From 9764c13d6d2977903f407761b27d847c0056e1c4 Mon Sep 17 00:00:00 2001 From: ReinUsesLisp Date: Wed, 30 Dec 2020 02:25:23 -0300 Subject: video_core: Rewrite the texture cache The current texture cache has several points that hurt maintainability and performance. It's easy to break unrelated parts of the cache when doing minor changes. The cache can easily forget valuable information about the cached textures by CPU writes or simply by its normal usage.The current texture cache has several points that hurt maintainability and performance. It's easy to break unrelated parts of the cache when doing minor changes. The cache can easily forget valuable information about the cached textures by CPU writes or simply by its normal usage. This commit aims to address those issues. --- .../renderer_opengl/gl_texture_cache.cpp | 1454 ++++++++++++-------- 1 file changed, 901 insertions(+), 553 deletions(-) (limited to 'src/video_core/renderer_opengl/gl_texture_cache.cpp') diff --git a/src/video_core/renderer_opengl/gl_texture_cache.cpp b/src/video_core/renderer_opengl/gl_texture_cache.cpp index daf352b50..4c690418c 100644 --- a/src/video_core/renderer_opengl/gl_texture_cache.cpp +++ b/src/video_core/renderer_opengl/gl_texture_cache.cpp @@ -2,173 +2,238 @@ // Licensed under GPLv2 or any later version // Refer to the license.txt file included. -#include "common/assert.h" -#include "common/bit_util.h" -#include "common/common_types.h" -#include "common/microprofile.h" -#include "common/scope_exit.h" -#include "core/core.h" -#include "video_core/morton.h" -#include "video_core/renderer_opengl/gl_resource_manager.h" +#include +#include +#include +#include + +#include + +#include "video_core/renderer_opengl/gl_device.h" +#include "video_core/renderer_opengl/gl_shader_manager.h" #include "video_core/renderer_opengl/gl_state_tracker.h" #include "video_core/renderer_opengl/gl_texture_cache.h" -#include "video_core/renderer_opengl/utils.h" -#include "video_core/texture_cache/surface_base.h" +#include "video_core/renderer_opengl/maxwell_to_gl.h" +#include "video_core/renderer_opengl/util_shaders.h" +#include "video_core/surface.h" +#include "video_core/texture_cache/format_lookup_table.h" +#include "video_core/texture_cache/samples_helper.h" #include "video_core/texture_cache/texture_cache.h" -#include "video_core/textures/convert.h" -#include "video_core/textures/texture.h" +#include "video_core/textures/decoders.h" namespace OpenGL { -using Tegra::Texture::SwizzleSource; -using VideoCore::MortonSwizzleMode; +namespace { +using Tegra::Texture::SwizzleSource; +using Tegra::Texture::TextureMipmapFilter; +using Tegra::Texture::TextureType; +using Tegra::Texture::TICEntry; +using Tegra::Texture::TSCEntry; +using VideoCommon::CalculateLevelStrideAlignment; +using VideoCommon::ImageCopy; +using VideoCommon::ImageFlagBits; +using VideoCommon::ImageType; +using VideoCommon::NUM_RT; +using VideoCommon::SamplesLog2; +using VideoCommon::SwizzleParameters; +using VideoCore::Surface::BytesPerBlock; +using VideoCore::Surface::IsPixelFormatASTC; +using VideoCore::Surface::IsPixelFormatSRGB; +using VideoCore::Surface::MaxPixelFormat; using VideoCore::Surface::PixelFormat; -using VideoCore::Surface::SurfaceTarget; using VideoCore::Surface::SurfaceType; -MICROPROFILE_DEFINE(OpenGL_Texture_Upload, "OpenGL", "Texture Upload", MP_RGB(128, 192, 128)); -MICROPROFILE_DEFINE(OpenGL_Texture_Download, "OpenGL", "Texture Download", MP_RGB(128, 192, 128)); -MICROPROFILE_DEFINE(OpenGL_Texture_Buffer_Copy, "OpenGL", "Texture Buffer Copy", - MP_RGB(128, 192, 128)); +struct CopyOrigin { + GLint level; + GLint x; + GLint y; + GLint z; +}; -namespace { +struct CopyRegion { + GLsizei width; + GLsizei height; + GLsizei depth; +}; struct FormatTuple { GLenum internal_format; GLenum format = GL_NONE; GLenum type = GL_NONE; + GLenum store_format = internal_format; }; -constexpr std::array tex_format_tuples = {{ - {GL_RGBA8, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8_REV}, // A8B8G8R8_UNORM - {GL_RGBA8_SNORM, GL_RGBA, GL_BYTE}, // A8B8G8R8_SNORM - {GL_RGBA8I, GL_RGBA_INTEGER, GL_BYTE}, // A8B8G8R8_SINT - {GL_RGBA8UI, GL_RGBA_INTEGER, GL_UNSIGNED_BYTE}, // A8B8G8R8_UINT - {GL_RGB565, GL_RGB, GL_UNSIGNED_SHORT_5_6_5}, // R5G6B5_UNORM - {GL_RGB565, GL_RGB, GL_UNSIGNED_SHORT_5_6_5_REV}, // B5G6R5_UNORM - {GL_RGB5_A1, GL_BGRA, GL_UNSIGNED_SHORT_1_5_5_5_REV}, // A1R5G5B5_UNORM - {GL_RGB10_A2, GL_RGBA, GL_UNSIGNED_INT_2_10_10_10_REV}, // A2B10G10R10_UNORM - {GL_RGB10_A2UI, GL_RGBA_INTEGER, GL_UNSIGNED_INT_2_10_10_10_REV}, // A2B10G10R10_UINT - {GL_RGB5_A1, GL_RGBA, GL_UNSIGNED_SHORT_1_5_5_5_REV}, // A1B5G5R5_UNORM - {GL_R8, GL_RED, GL_UNSIGNED_BYTE}, // R8_UNORM - {GL_R8_SNORM, GL_RED, GL_BYTE}, // R8_SNORM - {GL_R8I, GL_RED_INTEGER, GL_BYTE}, // R8_SINT - {GL_R8UI, GL_RED_INTEGER, GL_UNSIGNED_BYTE}, // R8_UINT - {GL_RGBA16F, GL_RGBA, GL_HALF_FLOAT}, // R16G16B16A16_FLOAT - {GL_RGBA16, GL_RGBA, GL_UNSIGNED_SHORT}, // R16G16B16A16_UNORM - {GL_RGBA16_SNORM, GL_RGBA, GL_SHORT}, // R16G16B16A16_SNORM - {GL_RGBA16I, GL_RGBA_INTEGER, GL_SHORT}, // R16G16B16A16_SINT - {GL_RGBA16UI, GL_RGBA_INTEGER, GL_UNSIGNED_SHORT}, // R16G16B16A16_UINT - {GL_R11F_G11F_B10F, GL_RGB, GL_UNSIGNED_INT_10F_11F_11F_REV}, // B10G11R11_FLOAT - {GL_RGBA32UI, GL_RGBA_INTEGER, GL_UNSIGNED_INT}, // R32G32B32A32_UINT - {GL_COMPRESSED_RGBA_S3TC_DXT1_EXT}, // BC1_RGBA_UNORM - {GL_COMPRESSED_RGBA_S3TC_DXT3_EXT}, // BC2_UNORM - {GL_COMPRESSED_RGBA_S3TC_DXT5_EXT}, // BC3_UNORM - {GL_COMPRESSED_RED_RGTC1}, // BC4_UNORM - {GL_COMPRESSED_SIGNED_RED_RGTC1}, // BC4_SNORM - {GL_COMPRESSED_RG_RGTC2}, // BC5_UNORM - {GL_COMPRESSED_SIGNED_RG_RGTC2}, // BC5_SNORM - {GL_COMPRESSED_RGBA_BPTC_UNORM}, // BC7_UNORM - {GL_COMPRESSED_RGB_BPTC_UNSIGNED_FLOAT}, // BC6H_UFLOAT - {GL_COMPRESSED_RGB_BPTC_SIGNED_FLOAT}, // BC6H_SFLOAT - {GL_COMPRESSED_RGBA_ASTC_4x4_KHR}, // ASTC_2D_4X4_UNORM - {GL_RGBA8, GL_BGRA, GL_UNSIGNED_BYTE}, // B8G8R8A8_UNORM - {GL_RGBA32F, GL_RGBA, GL_FLOAT}, // R32G32B32A32_FLOAT - {GL_RGBA32I, GL_RGBA_INTEGER, GL_INT}, // R32G32B32A32_SINT - {GL_RG32F, GL_RG, GL_FLOAT}, // R32G32_FLOAT - {GL_RG32I, GL_RG_INTEGER, GL_INT}, // R32G32_SINT - {GL_R32F, GL_RED, GL_FLOAT}, // R32_FLOAT - {GL_R16F, GL_RED, GL_HALF_FLOAT}, // R16_FLOAT - {GL_R16, GL_RED, GL_UNSIGNED_SHORT}, // R16_UNORM - {GL_R16_SNORM, GL_RED, GL_SHORT}, // R16_SNORM - {GL_R16UI, GL_RED_INTEGER, GL_UNSIGNED_SHORT}, // R16_UINT - {GL_R16I, GL_RED_INTEGER, GL_SHORT}, // R16_SINT - {GL_RG16, GL_RG, GL_UNSIGNED_SHORT}, // R16G16_UNORM - {GL_RG16F, GL_RG, GL_HALF_FLOAT}, // R16G16_FLOAT - {GL_RG16UI, GL_RG_INTEGER, GL_UNSIGNED_SHORT}, // R16G16_UINT - {GL_RG16I, GL_RG_INTEGER, GL_SHORT}, // R16G16_SINT - {GL_RG16_SNORM, GL_RG, GL_SHORT}, // R16G16_SNORM - {GL_RGB32F, GL_RGB, GL_FLOAT}, // R32G32B32_FLOAT - {GL_SRGB8_ALPHA8, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8_REV}, // A8B8G8R8_SRGB - {GL_RG8, GL_RG, GL_UNSIGNED_BYTE}, // R8G8_UNORM - {GL_RG8_SNORM, GL_RG, GL_BYTE}, // R8G8_SNORM - {GL_RG8I, GL_RG_INTEGER, GL_BYTE}, // R8G8_SINT - {GL_RG8UI, GL_RG_INTEGER, GL_UNSIGNED_BYTE}, // R8G8_UINT - {GL_RG32UI, GL_RG_INTEGER, GL_UNSIGNED_INT}, // R32G32_UINT - {GL_RGB16F, GL_RGBA, GL_HALF_FLOAT}, // R16G16B16X16_FLOAT - {GL_R32UI, GL_RED_INTEGER, GL_UNSIGNED_INT}, // R32_UINT - {GL_R32I, GL_RED_INTEGER, GL_INT}, // R32_SINT - {GL_COMPRESSED_RGBA_ASTC_8x8_KHR}, // ASTC_2D_8X8_UNORM - {GL_COMPRESSED_RGBA_ASTC_8x5_KHR}, // ASTC_2D_8X5_UNORM - {GL_COMPRESSED_RGBA_ASTC_5x4_KHR}, // ASTC_2D_5X4_UNORM - {GL_SRGB8_ALPHA8, GL_BGRA, GL_UNSIGNED_BYTE}, // B8G8R8A8_UNORM - // Compressed sRGB formats - {GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT1_EXT}, // BC1_RGBA_SRGB - {GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT3_EXT}, // BC2_SRGB - {GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT5_EXT}, // BC3_SRGB - {GL_COMPRESSED_SRGB_ALPHA_BPTC_UNORM}, // BC7_SRGB - {GL_RGBA4, GL_RGBA, GL_UNSIGNED_SHORT_4_4_4_4_REV}, // A4B4G4R4_UNORM - {GL_COMPRESSED_SRGB8_ALPHA8_ASTC_4x4_KHR}, // ASTC_2D_4X4_SRGB - {GL_COMPRESSED_SRGB8_ALPHA8_ASTC_8x8_KHR}, // ASTC_2D_8X8_SRGB - {GL_COMPRESSED_SRGB8_ALPHA8_ASTC_8x5_KHR}, // ASTC_2D_8X5_SRGB - {GL_COMPRESSED_SRGB8_ALPHA8_ASTC_5x4_KHR}, // ASTC_2D_5X4_SRGB - {GL_COMPRESSED_RGBA_ASTC_5x5_KHR}, // ASTC_2D_5X5_UNORM - {GL_COMPRESSED_SRGB8_ALPHA8_ASTC_5x5_KHR}, // ASTC_2D_5X5_SRGB - {GL_COMPRESSED_RGBA_ASTC_10x8_KHR}, // ASTC_2D_10X8_UNORM - {GL_COMPRESSED_SRGB8_ALPHA8_ASTC_10x8_KHR}, // ASTC_2D_10X8_SRGB - {GL_COMPRESSED_RGBA_ASTC_6x6_KHR}, // ASTC_2D_6X6_UNORM - {GL_COMPRESSED_SRGB8_ALPHA8_ASTC_6x6_KHR}, // ASTC_2D_6X6_SRGB - {GL_COMPRESSED_RGBA_ASTC_10x10_KHR}, // ASTC_2D_10X10_UNORM - {GL_COMPRESSED_SRGB8_ALPHA8_ASTC_10x10_KHR}, // ASTC_2D_10X10_SRGB - {GL_COMPRESSED_RGBA_ASTC_12x12_KHR}, // ASTC_2D_12X12_UNORM - {GL_COMPRESSED_SRGB8_ALPHA8_ASTC_12x12_KHR}, // ASTC_2D_12X12_SRGB - {GL_COMPRESSED_RGBA_ASTC_8x6_KHR}, // ASTC_2D_8X6_UNORM - {GL_COMPRESSED_SRGB8_ALPHA8_ASTC_8x6_KHR}, // ASTC_2D_8X6_SRGB - {GL_COMPRESSED_RGBA_ASTC_6x5_KHR}, // ASTC_2D_6X5_UNORM - {GL_COMPRESSED_SRGB8_ALPHA8_ASTC_6x5_KHR}, // ASTC_2D_6X5_SRGB - {GL_RGB9_E5, GL_RGB, GL_UNSIGNED_INT_5_9_9_9_REV}, // E5B9G9R9_FLOAT - - // Depth formats - {GL_DEPTH_COMPONENT32F, GL_DEPTH_COMPONENT, GL_FLOAT}, // D32_FLOAT - {GL_DEPTH_COMPONENT16, GL_DEPTH_COMPONENT, GL_UNSIGNED_SHORT}, // D16_UNORM - - // DepthStencil formats - {GL_DEPTH24_STENCIL8, GL_DEPTH_STENCIL, GL_UNSIGNED_INT_24_8}, // D24_UNORM_S8_UINT - {GL_DEPTH24_STENCIL8, GL_DEPTH_STENCIL, GL_UNSIGNED_INT_24_8}, // S8_UINT_D24_UNORM +constexpr std::array FORMAT_TABLE = {{ + {GL_RGBA8, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8_REV}, // A8B8G8R8_UNORM + {GL_RGBA8_SNORM, GL_RGBA, GL_BYTE}, // A8B8G8R8_SNORM + {GL_RGBA8I, GL_RGBA_INTEGER, GL_BYTE}, // A8B8G8R8_SINT + {GL_RGBA8UI, GL_RGBA_INTEGER, GL_UNSIGNED_BYTE}, // A8B8G8R8_UINT + {GL_RGB565, GL_RGB, GL_UNSIGNED_SHORT_5_6_5}, // R5G6B5_UNORM + {GL_RGB565, GL_RGB, GL_UNSIGNED_SHORT_5_6_5_REV}, // B5G6R5_UNORM + {GL_RGB5_A1, GL_BGRA, GL_UNSIGNED_SHORT_1_5_5_5_REV}, // A1R5G5B5_UNORM + {GL_RGB10_A2, GL_RGBA, GL_UNSIGNED_INT_2_10_10_10_REV}, // A2B10G10R10_UNORM + {GL_RGB10_A2UI, GL_RGBA_INTEGER, GL_UNSIGNED_INT_2_10_10_10_REV}, // A2B10G10R10_UINT + {GL_RGB5_A1, GL_RGBA, GL_UNSIGNED_SHORT_1_5_5_5_REV}, // A1B5G5R5_UNORM + {GL_R8, GL_RED, GL_UNSIGNED_BYTE}, // R8_UNORM + {GL_R8_SNORM, GL_RED, GL_BYTE}, // R8_SNORM + {GL_R8I, GL_RED_INTEGER, GL_BYTE}, // R8_SINT + {GL_R8UI, GL_RED_INTEGER, GL_UNSIGNED_BYTE}, // R8_UINT + {GL_RGBA16F, GL_RGBA, GL_HALF_FLOAT}, // R16G16B16A16_FLOAT + {GL_RGBA16, GL_RGBA, GL_UNSIGNED_SHORT}, // R16G16B16A16_UNORM + {GL_RGBA16_SNORM, GL_RGBA, GL_SHORT}, // R16G16B16A16_SNORM + {GL_RGBA16I, GL_RGBA_INTEGER, GL_SHORT}, // R16G16B16A16_SINT + {GL_RGBA16UI, GL_RGBA_INTEGER, GL_UNSIGNED_SHORT}, // R16G16B16A16_UINT + {GL_R11F_G11F_B10F, GL_RGB, GL_UNSIGNED_INT_10F_11F_11F_REV}, // B10G11R11_FLOAT + {GL_RGBA32UI, GL_RGBA_INTEGER, GL_UNSIGNED_INT}, // R32G32B32A32_UINT + {GL_COMPRESSED_RGBA_S3TC_DXT1_EXT}, // BC1_RGBA_UNORM + {GL_COMPRESSED_RGBA_S3TC_DXT3_EXT}, // BC2_UNORM + {GL_COMPRESSED_RGBA_S3TC_DXT5_EXT}, // BC3_UNORM + {GL_COMPRESSED_RED_RGTC1}, // BC4_UNORM + {GL_COMPRESSED_SIGNED_RED_RGTC1}, // BC4_SNORM + {GL_COMPRESSED_RG_RGTC2}, // BC5_UNORM + {GL_COMPRESSED_SIGNED_RG_RGTC2}, // BC5_SNORM + {GL_COMPRESSED_RGBA_BPTC_UNORM}, // BC7_UNORM + {GL_COMPRESSED_RGB_BPTC_UNSIGNED_FLOAT}, // BC6H_UFLOAT + {GL_COMPRESSED_RGB_BPTC_SIGNED_FLOAT}, // BC6H_SFLOAT + {GL_COMPRESSED_RGBA_ASTC_4x4_KHR}, // ASTC_2D_4X4_UNORM + {GL_RGBA8, GL_BGRA, GL_UNSIGNED_BYTE}, // B8G8R8A8_UNORM + {GL_RGBA32F, GL_RGBA, GL_FLOAT}, // R32G32B32A32_FLOAT + {GL_RGBA32I, GL_RGBA_INTEGER, GL_INT}, // R32G32B32A32_SINT + {GL_RG32F, GL_RG, GL_FLOAT}, // R32G32_FLOAT + {GL_RG32I, GL_RG_INTEGER, GL_INT}, // R32G32_SINT + {GL_R32F, GL_RED, GL_FLOAT}, // R32_FLOAT + {GL_R16F, GL_RED, GL_HALF_FLOAT}, // R16_FLOAT + {GL_R16, GL_RED, GL_UNSIGNED_SHORT}, // R16_UNORM + {GL_R16_SNORM, GL_RED, GL_SHORT}, // R16_SNORM + {GL_R16UI, GL_RED_INTEGER, GL_UNSIGNED_SHORT}, // R16_UINT + {GL_R16I, GL_RED_INTEGER, GL_SHORT}, // R16_SINT + {GL_RG16, GL_RG, GL_UNSIGNED_SHORT}, // R16G16_UNORM + {GL_RG16F, GL_RG, GL_HALF_FLOAT}, // R16G16_FLOAT + {GL_RG16UI, GL_RG_INTEGER, GL_UNSIGNED_SHORT}, // R16G16_UINT + {GL_RG16I, GL_RG_INTEGER, GL_SHORT}, // R16G16_SINT + {GL_RG16_SNORM, GL_RG, GL_SHORT}, // R16G16_SNORM + {GL_RGB32F, GL_RGB, GL_FLOAT}, // R32G32B32_FLOAT + {GL_SRGB8_ALPHA8, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8_REV, GL_RGBA8}, // A8B8G8R8_SRGB + {GL_RG8, GL_RG, GL_UNSIGNED_BYTE}, // R8G8_UNORM + {GL_RG8_SNORM, GL_RG, GL_BYTE}, // R8G8_SNORM + {GL_RG8I, GL_RG_INTEGER, GL_BYTE}, // R8G8_SINT + {GL_RG8UI, GL_RG_INTEGER, GL_UNSIGNED_BYTE}, // R8G8_UINT + {GL_RG32UI, GL_RG_INTEGER, GL_UNSIGNED_INT}, // R32G32_UINT + {GL_RGB16F, GL_RGBA, GL_HALF_FLOAT}, // R16G16B16X16_FLOAT + {GL_R32UI, GL_RED_INTEGER, GL_UNSIGNED_INT}, // R32_UINT + {GL_R32I, GL_RED_INTEGER, GL_INT}, // R32_SINT + {GL_COMPRESSED_RGBA_ASTC_8x8_KHR}, // ASTC_2D_8X8_UNORM + {GL_COMPRESSED_RGBA_ASTC_8x5_KHR}, // ASTC_2D_8X5_UNORM + {GL_COMPRESSED_RGBA_ASTC_5x4_KHR}, // ASTC_2D_5X4_UNORM + {GL_SRGB8_ALPHA8, GL_BGRA, GL_UNSIGNED_BYTE, GL_RGBA8}, // B8G8R8A8_UNORM + {GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT1_EXT}, // BC1_RGBA_SRGB + {GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT3_EXT}, // BC2_SRGB + {GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT5_EXT}, // BC3_SRGB + {GL_COMPRESSED_SRGB_ALPHA_BPTC_UNORM}, // BC7_SRGB + {GL_RGBA4, GL_RGBA, GL_UNSIGNED_SHORT_4_4_4_4_REV}, // A4B4G4R4_UNORM + {GL_COMPRESSED_SRGB8_ALPHA8_ASTC_4x4_KHR}, // ASTC_2D_4X4_SRGB + {GL_COMPRESSED_SRGB8_ALPHA8_ASTC_8x8_KHR}, // ASTC_2D_8X8_SRGB + {GL_COMPRESSED_SRGB8_ALPHA8_ASTC_8x5_KHR}, // ASTC_2D_8X5_SRGB + {GL_COMPRESSED_SRGB8_ALPHA8_ASTC_5x4_KHR}, // ASTC_2D_5X4_SRGB + {GL_COMPRESSED_RGBA_ASTC_5x5_KHR}, // ASTC_2D_5X5_UNORM + {GL_COMPRESSED_SRGB8_ALPHA8_ASTC_5x5_KHR}, // ASTC_2D_5X5_SRGB + {GL_COMPRESSED_RGBA_ASTC_10x8_KHR}, // ASTC_2D_10X8_UNORM + {GL_COMPRESSED_SRGB8_ALPHA8_ASTC_10x8_KHR}, // ASTC_2D_10X8_SRGB + {GL_COMPRESSED_RGBA_ASTC_6x6_KHR}, // ASTC_2D_6X6_UNORM + {GL_COMPRESSED_SRGB8_ALPHA8_ASTC_6x6_KHR}, // ASTC_2D_6X6_SRGB + {GL_COMPRESSED_RGBA_ASTC_10x10_KHR}, // ASTC_2D_10X10_UNORM + {GL_COMPRESSED_SRGB8_ALPHA8_ASTC_10x10_KHR}, // ASTC_2D_10X10_SRGB + {GL_COMPRESSED_RGBA_ASTC_12x12_KHR}, // ASTC_2D_12X12_UNORM + {GL_COMPRESSED_SRGB8_ALPHA8_ASTC_12x12_KHR}, // ASTC_2D_12X12_SRGB + {GL_COMPRESSED_RGBA_ASTC_8x6_KHR}, // ASTC_2D_8X6_UNORM + {GL_COMPRESSED_SRGB8_ALPHA8_ASTC_8x6_KHR}, // ASTC_2D_8X6_SRGB + {GL_COMPRESSED_RGBA_ASTC_6x5_KHR}, // ASTC_2D_6X5_UNORM + {GL_COMPRESSED_SRGB8_ALPHA8_ASTC_6x5_KHR}, // ASTC_2D_6X5_SRGB + {GL_RGB9_E5, GL_RGB, GL_UNSIGNED_INT_5_9_9_9_REV}, // E5B9G9R9_FLOAT + {GL_DEPTH_COMPONENT32F, GL_DEPTH_COMPONENT, GL_FLOAT}, // D32_FLOAT + {GL_DEPTH_COMPONENT16, GL_DEPTH_COMPONENT, GL_UNSIGNED_SHORT}, // D16_UNORM + {GL_DEPTH24_STENCIL8, GL_DEPTH_STENCIL, GL_UNSIGNED_INT_24_8}, // D24_UNORM_S8_UINT + {GL_DEPTH24_STENCIL8, GL_DEPTH_STENCIL, GL_UNSIGNED_INT_24_8}, // S8_UINT_D24_UNORM {GL_DEPTH32F_STENCIL8, GL_DEPTH_STENCIL, GL_FLOAT_32_UNSIGNED_INT_24_8_REV}, // D32_FLOAT_S8_UINT }}; +constexpr std::array ACCELERATED_FORMATS{ + GL_RGBA32F, GL_RGBA16F, GL_RG32F, GL_RG16F, GL_R11F_G11F_B10F, GL_R32F, + GL_R16F, GL_RGBA32UI, GL_RGBA16UI, GL_RGB10_A2UI, GL_RGBA8UI, GL_RG32UI, + GL_RG16UI, GL_RG8UI, GL_R32UI, GL_R16UI, GL_R8UI, GL_RGBA32I, + GL_RGBA16I, GL_RGBA8I, GL_RG32I, GL_RG16I, GL_RG8I, GL_R32I, + GL_R16I, GL_R8I, GL_RGBA16, GL_RGB10_A2, GL_RGBA8, GL_RG16, + GL_RG8, GL_R16, GL_R8, GL_RGBA16_SNORM, GL_RGBA8_SNORM, GL_RG16_SNORM, + GL_RG8_SNORM, GL_R16_SNORM, GL_R8_SNORM, +}; + const FormatTuple& GetFormatTuple(PixelFormat pixel_format) { - ASSERT(static_cast(pixel_format) < tex_format_tuples.size()); - return tex_format_tuples[static_cast(pixel_format)]; + ASSERT(static_cast(pixel_format) < FORMAT_TABLE.size()); + return FORMAT_TABLE[static_cast(pixel_format)]; } -GLenum GetTextureTarget(const SurfaceTarget& target) { - switch (target) { - case SurfaceTarget::TextureBuffer: +GLenum ImageTarget(const VideoCommon::ImageInfo& info) { + switch (info.type) { + case ImageType::e1D: + return GL_TEXTURE_1D_ARRAY; + case ImageType::e2D: + if (info.num_samples > 1) { + return GL_TEXTURE_2D_MULTISAMPLE_ARRAY; + } + return GL_TEXTURE_2D_ARRAY; + case ImageType::e3D: + return GL_TEXTURE_3D; + case ImageType::Linear: + return GL_TEXTURE_2D_ARRAY; + case ImageType::Buffer: return GL_TEXTURE_BUFFER; - case SurfaceTarget::Texture1D: + } + UNREACHABLE_MSG("Invalid image type={}", info.type); + return GL_NONE; +} + +GLenum ImageTarget(ImageViewType type, int num_samples = 1) { + const bool is_multisampled = num_samples > 1; + switch (type) { + case ImageViewType::e1D: return GL_TEXTURE_1D; - case SurfaceTarget::Texture2D: - return GL_TEXTURE_2D; - case SurfaceTarget::Texture3D: + case ImageViewType::e2D: + return is_multisampled ? GL_TEXTURE_2D_MULTISAMPLE : GL_TEXTURE_2D; + case ImageViewType::Cube: + return GL_TEXTURE_CUBE_MAP; + case ImageViewType::e3D: return GL_TEXTURE_3D; - case SurfaceTarget::Texture1DArray: + case ImageViewType::e1DArray: return GL_TEXTURE_1D_ARRAY; - case SurfaceTarget::Texture2DArray: - return GL_TEXTURE_2D_ARRAY; - case SurfaceTarget::TextureCubemap: - return GL_TEXTURE_CUBE_MAP; - case SurfaceTarget::TextureCubeArray: + case ImageViewType::e2DArray: + return is_multisampled ? GL_TEXTURE_2D_MULTISAMPLE_ARRAY : GL_TEXTURE_2D_ARRAY; + case ImageViewType::CubeArray: return GL_TEXTURE_CUBE_MAP_ARRAY; + case ImageViewType::Rect: + return GL_TEXTURE_RECTANGLE; + case ImageViewType::Buffer: + return GL_TEXTURE_BUFFER; } - UNREACHABLE(); - return {}; + UNREACHABLE_MSG("Invalid image view type={}", type); + return GL_NONE; } -GLint GetSwizzleSource(SwizzleSource source) { +GLenum TextureMode(PixelFormat format, bool is_first) { + switch (format) { + case PixelFormat::D24_UNORM_S8_UINT: + case PixelFormat::D32_FLOAT_S8_UINT: + return is_first ? GL_DEPTH_COMPONENT : GL_STENCIL_INDEX; + case PixelFormat::S8_UINT_D24_UNORM: + return is_first ? GL_STENCIL_INDEX : GL_DEPTH_COMPONENT; + default: + UNREACHABLE(); + return GL_DEPTH_COMPONENT; + } +} + +GLint Swizzle(SwizzleSource source) { switch (source) { case SwizzleSource::Zero: return GL_ZERO; @@ -184,530 +249,813 @@ GLint GetSwizzleSource(SwizzleSource source) { case SwizzleSource::OneFloat: return GL_ONE; } - UNREACHABLE(); + UNREACHABLE_MSG("Invalid swizzle source={}", source); return GL_NONE; } -GLenum GetComponent(PixelFormat format, bool is_first) { - switch (format) { - case PixelFormat::D24_UNORM_S8_UINT: - case PixelFormat::D32_FLOAT_S8_UINT: - return is_first ? GL_DEPTH_COMPONENT : GL_STENCIL_INDEX; - case PixelFormat::S8_UINT_D24_UNORM: - return is_first ? GL_STENCIL_INDEX : GL_DEPTH_COMPONENT; +GLenum AttachmentType(PixelFormat format) { + switch (const SurfaceType type = VideoCore::Surface::GetFormatType(format); type) { + case SurfaceType::Depth: + return GL_DEPTH_ATTACHMENT; + case SurfaceType::DepthStencil: + return GL_DEPTH_STENCIL_ATTACHMENT; default: - UNREACHABLE(); - return GL_DEPTH_COMPONENT; + UNIMPLEMENTED_MSG("Unimplemented type={}", type); + return GL_NONE; } } -void ApplyTextureDefaults(const SurfaceParams& params, GLuint texture) { - if (params.IsBuffer()) { - return; +[[nodiscard]] bool IsConverted(const Device& device, PixelFormat format, ImageType type) { + if (!device.HasASTC() && IsPixelFormatASTC(format)) { + return true; } - glTextureParameteri(texture, GL_TEXTURE_MIN_FILTER, GL_LINEAR); - glTextureParameteri(texture, GL_TEXTURE_MAG_FILTER, GL_LINEAR); - glTextureParameteri(texture, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE); - glTextureParameteri(texture, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE); - glTextureParameteri(texture, GL_TEXTURE_MAX_LEVEL, static_cast(params.num_levels - 1)); - if (params.num_levels == 1) { - glTextureParameterf(texture, GL_TEXTURE_LOD_BIAS, 1000.0f); + switch (format) { + case PixelFormat::BC4_UNORM: + case PixelFormat::BC5_UNORM: + return type == ImageType::e3D; + default: + break; } + return false; } -OGLTexture CreateTexture(const SurfaceParams& params, GLenum target, GLenum internal_format, - OGLBuffer& texture_buffer) { - OGLTexture texture; - texture.Create(target); +[[nodiscard]] constexpr SwizzleSource ConvertGreenRed(SwizzleSource value) { + switch (value) { + case SwizzleSource::G: + return SwizzleSource::R; + default: + return value; + } +} - switch (params.target) { - case SurfaceTarget::Texture1D: - glTextureStorage1D(texture.handle, params.emulated_levels, internal_format, params.width); - break; - case SurfaceTarget::TextureBuffer: - texture_buffer.Create(); - glNamedBufferStorage(texture_buffer.handle, params.width * params.GetBytesPerPixel(), - nullptr, GL_DYNAMIC_STORAGE_BIT); - glTextureBuffer(texture.handle, internal_format, texture_buffer.handle); +void ApplySwizzle(GLuint handle, PixelFormat format, std::array swizzle) { + switch (format) { + case PixelFormat::D24_UNORM_S8_UINT: + case PixelFormat::D32_FLOAT_S8_UINT: + case PixelFormat::S8_UINT_D24_UNORM: + UNIMPLEMENTED_IF(swizzle[0] != SwizzleSource::R && swizzle[0] != SwizzleSource::G); + glTextureParameteri(handle, GL_DEPTH_STENCIL_TEXTURE_MODE, + TextureMode(format, swizzle[0] == SwizzleSource::R)); + std::ranges::transform(swizzle, swizzle.begin(), ConvertGreenRed); break; - case SurfaceTarget::Texture2D: - case SurfaceTarget::TextureCubemap: - glTextureStorage2D(texture.handle, params.emulated_levels, internal_format, params.width, - params.height); + default: break; - case SurfaceTarget::Texture3D: - case SurfaceTarget::Texture2DArray: - case SurfaceTarget::TextureCubeArray: - glTextureStorage3D(texture.handle, params.emulated_levels, internal_format, params.width, - params.height, params.depth); + } + std::array gl_swizzle; + std::ranges::transform(swizzle, gl_swizzle.begin(), Swizzle); + glTextureParameteriv(handle, GL_TEXTURE_SWIZZLE_RGBA, gl_swizzle.data()); +} + +[[nodiscard]] bool CanBeAccelerated(const TextureCacheRuntime& runtime, + const VideoCommon::ImageInfo& info) { + // Disable accelerated uploads for now as they don't implement swizzled uploads + return false; + switch (info.type) { + case ImageType::e2D: + case ImageType::e3D: + case ImageType::Linear: break; default: - UNREACHABLE(); + return false; + } + const GLenum internal_format = GetFormatTuple(info.format).internal_format; + const auto& format_info = runtime.FormatInfo(info.type, internal_format); + if (format_info.is_compressed) { + return false; + } + if (std::ranges::find(ACCELERATED_FORMATS, internal_format) == ACCELERATED_FORMATS.end()) { + return false; } + if (format_info.compatibility_by_size) { + return true; + } + const GLenum store_format = StoreFormat(BytesPerBlock(info.format)); + const GLenum store_class = runtime.FormatInfo(info.type, store_format).compatibility_class; + return format_info.compatibility_class == store_class; +} - ApplyTextureDefaults(params, texture.handle); +[[nodiscard]] CopyOrigin MakeCopyOrigin(VideoCommon::Offset3D offset, + VideoCommon::SubresourceLayers subresource, GLenum target) { + switch (target) { + case GL_TEXTURE_2D_ARRAY: + case GL_TEXTURE_2D_MULTISAMPLE_ARRAY: + return CopyOrigin{ + .level = static_cast(subresource.base_level), + .x = static_cast(offset.x), + .y = static_cast(offset.y), + .z = static_cast(subresource.base_layer), + }; + case GL_TEXTURE_3D: + return CopyOrigin{ + .level = static_cast(subresource.base_level), + .x = static_cast(offset.x), + .y = static_cast(offset.y), + .z = static_cast(offset.z), + }; + default: + UNIMPLEMENTED_MSG("Unimplemented copy target={}", target); + return CopyOrigin{.level = 0, .x = 0, .y = 0, .z = 0}; + } +} - return texture; +[[nodiscard]] CopyRegion MakeCopyRegion(VideoCommon::Extent3D extent, + VideoCommon::SubresourceLayers dst_subresource, + GLenum target) { + switch (target) { + case GL_TEXTURE_2D_ARRAY: + case GL_TEXTURE_2D_MULTISAMPLE_ARRAY: + return CopyRegion{ + .width = static_cast(extent.width), + .height = static_cast(extent.height), + .depth = static_cast(dst_subresource.num_layers), + }; + case GL_TEXTURE_3D: + return CopyRegion{ + .width = static_cast(extent.width), + .height = static_cast(extent.height), + .depth = static_cast(extent.depth), + }; + default: + UNIMPLEMENTED_MSG("Unimplemented copy target={}", target); + return CopyRegion{.width = 0, .height = 0, .depth = 0}; + } } -constexpr u32 EncodeSwizzle(SwizzleSource x_source, SwizzleSource y_source, SwizzleSource z_source, - SwizzleSource w_source) { - return (static_cast(x_source) << 24) | (static_cast(y_source) << 16) | - (static_cast(z_source) << 8) | static_cast(w_source); +void AttachTexture(GLuint fbo, GLenum attachment, const ImageView* image_view) { + if (False(image_view->flags & VideoCommon::ImageViewFlagBits::Slice)) { + const GLuint texture = image_view->DefaultHandle(); + glNamedFramebufferTexture(fbo, attachment, texture, 0); + return; + } + const GLuint texture = image_view->Handle(ImageViewType::e3D); + if (image_view->range.extent.layers > 1) { + // TODO: OpenGL doesn't support rendering to a fixed number of slices + glNamedFramebufferTexture(fbo, attachment, texture, 0); + } else { + const u32 slice = image_view->range.base.layer; + glNamedFramebufferTextureLayer(fbo, attachment, texture, 0, slice); + } } } // Anonymous namespace -CachedSurface::CachedSurface(const GPUVAddr gpu_addr_, const SurfaceParams& params_, - bool is_astc_supported_) - : SurfaceBase{gpu_addr_, params_, is_astc_supported_} { - if (is_converted) { - internal_format = params.srgb_conversion ? GL_SRGB8_ALPHA8 : GL_RGBA8; - format = GL_RGBA; - type = GL_UNSIGNED_BYTE; - } else { - const auto& tuple{GetFormatTuple(params.pixel_format)}; - internal_format = tuple.internal_format; - format = tuple.format; - type = tuple.type; - is_compressed = params.IsCompressed(); - } - target = GetTextureTarget(params.target); - texture = CreateTexture(params, target, internal_format, texture_buffer); - DecorateSurfaceName(); +ImageBufferMap::ImageBufferMap(GLuint handle_, u8* map, size_t size, OGLSync* sync_) + : span(map, size), sync{sync_}, handle{handle_} {} - u32 num_layers = 1; - if (params.is_layered || params.target == SurfaceTarget::Texture3D) { - num_layers = params.depth; +ImageBufferMap::~ImageBufferMap() { + if (sync) { + sync->Create(); } - - main_view = - CreateViewInner(ViewParams(params.target, 0, num_layers, 0, params.num_levels), true); } -CachedSurface::~CachedSurface() = default; +TextureCacheRuntime::TextureCacheRuntime(const Device& device_, ProgramManager& program_manager, + StateTracker& state_tracker_) + : device{device_}, state_tracker{state_tracker_}, util_shaders(program_manager) { + static constexpr std::array TARGETS{GL_TEXTURE_1D_ARRAY, GL_TEXTURE_2D_ARRAY, GL_TEXTURE_3D}; + for (size_t i = 0; i < TARGETS.size(); ++i) { + const GLenum target = TARGETS[i]; + for (const FormatTuple& tuple : FORMAT_TABLE) { + const GLenum format = tuple.internal_format; + GLint compat_class; + GLint compat_type; + GLint is_compressed; + glGetInternalformativ(target, format, GL_IMAGE_COMPATIBILITY_CLASS, 1, &compat_class); + glGetInternalformativ(target, format, GL_IMAGE_FORMAT_COMPATIBILITY_TYPE, 1, + &compat_type); + glGetInternalformativ(target, format, GL_TEXTURE_COMPRESSED, 1, &is_compressed); + const FormatProperties properties{ + .compatibility_class = static_cast(compat_class), + .compatibility_by_size = compat_type == GL_IMAGE_FORMAT_COMPATIBILITY_BY_SIZE, + .is_compressed = is_compressed == GL_TRUE, + }; + format_properties[i].emplace(format, properties); + } + } + null_image_1d_array.Create(GL_TEXTURE_1D_ARRAY); + null_image_cube_array.Create(GL_TEXTURE_CUBE_MAP_ARRAY); + null_image_3d.Create(GL_TEXTURE_3D); + null_image_rect.Create(GL_TEXTURE_RECTANGLE); + glTextureStorage2D(null_image_1d_array.handle, 1, GL_R8, 1, 1); + glTextureStorage3D(null_image_cube_array.handle, 1, GL_R8, 1, 1, 6); + glTextureStorage3D(null_image_3d.handle, 1, GL_R8, 1, 1, 1); + glTextureStorage2D(null_image_rect.handle, 1, GL_R8, 1, 1); + + std::array new_handles; + glGenTextures(static_cast(new_handles.size()), new_handles.data()); + null_image_view_1d.handle = new_handles[0]; + null_image_view_2d.handle = new_handles[1]; + null_image_view_2d_array.handle = new_handles[2]; + null_image_view_cube.handle = new_handles[3]; + glTextureView(null_image_view_1d.handle, GL_TEXTURE_1D, null_image_1d_array.handle, GL_R8, 0, 1, + 0, 1); + glTextureView(null_image_view_2d.handle, GL_TEXTURE_2D, null_image_cube_array.handle, GL_R8, 0, + 1, 0, 1); + glTextureView(null_image_view_2d_array.handle, GL_TEXTURE_2D_ARRAY, + null_image_cube_array.handle, GL_R8, 0, 1, 0, 1); + glTextureView(null_image_view_cube.handle, GL_TEXTURE_CUBE_MAP, null_image_cube_array.handle, + GL_R8, 0, 1, 0, 6); + const std::array texture_handles{ + null_image_1d_array.handle, null_image_cube_array.handle, null_image_3d.handle, + null_image_rect.handle, null_image_view_1d.handle, null_image_view_2d.handle, + null_image_view_2d_array.handle, null_image_view_cube.handle, + }; + for (const GLuint handle : texture_handles) { + static constexpr std::array NULL_SWIZZLE{GL_ZERO, GL_ZERO, GL_ZERO, GL_ZERO}; + glTextureParameteriv(handle, GL_TEXTURE_SWIZZLE_RGBA, NULL_SWIZZLE.data()); + } + const auto set_view = [this](ImageViewType type, GLuint handle) { + if (device.HasDebuggingToolAttached()) { + const std::string name = fmt::format("NullImage {}", type); + glObjectLabel(GL_TEXTURE, handle, static_cast(name.size()), name.data()); + } + null_image_views[static_cast(type)] = handle; + }; + set_view(ImageViewType::e1D, null_image_view_1d.handle); + set_view(ImageViewType::e2D, null_image_view_2d.handle); + set_view(ImageViewType::Cube, null_image_view_cube.handle); + set_view(ImageViewType::e3D, null_image_3d.handle); + set_view(ImageViewType::e1DArray, null_image_1d_array.handle); + set_view(ImageViewType::e2DArray, null_image_view_2d_array.handle); + set_view(ImageViewType::CubeArray, null_image_cube_array.handle); + set_view(ImageViewType::Rect, null_image_rect.handle); +} -void CachedSurface::DownloadTexture(std::vector& staging_buffer) { - MICROPROFILE_SCOPE(OpenGL_Texture_Download); +TextureCacheRuntime::~TextureCacheRuntime() = default; - if (params.IsBuffer()) { - glGetNamedBufferSubData(texture_buffer.handle, 0, - static_cast(params.GetHostSizeInBytes(false)), - staging_buffer.data()); - return; - } +void TextureCacheRuntime::Finish() { + glFinish(); +} - SCOPE_EXIT({ glPixelStorei(GL_PACK_ROW_LENGTH, 0); }); +ImageBufferMap TextureCacheRuntime::MapUploadBuffer(size_t size) { + return upload_buffers.RequestMap(size, true); +} - for (u32 level = 0; level < params.emulated_levels; ++level) { - glPixelStorei(GL_PACK_ALIGNMENT, std::min(8U, params.GetRowAlignment(level, is_converted))); - glPixelStorei(GL_PACK_ROW_LENGTH, static_cast(params.GetMipWidth(level))); - const std::size_t mip_offset = params.GetHostMipmapLevelOffset(level, is_converted); +ImageBufferMap TextureCacheRuntime::MapDownloadBuffer(size_t size) { + return download_buffers.RequestMap(size, false); +} - u8* const mip_data = staging_buffer.data() + mip_offset; - const GLsizei size = static_cast(params.GetHostMipmapSize(level)); - if (is_compressed) { - glGetCompressedTextureImage(texture.handle, level, size, mip_data); - } else { - glGetTextureImage(texture.handle, level, format, type, size, mip_data); - } +void TextureCacheRuntime::CopyImage(Image& dst_image, Image& src_image, + std::span copies) { + const GLuint dst_name = dst_image.Handle(); + const GLuint src_name = src_image.Handle(); + const GLenum dst_target = ImageTarget(dst_image.info); + const GLenum src_target = ImageTarget(src_image.info); + for (const ImageCopy& copy : copies) { + const auto src_origin = MakeCopyOrigin(copy.src_offset, copy.src_subresource, src_target); + const auto dst_origin = MakeCopyOrigin(copy.dst_offset, copy.dst_subresource, dst_target); + const auto region = MakeCopyRegion(copy.extent, copy.dst_subresource, dst_target); + glCopyImageSubData(src_name, src_target, src_origin.level, src_origin.x, src_origin.y, + src_origin.z, dst_name, dst_target, dst_origin.level, dst_origin.x, + dst_origin.y, dst_origin.z, region.width, region.height, region.depth); } } -void CachedSurface::UploadTexture(const std::vector& staging_buffer) { - MICROPROFILE_SCOPE(OpenGL_Texture_Upload); - SCOPE_EXIT({ glPixelStorei(GL_UNPACK_ROW_LENGTH, 0); }); - for (u32 level = 0; level < params.emulated_levels; ++level) { - UploadTextureMipmap(level, staging_buffer); +bool TextureCacheRuntime::CanImageBeCopied(const Image& dst, const Image& src) { + if (dst.info.type == ImageType::e3D && dst.info.format == PixelFormat::BC4_UNORM) { + return false; } + return true; } -void CachedSurface::UploadTextureMipmap(u32 level, const std::vector& staging_buffer) { - glPixelStorei(GL_UNPACK_ALIGNMENT, std::min(8U, params.GetRowAlignment(level, is_converted))); - glPixelStorei(GL_UNPACK_ROW_LENGTH, static_cast(params.GetMipWidth(level))); - - const std::size_t mip_offset = params.GetHostMipmapLevelOffset(level, is_converted); - const u8* buffer{staging_buffer.data() + mip_offset}; - if (is_compressed) { - const auto image_size{static_cast(params.GetHostMipmapSize(level))}; - switch (params.target) { - case SurfaceTarget::Texture2D: - glCompressedTextureSubImage2D(texture.handle, level, 0, 0, - static_cast(params.GetMipWidth(level)), - static_cast(params.GetMipHeight(level)), - internal_format, image_size, buffer); - break; - case SurfaceTarget::Texture3D: - case SurfaceTarget::Texture2DArray: - case SurfaceTarget::TextureCubeArray: - glCompressedTextureSubImage3D(texture.handle, level, 0, 0, 0, - static_cast(params.GetMipWidth(level)), - static_cast(params.GetMipHeight(level)), - static_cast(params.GetMipDepth(level)), - internal_format, image_size, buffer); - break; - case SurfaceTarget::TextureCubemap: { - const std::size_t host_layer_size{params.GetHostLayerSize(level)}; - for (std::size_t face = 0; face < params.depth; ++face) { - glCompressedTextureSubImage3D(texture.handle, level, 0, 0, static_cast(face), - static_cast(params.GetMipWidth(level)), - static_cast(params.GetMipHeight(level)), 1, - internal_format, - static_cast(host_layer_size), buffer); - buffer += host_layer_size; - } - break; - } - default: - UNREACHABLE(); - } +void TextureCacheRuntime::EmulateCopyImage(Image& dst, Image& src, + std::span copies) { + if (dst.info.type == ImageType::e3D && dst.info.format == PixelFormat::BC4_UNORM) { + ASSERT(src.info.type == ImageType::e3D); + util_shaders.CopyBC4(dst, src, copies); } else { - switch (params.target) { - case SurfaceTarget::Texture1D: - glTextureSubImage1D(texture.handle, level, 0, params.GetMipWidth(level), format, type, - buffer); - break; - case SurfaceTarget::TextureBuffer: - ASSERT(level == 0); - glNamedBufferSubData(texture_buffer.handle, 0, - params.GetMipWidth(level) * params.GetBytesPerPixel(), buffer); - break; - case SurfaceTarget::Texture1DArray: - case SurfaceTarget::Texture2D: - glTextureSubImage2D(texture.handle, level, 0, 0, params.GetMipWidth(level), - params.GetMipHeight(level), format, type, buffer); - break; - case SurfaceTarget::Texture3D: - case SurfaceTarget::Texture2DArray: - case SurfaceTarget::TextureCubeArray: - glTextureSubImage3D( - texture.handle, level, 0, 0, 0, static_cast(params.GetMipWidth(level)), - static_cast(params.GetMipHeight(level)), - static_cast(params.GetMipDepth(level)), format, type, buffer); - break; - case SurfaceTarget::TextureCubemap: - for (std::size_t face = 0; face < params.depth; ++face) { - glTextureSubImage3D(texture.handle, level, 0, 0, static_cast(face), - params.GetMipWidth(level), params.GetMipHeight(level), 1, - format, type, buffer); - buffer += params.GetHostLayerSize(level); - } - break; - default: - UNREACHABLE(); - } + UNREACHABLE(); } } -void CachedSurface::DecorateSurfaceName() { - LabelGLObject(GL_TEXTURE, texture.handle, GetGpuAddr(), params.TargetName()); -} +void TextureCacheRuntime::BlitFramebuffer(Framebuffer* dst, Framebuffer* src, + const std::array& dst_region, + const std::array& src_region, + Tegra::Engines::Fermi2D::Filter filter, + Tegra::Engines::Fermi2D::Operation operation) { + state_tracker.NotifyScissor0(); + state_tracker.NotifyRasterizeEnable(); + state_tracker.NotifyFramebufferSRGB(); -void CachedSurfaceView::DecorateViewName(GPUVAddr gpu_addr, const std::string& prefix) { - LabelGLObject(GL_TEXTURE, main_view.handle, gpu_addr, prefix); + ASSERT(dst->BufferBits() == src->BufferBits()); + + glEnable(GL_FRAMEBUFFER_SRGB); + glDisable(GL_RASTERIZER_DISCARD); + glDisablei(GL_SCISSOR_TEST, 0); + + const GLbitfield buffer_bits = dst->BufferBits(); + const bool has_depth = (buffer_bits & ~GL_COLOR_BUFFER_BIT) != 0; + const bool is_linear = !has_depth && filter == Tegra::Engines::Fermi2D::Filter::Bilinear; + glBlitNamedFramebuffer(src->Handle(), dst->Handle(), src_region[0].x, src_region[0].y, + src_region[1].x, src_region[1].y, dst_region[0].x, dst_region[0].y, + dst_region[1].x, dst_region[1].y, buffer_bits, + is_linear ? GL_LINEAR : GL_NEAREST); } -View CachedSurface::CreateView(const ViewParams& view_key) { - return CreateViewInner(view_key, false); +void TextureCacheRuntime::AccelerateImageUpload(Image& image, const ImageBufferMap& map, + size_t buffer_offset, + std::span swizzles) { + switch (image.info.type) { + case ImageType::e2D: + return util_shaders.BlockLinearUpload2D(image, map, buffer_offset, swizzles); + case ImageType::e3D: + return util_shaders.BlockLinearUpload3D(image, map, buffer_offset, swizzles); + case ImageType::Linear: + return util_shaders.PitchUpload(image, map, buffer_offset, swizzles); + default: + UNREACHABLE(); + break; + } } -View CachedSurface::CreateViewInner(const ViewParams& view_key, const bool is_proxy) { - auto view = std::make_shared(*this, view_key, is_proxy); - views[view_key] = view; - if (!is_proxy) - view->DecorateViewName(gpu_addr, params.TargetName() + "V:" + std::to_string(view_count++)); - return view; +void TextureCacheRuntime::InsertUploadMemoryBarrier() { + glMemoryBarrier(GL_TEXTURE_FETCH_BARRIER_BIT | GL_SHADER_IMAGE_ACCESS_BARRIER_BIT); } -CachedSurfaceView::CachedSurfaceView(CachedSurface& surface_, const ViewParams& params_, - bool is_proxy_) - : ViewBase{params_}, surface{surface_}, format{surface_.internal_format}, - target{GetTextureTarget(params_.target)}, is_proxy{is_proxy_} { - if (!is_proxy_) { - main_view = CreateTextureView(); +FormatProperties TextureCacheRuntime::FormatInfo(ImageType type, GLenum internal_format) const { + switch (type) { + case ImageType::e1D: + return format_properties[0].at(internal_format); + case ImageType::e2D: + case ImageType::Linear: + return format_properties[1].at(internal_format); + case ImageType::e3D: + return format_properties[2].at(internal_format); + default: + UNREACHABLE(); + return FormatProperties{}; } } -CachedSurfaceView::~CachedSurfaceView() = default; +TextureCacheRuntime::StagingBuffers::StagingBuffers(GLenum storage_flags_, GLenum map_flags_) + : storage_flags{storage_flags_}, map_flags{map_flags_} {} -void CachedSurfaceView::Attach(GLenum attachment, GLenum fb_target) const { - ASSERT(params.num_levels == 1); +TextureCacheRuntime::StagingBuffers::~StagingBuffers() = default; - if (params.target == SurfaceTarget::Texture3D) { - if (params.num_layers > 1) { - ASSERT(params.base_layer == 0); - glFramebufferTexture(fb_target, attachment, surface.texture.handle, params.base_level); - } else { - glFramebufferTexture3D(fb_target, attachment, target, surface.texture.handle, - params.base_level, params.base_layer); - } - return; +ImageBufferMap TextureCacheRuntime::StagingBuffers::RequestMap(size_t requested_size, + bool insert_fence) { + const size_t index = RequestBuffer(requested_size); + OGLSync* const sync = insert_fence ? &syncs[index] : nullptr; + return ImageBufferMap(buffers[index].handle, maps[index], requested_size, sync); +} + +size_t TextureCacheRuntime::StagingBuffers::RequestBuffer(size_t requested_size) { + if (const std::optional index = FindBuffer(requested_size); index) { + return *index; } - if (params.num_layers > 1) { - UNIMPLEMENTED_IF(params.base_layer != 0); - glFramebufferTexture(fb_target, attachment, GetTexture(), 0); - return; + OGLBuffer& buffer = buffers.emplace_back(); + buffer.Create(); + glNamedBufferStorage(buffer.handle, requested_size, nullptr, + storage_flags | GL_MAP_PERSISTENT_BIT); + maps.push_back(static_cast(glMapNamedBufferRange(buffer.handle, 0, requested_size, + map_flags | GL_MAP_PERSISTENT_BIT))); + + syncs.emplace_back(); + sizes.push_back(requested_size); + + ASSERT(syncs.size() == buffers.size() && buffers.size() == maps.size() && + maps.size() == sizes.size()); + + return buffers.size() - 1; +} + +std::optional TextureCacheRuntime::StagingBuffers::FindBuffer(size_t requested_size) { + size_t smallest_buffer = std::numeric_limits::max(); + std::optional found; + const size_t num_buffers = sizes.size(); + for (size_t index = 0; index < num_buffers; ++index) { + const size_t buffer_size = sizes[index]; + if (buffer_size < requested_size || buffer_size >= smallest_buffer) { + continue; + } + if (syncs[index].handle != 0) { + GLint status; + glGetSynciv(syncs[index].handle, GL_SYNC_STATUS, 1, nullptr, &status); + if (status != GL_SIGNALED) { + continue; + } + syncs[index].Release(); + } + smallest_buffer = buffer_size; + found = index; } + return found; +} - const GLenum view_target = surface.GetTarget(); - const GLuint texture = surface.GetTexture(); - switch (surface.GetSurfaceParams().target) { - case SurfaceTarget::Texture1D: - glFramebufferTexture1D(fb_target, attachment, view_target, texture, params.base_level); +Image::Image(TextureCacheRuntime& runtime, const VideoCommon::ImageInfo& info_, GPUVAddr gpu_addr_, + VAddr cpu_addr_) + : VideoCommon::ImageBase(info_, gpu_addr_, cpu_addr_) { + if (CanBeAccelerated(runtime, info)) { + flags |= ImageFlagBits::AcceleratedUpload; + } + if (IsConverted(runtime.device, info.format, info.type)) { + flags |= ImageFlagBits::Converted; + gl_internal_format = IsPixelFormatSRGB(info.format) ? GL_SRGB8_ALPHA8 : GL_RGBA8; + gl_store_format = GL_RGBA8; + gl_format = GL_RGBA; + gl_type = GL_UNSIGNED_INT_8_8_8_8_REV; + } else { + const auto& tuple = GetFormatTuple(info.format); + gl_internal_format = tuple.internal_format; + gl_store_format = tuple.store_format; + gl_format = tuple.format; + gl_type = tuple.type; + } + const GLenum target = ImageTarget(info); + const GLsizei width = info.size.width; + const GLsizei height = info.size.height; + const GLsizei depth = info.size.depth; + const int max_host_mip_levels = std::bit_width(info.size.width); + const GLsizei num_levels = std::min(info.resources.levels, max_host_mip_levels); + const GLsizei num_layers = info.resources.layers; + const GLsizei num_samples = info.num_samples; + + GLuint handle = 0; + if (target != GL_TEXTURE_BUFFER) { + texture.Create(target); + handle = texture.handle; + } + switch (target) { + case GL_TEXTURE_1D_ARRAY: + glTextureStorage2D(handle, num_levels, gl_store_format, width, num_layers); break; - case SurfaceTarget::Texture2D: - glFramebufferTexture2D(fb_target, attachment, view_target, texture, params.base_level); + case GL_TEXTURE_2D_ARRAY: + glTextureStorage3D(handle, num_levels, gl_store_format, width, height, num_layers); break; - case SurfaceTarget::Texture1DArray: - case SurfaceTarget::Texture2DArray: - case SurfaceTarget::TextureCubemap: - case SurfaceTarget::TextureCubeArray: - glFramebufferTextureLayer(fb_target, attachment, texture, params.base_level, - params.base_layer); + case GL_TEXTURE_2D_MULTISAMPLE_ARRAY: { + // TODO: Where should 'fixedsamplelocations' come from? + const auto [samples_x, samples_y] = SamplesLog2(info.num_samples); + glTextureStorage3DMultisample(handle, num_samples, gl_store_format, width >> samples_x, + height >> samples_y, num_layers, GL_FALSE); + break; + } + case GL_TEXTURE_RECTANGLE: + glTextureStorage2D(handle, num_levels, gl_store_format, width, height); + break; + case GL_TEXTURE_3D: + glTextureStorage3D(handle, num_levels, gl_store_format, width, height, depth); + break; + case GL_TEXTURE_BUFFER: + buffer.Create(); + glNamedBufferStorage(buffer.handle, guest_size_bytes, nullptr, 0); break; default: - UNIMPLEMENTED(); + UNREACHABLE_MSG("Invalid target=0x{:x}", target); + break; + } + if (runtime.device.HasDebuggingToolAttached()) { + const std::string name = VideoCommon::Name(*this); + glObjectLabel(target == GL_TEXTURE_BUFFER ? GL_BUFFER : GL_TEXTURE, handle, + static_cast(name.size()), name.data()); } } -GLuint CachedSurfaceView::GetTexture(SwizzleSource x_source, SwizzleSource y_source, - SwizzleSource z_source, SwizzleSource w_source) { - if (GetSurfaceParams().IsBuffer()) { - return GetTexture(); - } - const u32 new_swizzle = EncodeSwizzle(x_source, y_source, z_source, w_source); - if (current_swizzle == new_swizzle) { - return current_view; - } - current_swizzle = new_swizzle; +void Image::UploadMemory(const ImageBufferMap& map, size_t buffer_offset, + std::span copies) { + glBindBuffer(GL_PIXEL_UNPACK_BUFFER, map.Handle()); + glFlushMappedBufferRange(GL_PIXEL_UNPACK_BUFFER, buffer_offset, unswizzled_size_bytes); - const auto [entry, is_cache_miss] = view_cache.try_emplace(new_swizzle); - OGLTextureView& view = entry->second; - if (!is_cache_miss) { - current_view = view.handle; - return view.handle; - } - view = CreateTextureView(); - current_view = view.handle; + glPixelStorei(GL_UNPACK_ALIGNMENT, 1); - std::array swizzle{x_source, y_source, z_source, w_source}; + u32 current_row_length = std::numeric_limits::max(); + u32 current_image_height = std::numeric_limits::max(); - switch (const PixelFormat pixel_format = GetSurfaceParams().pixel_format) { - case PixelFormat::D24_UNORM_S8_UINT: - case PixelFormat::D32_FLOAT_S8_UINT: - case PixelFormat::S8_UINT_D24_UNORM: - UNIMPLEMENTED_IF(x_source != SwizzleSource::R && x_source != SwizzleSource::G); - glTextureParameteri(view.handle, GL_DEPTH_STENCIL_TEXTURE_MODE, - GetComponent(pixel_format, x_source == SwizzleSource::R)); - - // Make sure we sample the first component - std::transform(swizzle.begin(), swizzle.end(), swizzle.begin(), [](SwizzleSource value) { - return value == SwizzleSource::G ? SwizzleSource::R : value; - }); - [[fallthrough]]; - default: { - const std::array gl_swizzle = {GetSwizzleSource(swizzle[0]), GetSwizzleSource(swizzle[1]), - GetSwizzleSource(swizzle[2]), GetSwizzleSource(swizzle[3])}; - glTextureParameteriv(view.handle, GL_TEXTURE_SWIZZLE_RGBA, gl_swizzle.data()); - break; - } + for (const VideoCommon::BufferImageCopy& copy : copies) { + if (current_row_length != copy.buffer_row_length) { + current_row_length = copy.buffer_row_length; + glPixelStorei(GL_UNPACK_ROW_LENGTH, current_row_length); + } + if (current_image_height != copy.buffer_image_height) { + current_image_height = copy.buffer_image_height; + glPixelStorei(GL_UNPACK_IMAGE_HEIGHT, current_image_height); + } + CopyBufferToImage(copy, buffer_offset); } - return view.handle; } -OGLTextureView CachedSurfaceView::CreateTextureView() const { - OGLTextureView texture_view; - texture_view.Create(); - - if (target == GL_TEXTURE_3D) { - glTextureView(texture_view.handle, target, surface.texture.handle, format, - params.base_level, params.num_levels, 0, 1); - } else { - glTextureView(texture_view.handle, target, surface.texture.handle, format, - params.base_level, params.num_levels, params.base_layer, params.num_layers); +void Image::UploadMemory(const ImageBufferMap& map, size_t buffer_offset, + std::span copies) { + for (const VideoCommon::BufferCopy& copy : copies) { + glCopyNamedBufferSubData(map.Handle(), buffer.handle, copy.src_offset + buffer_offset, + copy.dst_offset, copy.size); } - ApplyTextureDefaults(surface.GetSurfaceParams(), texture_view.handle); - - return texture_view; } -TextureCacheOpenGL::TextureCacheOpenGL(VideoCore::RasterizerInterface& rasterizer_, - Tegra::Engines::Maxwell3D& maxwell3d_, - Tegra::MemoryManager& gpu_memory_, const Device& device_, - StateTracker& state_tracker_) - : TextureCacheBase{rasterizer_, maxwell3d_, gpu_memory_, device_.HasASTC()}, - state_tracker{state_tracker_} { - src_framebuffer.Create(); - dst_framebuffer.Create(); -} +void Image::DownloadMemory(ImageBufferMap& map, size_t buffer_offset, + std::span copies) { + glMemoryBarrier(GL_PIXEL_BUFFER_BARRIER_BIT); // TODO: Move this to its own API -TextureCacheOpenGL::~TextureCacheOpenGL() = default; + glBindBuffer(GL_PIXEL_PACK_BUFFER, map.Handle()); + glPixelStorei(GL_PACK_ALIGNMENT, 1); -Surface TextureCacheOpenGL::CreateSurface(GPUVAddr gpu_addr, const SurfaceParams& params) { - return std::make_shared(gpu_addr, params, is_astc_supported); -} + u32 current_row_length = std::numeric_limits::max(); + u32 current_image_height = std::numeric_limits::max(); -void TextureCacheOpenGL::ImageCopy(Surface& src_surface, Surface& dst_surface, - const VideoCommon::CopyParams& copy_params) { - const auto& src_params = src_surface->GetSurfaceParams(); - const auto& dst_params = dst_surface->GetSurfaceParams(); - if (src_params.type != dst_params.type) { - // A fallback is needed - return; + for (const VideoCommon::BufferImageCopy& copy : copies) { + if (current_row_length != copy.buffer_row_length) { + current_row_length = copy.buffer_row_length; + glPixelStorei(GL_PACK_ROW_LENGTH, current_row_length); + } + if (current_image_height != copy.buffer_image_height) { + current_image_height = copy.buffer_image_height; + glPixelStorei(GL_PACK_IMAGE_HEIGHT, current_image_height); + } + CopyImageToBuffer(copy, buffer_offset); } - const auto src_handle = src_surface->GetTexture(); - const auto src_target = src_surface->GetTarget(); - const auto dst_handle = dst_surface->GetTexture(); - const auto dst_target = dst_surface->GetTarget(); - glCopyImageSubData(src_handle, src_target, copy_params.source_level, copy_params.source_x, - copy_params.source_y, copy_params.source_z, dst_handle, dst_target, - copy_params.dest_level, copy_params.dest_x, copy_params.dest_y, - copy_params.dest_z, copy_params.width, copy_params.height, - copy_params.depth); } -void TextureCacheOpenGL::ImageBlit(View& src_view, View& dst_view, - const Tegra::Engines::Fermi2D::Config& copy_config) { - const auto& src_params{src_view->GetSurfaceParams()}; - const auto& dst_params{dst_view->GetSurfaceParams()}; - UNIMPLEMENTED_IF(src_params.depth != 1); - UNIMPLEMENTED_IF(dst_params.depth != 1); - - state_tracker.NotifyScissor0(); - state_tracker.NotifyFramebuffer(); - state_tracker.NotifyRasterizeEnable(); - state_tracker.NotifyFramebufferSRGB(); +void Image::CopyBufferToImage(const VideoCommon::BufferImageCopy& copy, size_t buffer_offset) { + // Compressed formats don't have a pixel format or type + const bool is_compressed = gl_format == GL_NONE; + const void* const offset = reinterpret_cast(copy.buffer_offset + buffer_offset); - if (dst_params.srgb_conversion) { - glEnable(GL_FRAMEBUFFER_SRGB); - } else { - glDisable(GL_FRAMEBUFFER_SRGB); + switch (info.type) { + case ImageType::e1D: + if (is_compressed) { + glCompressedTextureSubImage2D(texture.handle, copy.image_subresource.base_level, + copy.image_offset.x, copy.image_subresource.base_layer, + copy.image_extent.width, + copy.image_subresource.num_layers, gl_internal_format, + static_cast(copy.buffer_size), offset); + } else { + glTextureSubImage2D(texture.handle, copy.image_subresource.base_level, + copy.image_offset.x, copy.image_subresource.base_layer, + copy.image_extent.width, copy.image_subresource.num_layers, + gl_format, gl_type, offset); + } + break; + case ImageType::e2D: + case ImageType::Linear: + if (is_compressed) { + glCompressedTextureSubImage3D( + texture.handle, copy.image_subresource.base_level, copy.image_offset.x, + copy.image_offset.y, copy.image_subresource.base_layer, copy.image_extent.width, + copy.image_extent.height, copy.image_subresource.num_layers, gl_internal_format, + static_cast(copy.buffer_size), offset); + } else { + glTextureSubImage3D(texture.handle, copy.image_subresource.base_level, + copy.image_offset.x, copy.image_offset.y, + copy.image_subresource.base_layer, copy.image_extent.width, + copy.image_extent.height, copy.image_subresource.num_layers, + gl_format, gl_type, offset); + } + break; + case ImageType::e3D: + if (is_compressed) { + glCompressedTextureSubImage3D( + texture.handle, copy.image_subresource.base_level, copy.image_offset.x, + copy.image_offset.y, copy.image_offset.z, copy.image_extent.width, + copy.image_extent.height, copy.image_extent.depth, gl_internal_format, + static_cast(copy.buffer_size), offset); + } else { + glTextureSubImage3D(texture.handle, copy.image_subresource.base_level, + copy.image_offset.x, copy.image_offset.y, copy.image_offset.z, + copy.image_extent.width, copy.image_extent.height, + copy.image_extent.depth, gl_format, gl_type, offset); + } + break; + default: + UNREACHABLE(); } - glDisable(GL_RASTERIZER_DISCARD); - glDisablei(GL_SCISSOR_TEST, 0); - - glBindFramebuffer(GL_READ_FRAMEBUFFER, src_framebuffer.handle); - glBindFramebuffer(GL_DRAW_FRAMEBUFFER, dst_framebuffer.handle); - - GLenum buffers = 0; - if (src_params.type == SurfaceType::ColorTexture) { - src_view->Attach(GL_COLOR_ATTACHMENT0, GL_READ_FRAMEBUFFER); - glFramebufferTexture2D(GL_READ_FRAMEBUFFER, GL_DEPTH_STENCIL_ATTACHMENT, GL_TEXTURE_2D, 0, - 0); - - dst_view->Attach(GL_COLOR_ATTACHMENT0, GL_DRAW_FRAMEBUFFER); - glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_DEPTH_STENCIL_ATTACHMENT, GL_TEXTURE_2D, 0, - 0); - - buffers = GL_COLOR_BUFFER_BIT; - } else if (src_params.type == SurfaceType::Depth) { - glFramebufferTexture2D(GL_READ_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D, 0, 0); - src_view->Attach(GL_DEPTH_ATTACHMENT, GL_READ_FRAMEBUFFER); - glFramebufferTexture2D(GL_READ_FRAMEBUFFER, GL_STENCIL_ATTACHMENT, GL_TEXTURE_2D, 0, 0); +} - glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D, 0, 0); - dst_view->Attach(GL_DEPTH_ATTACHMENT, GL_DRAW_FRAMEBUFFER); - glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_STENCIL_ATTACHMENT, GL_TEXTURE_2D, 0, 0); +void Image::CopyImageToBuffer(const VideoCommon::BufferImageCopy& copy, size_t buffer_offset) { + const GLint x_offset = copy.image_offset.x; + const GLsizei width = copy.image_extent.width; - buffers = GL_DEPTH_BUFFER_BIT; - } else if (src_params.type == SurfaceType::DepthStencil) { - glFramebufferTexture2D(GL_READ_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D, 0, 0); - src_view->Attach(GL_DEPTH_STENCIL_ATTACHMENT, GL_READ_FRAMEBUFFER); + const GLint level = copy.image_subresource.base_level; + const GLsizei buffer_size = static_cast(copy.buffer_size); + void* const offset = reinterpret_cast(copy.buffer_offset + buffer_offset); - glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D, 0, 0); - dst_view->Attach(GL_DEPTH_STENCIL_ATTACHMENT, GL_DRAW_FRAMEBUFFER); + GLint y_offset = 0; + GLint z_offset = 0; + GLsizei height = 1; + GLsizei depth = 1; - buffers = GL_DEPTH_BUFFER_BIT | GL_STENCIL_BUFFER_BIT; + switch (info.type) { + case ImageType::e1D: + y_offset = copy.image_subresource.base_layer; + height = copy.image_subresource.num_layers; + break; + case ImageType::e2D: + case ImageType::Linear: + y_offset = copy.image_offset.y; + z_offset = copy.image_subresource.base_layer; + height = copy.image_extent.height; + depth = copy.image_subresource.num_layers; + break; + case ImageType::e3D: + y_offset = copy.image_offset.y; + z_offset = copy.image_offset.z; + height = copy.image_extent.height; + depth = copy.image_extent.depth; + break; + default: + UNREACHABLE(); + } + // Compressed formats don't have a pixel format or type + const bool is_compressed = gl_format == GL_NONE; + if (is_compressed) { + glGetCompressedTextureSubImage(texture.handle, level, x_offset, y_offset, z_offset, width, + height, depth, buffer_size, offset); + } else { + glGetTextureSubImage(texture.handle, level, x_offset, y_offset, z_offset, width, height, + depth, gl_format, gl_type, buffer_size, offset); } - - const Common::Rectangle& src_rect = copy_config.src_rect; - const Common::Rectangle& dst_rect = copy_config.dst_rect; - const bool is_linear = copy_config.filter == Tegra::Engines::Fermi2D::Filter::Linear; - - glBlitFramebuffer(static_cast(src_rect.left), static_cast(src_rect.top), - static_cast(src_rect.right), static_cast(src_rect.bottom), - static_cast(dst_rect.left), static_cast(dst_rect.top), - static_cast(dst_rect.right), static_cast(dst_rect.bottom), - buffers, - is_linear && (buffers == GL_COLOR_BUFFER_BIT) ? GL_LINEAR : GL_NEAREST); } -void TextureCacheOpenGL::BufferCopy(Surface& src_surface, Surface& dst_surface) { - MICROPROFILE_SCOPE(OpenGL_Texture_Buffer_Copy); - const auto& src_params = src_surface->GetSurfaceParams(); - const auto& dst_params = dst_surface->GetSurfaceParams(); - UNIMPLEMENTED_IF(src_params.num_levels > 1 || dst_params.num_levels > 1); +ImageView::ImageView(TextureCacheRuntime& runtime, const VideoCommon::ImageViewInfo& info, + ImageId image_id_, Image& image) + : VideoCommon::ImageViewBase{info, image.info, image_id_}, views{runtime.null_image_views} { + const Device& device = runtime.device; + if (True(image.flags & ImageFlagBits::Converted)) { + internal_format = IsPixelFormatSRGB(info.format) ? GL_SRGB8_ALPHA8 : GL_RGBA8; + } else { + internal_format = GetFormatTuple(format).internal_format; + } + VideoCommon::SubresourceRange flatten_range = info.range; + std::array handles; + stored_views.reserve(2); - const auto source_format = GetFormatTuple(src_params.pixel_format); - const auto dest_format = GetFormatTuple(dst_params.pixel_format); + switch (info.type) { + case ImageViewType::e1DArray: + flatten_range.extent.layers = 1; + [[fallthrough]]; + case ImageViewType::e1D: + glGenTextures(2, handles.data()); + SetupView(device, image, ImageViewType::e1D, handles[0], info, flatten_range); + SetupView(device, image, ImageViewType::e1DArray, handles[1], info, info.range); + break; + case ImageViewType::e2DArray: + flatten_range.extent.layers = 1; + [[fallthrough]]; + case ImageViewType::e2D: + if (True(flags & VideoCommon::ImageViewFlagBits::Slice)) { + // 2D and 2D array views on a 3D textures are used exclusively for render targets + ASSERT(info.range.extent.levels == 1); + const VideoCommon::SubresourceRange slice_range{ + .base = {.level = info.range.base.level, .layer = 0}, + .extent = {.levels = 1, .layers = 1}, + }; + glGenTextures(1, handles.data()); + SetupView(device, image, ImageViewType::e3D, handles[0], info, slice_range); + break; + } + glGenTextures(2, handles.data()); + SetupView(device, image, ImageViewType::e2D, handles[0], info, flatten_range); + SetupView(device, image, ImageViewType::e2DArray, handles[1], info, info.range); + break; + case ImageViewType::e3D: + glGenTextures(1, handles.data()); + SetupView(device, image, ImageViewType::e3D, handles[0], info, info.range); + break; + case ImageViewType::CubeArray: + flatten_range.extent.layers = 6; + [[fallthrough]]; + case ImageViewType::Cube: + glGenTextures(2, handles.data()); + SetupView(device, image, ImageViewType::Cube, handles[0], info, flatten_range); + SetupView(device, image, ImageViewType::CubeArray, handles[1], info, info.range); + break; + case ImageViewType::Rect: + glGenTextures(1, handles.data()); + SetupView(device, image, ImageViewType::Rect, handles[0], info, info.range); + break; + case ImageViewType::Buffer: + glCreateTextures(GL_TEXTURE_BUFFER, 1, handles.data()); + SetupView(device, image, ImageViewType::Buffer, handles[0], info, info.range); + break; + } + default_handle = Handle(info.type); +} - const std::size_t source_size = src_surface->GetHostSizeInBytes(); - const std::size_t dest_size = dst_surface->GetHostSizeInBytes(); +ImageView::ImageView(TextureCacheRuntime& runtime, const VideoCommon::NullImageParams& params) + : VideoCommon::ImageViewBase{params}, views{runtime.null_image_views} {} - const std::size_t buffer_size = std::max(source_size, dest_size); +void ImageView::SetupView(const Device& device, Image& image, ImageViewType view_type, + GLuint handle, const VideoCommon::ImageViewInfo& info, + VideoCommon::SubresourceRange view_range) { + if (info.type == ImageViewType::Buffer) { + // TODO: Take offset from buffer cache + glTextureBufferRange(handle, internal_format, image.buffer.handle, 0, + image.guest_size_bytes); + } else { + const GLuint parent = image.texture.handle; + const GLenum target = ImageTarget(view_type, image.info.num_samples); + glTextureView(handle, target, parent, internal_format, view_range.base.level, + view_range.extent.levels, view_range.base.layer, view_range.extent.layers); + if (!info.IsRenderTarget()) { + ApplySwizzle(handle, format, info.Swizzle()); + } + } + if (device.HasDebuggingToolAttached()) { + const std::string name = VideoCommon::Name(*this, view_type); + glObjectLabel(GL_TEXTURE, handle, static_cast(name.size()), name.data()); + } + stored_views.emplace_back().handle = handle; + views[static_cast(view_type)] = handle; +} - GLuint copy_pbo_handle = FetchPBO(buffer_size); +Sampler::Sampler(TextureCacheRuntime& runtime, const TSCEntry& config) { + const GLenum compare_mode = config.depth_compare_enabled ? GL_COMPARE_REF_TO_TEXTURE : GL_NONE; + const GLenum compare_func = MaxwellToGL::DepthCompareFunc(config.depth_compare_func); + const GLenum mag = MaxwellToGL::TextureFilterMode(config.mag_filter, TextureMipmapFilter::None); + const GLenum min = MaxwellToGL::TextureFilterMode(config.min_filter, config.mipmap_filter); + const GLenum reduction_filter = MaxwellToGL::ReductionFilter(config.reduction_filter); + const GLint seamless = config.cubemap_interface_filtering ? GL_TRUE : GL_FALSE; + + UNIMPLEMENTED_IF(config.cubemap_anisotropy != 1); + UNIMPLEMENTED_IF(config.float_coord_normalization != 0); + + sampler.Create(); + const GLuint handle = sampler.handle; + glSamplerParameteri(handle, GL_TEXTURE_WRAP_S, MaxwellToGL::WrapMode(config.wrap_u)); + glSamplerParameteri(handle, GL_TEXTURE_WRAP_T, MaxwellToGL::WrapMode(config.wrap_v)); + glSamplerParameteri(handle, GL_TEXTURE_WRAP_R, MaxwellToGL::WrapMode(config.wrap_p)); + glSamplerParameteri(handle, GL_TEXTURE_COMPARE_MODE, compare_mode); + glSamplerParameteri(handle, GL_TEXTURE_COMPARE_FUNC, compare_func); + glSamplerParameteri(handle, GL_TEXTURE_MAG_FILTER, mag); + glSamplerParameteri(handle, GL_TEXTURE_MIN_FILTER, min); + glSamplerParameterf(handle, GL_TEXTURE_LOD_BIAS, config.LodBias()); + glSamplerParameterf(handle, GL_TEXTURE_MIN_LOD, config.MinLod()); + glSamplerParameterf(handle, GL_TEXTURE_MAX_LOD, config.MaxLod()); + glSamplerParameterfv(handle, GL_TEXTURE_BORDER_COLOR, config.BorderColor().data()); + + if (GLAD_GL_ARB_texture_filter_anisotropic || GLAD_GL_EXT_texture_filter_anisotropic) { + glSamplerParameterf(handle, GL_TEXTURE_MAX_ANISOTROPY, config.MaxAnisotropy()); + } else { + LOG_WARNING(Render_OpenGL, "GL_ARB_texture_filter_anisotropic is required"); + } + if (GLAD_GL_ARB_texture_filter_minmax || GLAD_GL_EXT_texture_filter_minmax) { + glSamplerParameteri(handle, GL_TEXTURE_REDUCTION_MODE_ARB, reduction_filter); + } else if (reduction_filter != GL_WEIGHTED_AVERAGE_ARB) { + LOG_WARNING(Render_OpenGL, "GL_ARB_texture_filter_minmax is required"); + } + if (GLAD_GL_ARB_seamless_cubemap_per_texture || GLAD_GL_AMD_seamless_cubemap_per_texture) { + glSamplerParameteri(handle, GL_TEXTURE_CUBE_MAP_SEAMLESS, seamless); + } else if (seamless == GL_FALSE) { + // We default to false because it's more common + LOG_WARNING(Render_OpenGL, "GL_ARB_seamless_cubemap_per_texture is required"); + } +} - glBindBuffer(GL_PIXEL_PACK_BUFFER, copy_pbo_handle); +Framebuffer::Framebuffer(TextureCacheRuntime& runtime, std::span color_buffers, + ImageView* depth_buffer, const VideoCommon::RenderTargets& key) { + // Bind to READ_FRAMEBUFFER to stop Nvidia's driver from creating an EXT_framebuffer instead of + // a core framebuffer. EXT framebuffer attachments have to match in size and can be shared + // across contexts. yuzu doesn't share framebuffers across contexts and we need attachments with + // mismatching size, this is why core framebuffers are preferred. + GLuint handle; + glGenFramebuffers(1, &handle); + glBindFramebuffer(GL_READ_FRAMEBUFFER, handle); + + GLsizei num_buffers = 0; + std::array gl_draw_buffers; + gl_draw_buffers.fill(GL_NONE); + + for (size_t index = 0; index < color_buffers.size(); ++index) { + const ImageView* const image_view = color_buffers[index]; + if (!image_view) { + continue; + } + buffer_bits |= GL_COLOR_BUFFER_BIT; + gl_draw_buffers[index] = GL_COLOR_ATTACHMENT0 + key.draw_buffers[index]; + num_buffers = static_cast(index + 1); - if (src_surface->IsCompressed()) { - glGetCompressedTextureImage(src_surface->GetTexture(), 0, static_cast(source_size), - nullptr); - } else { - glGetTextureImage(src_surface->GetTexture(), 0, source_format.format, source_format.type, - static_cast(source_size), nullptr); + const GLenum attachment = static_cast(GL_COLOR_ATTACHMENT0 + index); + AttachTexture(handle, attachment, image_view); } - glBindBuffer(GL_PIXEL_PACK_BUFFER, 0); - glBindBuffer(GL_PIXEL_UNPACK_BUFFER, copy_pbo_handle); + if (const ImageView* const image_view = depth_buffer; image_view) { + if (GetFormatType(image_view->format) == SurfaceType::DepthStencil) { + buffer_bits |= GL_DEPTH_BUFFER_BIT | GL_STENCIL_BUFFER_BIT; + } else { + buffer_bits |= GL_DEPTH_BUFFER_BIT; + } + const GLenum attachment = AttachmentType(image_view->format); + AttachTexture(handle, attachment, image_view); + } - const GLsizei width = static_cast(dst_params.width); - const GLsizei height = static_cast(dst_params.height); - const GLsizei depth = static_cast(dst_params.depth); - if (dst_surface->IsCompressed()) { - LOG_CRITICAL(HW_GPU, "Compressed buffer copy is unimplemented!"); - UNREACHABLE(); + if (num_buffers > 1) { + glNamedFramebufferDrawBuffers(handle, num_buffers, gl_draw_buffers.data()); + } else if (num_buffers > 0) { + glNamedFramebufferDrawBuffer(handle, gl_draw_buffers[0]); } else { - switch (dst_params.target) { - case SurfaceTarget::Texture1D: - glTextureSubImage1D(dst_surface->GetTexture(), 0, 0, width, dest_format.format, - dest_format.type, nullptr); - break; - case SurfaceTarget::Texture2D: - glTextureSubImage2D(dst_surface->GetTexture(), 0, 0, 0, width, height, - dest_format.format, dest_format.type, nullptr); - break; - case SurfaceTarget::Texture3D: - case SurfaceTarget::Texture2DArray: - case SurfaceTarget::TextureCubeArray: - glTextureSubImage3D(dst_surface->GetTexture(), 0, 0, 0, 0, width, height, depth, - dest_format.format, dest_format.type, nullptr); - break; - case SurfaceTarget::TextureCubemap: - glTextureSubImage3D(dst_surface->GetTexture(), 0, 0, 0, 0, width, height, depth, - dest_format.format, dest_format.type, nullptr); - break; - default: - LOG_CRITICAL(Render_OpenGL, "Unimplemented surface target={}", dst_params.target); - UNREACHABLE(); - } + glNamedFramebufferDrawBuffer(handle, GL_NONE); } - glBindBuffer(GL_PIXEL_UNPACK_BUFFER, 0); - glTextureBarrier(); -} + glNamedFramebufferParameteri(handle, GL_FRAMEBUFFER_DEFAULT_WIDTH, key.size.width); + glNamedFramebufferParameteri(handle, GL_FRAMEBUFFER_DEFAULT_HEIGHT, key.size.height); + // TODO + // glNamedFramebufferParameteri(handle, GL_FRAMEBUFFER_DEFAULT_LAYERS, ...); + // glNamedFramebufferParameteri(handle, GL_FRAMEBUFFER_DEFAULT_SAMPLES, ...); + // glNamedFramebufferParameteri(handle, GL_FRAMEBUFFER_DEFAULT_FIXED_SAMPLE_LOCATIONS, ...); -GLuint TextureCacheOpenGL::FetchPBO(std::size_t buffer_size) { - ASSERT_OR_EXECUTE(buffer_size > 0, { return 0; }); - const u32 l2 = Common::Log2Ceil64(static_cast(buffer_size)); - OGLBuffer& cp = copy_pbo_cache[l2]; - if (cp.handle == 0) { - const std::size_t ceil_size = 1ULL << l2; - cp.Create(); - cp.MakeStreamCopy(ceil_size); + if (runtime.device.HasDebuggingToolAttached()) { + const std::string name = VideoCommon::Name(key); + glObjectLabel(GL_FRAMEBUFFER, handle, static_cast(name.size()), name.data()); } - return cp.handle; + framebuffer.handle = handle; } } // namespace OpenGL -- cgit v1.2.3