl.gl_major >= 3 && mipmapped)
 				glTexParameterf(GL_TEXTURE_2D, GL_TEXTURE_LOD_BIAS, D_Adjust_LoD_Bias[gp->tsp.MipMapD]);
 #endif
+			if (gl.max_anisotropy > 1.f)
+			{
+				if (config::AnisotropicFiltering > 1)
+				{
+					glTexParameterf(GL_TEXTURE_2D, GL_TEXTURE_MAX_ANISOTROPY,
+							std::min<float>(config::AnisotropicFiltering, gl.max_anisotropy));
+					// Set the recommended minification filter for best results
+					if (mipmapped)
+						glcache.TexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_LINEAR_MIPMAP_LINEAR);
+				}
+				else
+					glTexParameterf(GL_TEXTURE_2D, GL_TEXTURE_MAX_ANISOTROPY, 1.f);
+			}
 		}
 	}
 

core/rend/gles/gles.cpp

@@ -466,7 +466,7 @@ void findGLVersion()
     	gl.highp_float_supported = true;
 	}
 	gl.max_anisotropy = 1.f;
-#if !defined(GLES2) && defined(GL_MAX_TEXTURE_MAX_ANISOTROPY)
+#if !defined(GLES2)
 	if (gl.gl_major >= 3)
 	{
 		for (u32 i = 0; ; i++)

core/rend/gles/gles.h

@@ -8,6 +8,13 @@
 #include <unordered_map>
 #include <glm/glm.hpp>
 
+#ifndef GL_TEXTURE_MAX_ANISOTROPY
+#define GL_TEXTURE_MAX_ANISOTROPY         0x84FE
+#endif
+#ifndef GL_MAX_TEXTURE_MAX_ANISOTROPY
+#define GL_MAX_TEXTURE_MAX_ANISOTROPY     0x84FF
+#endif
+
 #define glCheck() do { if (unlikely(config::OpenGlChecks)) { verify(glGetError()==GL_NO_ERROR); } } while(0)
 
 #define VERTEX_POS_ARRAY 0

core/rend/vulkan/oit/oit_buffer.h

@@ -24,8 +24,6 @@
 
 #include <memory>
 
-const vk::DeviceSize PixelBufferSize = 512 * 1024 * 1024;
-
 class OITBuffers
 {
 public:
@@ -51,7 +49,7 @@ public:
 
 		if (!pixelBuffer)
 		{
-			pixelBuffer = std::unique_ptr<BufferData>(new BufferData(std::min(PixelBufferSize, context->GetMaxMemoryAllocationSize()),
+			pixelBuffer = std::unique_ptr<BufferData>(new BufferData(std::min<vk::DeviceSize>(config::PixelBufferSize, context->GetMaxMemoryAllocationSize()),
 					vk::BufferUsageFlagBits::eStorageBuffer, vk::MemoryPropertyFlagBits::eDeviceLocal));
 		}
 		if (!pixelCounter)

core/rend/vulkan/texture.h

@@ -91,11 +91,13 @@ public:
 		vk::SamplerAddressMode vRepeat = tsp.ClampV ? vk::SamplerAddressMode::eClampToEdge
 				: tsp.FlipV ? vk::SamplerAddressMode::eMirroredRepeat : vk::SamplerAddressMode::eRepeat;
 
+		bool anisotropicFiltering = config::AnisotropicFiltering > 1 && VulkanContext::Instance()->SupportsSamplerAnisotropy()
+				&& filter == vk::Filter::eLinear;
 		return samplers.emplace(
 					std::make_pair(samplerHash, VulkanContext::Instance()->GetDevice().createSamplerUnique(
 						vk::SamplerCreateInfo(vk::SamplerCreateFlags(), filter, filter,
 							vk::SamplerMipmapMode::eNearest, uRepeat, vRepeat, vk::SamplerAddressMode::eClampToEdge, D_Adjust_LoD_Bias[tsp.MipMapD],
-							VulkanContext::Instance()->SupportsSamplerAnisotropy() && filter == vk::Filter::eLinear, 4.0f,
+							anisotropicFiltering, std::min<float>(config::AnisotropicFiltering, VulkanContext::Instance()->GetMaxSamplerAnisotropy()),
 							false, vk::CompareOp::eNever,
 							0.0f, 256.0f, vk::BorderColor::eFloatOpaqueBlack)))).first->second.get();
 	}

core/rend/vulkan/vulkan_context.cpp

@@ -222,6 +222,7 @@ bool VulkanContext::InitInstance(const char** extensions, uint32_t extensions_co
 		uniformBufferAlignment = properties->limits.minUniformBufferOffsetAlignment;
 		storageBufferAlignment = properties->limits.minStorageBufferOffsetAlignment;
 		maxStorageBufferRange = properties->limits.maxStorageBufferRange;
+		maxSamplerAnisotropy =  properties->limits.maxSamplerAnisotropy;
 		unifiedMemory = properties->deviceType == vk::PhysicalDeviceType::eIntegratedGpu;
 		vendorID = properties->vendorID;
 		NOTICE_LOG(RENDERER, "Vulkan API %s. Device %s", vulkan11 ? "1.1" : "1.0", properties->deviceName);

core/rend/vulkan/vulkan_context.h

@@ -98,6 +98,7 @@ public:
 	static VulkanContext *Instance() { return contextInstance; }
 	bool SupportsFragmentShaderStoresAndAtomics() const { return fragmentStoresAndAtomics; }
 	bool SupportsSamplerAnisotropy() const { return samplerAnisotropy; }
+	float GetMaxSamplerAnisotropy() const { return samplerAnisotropy ? maxSamplerAnisotropy : 1.f; }
 	bool SupportsDedicatedAllocation() const { return dedicatedAllocationSupported; }
 	const VMAllocator& GetAllocator() const { return allocator; }
 	bool IsUnifiedMemory() const { return unifiedMemory; }
@@ -159,6 +160,7 @@ private:
 	bool optimalTilingSupported4444 = false;
 	bool fragmentStoresAndAtomics = false;
 	bool samplerAnisotropy = false;
+	float maxSamplerAnisotropy = 0.f;
 	bool dedicatedAllocationSupported = false;
 	bool unifiedMemory = false;
 	u32 vendorID = 0;

shell/libretro/option_lr.h

@@ -189,8 +189,5 @@ protected:
 
 // TODO
 extern Option<bool> ThreadedRendering;
-extern Option<int> AnisotropicFiltering;
 extern Option<bool> UpdateMode;
 extern Option<bool> PowerVR2Filter;
-extern Option<u64> PixelBufferSize;
-extern Option<bool> FastGDRomLoad;