summaryrefslogtreecommitdiffstats
path: root/src/video_core/renderer_opengl
diff options
context:
space:
mode:
authorFernando Sahmkow <fsahmkow27@gmail.com>2019-06-13 16:26:17 +0200
committerReinUsesLisp <reinuseslisp@airmail.cc>2019-06-21 02:38:34 +0200
commit4db28f72f617b1500581e621719928fa0807d9ac (patch)
treec282bb335ed9f714610db3f010561c71506b1647 /src/video_core/renderer_opengl
parenttexture_cache: Implement siblings texture formats. (diff)
downloadyuzu-4db28f72f617b1500581e621719928fa0807d9ac.tar
yuzu-4db28f72f617b1500581e621719928fa0807d9ac.tar.gz
yuzu-4db28f72f617b1500581e621719928fa0807d9ac.tar.bz2
yuzu-4db28f72f617b1500581e621719928fa0807d9ac.tar.lz
yuzu-4db28f72f617b1500581e621719928fa0807d9ac.tar.xz
yuzu-4db28f72f617b1500581e621719928fa0807d9ac.tar.zst
yuzu-4db28f72f617b1500581e621719928fa0807d9ac.zip
Diffstat (limited to 'src/video_core/renderer_opengl')
-rw-r--r--src/video_core/renderer_opengl/gl_rasterizer_cache.cpp1381
-rw-r--r--src/video_core/renderer_opengl/gl_rasterizer_cache.h575
2 files changed, 0 insertions, 1956 deletions
diff --git a/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp b/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp
deleted file mode 100644
index e27da1fa7..000000000
--- a/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp
+++ /dev/null
@@ -1,1381 +0,0 @@
-// Copyright 2018 yuzu Emulator Project
-// Licensed under GPLv2 or any later version
-// Refer to the license.txt file included.
-
-#include <algorithm>
-#include <optional>
-#include <glad/glad.h>
-
-#include "common/alignment.h"
-#include "common/assert.h"
-#include "common/logging/log.h"
-#include "common/microprofile.h"
-#include "common/scope_exit.h"
-#include "core/core.h"
-#include "core/hle/kernel/process.h"
-#include "core/settings.h"
-#include "video_core/engines/maxwell_3d.h"
-#include "video_core/memory_manager.h"
-#include "video_core/morton.h"
-#include "video_core/renderer_opengl/gl_rasterizer.h"
-#include "video_core/renderer_opengl/gl_rasterizer_cache.h"
-#include "video_core/renderer_opengl/utils.h"
-#include "video_core/surface.h"
-#include "video_core/textures/convert.h"
-#include "video_core/textures/decoders.h"
-
-namespace OpenGL {
-
-using VideoCore::MortonSwizzle;
-using VideoCore::MortonSwizzleMode;
-using VideoCore::Surface::ComponentTypeFromDepthFormat;
-using VideoCore::Surface::ComponentTypeFromRenderTarget;
-using VideoCore::Surface::ComponentTypeFromTexture;
-using VideoCore::Surface::PixelFormatFromDepthFormat;
-using VideoCore::Surface::PixelFormatFromRenderTargetFormat;
-using VideoCore::Surface::PixelFormatFromTextureFormat;
-using VideoCore::Surface::SurfaceTargetFromTextureType;
-
-struct FormatTuple {
- GLint internal_format;
- GLenum format;
- GLenum type;
- ComponentType component_type;
- bool compressed;
-};
-
-static void ApplyTextureDefaults(GLuint texture, u32 max_mip_level) {
- 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, max_mip_level - 1);
- if (max_mip_level == 1) {
- glTextureParameterf(texture, GL_TEXTURE_LOD_BIAS, 1000.0);
- }
-}
-
-void SurfaceParams::InitCacheParameters(GPUVAddr gpu_addr_) {
- auto& memory_manager{Core::System::GetInstance().GPU().MemoryManager()};
-
- gpu_addr = gpu_addr_;
- host_ptr = memory_manager.GetPointer(gpu_addr_);
- size_in_bytes = SizeInBytesRaw();
-
- if (IsPixelFormatASTC(pixel_format)) {
- // ASTC is uncompressed in software, in emulated as RGBA8
- size_in_bytes_gl = width * height * depth * 4;
- } else {
- size_in_bytes_gl = SizeInBytesGL();
- }
-}
-
-std::size_t SurfaceParams::InnerMipmapMemorySize(u32 mip_level, bool force_gl, bool layer_only,
- bool uncompressed) const {
- const u32 tile_x{GetDefaultBlockWidth(pixel_format)};
- const u32 tile_y{GetDefaultBlockHeight(pixel_format)};
- const u32 bytes_per_pixel{GetBytesPerPixel(pixel_format)};
- u32 m_depth = (layer_only ? 1U : depth);
- u32 m_width = MipWidth(mip_level);
- u32 m_height = MipHeight(mip_level);
- m_width = uncompressed ? m_width : std::max(1U, (m_width + tile_x - 1) / tile_x);
- m_height = uncompressed ? m_height : std::max(1U, (m_height + tile_y - 1) / tile_y);
- m_depth = std::max(1U, m_depth >> mip_level);
- u32 m_block_height = MipBlockHeight(mip_level);
- u32 m_block_depth = MipBlockDepth(mip_level);
- return Tegra::Texture::CalculateSize(force_gl ? false : is_tiled, bytes_per_pixel, m_width,
- m_height, m_depth, m_block_height, m_block_depth);
-}
-
-std::size_t SurfaceParams::InnerMemorySize(bool force_gl, bool layer_only,
- bool uncompressed) const {
- std::size_t block_size_bytes = Tegra::Texture::GetGOBSize() * block_height * block_depth;
- std::size_t size = 0;
- for (u32 i = 0; i < max_mip_level; i++) {
- size += InnerMipmapMemorySize(i, force_gl, layer_only, uncompressed);
- }
- if (!force_gl && is_tiled) {
- size = Common::AlignUp(size, block_size_bytes);
- }
- return size;
-}
-
-/*static*/ SurfaceParams SurfaceParams::CreateForTexture(
- const Tegra::Texture::FullTextureInfo& config, const GLShader::SamplerEntry& entry) {
- SurfaceParams params{};
- params.is_tiled = config.tic.IsTiled();
- params.block_width = params.is_tiled ? config.tic.BlockWidth() : 0,
- params.block_height = params.is_tiled ? config.tic.BlockHeight() : 0,
- params.block_depth = params.is_tiled ? config.tic.BlockDepth() : 0,
- params.tile_width_spacing = params.is_tiled ? (1 << config.tic.tile_width_spacing.Value()) : 1;
- params.srgb_conversion = config.tic.IsSrgbConversionEnabled();
- params.pixel_format = PixelFormatFromTextureFormat(config.tic.format, config.tic.r_type.Value(),
- params.srgb_conversion);
-
- if (config.tsc.depth_compare_enabled) {
- // Some titles create a 'R16U' (normalized 16-bit) texture with depth_compare enabled,
- // then attempt to sample from it via a shadow sampler. Convert format to Z16 (which also
- // causes GetFormatType to properly return 'Depth' below).
- if (GetFormatType(params.pixel_format) == SurfaceType::ColorTexture) {
- switch (params.pixel_format) {
- case PixelFormat::R16S:
- case PixelFormat::R16U:
- case PixelFormat::R16F:
- params.pixel_format = PixelFormat::Z16;
- break;
- case PixelFormat::R32F:
- params.pixel_format = PixelFormat::Z32F;
- break;
- default:
- LOG_WARNING(HW_GPU, "Color texture format being used with depth compare: {}",
- static_cast<u32>(params.pixel_format));
- break;
- }
- }
- }
-
- params.component_type = ComponentTypeFromTexture(config.tic.r_type.Value());
- params.type = GetFormatType(params.pixel_format);
- UNIMPLEMENTED_IF(params.type == SurfaceType::ColorTexture && config.tsc.depth_compare_enabled);
-
- params.width = Common::AlignUp(config.tic.Width(), GetCompressionFactor(params.pixel_format));
- params.height = Common::AlignUp(config.tic.Height(), GetCompressionFactor(params.pixel_format));
- if (config.tic.IsLineal()) {
- params.pitch = config.tic.Pitch();
- }
- params.unaligned_height = config.tic.Height();
- params.target = SurfaceTargetFromTextureType(config.tic.texture_type);
- params.identity = SurfaceClass::Uploaded;
-
- switch (params.target) {
- case SurfaceTarget::Texture1D:
- case SurfaceTarget::TextureBuffer:
- case SurfaceTarget::Texture2D:
- params.depth = 1;
- break;
- case SurfaceTarget::TextureCubemap:
- params.depth = config.tic.Depth() * 6;
- break;
- case SurfaceTarget::Texture3D:
- params.depth = config.tic.Depth();
- break;
- case SurfaceTarget::Texture2DArray:
- params.depth = config.tic.Depth();
- if (!entry.IsArray()) {
- // TODO(bunnei): We have seen games re-use a Texture2D as Texture2DArray with depth of
- // one, but sample the texture in the shader as if it were not an array texture. This
- // probably is valid on hardware, but we still need to write a test to confirm this. In
- // emulation, the workaround here is to continue to treat this as a Texture2D. An
- // example game that does this is Super Mario Odyssey (in Cloud Kingdom).
- ASSERT(params.depth == 1);
- params.target = SurfaceTarget::Texture2D;
- }
- break;
- case SurfaceTarget::TextureCubeArray:
- params.depth = config.tic.Depth() * 6;
- if (!entry.IsArray()) {
- ASSERT(params.depth == 6);
- params.target = SurfaceTarget::TextureCubemap;
- }
- break;
- default:
- LOG_CRITICAL(HW_GPU, "Unknown depth for target={}", static_cast<u32>(params.target));
- UNREACHABLE();
- params.depth = 1;
- break;
- }
-
- params.is_layered = SurfaceTargetIsLayered(params.target);
- params.is_array = SurfaceTargetIsArray(params.target);
- params.max_mip_level = config.tic.max_mip_level + 1;
- params.rt = {};
-
- params.InitCacheParameters(config.tic.Address());
-
- return params;
-}
-
-/*static*/ SurfaceParams SurfaceParams::CreateForFramebuffer(std::size_t index) {
- const auto& config{Core::System::GetInstance().GPU().Maxwell3D().regs.rt[index]};
- SurfaceParams params{};
-
- params.is_tiled =
- config.memory_layout.type == Tegra::Engines::Maxwell3D::Regs::InvMemoryLayout::BlockLinear;
- params.block_width = 1 << config.memory_layout.block_width;
- params.block_height = 1 << config.memory_layout.block_height;
- params.block_depth = 1 << config.memory_layout.block_depth;
- params.tile_width_spacing = 1;
- params.pixel_format = PixelFormatFromRenderTargetFormat(config.format);
- params.srgb_conversion = config.format == Tegra::RenderTargetFormat::BGRA8_SRGB ||
- config.format == Tegra::RenderTargetFormat::RGBA8_SRGB;
- params.component_type = ComponentTypeFromRenderTarget(config.format);
- params.type = GetFormatType(params.pixel_format);
- if (params.is_tiled) {
- params.width = config.width;
- } else {
- params.pitch = config.width;
- const u32 bpp = params.GetFormatBpp() / 8;
- params.width = params.pitch / bpp;
- }
- params.height = config.height;
- params.unaligned_height = config.height;
- params.target = SurfaceTarget::Texture2D;
- params.identity = SurfaceClass::RenderTarget;
- params.depth = 1;
- params.max_mip_level = 1;
- params.is_layered = false;
-
- // Render target specific parameters, not used for caching
- params.rt.index = static_cast<u32>(index);
- params.rt.array_mode = config.array_mode;
- params.rt.layer_stride = config.layer_stride;
- params.rt.volume = config.volume;
- params.rt.base_layer = config.base_layer;
-
- params.InitCacheParameters(config.Address());
-
- return params;
-}
-
-/*static*/ SurfaceParams SurfaceParams::CreateForDepthBuffer(
- u32 zeta_width, u32 zeta_height, GPUVAddr zeta_address, Tegra::DepthFormat format,
- u32 block_width, u32 block_height, u32 block_depth,
- Tegra::Engines::Maxwell3D::Regs::InvMemoryLayout type) {
- SurfaceParams params{};
-
- params.is_tiled = type == Tegra::Engines::Maxwell3D::Regs::InvMemoryLayout::BlockLinear;
- params.block_width = 1 << std::min(block_width, 5U);
- params.block_height = 1 << std::min(block_height, 5U);
- params.block_depth = 1 << std::min(block_depth, 5U);
- params.tile_width_spacing = 1;
- params.pixel_format = PixelFormatFromDepthFormat(format);
- params.component_type = ComponentTypeFromDepthFormat(format);
- params.type = GetFormatType(params.pixel_format);
- params.srgb_conversion = false;
- params.width = zeta_width;
- params.height = zeta_height;
- params.unaligned_height = zeta_height;
- params.target = SurfaceTarget::Texture2D;
- params.identity = SurfaceClass::DepthBuffer;
- params.depth = 1;
- params.max_mip_level = 1;
- params.is_layered = false;
- params.rt = {};
-
- params.InitCacheParameters(zeta_address);
-
- return params;
-}
-
-/*static*/ SurfaceParams SurfaceParams::CreateForFermiCopySurface(
- const Tegra::Engines::Fermi2D::Regs::Surface& config) {
- SurfaceParams params{};
-
- params.is_tiled = !config.linear;
- params.block_width = params.is_tiled ? std::min(config.BlockWidth(), 32U) : 0,
- params.block_height = params.is_tiled ? std::min(config.BlockHeight(), 32U) : 0,
- params.block_depth = params.is_tiled ? std::min(config.BlockDepth(), 32U) : 0,
- params.tile_width_spacing = 1;
- params.pixel_format = PixelFormatFromRenderTargetFormat(config.format);
- params.srgb_conversion = config.format == Tegra::RenderTargetFormat::BGRA8_SRGB ||
- config.format == Tegra::RenderTargetFormat::RGBA8_SRGB;
- params.component_type = ComponentTypeFromRenderTarget(config.format);
- params.type = GetFormatType(params.pixel_format);
- params.width = config.width;
- params.pitch = config.pitch;
- params.height = config.height;
- params.unaligned_height = config.height;
- params.target = SurfaceTarget::Texture2D;
- params.identity = SurfaceClass::Copy;
- params.depth = 1;
- params.max_mip_level = 1;
- params.rt = {};
-
- params.InitCacheParameters(config.Address());
-
- return params;
-}
-
-static constexpr std::array<FormatTuple, VideoCore::Surface::MaxPixelFormat> tex_format_tuples = {{
- {GL_RGBA8, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8_REV, ComponentType::UNorm, false}, // ABGR8U
- {GL_RGBA8, GL_RGBA, GL_BYTE, ComponentType::SNorm, false}, // ABGR8S
- {GL_RGBA8UI, GL_RGBA_INTEGER, GL_UNSIGNED_BYTE, ComponentType::UInt, false}, // ABGR8UI
- {GL_RGB8, GL_RGB, GL_UNSIGNED_SHORT_5_6_5_REV, ComponentType::UNorm, false}, // B5G6R5U
- {GL_RGB10_A2, GL_RGBA, GL_UNSIGNED_INT_2_10_10_10_REV, ComponentType::UNorm,
- false}, // A2B10G10R10U
- {GL_RGB5_A1, GL_RGBA, GL_UNSIGNED_SHORT_1_5_5_5_REV, ComponentType::UNorm, false}, // A1B5G5R5U
- {GL_R8, GL_RED, GL_UNSIGNED_BYTE, ComponentType::UNorm, false}, // R8U
- {GL_R8UI, GL_RED_INTEGER, GL_UNSIGNED_BYTE, ComponentType::UInt, false}, // R8UI
- {GL_RGBA16F, GL_RGBA, GL_HALF_FLOAT, ComponentType::Float, false}, // RGBA16F
- {GL_RGBA16, GL_RGBA, GL_UNSIGNED_SHORT, ComponentType::UNorm, false}, // RGBA16U
- {GL_RGBA16UI, GL_RGBA_INTEGER, GL_UNSIGNED_SHORT, ComponentType::UInt, false}, // RGBA16UI
- {GL_R11F_G11F_B10F, GL_RGB, GL_UNSIGNED_INT_10F_11F_11F_REV, ComponentType::Float,
- false}, // R11FG11FB10F
- {GL_RGBA32UI, GL_RGBA_INTEGER, GL_UNSIGNED_INT, ComponentType::UInt, false}, // RGBA32UI
- {GL_COMPRESSED_RGBA_S3TC_DXT1_EXT, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8, ComponentType::UNorm,
- true}, // DXT1
- {GL_COMPRESSED_RGBA_S3TC_DXT3_EXT, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8, ComponentType::UNorm,
- true}, // DXT23
- {GL_COMPRESSED_RGBA_S3TC_DXT5_EXT, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8, ComponentType::UNorm,
- true}, // DXT45
- {GL_COMPRESSED_RED_RGTC1, GL_RED, GL_UNSIGNED_INT_8_8_8_8, ComponentType::UNorm, true}, // DXN1
- {GL_COMPRESSED_RG_RGTC2, GL_RG, GL_UNSIGNED_INT_8_8_8_8, ComponentType::UNorm,
- true}, // DXN2UNORM
- {GL_COMPRESSED_SIGNED_RG_RGTC2, GL_RG, GL_INT, ComponentType::SNorm, true}, // DXN2SNORM
- {GL_COMPRESSED_RGBA_BPTC_UNORM, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8, ComponentType::UNorm,
- true}, // BC7U
- {GL_COMPRESSED_RGB_BPTC_UNSIGNED_FLOAT, GL_RGB, GL_UNSIGNED_INT_8_8_8_8, ComponentType::Float,
- true}, // BC6H_UF16
- {GL_COMPRESSED_RGB_BPTC_SIGNED_FLOAT, GL_RGB, GL_UNSIGNED_INT_8_8_8_8, ComponentType::Float,
- true}, // BC6H_SF16
- {GL_RGBA8, GL_RGBA, GL_UNSIGNED_BYTE, ComponentType::UNorm, false}, // ASTC_2D_4X4
- {GL_RGBA8, GL_BGRA, GL_UNSIGNED_BYTE, ComponentType::UNorm, false}, // BGRA8
- {GL_RGBA32F, GL_RGBA, GL_FLOAT, ComponentType::Float, false}, // RGBA32F
- {GL_RG32F, GL_RG, GL_FLOAT, ComponentType::Float, false}, // RG32F
- {GL_R32F, GL_RED, GL_FLOAT, ComponentType::Float, false}, // R32F
- {GL_R16F, GL_RED, GL_HALF_FLOAT, ComponentType::Float, false}, // R16F
- {GL_R16, GL_RED, GL_UNSIGNED_SHORT, ComponentType::UNorm, false}, // R16U
- {GL_R16_SNORM, GL_RED, GL_SHORT, ComponentType::SNorm, false}, // R16S
- {GL_R16UI, GL_RED_INTEGER, GL_UNSIGNED_SHORT, ComponentType::UInt, false}, // R16UI
- {GL_R16I, GL_RED_INTEGER, GL_SHORT, ComponentType::SInt, false}, // R16I
- {GL_RG16, GL_RG, GL_UNSIGNED_SHORT, ComponentType::UNorm, false}, // RG16
- {GL_RG16F, GL_RG, GL_HALF_FLOAT, ComponentType::Float, false}, // RG16F
- {GL_RG16UI, GL_RG_INTEGER, GL_UNSIGNED_SHORT, ComponentType::UInt, false}, // RG16UI
- {GL_RG16I, GL_RG_INTEGER, GL_SHORT, ComponentType::SInt, false}, // RG16I
- {GL_RG16_SNORM, GL_RG, GL_SHORT, ComponentType::SNorm, false}, // RG16S
- {GL_RGB32F, GL_RGB, GL_FLOAT, ComponentType::Float, false}, // RGB32F
- {GL_SRGB8_ALPHA8, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8_REV, ComponentType::UNorm,
- false}, // RGBA8_SRGB
- {GL_RG8, GL_RG, GL_UNSIGNED_BYTE, ComponentType::UNorm, false}, // RG8U
- {GL_RG8, GL_RG, GL_BYTE, ComponentType::SNorm, false}, // RG8S
- {GL_RG32UI, GL_RG_INTEGER, GL_UNSIGNED_INT, ComponentType::UInt, false}, // RG32UI
- {GL_R32UI, GL_RED_INTEGER, GL_UNSIGNED_INT, ComponentType::UInt, false}, // R32UI
- {GL_RGBA8, GL_RGBA, GL_UNSIGNED_BYTE, ComponentType::UNorm, false}, // ASTC_2D_8X8
- {GL_RGBA8, GL_RGBA, GL_UNSIGNED_BYTE, ComponentType::UNorm, false}, // ASTC_2D_8X5
- {GL_RGBA8, GL_RGBA, GL_UNSIGNED_BYTE, ComponentType::UNorm, false}, // ASTC_2D_5X4
- {GL_SRGB8_ALPHA8, GL_BGRA, GL_UNSIGNED_BYTE, ComponentType::UNorm, false}, // BGRA8
- // Compressed sRGB formats
- {GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT1_EXT, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8, ComponentType::UNorm,
- true}, // DXT1_SRGB
- {GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT3_EXT, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8, ComponentType::UNorm,
- true}, // DXT23_SRGB
- {GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT5_EXT, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8, ComponentType::UNorm,
- true}, // DXT45_SRGB
- {GL_COMPRESSED_SRGB_ALPHA_BPTC_UNORM, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8, ComponentType::UNorm,
- true}, // BC7U_SRGB
- {GL_SRGB8_ALPHA8, GL_RGBA, GL_UNSIGNED_BYTE, ComponentType::UNorm, false}, // ASTC_2D_4X4_SRGB
- {GL_SRGB8_ALPHA8, GL_RGBA, GL_UNSIGNED_BYTE, ComponentType::UNorm, false}, // ASTC_2D_8X8_SRGB
- {GL_SRGB8_ALPHA8, GL_RGBA, GL_UNSIGNED_BYTE, ComponentType::UNorm, false}, // ASTC_2D_8X5_SRGB
- {GL_SRGB8_ALPHA8, GL_RGBA, GL_UNSIGNED_BYTE, ComponentType::UNorm, false}, // ASTC_2D_5X4_SRGB
- {GL_RGBA8, GL_RGBA, GL_UNSIGNED_BYTE, ComponentType::UNorm, false}, // ASTC_2D_5X5
- {GL_SRGB8_ALPHA8, GL_RGBA, GL_UNSIGNED_BYTE, ComponentType::UNorm, false}, // ASTC_2D_5X5_SRGB
- {GL_RGBA8, GL_RGBA, GL_UNSIGNED_BYTE, ComponentType::UNorm, false}, // ASTC_2D_10X8
- {GL_SRGB8_ALPHA8, GL_RGBA, GL_UNSIGNED_BYTE, ComponentType::UNorm, false}, // ASTC_2D_10X8_SRGB
-
- // Depth formats
- {GL_DEPTH_COMPONENT32F, GL_DEPTH_COMPONENT, GL_FLOAT, ComponentType::Float, false}, // Z32F
- {GL_DEPTH_COMPONENT16, GL_DEPTH_COMPONENT, GL_UNSIGNED_SHORT, ComponentType::UNorm,
- false}, // Z16
-
- // DepthStencil formats
- {GL_DEPTH24_STENCIL8, GL_DEPTH_STENCIL, GL_UNSIGNED_INT_24_8, ComponentType::UNorm,
- false}, // Z24S8
- {GL_DEPTH24_STENCIL8, GL_DEPTH_STENCIL, GL_UNSIGNED_INT_24_8, ComponentType::UNorm,
- false}, // S8Z24
- {GL_DEPTH32F_STENCIL8, GL_DEPTH_STENCIL, GL_FLOAT_32_UNSIGNED_INT_24_8_REV,
- ComponentType::Float, false}, // Z32FS8
-}};
-
-static GLenum SurfaceTargetToGL(SurfaceTarget target) {
- switch (target) {
- case SurfaceTarget::Texture1D:
- return GL_TEXTURE_1D;
- case SurfaceTarget::TextureBuffer:
- return GL_TEXTURE_BUFFER;
- case SurfaceTarget::Texture2D:
- return GL_TEXTURE_2D;
- case SurfaceTarget::Texture3D:
- return GL_TEXTURE_3D;
- case SurfaceTarget::Texture1DArray:
- return GL_TEXTURE_1D_ARRAY;
- case SurfaceTarget::Texture2DArray:
- return GL_TEXTURE_2D_ARRAY;
- case SurfaceTarget::TextureCubemap:
- return GL_TEXTURE_CUBE_MAP;
- case SurfaceTarget::TextureCubeArray:
- return GL_TEXTURE_CUBE_MAP_ARRAY;
- }
- LOG_CRITICAL(Render_OpenGL, "Unimplemented texture target={}", static_cast<u32>(target));
- UNREACHABLE();
- return {};
-}
-
-static const FormatTuple& GetFormatTuple(PixelFormat pixel_format, ComponentType component_type) {
- ASSERT(static_cast<std::size_t>(pixel_format) < tex_format_tuples.size());
- auto& format = tex_format_tuples[static_cast<unsigned int>(pixel_format)];
- ASSERT(component_type == format.component_type);
-
- return format;
-}
-
-/// Returns the discrepant array target
-constexpr GLenum GetArrayDiscrepantTarget(SurfaceTarget target) {
- switch (target) {
- case SurfaceTarget::Texture1D:
- return GL_TEXTURE_1D_ARRAY;
- case SurfaceTarget::Texture2D:
- return GL_TEXTURE_2D_ARRAY;
- case SurfaceTarget::Texture3D:
- return GL_NONE;
- case SurfaceTarget::Texture1DArray:
- return GL_TEXTURE_1D;
- case SurfaceTarget::Texture2DArray:
- return GL_TEXTURE_2D;
- case SurfaceTarget::TextureCubemap:
- return GL_TEXTURE_CUBE_MAP_ARRAY;
- case SurfaceTarget::TextureCubeArray:
- return GL_TEXTURE_CUBE_MAP;
- }
- return GL_NONE;
-}
-
-Common::Rectangle<u32> SurfaceParams::GetRect(u32 mip_level) const {
- u32 actual_height{std::max(1U, unaligned_height >> mip_level)};
- if (IsPixelFormatASTC(pixel_format)) {
- // ASTC formats must stop at the ATSC block size boundary
- actual_height = Common::AlignDown(actual_height, GetASTCBlockSize(pixel_format).second);
- }
- return {0, actual_height, MipWidth(mip_level), 0};
-}
-
-void SwizzleFunc(const MortonSwizzleMode& mode, const SurfaceParams& params,
- std::vector<u8>& gl_buffer, u32 mip_level) {
- u32 depth = params.MipDepth(mip_level);
- if (params.target == SurfaceTarget::Texture2D) {
- // TODO(Blinkhawk): Eliminate this condition once all texture types are implemented.
- depth = 1U;
- }
- if (params.is_layered) {
- u64 offset = params.GetMipmapLevelOffset(mip_level);
- u64 offset_gl = 0;
- const u64 layer_size = params.LayerMemorySize();
- const u64 gl_size = params.LayerSizeGL(mip_level);
- for (u32 i = 0; i < params.depth; i++) {
- MortonSwizzle(mode, params.pixel_format, params.MipWidth(mip_level),
- params.MipBlockHeight(mip_level), params.MipHeight(mip_level),
- params.MipBlockDepth(mip_level), 1, params.tile_width_spacing,
- gl_buffer.data() + offset_gl, params.host_ptr + offset);
- offset += layer_size;
- offset_gl += gl_size;
- }
- } else {
- const u64 offset = params.GetMipmapLevelOffset(mip_level);
- MortonSwizzle(mode, params.pixel_format, params.MipWidth(mip_level),
- params.MipBlockHeight(mip_level), params.MipHeight(mip_level),
- params.MipBlockDepth(mip_level), depth, params.tile_width_spacing,
- gl_buffer.data(), params.host_ptr + offset);
- }
-}
-
-void RasterizerCacheOpenGL::FastCopySurface(const Surface& src_surface,
- const Surface& dst_surface) {
- const auto& src_params{src_surface->GetSurfaceParams()};
- const auto& dst_params{dst_surface->GetSurfaceParams()};
-
- const u32 width{std::min(src_params.width, dst_params.width)};
- const u32 height{std::min(src_params.height, dst_params.height)};
-
- glCopyImageSubData(src_surface->Texture().handle, SurfaceTargetToGL(src_params.target), 0, 0, 0,
- 0, dst_surface->Texture().handle, SurfaceTargetToGL(dst_params.target), 0, 0,
- 0, 0, width, height, 1);
-
- dst_surface->MarkAsModified(true, *this);
-}
-
-MICROPROFILE_DEFINE(OpenGL_CopySurface, "OpenGL", "CopySurface", MP_RGB(128, 192, 64));
-void RasterizerCacheOpenGL::CopySurface(const Surface& src_surface, const Surface& dst_surface,
- const GLuint copy_pbo_handle, const GLenum src_attachment,
- const GLenum dst_attachment,
- const std::size_t cubemap_face) {
- MICROPROFILE_SCOPE(OpenGL_CopySurface);
- ASSERT_MSG(dst_attachment == 0, "Unimplemented");
-
- const auto& src_params{src_surface->GetSurfaceParams()};
- const auto& dst_params{dst_surface->GetSurfaceParams()};
-
- const auto source_format = GetFormatTuple(src_params.pixel_format, src_params.component_type);
- const auto dest_format = GetFormatTuple(dst_params.pixel_format, dst_params.component_type);
-
- const std::size_t buffer_size = std::max(src_params.size_in_bytes, dst_params.size_in_bytes);
-
- glBindBuffer(GL_PIXEL_PACK_BUFFER, copy_pbo_handle);
- glBufferData(GL_PIXEL_PACK_BUFFER, buffer_size, nullptr, GL_STREAM_COPY);
- if (source_format.compressed) {
- glGetCompressedTextureImage(src_surface->Texture().handle, src_attachment,
- static_cast<GLsizei>(src_params.size_in_bytes), nullptr);
- } else {
- glGetTextureImage(src_surface->Texture().handle, src_attachment, source_format.format,
- source_format.type, static_cast<GLsizei>(src_params.size_in_bytes),
- nullptr);
- }
- // If the new texture is bigger than the previous one, we need to fill in the rest with data
- // from the CPU.
- if (src_params.size_in_bytes < dst_params.size_in_bytes) {
- // Upload the rest of the memory.
- if (dst_params.is_tiled) {
- // TODO(Subv): We might have to de-tile the subtexture and re-tile it with the rest
- // of the data in this case. Games like Super Mario Odyssey seem to hit this case
- // when drawing, it re-uses the memory of a previous texture as a bigger framebuffer
- // but it doesn't clear it beforehand, the texture is already full of zeros.
- LOG_DEBUG(HW_GPU, "Trying to upload extra texture data from the CPU during "
- "reinterpretation but the texture is tiled.");
- }
- const std::size_t remaining_size = dst_params.size_in_bytes - src_params.size_in_bytes;
- auto& memory_manager{Core::System::GetInstance().GPU().MemoryManager()};
- glBufferSubData(GL_PIXEL_PACK_BUFFER, src_params.size_in_bytes, remaining_size,
- memory_manager.GetPointer(dst_params.gpu_addr + src_params.size_in_bytes));
- }
-
- glBindBuffer(GL_PIXEL_PACK_BUFFER, 0);
-
- const GLsizei width{static_cast<GLsizei>(
- std::min(src_params.GetRect().GetWidth(), dst_params.GetRect().GetWidth()))};
- const GLsizei height{static_cast<GLsizei>(
- std::min(src_params.GetRect().GetHeight(), dst_params.GetRect().GetHeight()))};
-
- glBindBuffer(GL_PIXEL_UNPACK_BUFFER, copy_pbo_handle);
- if (dest_format.compressed) {
- LOG_CRITICAL(HW_GPU, "Compressed copy is unimplemented!");
- UNREACHABLE();
- } else {
- switch (dst_params.target) {
- case SurfaceTarget::Texture1D:
- glTextureSubImage1D(dst_surface->Texture().handle, 0, 0, width, dest_format.format,
- dest_format.type, nullptr);
- break;
- case SurfaceTarget::Texture2D:
- glTextureSubImage2D(dst_surface->Texture().handle, 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->Texture().handle, 0, 0, 0, 0, width, height,
- static_cast<GLsizei>(dst_params.depth), dest_format.format,
- dest_format.type, nullptr);
- break;
- case SurfaceTarget::TextureCubemap:
- glTextureSubImage3D(dst_surface->Texture().handle, 0, 0, 0,
- static_cast<GLint>(cubemap_face), width, height, 1,
- dest_format.format, dest_format.type, nullptr);
- break;
- default:
- LOG_CRITICAL(Render_OpenGL, "Unimplemented surface target={}",
- static_cast<u32>(dst_params.target));
- UNREACHABLE();
- }
- glBindBuffer(GL_PIXEL_UNPACK_BUFFER, 0);
- }
-
- dst_surface->MarkAsModified(true, *this);
-}
-
-CachedSurface::CachedSurface(const SurfaceParams& params)
- : RasterizerCacheObject{params.host_ptr}, params{params},
- gl_target{SurfaceTargetToGL(params.target)}, cached_size_in_bytes{params.size_in_bytes} {
-
- const auto optional_cpu_addr{
- Core::System::GetInstance().GPU().MemoryManager().GpuToCpuAddress(params.gpu_addr)};
- ASSERT_MSG(optional_cpu_addr, "optional_cpu_addr is invalid");
- cpu_addr = *optional_cpu_addr;
-
- texture.Create(gl_target);
-
- // TODO(Rodrigo): Using params.GetRect() returns a different size than using its Mip*(0)
- // alternatives. This signals a bug on those functions.
- const auto width = static_cast<GLsizei>(params.MipWidth(0));
- const auto height = static_cast<GLsizei>(params.MipHeight(0));
- memory_size = params.MemorySize();
- reinterpreted = false;
-
- const auto& format_tuple = GetFormatTuple(params.pixel_format, params.component_type);
- gl_internal_format = format_tuple.internal_format;
-
- switch (params.target) {
- case SurfaceTarget::Texture1D:
- glTextureStorage1D(texture.handle, params.max_mip_level, gl_internal_format, width);
- break;
- case SurfaceTarget::TextureBuffer:
- texture_buffer.Create();
- glNamedBufferStorage(texture_buffer.handle,
- params.width * GetBytesPerPixel(params.pixel_format), nullptr,
- GL_DYNAMIC_STORAGE_BIT);
- glTextureBuffer(texture.handle, gl_internal_format, texture_buffer.handle);
- break;
- case SurfaceTarget::Texture2D:
- case SurfaceTarget::TextureCubemap:
- glTextureStorage2D(texture.handle, params.max_mip_level, gl_internal_format, width, height);
- break;
- case SurfaceTarget::Texture3D:
- case SurfaceTarget::Texture2DArray:
- case SurfaceTarget::TextureCubeArray:
- glTextureStorage3D(texture.handle, params.max_mip_level, gl_internal_format, width, height,
- params.depth);
- break;
- default:
- LOG_CRITICAL(Render_OpenGL, "Unimplemented surface target={}",
- static_cast<u32>(params.target));
- UNREACHABLE();
- glTextureStorage2D(texture.handle, params.max_mip_level, gl_internal_format, width, height);
- }
-
- if (params.target != SurfaceTarget::TextureBuffer) {
- ApplyTextureDefaults(texture.handle, params.max_mip_level);
- }
-
- OpenGL::LabelGLObject(GL_TEXTURE, texture.handle, params.gpu_addr, params.IdentityString());
-}
-
-MICROPROFILE_DEFINE(OpenGL_SurfaceLoad, "OpenGL", "Surface Load", MP_RGB(128, 192, 64));
-void CachedSurface::LoadGLBuffer(RasterizerTemporaryMemory& res_cache_tmp_mem) {
- MICROPROFILE_SCOPE(OpenGL_SurfaceLoad);
- auto& gl_buffer = res_cache_tmp_mem.gl_buffer;
- if (gl_buffer.size() < params.max_mip_level)
- gl_buffer.resize(params.max_mip_level);
- for (u32 i = 0; i < params.max_mip_level; i++)
- gl_buffer[i].resize(params.GetMipmapSizeGL(i));
- if (params.is_tiled) {
- ASSERT_MSG(params.block_width == 1, "Block width is defined as {} on texture type {}",
- params.block_width, static_cast<u32>(params.target));
- for (u32 i = 0; i < params.max_mip_level; i++)
- SwizzleFunc(MortonSwizzleMode::MortonToLinear, params, gl_buffer[i], i);
- } else {
- const u32 bpp = params.GetFormatBpp() / 8;
- const u32 copy_size = (params.width * bpp + GetDefaultBlockWidth(params.pixel_format) - 1) /
- GetDefaultBlockWidth(params.pixel_format);
- if (params.pitch == copy_size) {
- std::memcpy(gl_buffer[0].data(), params.host_ptr, params.size_in_bytes_gl);
- } else {
- const u32 height = (params.height + GetDefaultBlockHeight(params.pixel_format) - 1) /
- GetDefaultBlockHeight(params.pixel_format);
- const u8* start{params.host_ptr};
- u8* write_to = gl_buffer[0].data();
- for (u32 h = height; h > 0; h--) {
- std::memcpy(write_to, start, copy_size);
- start += params.pitch;
- write_to += copy_size;
- }
- }
- }
- for (u32 i = 0; i < params.max_mip_level; i++) {
- const u32 width = params.MipWidth(i);
- const u32 height = params.MipHeight(i);
- const u32 depth = params.MipDepth(i);
- if (VideoCore::Surface::IsPixelFormatASTC(params.pixel_format)) {
- // Reserve size for RGBA8 conversion
- constexpr std::size_t rgba_bpp = 4;
- gl_buffer[i].resize(std::max(gl_buffer[i].size(), width * height * depth * rgba_bpp));
- }
- Tegra::Texture::ConvertFromGuestToHost(gl_buffer[i].data(), params.pixel_format, width,
- height, depth, true, true);
- }
-}
-
-MICROPROFILE_DEFINE(OpenGL_SurfaceFlush, "OpenGL", "Surface Flush", MP_RGB(128, 192, 64));
-void CachedSurface::FlushGLBuffer(RasterizerTemporaryMemory& res_cache_tmp_mem) {
- MICROPROFILE_SCOPE(OpenGL_SurfaceFlush);
-
- ASSERT_MSG(!IsPixelFormatASTC(params.pixel_format), "Unimplemented");
-
- auto& gl_buffer = res_cache_tmp_mem.gl_buffer;
- // OpenGL temporary buffer needs to be big enough to store raw texture size
- gl_buffer[0].resize(GetSizeInBytes());
-
- const FormatTuple& tuple = GetFormatTuple(params.pixel_format, params.component_type);
- const u32 align = std::clamp(params.RowAlign(0), 1U, 8U);
- glPixelStorei(GL_PACK_ALIGNMENT, align);
- glPixelStorei(GL_PACK_ROW_LENGTH, static_cast<GLint>(params.width));
- ASSERT(!tuple.compressed);
- glBindBuffer(GL_PIXEL_PACK_BUFFER, 0);
- glGetTextureImage(texture.handle, 0, tuple.format, tuple.type,
- static_cast<GLsizei>(gl_buffer[0].size()), gl_buffer[0].data());
- glPixelStorei(GL_PACK_ROW_LENGTH, 0);
- Tegra::Texture::ConvertFromHostToGuest(gl_buffer[0].data(), params.pixel_format, params.width,
- params.height, params.depth, true, true);
- if (params.is_tiled) {
- ASSERT_MSG(params.block_width == 1, "Block width is defined as {} on texture type {}",
- params.block_width, static_cast<u32>(params.target));
-
- SwizzleFunc(MortonSwizzleMode::LinearToMorton, params, gl_buffer[0], 0);
- } else {
- const u32 bpp = params.GetFormatBpp() / 8;
- const u32 copy_size = params.width * bpp;
- if (params.pitch == copy_size) {
- std::memcpy(params.host_ptr, gl_buffer[0].data(), GetSizeInBytes());
- } else {
- u8* start{params.host_ptr};
- const u8* read_to = gl_buffer[0].data();
- for (u32 h = params.height; h > 0; h--) {
- std::memcpy(start, read_to, copy_size);
- start += params.pitch;
- read_to += copy_size;
- }
- }
- }
-}
-
-void CachedSurface::UploadGLMipmapTexture(RasterizerTemporaryMemory& res_cache_tmp_mem, u32 mip_map,
- GLuint read_fb_handle, GLuint draw_fb_handle) {
- const auto& rect{params.GetRect(mip_map)};
-
- auto& gl_buffer = res_cache_tmp_mem.gl_buffer;
-
- // Load data from memory to the surface
- const auto x0 = static_cast<GLint>(rect.left);
- const auto y0 = static_cast<GLint>(rect.bottom);
- auto buffer_offset =
- static_cast<std::size_t>(static_cast<std::size_t>(y0) * params.MipWidth(mip_map) +
- static_cast<std::size_t>(x0)) *
- GetBytesPerPixel(params.pixel_format);
-
- const FormatTuple& tuple = GetFormatTuple(params.pixel_format, params.component_type);
-
- const u32 align = std::clamp(params.RowAlign(mip_map), 1U, 8U);
- glPixelStorei(GL_UNPACK_ALIGNMENT, align);
- glPixelStorei(GL_UNPACK_ROW_LENGTH, static_cast<GLint>(params.MipWidth(mip_map)));
-
- const auto image_size = static_cast<GLsizei>(params.GetMipmapSizeGL(mip_map, false));
- if (tuple.compressed) {
- switch (params.target) {
- case SurfaceTarget::Texture2D:
- glCompressedTextureSubImage2D(
- texture.handle, mip_map, 0, 0, static_cast<GLsizei>(params.MipWidth(mip_map)),
- static_cast<GLsizei>(params.MipHeight(mip_map)), tuple.internal_format, image_size,
- &gl_buffer[mip_map][buffer_offset]);
- break;
- case SurfaceTarget::Texture3D:
- glCompressedTextureSubImage3D(
- texture.handle, mip_map, 0, 0, 0, static_cast<GLsizei>(params.MipWidth(mip_map)),
- static_cast<GLsizei>(params.MipHeight(mip_map)),
- static_cast<GLsizei>(params.MipDepth(mip_map)), tuple.internal_format, image_size,
- &gl_buffer[mip_map][buffer_offset]);
- break;
- case SurfaceTarget::Texture2DArray:
- case SurfaceTarget::TextureCubeArray:
- glCompressedTextureSubImage3D(
- texture.handle, mip_map, 0, 0, 0, static_cast<GLsizei>(params.MipWidth(mip_map)),
- static_cast<GLsizei>(params.MipHeight(mip_map)), static_cast<GLsizei>(params.depth),
- tuple.internal_format, image_size, &gl_buffer[mip_map][buffer_offset]);
- break;
- case SurfaceTarget::TextureCubemap: {
- const auto layer_size = static_cast<GLsizei>(params.LayerSizeGL(mip_map));
- for (std::size_t face = 0; face < params.depth; ++face) {
- glCompressedTextureSubImage3D(
- texture.handle, mip_map, 0, 0, static_cast<GLint>(face),
- static_cast<GLsizei>(params.MipWidth(mip_map)),
- static_cast<GLsizei>(params.MipHeight(mip_map)), 1, tuple.internal_format,
- layer_size, &gl_buffer[mip_map][buffer_offset]);
- buffer_offset += layer_size;
- }
- break;
- }
- default:
- LOG_CRITICAL(Render_OpenGL, "Unimplemented surface target={}",
- static_cast<u32>(params.target));
- UNREACHABLE();
- glCompressedTextureSubImage2D(
- texture.handle, mip_map, 0, 0, static_cast<GLsizei>(params.MipWidth(mip_map)),
- static_cast<GLsizei>(params.MipHeight(mip_map)), tuple.internal_format,
- static_cast<GLsizei>(params.size_in_bytes_gl), &gl_buffer[mip_map][buffer_offset]);
- }
- } else {
- switch (params.target) {
- case SurfaceTarget::Texture1D:
- glTextureSubImage1D(texture.handle, mip_map, x0, static_cast<GLsizei>(rect.GetWidth()),
- tuple.format, tuple.type, &gl_buffer[mip_map][buffer_offset]);
- break;
- case SurfaceTarget::TextureBuffer:
- ASSERT(mip_map == 0);
- glNamedBufferSubData(texture_buffer.handle, x0,
- static_cast<GLsizeiptr>(rect.GetWidth()) *
- GetBytesPerPixel(params.pixel_format),
- &gl_buffer[mip_map][buffer_offset]);
- break;
- case SurfaceTarget::Texture2D:
- glTextureSubImage2D(texture.handle, mip_map, x0, y0,
- static_cast<GLsizei>(rect.GetWidth()),
- static_cast<GLsizei>(rect.GetHeight()), tuple.format, tuple.type,
- &gl_buffer[mip_map][buffer_offset]);
- break;
- case SurfaceTarget::Texture3D:
- glTextureSubImage3D(texture.handle, mip_map, x0, y0, 0,
- static_cast<GLsizei>(rect.GetWidth()),
- static_cast<GLsizei>(rect.GetHeight()), params.MipDepth(mip_map),
- tuple.format, tuple.type, &gl_buffer[mip_map][buffer_offset]);
- break;
- case SurfaceTarget::Texture2DArray:
- case SurfaceTarget::TextureCubeArray:
- glTextureSubImage3D(texture.handle, mip_map, x0, y0, 0,
- static_cast<GLsizei>(rect.GetWidth()),
- static_cast<GLsizei>(rect.GetHeight()), params.depth, tuple.format,
- tuple.type, &gl_buffer[mip_map][buffer_offset]);
- break;
- case SurfaceTarget::TextureCubemap: {
- for (std::size_t face = 0; face < params.depth; ++face) {
- glTextureSubImage3D(texture.handle, mip_map, x0, y0, static_cast<GLint>(face),
- static_cast<GLsizei>(rect.GetWidth()),
- static_cast<GLsizei>(rect.GetHeight()), 1, tuple.format,
- tuple.type, &gl_buffer[mip_map][buffer_offset]);
- buffer_offset += params.LayerSizeGL(mip_map);
- }
- break;
- }
- default:
- LOG_CRITICAL(Render_OpenGL, "Unimplemented surface target={}",
- static_cast<u32>(params.target));
- UNREACHABLE();
- glTextureSubImage2D(texture.handle, mip_map, x0, y0,
- static_cast<GLsizei>(rect.GetWidth()),
- static_cast<GLsizei>(rect.GetHeight()), tuple.format, tuple.type,
- &gl_buffer[mip_map][buffer_offset]);
- }
- }
-
- glPixelStorei(GL_UNPACK_ROW_LENGTH, 0);
-}
-
-void CachedSurface::EnsureTextureDiscrepantView() {
- if (discrepant_view.handle != 0)
- return;
-
- const GLenum target{GetArrayDiscrepantTarget(params.target)};
- ASSERT(target != GL_NONE);
-
- const GLuint num_layers{target == GL_TEXTURE_CUBE_MAP_ARRAY ? 6u : 1u};
- constexpr GLuint min_layer = 0;
- constexpr GLuint min_level = 0;
-
- glGenTextures(1, &discrepant_view.handle);
- glTextureView(discrepant_view.handle, target, texture.handle, gl_internal_format, min_level,
- params.max_mip_level, min_layer, num_layers);
- ApplyTextureDefaults(discrepant_view.handle, params.max_mip_level);
- glTextureParameteriv(discrepant_view.handle, GL_TEXTURE_SWIZZLE_RGBA,
- reinterpret_cast<const GLint*>(swizzle.data()));
-}
-
-MICROPROFILE_DEFINE(OpenGL_TextureUL, "OpenGL", "Texture Upload", MP_RGB(128, 192, 64));
-void CachedSurface::UploadGLTexture(RasterizerTemporaryMemory& res_cache_tmp_mem,
- GLuint read_fb_handle, GLuint draw_fb_handle) {
- MICROPROFILE_SCOPE(OpenGL_TextureUL);
-
- for (u32 i = 0; i < params.max_mip_level; i++)
- UploadGLMipmapTexture(res_cache_tmp_mem, i, read_fb_handle, draw_fb_handle);
-}
-
-void CachedSurface::UpdateSwizzle(Tegra::Texture::SwizzleSource swizzle_x,
- Tegra::Texture::SwizzleSource swizzle_y,
- Tegra::Texture::SwizzleSource swizzle_z,
- Tegra::Texture::SwizzleSource swizzle_w) {
- if (params.target == SurfaceTarget::TextureBuffer) {
- return;
- }
- const GLenum new_x = MaxwellToGL::SwizzleSource(swizzle_x);
- const GLenum new_y = MaxwellToGL::SwizzleSource(swizzle_y);
- const GLenum new_z = MaxwellToGL::SwizzleSource(swizzle_z);
- const GLenum new_w = MaxwellToGL::SwizzleSource(swizzle_w);
- if (swizzle[0] == new_x && swizzle[1] == new_y && swizzle[2] == new_z && swizzle[3] == new_w) {
- return;
- }
- swizzle = {new_x, new_y, new_z, new_w};
- const auto swizzle_data = reinterpret_cast<const GLint*>(swizzle.data());
- glTextureParameteriv(texture.handle, GL_TEXTURE_SWIZZLE_RGBA, swizzle_data);
- if (discrepant_view.handle != 0) {
- glTextureParameteriv(discrepant_view.handle, GL_TEXTURE_SWIZZLE_RGBA, swizzle_data);
- }
-}
-
-RasterizerCacheOpenGL::RasterizerCacheOpenGL(RasterizerOpenGL& rasterizer)
- : RasterizerCache{rasterizer} {
- read_framebuffer.Create();
- draw_framebuffer.Create();
- copy_pbo.Create();
-}
-
-Surface RasterizerCacheOpenGL::GetTextureSurface(const Tegra::Texture::FullTextureInfo& config,
- const GLShader::SamplerEntry& entry) {
- return GetSurface(SurfaceParams::CreateForTexture(config, entry));
-}
-
-Surface RasterizerCacheOpenGL::GetDepthBufferSurface(bool preserve_contents) {
- auto& gpu{Core::System::GetInstance().GPU().Maxwell3D()};
- const auto& regs{gpu.regs};
-
- if (!gpu.dirty_flags.zeta_buffer) {
- return last_depth_buffer;
- }
- gpu.dirty_flags.zeta_buffer = false;
-
- if (!regs.zeta.Address() || !regs.zeta_enable) {
- return last_depth_buffer = {};
- }
-
- SurfaceParams depth_params{SurfaceParams::CreateForDepthBuffer(
- regs.zeta_width, regs.zeta_height, regs.zeta.Address(), regs.zeta.format,
- regs.zeta.memory_layout.block_width, regs.zeta.memory_layout.block_height,
- regs.zeta.memory_layout.block_depth, regs.zeta.memory_layout.type)};
-
- return last_depth_buffer = GetSurface(depth_params, preserve_contents);
-}
-
-Surface RasterizerCacheOpenGL::GetColorBufferSurface(std::size_t index, bool preserve_contents) {
- auto& gpu{Core::System::GetInstance().GPU().Maxwell3D()};
- const auto& regs{gpu.regs};
-
- if (!gpu.dirty_flags.color_buffer[index]) {
- return current_color_buffers[index];
- }
- gpu.dirty_flags.color_buffer.reset(index);
-
- ASSERT(index < Tegra::Engines::Maxwell3D::Regs::NumRenderTargets);
-
- if (index >= regs.rt_control.count) {
- return current_color_buffers[index] = {};
- }
-
- if (regs.rt[index].Address() == 0 || regs.rt[index].format == Tegra::RenderTargetFormat::NONE) {
- return current_color_buffers[index] = {};
- }
-
- const SurfaceParams color_params{SurfaceParams::CreateForFramebuffer(index)};
-
- return current_color_buffers[index] = GetSurface(color_params, preserve_contents);
-}
-
-void RasterizerCacheOpenGL::LoadSurface(const Surface& surface) {
- surface->LoadGLBuffer(temporal_memory);
- surface->UploadGLTexture(temporal_memory, read_framebuffer.handle, draw_framebuffer.handle);
- surface->MarkAsModified(false, *this);
- surface->MarkForReload(false);
-}
-
-Surface RasterizerCacheOpenGL::GetSurface(const SurfaceParams& params, bool preserve_contents) {
- if (!params.IsValid()) {
- return {};
- }
-
- // Look up surface in the cache based on address
- Surface surface{TryGet(params.host_ptr)};
- if (surface) {
- if (surface->GetSurfaceParams().IsCompatibleSurface(params)) {
- // Use the cached surface as-is unless it's not synced with memory
- if (surface->MustReload())
- LoadSurface(surface);
- return surface;
- } else if (preserve_contents) {
- // If surface parameters changed and we care about keeping the previous data, recreate
- // the surface from the old one
- Surface new_surface{RecreateSurface(surface, params)};
- Unregister(surface);
- Register(new_surface);
- if (new_surface->IsUploaded()) {
- RegisterReinterpretSurface(new_surface);
- }
- return new_surface;
- } else {
- // Delete the old surface before creating a new one to prevent collisions.
- Unregister(surface);
- }
- }
-
- // No cached surface found - get a new one
- surface = GetUncachedSurface(params);
- Register(surface);
-
- // Only load surface from memory if we care about the contents
- if (preserve_contents) {
- LoadSurface(surface);
- }
-
- return surface;
-}
-
-Surface RasterizerCacheOpenGL::GetUncachedSurface(const SurfaceParams& params) {
- Surface surface{TryGetReservedSurface(params)};
- if (!surface) {
- // No reserved surface available, create a new one and reserve it
- surface = std::make_shared<CachedSurface>(params);
- ReserveSurface(surface);
- }
- return surface;
-}
-
-void RasterizerCacheOpenGL::FastLayeredCopySurface(const Surface& src_surface,
- const Surface& dst_surface) {
- const auto& init_params{src_surface->GetSurfaceParams()};
- const auto& dst_params{dst_surface->GetSurfaceParams()};
- auto& memory_manager{Core::System::GetInstance().GPU().MemoryManager()};
- GPUVAddr address{init_params.gpu_addr};
- const std::size_t layer_size{dst_params.LayerMemorySize()};
- for (u32 layer = 0; layer < dst_params.depth; layer++) {
- for (u32 mipmap = 0; mipmap < dst_params.max_mip_level; mipmap++) {
- const GPUVAddr sub_address{address + dst_params.GetMipmapLevelOffset(mipmap)};
- const Surface& copy{TryGet(memory_manager.GetPointer(sub_address))};
- if (!copy) {
- continue;
- }
- const auto& src_params{copy->GetSurfaceParams()};
- const u32 width{std::min(src_params.width, dst_params.MipWidth(mipmap))};
- const u32 height{std::min(src_params.height, dst_params.MipHeight(mipmap))};
-
- glCopyImageSubData(copy->Texture().handle, SurfaceTargetToGL(src_params.target), 0, 0,
- 0, 0, dst_surface->Texture().handle,
- SurfaceTargetToGL(dst_params.target), mipmap, 0, 0, layer, width,
- height, 1);
- }
- address += layer_size;
- }
-
- dst_surface->MarkAsModified(true, *this);
-}
-
-static bool BlitSurface(const Surface& src_surface, const Surface& dst_surface,
- const Common::Rectangle<u32>& src_rect,
- const Common::Rectangle<u32>& dst_rect, GLuint read_fb_handle,
- GLuint draw_fb_handle, GLenum src_attachment = 0, GLenum dst_attachment = 0,
- std::size_t cubemap_face = 0) {
-
- const auto& src_params{src_surface->GetSurfaceParams()};
- const auto& dst_params{dst_surface->GetSurfaceParams()};
-
- OpenGLState prev_state{OpenGLState::GetCurState()};
- SCOPE_EXIT({ prev_state.Apply(); });
-
- OpenGLState state;
- state.draw.read_framebuffer = read_fb_handle;
- state.draw.draw_framebuffer = draw_fb_handle;
- state.Apply();
-
- u32 buffers{};
-
- if (src_params.type == SurfaceType::ColorTexture) {
- switch (src_params.target) {
- case SurfaceTarget::Texture2D:
- glFramebufferTexture2D(GL_READ_FRAMEBUFFER, GL_COLOR_ATTACHMENT0 + src_attachment,
- GL_TEXTURE_2D, src_surface->Texture().handle, 0);
- glFramebufferTexture2D(GL_READ_FRAMEBUFFER, GL_DEPTH_STENCIL_ATTACHMENT, GL_TEXTURE_2D,
- 0, 0);
- break;
- case SurfaceTarget::TextureCubemap:
- glFramebufferTexture2D(
- GL_READ_FRAMEBUFFER, GL_COLOR_ATTACHMENT0 + src_attachment,
- static_cast<GLenum>(GL_TEXTURE_CUBE_MAP_POSITIVE_X + cubemap_face),
- src_surface->Texture().handle, 0);
- glFramebufferTexture2D(
- GL_READ_FRAMEBUFFER, GL_DEPTH_STENCIL_ATTACHMENT,
- static_cast<GLenum>(GL_TEXTURE_CUBE_MAP_POSITIVE_X + cubemap_face), 0, 0);
- break;
- case SurfaceTarget::Texture2DArray:
- glFramebufferTextureLayer(GL_READ_FRAMEBUFFER, GL_COLOR_ATTACHMENT0 + src_attachment,
- src_surface->Texture().handle, 0, 0);
- glFramebufferTextureLayer(GL_READ_FRAMEBUFFER, GL_DEPTH_STENCIL_ATTACHMENT, 0, 0, 0);
- break;
- case SurfaceTarget::Texture3D:
- glFramebufferTexture3D(GL_READ_FRAMEBUFFER, GL_COLOR_ATTACHMENT0 + src_attachment,
- SurfaceTargetToGL(src_params.target),
- src_surface->Texture().handle, 0, 0);
- glFramebufferTexture3D(GL_READ_FRAMEBUFFER, GL_DEPTH_STENCIL_ATTACHMENT,
- SurfaceTargetToGL(src_params.target), 0, 0, 0);
- break;
- default:
- glFramebufferTexture2D(GL_READ_FRAMEBUFFER, GL_COLOR_ATTACHMENT0 + src_attachment,
- GL_TEXTURE_2D, src_surface->Texture().handle, 0);
- glFramebufferTexture2D(GL_READ_FRAMEBUFFER, GL_DEPTH_STENCIL_ATTACHMENT, GL_TEXTURE_2D,
- 0, 0);
- break;
- }
-
- switch (dst_params.target) {
- case SurfaceTarget::Texture2D:
- glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0 + dst_attachment,
- GL_TEXTURE_2D, dst_surface->Texture().handle, 0);
- glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_DEPTH_STENCIL_ATTACHMENT, GL_TEXTURE_2D,
- 0, 0);
- break;
- case SurfaceTarget::TextureCubemap:
- glFramebufferTexture2D(
- GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0 + dst_attachment,
- static_cast<GLenum>(GL_TEXTURE_CUBE_MAP_POSITIVE_X + cubemap_face),
- dst_surface->Texture().handle, 0);
- glFramebufferTexture2D(
- GL_DRAW_FRAMEBUFFER, GL_DEPTH_STENCIL_ATTACHMENT,
- static_cast<GLenum>(GL_TEXTURE_CUBE_MAP_POSITIVE_X + cubemap_face), 0, 0);
- break;
- case SurfaceTarget::Texture2DArray:
- glFramebufferTextureLayer(GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0 + dst_attachment,
- dst_surface->Texture().handle, 0, 0);
- glFramebufferTextureLayer(GL_DRAW_FRAMEBUFFER, GL_DEPTH_STENCIL_ATTACHMENT, 0, 0, 0);
- break;
-
- case SurfaceTarget::Texture3D:
- glFramebufferTexture3D(GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0 + dst_attachment,
- SurfaceTargetToGL(dst_params.target),
- dst_surface->Texture().handle, 0, 0);
- glFramebufferTexture3D(GL_DRAW_FRAMEBUFFER, GL_DEPTH_STENCIL_ATTACHMENT,
- SurfaceTargetToGL(dst_params.target), 0, 0, 0);
- break;
- default:
- glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0 + dst_attachment,
- GL_TEXTURE_2D, dst_surface->Texture().handle, 0);
- glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_DEPTH_STENCIL_ATTACHMENT, GL_TEXTURE_2D,
- 0, 0);
- break;
- }
-
- buffers = GL_COLOR_BUFFER_BIT;
- } else if (src_params.type == SurfaceType::Depth) {
- glFramebufferTexture2D(GL_READ_FRAMEBUFFER, GL_COLOR_ATTACHMENT0 + src_attachment,
- GL_TEXTURE_2D, 0, 0);
- glFramebufferTexture2D(GL_READ_FRAMEBUFFER, GL_DEPTH_ATTACHMENT, GL_TEXTURE_2D,
- src_surface->Texture().handle, 0);
- glFramebufferTexture2D(GL_READ_FRAMEBUFFER, GL_STENCIL_ATTACHMENT, GL_TEXTURE_2D, 0, 0);
-
- glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0 + dst_attachment,
- GL_TEXTURE_2D, 0, 0);
- glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_DEPTH_ATTACHMENT, GL_TEXTURE_2D,
- dst_surface->Texture().handle, 0);
- glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_STENCIL_ATTACHMENT, GL_TEXTURE_2D, 0, 0);
-
- buffers = GL_DEPTH_BUFFER_BIT;
- } else if (src_params.type == SurfaceType::DepthStencil) {
- glFramebufferTexture2D(GL_READ_FRAMEBUFFER, GL_COLOR_ATTACHMENT0 + src_attachment,
- GL_TEXTURE_2D, 0, 0);
- glFramebufferTexture2D(GL_READ_FRAMEBUFFER, GL_DEPTH_STENCIL_ATTACHMENT, GL_TEXTURE_2D,
- src_surface->Texture().handle, 0);
-
- glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0 + dst_attachment,
- GL_TEXTURE_2D, 0, 0);
- glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_DEPTH_STENCIL_ATTACHMENT, GL_TEXTURE_2D,
- dst_surface->Texture().handle, 0);
-
- buffers = GL_DEPTH_BUFFER_BIT | GL_STENCIL_BUFFER_BIT;
- }
-
- glBlitFramebuffer(src_rect.left, src_rect.top, src_rect.right, src_rect.bottom, dst_rect.left,
- dst_rect.top, dst_rect.right, dst_rect.bottom, buffers,
- buffers == GL_COLOR_BUFFER_BIT ? GL_LINEAR : GL_NEAREST);
-
- return true;
-}
-
-void RasterizerCacheOpenGL::FermiCopySurface(
- const Tegra::Engines::Fermi2D::Regs::Surface& src_config,
- const Tegra::Engines::Fermi2D::Regs::Surface& dst_config,
- const Common::Rectangle<u32>& src_rect, const Common::Rectangle<u32>& dst_rect) {
-
- const auto& src_params = SurfaceParams::CreateForFermiCopySurface(src_config);
- const auto& dst_params = SurfaceParams::CreateForFermiCopySurface(dst_config);
-
- ASSERT(src_params.pixel_format == dst_params.pixel_format);
- ASSERT(src_params.block_height == dst_params.block_height);
- ASSERT(src_params.is_tiled == dst_params.is_tiled);
- ASSERT(src_params.depth == dst_params.depth);
- ASSERT(src_params.target == dst_params.target);
- ASSERT(src_params.rt.index == dst_params.rt.index);
-
- auto src_surface = GetSurface(src_params, true);
- auto dst_surface = GetSurface(dst_params, true);
-
- BlitSurface(src_surface, dst_surface, src_rect, dst_rect, read_framebuffer.handle,
- draw_framebuffer.handle);
-
- dst_surface->MarkAsModified(true, *this);
-}
-
-void RasterizerCacheOpenGL::AccurateCopySurface(const Surface& src_surface,
- const Surface& dst_surface) {
- const auto& src_params{src_surface->GetSurfaceParams()};
- const auto& dst_params{dst_surface->GetSurfaceParams()};
-
- // Flush enough memory for both the source and destination surface
- FlushRegion(ToCacheAddr(src_params.host_ptr),
- std::max(src_params.MemorySize(), dst_params.MemorySize()));
-
- LoadSurface(dst_surface);
-}
-
-Surface RasterizerCacheOpenGL::RecreateSurface(const Surface& old_surface,
- const SurfaceParams& new_params) {
- // Verify surface is compatible for blitting
- auto old_params{old_surface->GetSurfaceParams()};
-
- // Get a new surface with the new parameters, and blit the previous surface to it
- Surface new_surface{GetUncachedSurface(new_params)};
-
- // With use_accurate_gpu_emulation enabled, do an accurate surface copy
- if (Settings::values.use_accurate_gpu_emulation) {
- AccurateCopySurface(old_surface, new_surface);
- return new_surface;
- }
-
- const bool old_compressed =
- GetFormatTuple(old_params.pixel_format, old_params.component_type).compressed;
- const bool new_compressed =
- GetFormatTuple(new_params.pixel_format, new_params.component_type).compressed;
- const bool compatible_formats =
- GetFormatBpp(old_params.pixel_format) == GetFormatBpp(new_params.pixel_format) &&
- !(old_compressed || new_compressed);
- // For compatible surfaces, we can just do fast glCopyImageSubData based copy
- if (old_params.target == new_params.target && old_params.depth == new_params.depth &&
- old_params.depth == 1 && compatible_formats) {
- FastCopySurface(old_surface, new_surface);
- return new_surface;
- }
-
- switch (new_params.target) {
- case SurfaceTarget::Texture2D:
- CopySurface(old_surface, new_surface, copy_pbo.handle);
- break;
- case SurfaceTarget::Texture3D:
- AccurateCopySurface(old_surface, new_surface);
- break;
- case SurfaceTarget::TextureCubemap:
- case SurfaceTarget::Texture2DArray:
- case SurfaceTarget::TextureCubeArray:
- if (compatible_formats)
- FastLayeredCopySurface(old_surface, new_surface);
- else {
- AccurateCopySurface(old_surface, new_surface);
- }
- break;
- default:
- LOG_CRITICAL(Render_OpenGL, "Unimplemented surface target={}",
- static_cast<u32>(new_params.target));
- UNREACHABLE();
- }
-
- return new_surface;
-}
-
-Surface RasterizerCacheOpenGL::TryFindFramebufferSurface(const u8* host_ptr) const {
- return TryGet(host_ptr);
-}
-
-void RasterizerCacheOpenGL::ReserveSurface(const Surface& surface) {
- const auto& surface_reserve_key{SurfaceReserveKey::Create(surface->GetSurfaceParams())};
- surface_reserve[surface_reserve_key] = surface;
-}
-
-Surface RasterizerCacheOpenGL::TryGetReservedSurface(const SurfaceParams& params) {
- const auto& surface_reserve_key{SurfaceReserveKey::Create(params)};
- auto search{surface_reserve.find(surface_reserve_key)};
- if (search != surface_reserve.end()) {
- return search->second;
- }
- return {};
-}
-
-static std::optional<u32> TryFindBestMipMap(std::size_t memory, const SurfaceParams params,
- u32 height) {
- for (u32 i = 0; i < params.max_mip_level; i++) {
- if (memory == params.GetMipmapSingleSize(i) && params.MipHeight(i) == height) {
- return {i};
- }
- }
- return {};
-}
-
-static std::optional<u32> TryFindBestLayer(GPUVAddr addr, const SurfaceParams params, u32 mipmap) {
- const std::size_t size{params.LayerMemorySize()};
- GPUVAddr start{params.gpu_addr + params.GetMipmapLevelOffset(mipmap)};
- for (u32 i = 0; i < params.depth; i++) {
- if (start == addr) {
- return {i};
- }
- start += size;
- }
- return {};
-}
-
-static bool LayerFitReinterpretSurface(RasterizerCacheOpenGL& cache, const Surface render_surface,
- const Surface blitted_surface) {
- const auto& dst_params = blitted_surface->GetSurfaceParams();
- const auto& src_params = render_surface->GetSurfaceParams();
- const std::size_t src_memory_size = src_params.size_in_bytes;
- const std::optional<u32> level =
- TryFindBestMipMap(src_memory_size, dst_params, src_params.height);
- if (level.has_value()) {
- if (src_params.width == dst_params.MipWidthGobAligned(*level) &&
- src_params.height == dst_params.MipHeight(*level) &&
- src_params.block_height >= dst_params.MipBlockHeight(*level)) {
- const std::optional<u32> slot =
- TryFindBestLayer(render_surface->GetSurfaceParams().gpu_addr, dst_params, *level);
- if (slot.has_value()) {
- glCopyImageSubData(render_surface->Texture().handle,
- SurfaceTargetToGL(src_params.target), 0, 0, 0, 0,
- blitted_surface->Texture().handle,
- SurfaceTargetToGL(dst_params.target), *level, 0, 0, *slot,
- dst_params.MipWidth(*level), dst_params.MipHeight(*level), 1);
- blitted_surface->MarkAsModified(true, cache);
- return true;
- }
- }
- }
- return false;
-}
-
-static bool IsReinterpretInvalid(const Surface render_surface, const Surface blitted_surface) {
- const VAddr bound1 = blitted_surface->GetCpuAddr() + blitted_surface->GetMemorySize();
- const VAddr bound2 = render_surface->GetCpuAddr() + render_surface->GetMemorySize();
- if (bound2 > bound1)
- return true;
- const auto& dst_params = blitted_surface->GetSurfaceParams();
- const auto& src_params = render_surface->GetSurfaceParams();
- return (dst_params.component_type != src_params.component_type);
-}
-
-static bool IsReinterpretInvalidSecond(const Surface render_surface,
- const Surface blitted_surface) {
- const auto& dst_params = blitted_surface->GetSurfaceParams();
- const auto& src_params = render_surface->GetSurfaceParams();
- return (dst_params.height > src_params.height && dst_params.width > src_params.width);
-}
-
-bool RasterizerCacheOpenGL::PartialReinterpretSurface(Surface triggering_surface,
- Surface intersect) {
- if (IsReinterpretInvalid(triggering_surface, intersect)) {
- Unregister(intersect);
- return false;
- }
- if (!LayerFitReinterpretSurface(*this, triggering_surface, intersect)) {
- if (IsReinterpretInvalidSecond(triggering_surface, intersect)) {
- Unregister(intersect);
- return false;
- }
- FlushObject(intersect);
- FlushObject(triggering_surface);
- intersect->MarkForReload(true);
- }
- return true;
-}
-
-void RasterizerCacheOpenGL::SignalPreDrawCall() {
- if (texception && GLAD_GL_ARB_texture_barrier) {
- glTextureBarrier();
- }
- texception = false;
-}
-
-void RasterizerCacheOpenGL::SignalPostDrawCall() {
- for (u32 i = 0; i < Maxwell::NumRenderTargets; i++) {
- if (current_color_buffers[i] != nullptr) {
- Surface intersect =
- CollideOnReinterpretedSurface(current_color_buffers[i]->GetCacheAddr());
- if (intersect != nullptr) {
- PartialReinterpretSurface(current_color_buffers[i], intersect);
- texception = true;
- }
- }
- }
-}
-
-} // namespace OpenGL
diff --git a/src/video_core/renderer_opengl/gl_rasterizer_cache.h b/src/video_core/renderer_opengl/gl_rasterizer_cache.h
deleted file mode 100644
index bbab79575..000000000
--- a/src/video_core/renderer_opengl/gl_rasterizer_cache.h
+++ /dev/null
@@ -1,575 +0,0 @@
-// Copyright 2018 yuzu Emulator Project
-// Licensed under GPLv2 or any later version
-// Refer to the license.txt file included.
-
-#pragma once
-
-#include <array>
-#include <memory>
-#include <string>
-#include <tuple>
-#include <vector>
-
-#include "common/alignment.h"
-#include "common/bit_util.h"
-#include "common/common_types.h"
-#include "common/hash.h"
-#include "common/math_util.h"
-#include "video_core/engines/fermi_2d.h"
-#include "video_core/engines/maxwell_3d.h"
-#include "video_core/rasterizer_cache.h"
-#include "video_core/renderer_opengl/gl_resource_manager.h"
-#include "video_core/renderer_opengl/gl_shader_gen.h"
-#include "video_core/surface.h"
-#include "video_core/textures/decoders.h"
-#include "video_core/textures/texture.h"
-
-namespace OpenGL {
-
-class CachedSurface;
-using Surface = std::shared_ptr<CachedSurface>;
-using SurfaceSurfaceRect_Tuple = std::tuple<Surface, Surface, Common::Rectangle<u32>>;
-
-using SurfaceTarget = VideoCore::Surface::SurfaceTarget;
-using SurfaceType = VideoCore::Surface::SurfaceType;
-using PixelFormat = VideoCore::Surface::PixelFormat;
-using ComponentType = VideoCore::Surface::ComponentType;
-using Maxwell = Tegra::Engines::Maxwell3D::Regs;
-
-struct SurfaceParams {
- enum class SurfaceClass {
- Uploaded,
- RenderTarget,
- DepthBuffer,
- Copy,
- };
-
- static std::string SurfaceTargetName(SurfaceTarget target) {
- switch (target) {
- case SurfaceTarget::Texture1D:
- return "Texture1D";
- case SurfaceTarget::Texture2D:
- return "Texture2D";
- case SurfaceTarget::Texture3D:
- return "Texture3D";
- case SurfaceTarget::Texture1DArray:
- return "Texture1DArray";
- case SurfaceTarget::Texture2DArray:
- return "Texture2DArray";
- case SurfaceTarget::TextureCubemap:
- return "TextureCubemap";
- case SurfaceTarget::TextureCubeArray:
- return "TextureCubeArray";
- default:
- LOG_CRITICAL(HW_GPU, "Unimplemented surface_target={}", static_cast<u32>(target));
- UNREACHABLE();
- return fmt::format("TextureUnknown({})", static_cast<u32>(target));
- }
- }
-
- u32 GetFormatBpp() const {
- return VideoCore::Surface::GetFormatBpp(pixel_format);
- }
-
- /// Returns the rectangle corresponding to this surface
- Common::Rectangle<u32> GetRect(u32 mip_level = 0) const;
-
- /// Returns the total size of this surface in bytes, adjusted for compression
- std::size_t SizeInBytesRaw(bool ignore_tiled = false) const {
- const u32 compression_factor{GetCompressionFactor(pixel_format)};
- const u32 bytes_per_pixel{GetBytesPerPixel(pixel_format)};
- const size_t uncompressed_size{
- Tegra::Texture::CalculateSize((ignore_tiled ? false : is_tiled), bytes_per_pixel, width,
- height, depth, block_height, block_depth)};
-
- // Divide by compression_factor^2, as height and width are factored by this
- return uncompressed_size / (compression_factor * compression_factor);
- }
-
- /// Returns the size of this surface as an OpenGL texture in bytes
- std::size_t SizeInBytesGL() const {
- return SizeInBytesRaw(true);
- }
-
- /// Returns the size of this surface as a cube face in bytes
- std::size_t SizeInBytesCubeFace() const {
- return size_in_bytes / 6;
- }
-
- /// Returns the size of this surface as an OpenGL cube face in bytes
- std::size_t SizeInBytesCubeFaceGL() const {
- return size_in_bytes_gl / 6;
- }
-
- /// Returns the exact size of memory occupied by the texture in VRAM, including mipmaps.
- std::size_t MemorySize() const {
- std::size_t size = InnerMemorySize(false, is_layered);
- if (is_layered)
- return size * depth;
- return size;
- }
-
- /// Returns true if the parameters constitute a valid rasterizer surface.
- bool IsValid() const {
- return gpu_addr && host_ptr && height && width;
- }
-
- /// Returns the exact size of the memory occupied by a layer in a texture in VRAM, including
- /// mipmaps.
- std::size_t LayerMemorySize() const {
- return InnerMemorySize(false, true);
- }
-
- /// Returns the size of a layer of this surface in OpenGL.
- std::size_t LayerSizeGL(u32 mip_level) const {
- return InnerMipmapMemorySize(mip_level, true, is_layered, false);
- }
-
- std::size_t GetMipmapSizeGL(u32 mip_level, bool ignore_compressed = true) const {
- std::size_t size = InnerMipmapMemorySize(mip_level, true, is_layered, ignore_compressed);
- if (is_layered)
- return size * depth;
- return size;
- }
-
- std::size_t GetMipmapLevelOffset(u32 mip_level) const {
- std::size_t offset = 0;
- for (u32 i = 0; i < mip_level; i++)
- offset += InnerMipmapMemorySize(i, false, is_layered);
- return offset;
- }
-
- std::size_t GetMipmapLevelOffsetGL(u32 mip_level) const {
- std::size_t offset = 0;
- for (u32 i = 0; i < mip_level; i++)
- offset += InnerMipmapMemorySize(i, true, is_layered);
- return offset;
- }
-
- std::size_t GetMipmapSingleSize(u32 mip_level) const {
- return InnerMipmapMemorySize(mip_level, false, is_layered);
- }
-
- u32 MipWidth(u32 mip_level) const {
- return std::max(1U, width >> mip_level);
- }
-
- u32 MipWidthGobAligned(u32 mip_level) const {
- return Common::AlignUp(std::max(1U, width >> mip_level), 64U * 8U / GetFormatBpp());
- }
-
- u32 MipHeight(u32 mip_level) const {
- return std::max(1U, height >> mip_level);
- }
-
- u32 MipDepth(u32 mip_level) const {
- return is_layered ? depth : std::max(1U, depth >> mip_level);
- }
-
- // Auto block resizing algorithm from:
- // https://cgit.freedesktop.org/mesa/mesa/tree/src/gallium/drivers/nouveau/nv50/nv50_miptree.c
- u32 MipBlockHeight(u32 mip_level) const {
- if (mip_level == 0)
- return block_height;
- u32 alt_height = MipHeight(mip_level);
- u32 h = GetDefaultBlockHeight(pixel_format);
- u32 blocks_in_y = (alt_height + h - 1) / h;
- u32 bh = 16;
- while (bh > 1 && blocks_in_y <= bh * 4) {
- bh >>= 1;
- }
- return bh;
- }
-
- u32 MipBlockDepth(u32 mip_level) const {
- if (mip_level == 0) {
- return block_depth;
- }
-
- if (is_layered) {
- return 1;
- }
-
- const u32 mip_depth = MipDepth(mip_level);
- u32 bd = 32;
- while (bd > 1 && mip_depth * 2 <= bd) {
- bd >>= 1;
- }
-
- if (bd == 32) {
- const u32 bh = MipBlockHeight(mip_level);
- if (bh >= 4) {
- return 16;
- }
- }
-
- return bd;
- }
-
- u32 RowAlign(u32 mip_level) const {
- const u32 m_width = MipWidth(mip_level);
- const u32 bytes_per_pixel = GetBytesPerPixel(pixel_format);
- const u32 l2 = Common::CountTrailingZeroes32(m_width * bytes_per_pixel);
- return (1U << l2);
- }
-
- /// Creates SurfaceParams from a texture configuration
- static SurfaceParams CreateForTexture(const Tegra::Texture::FullTextureInfo& config,
- const GLShader::SamplerEntry& entry);
-
- /// Creates SurfaceParams from a framebuffer configuration
- static SurfaceParams CreateForFramebuffer(std::size_t index);
-
- /// Creates SurfaceParams for a depth buffer configuration
- static SurfaceParams CreateForDepthBuffer(
- u32 zeta_width, u32 zeta_height, GPUVAddr zeta_address, Tegra::DepthFormat format,
- u32 block_width, u32 block_height, u32 block_depth,
- Tegra::Engines::Maxwell3D::Regs::InvMemoryLayout type);
-
- /// Creates SurfaceParams for a Fermi2D surface copy
- static SurfaceParams CreateForFermiCopySurface(
- const Tegra::Engines::Fermi2D::Regs::Surface& config);
-
- /// Checks if surfaces are compatible for caching
- bool IsCompatibleSurface(const SurfaceParams& other) const {
- if (std::tie(pixel_format, type, width, height, target, depth, is_tiled) ==
- std::tie(other.pixel_format, other.type, other.width, other.height, other.target,
- other.depth, other.is_tiled)) {
- if (!is_tiled)
- return true;
- return std::tie(block_height, block_depth, tile_width_spacing) ==
- std::tie(other.block_height, other.block_depth, other.tile_width_spacing);
- }
- return false;
- }
-
- /// Initializes parameters for caching, should be called after everything has been initialized
- void InitCacheParameters(GPUVAddr gpu_addr);
-
- std::string TargetName() const {
- switch (target) {
- case SurfaceTarget::Texture1D:
- return "1D";
- case SurfaceTarget::TextureBuffer:
- return "Buffer";
- case SurfaceTarget::Texture2D:
- return "2D";
- case SurfaceTarget::Texture3D:
- return "3D";
- case SurfaceTarget::Texture1DArray:
- return "1DArray";
- case SurfaceTarget::Texture2DArray:
- return "2DArray";
- case SurfaceTarget::TextureCubemap:
- return "Cube";
- default:
- LOG_CRITICAL(HW_GPU, "Unimplemented surface_target={}", static_cast<u32>(target));
- UNREACHABLE();
- return fmt::format("TUK({})", static_cast<u32>(target));
- }
- }
-
- std::string ClassName() const {
- switch (identity) {
- case SurfaceClass::Uploaded:
- return "UP";
- case SurfaceClass::RenderTarget:
- return "RT";
- case SurfaceClass::DepthBuffer:
- return "DB";
- case SurfaceClass::Copy:
- return "CP";
- default:
- LOG_CRITICAL(HW_GPU, "Unimplemented surface_class={}", static_cast<u32>(identity));
- UNREACHABLE();
- return fmt::format("CUK({})", static_cast<u32>(identity));
- }
- }
-
- std::string IdentityString() const {
- return ClassName() + '_' + TargetName() + '_' + (is_tiled ? 'T' : 'L');
- }
-
- bool is_tiled;
- u32 block_width;
- u32 block_height;
- u32 block_depth;
- u32 tile_width_spacing;
- PixelFormat pixel_format;
- ComponentType component_type;
- SurfaceType type;
- u32 width;
- u32 height;
- u32 depth;
- u32 unaligned_height;
- u32 pitch;
- SurfaceTarget target;
- SurfaceClass identity;
- u32 max_mip_level;
- bool is_layered;
- bool is_array;
- bool srgb_conversion;
- // Parameters used for caching
- u8* host_ptr;
- GPUVAddr gpu_addr;
- std::size_t size_in_bytes;
- std::size_t size_in_bytes_gl;
-
- // Render target specific parameters, not used in caching
- struct {
- u32 index;
- u32 array_mode;
- u32 volume;
- u32 layer_stride;
- u32 base_layer;
- } rt;
-
-private:
- std::size_t InnerMipmapMemorySize(u32 mip_level, bool force_gl = false, bool layer_only = false,
- bool uncompressed = false) const;
- std::size_t InnerMemorySize(bool force_gl = false, bool layer_only = false,
- bool uncompressed = false) const;
-};
-
-}; // namespace OpenGL
-
-/// Hashable variation of SurfaceParams, used for a key in the surface cache
-struct SurfaceReserveKey : Common::HashableStruct<OpenGL::SurfaceParams> {
- static SurfaceReserveKey Create(const OpenGL::SurfaceParams& params) {
- SurfaceReserveKey res;
- res.state = params;
- res.state.identity = {}; // Ignore the origin of the texture
- res.state.gpu_addr = {}; // Ignore GPU vaddr in caching
- res.state.rt = {}; // Ignore rt config in caching
- return res;
- }
-};
-namespace std {
-template <>
-struct hash<SurfaceReserveKey> {
- std::size_t operator()(const SurfaceReserveKey& k) const {
- return k.Hash();
- }
-};
-} // namespace std
-
-namespace OpenGL {
-
-class RasterizerOpenGL;
-
-// This is used to store temporary big buffers,
-// instead of creating/destroying all the time
-struct RasterizerTemporaryMemory {
- std::vector<std::vector<u8>> gl_buffer;
-};
-
-class CachedSurface final : public RasterizerCacheObject {
-public:
- explicit CachedSurface(const SurfaceParams& params);
-
- VAddr GetCpuAddr() const override {
- return cpu_addr;
- }
-
- std::size_t GetSizeInBytes() const override {
- return cached_size_in_bytes;
- }
-
- std::size_t GetMemorySize() const {
- return memory_size;
- }
-
- const OGLTexture& Texture() const {
- return texture;
- }
-
- const OGLTexture& Texture(bool as_array) {
- if (params.is_array == as_array) {
- return texture;
- } else {
- EnsureTextureDiscrepantView();
- return discrepant_view;
- }
- }
-
- GLenum Target() const {
- return gl_target;
- }
-
- const SurfaceParams& GetSurfaceParams() const {
- return params;
- }
-
- // Read/Write data in Switch memory to/from gl_buffer
- void LoadGLBuffer(RasterizerTemporaryMemory& res_cache_tmp_mem);
- void FlushGLBuffer(RasterizerTemporaryMemory& res_cache_tmp_mem);
-
- // Upload data in gl_buffer to this surface's texture
- void UploadGLTexture(RasterizerTemporaryMemory& res_cache_tmp_mem, GLuint read_fb_handle,
- GLuint draw_fb_handle);
-
- void UpdateSwizzle(Tegra::Texture::SwizzleSource swizzle_x,
- Tegra::Texture::SwizzleSource swizzle_y,
- Tegra::Texture::SwizzleSource swizzle_z,
- Tegra::Texture::SwizzleSource swizzle_w);
-
- void MarkReinterpreted() {
- reinterpreted = true;
- }
-
- bool IsReinterpreted() const {
- return reinterpreted;
- }
-
- void MarkForReload(bool reload) {
- must_reload = reload;
- }
-
- bool MustReload() const {
- return must_reload;
- }
-
- bool IsUploaded() const {
- return params.identity == SurfaceParams::SurfaceClass::Uploaded;
- }
-
-private:
- void UploadGLMipmapTexture(RasterizerTemporaryMemory& res_cache_tmp_mem, u32 mip_map,
- GLuint read_fb_handle, GLuint draw_fb_handle);
-
- void EnsureTextureDiscrepantView();
-
- OGLTexture texture;
- OGLTexture discrepant_view;
- OGLBuffer texture_buffer;
- SurfaceParams params{};
- GLenum gl_target{};
- GLenum gl_internal_format{};
- std::size_t cached_size_in_bytes{};
- std::array<GLenum, 4> swizzle{GL_RED, GL_GREEN, GL_BLUE, GL_ALPHA};
- std::size_t memory_size;
- bool reinterpreted = false;
- bool must_reload = false;
- VAddr cpu_addr{};
-};
-
-class RasterizerCacheOpenGL final : public RasterizerCache<Surface> {
-public:
- explicit RasterizerCacheOpenGL(RasterizerOpenGL& rasterizer);
-
- /// Get a surface based on the texture configuration
- Surface GetTextureSurface(const Tegra::Texture::FullTextureInfo& config,
- const GLShader::SamplerEntry& entry);
-
- /// Get the depth surface based on the framebuffer configuration
- Surface GetDepthBufferSurface(bool preserve_contents);
-
- /// Get the color surface based on the framebuffer configuration and the specified render target
- Surface GetColorBufferSurface(std::size_t index, bool preserve_contents);
-
- /// Tries to find a framebuffer using on the provided CPU address
- Surface TryFindFramebufferSurface(const u8* host_ptr) const;
-
- /// Copies the contents of one surface to another
- void FermiCopySurface(const Tegra::Engines::Fermi2D::Regs::Surface& src_config,
- const Tegra::Engines::Fermi2D::Regs::Surface& dst_config,
- const Common::Rectangle<u32>& src_rect,
- const Common::Rectangle<u32>& dst_rect);
-
- void SignalPreDrawCall();
- void SignalPostDrawCall();
-
-protected:
- void FlushObjectInner(const Surface& object) override {
- object->FlushGLBuffer(temporal_memory);
- }
-
-private:
- void LoadSurface(const Surface& surface);
- Surface GetSurface(const SurfaceParams& params, bool preserve_contents = true);
-
- /// Gets an uncached surface, creating it if need be
- Surface GetUncachedSurface(const SurfaceParams& params);
-
- /// Recreates a surface with new parameters
- Surface RecreateSurface(const Surface& old_surface, const SurfaceParams& new_params);
-
- /// Reserves a unique surface that can be reused later
- void ReserveSurface(const Surface& surface);
-
- /// Tries to get a reserved surface for the specified parameters
- Surface TryGetReservedSurface(const SurfaceParams& params);
-
- // Partialy reinterpret a surface based on a triggering_surface that collides with it.
- // returns true if the reinterpret was successful, false in case it was not.
- bool PartialReinterpretSurface(Surface triggering_surface, Surface intersect);
-
- /// Performs a slow but accurate surface copy, flushing to RAM and reinterpreting the data
- void AccurateCopySurface(const Surface& src_surface, const Surface& dst_surface);
- void FastLayeredCopySurface(const Surface& src_surface, const Surface& dst_surface);
- void FastCopySurface(const Surface& src_surface, const Surface& dst_surface);
- void CopySurface(const Surface& src_surface, const Surface& dst_surface,
- const GLuint copy_pbo_handle, const GLenum src_attachment = 0,
- const GLenum dst_attachment = 0, const std::size_t cubemap_face = 0);
-
- /// The surface reserve is a "backup" cache, this is where we put unique surfaces that have
- /// previously been used. This is to prevent surfaces from being constantly created and
- /// destroyed when used with different surface parameters.
- std::unordered_map<SurfaceReserveKey, Surface> surface_reserve;
-
- OGLFramebuffer read_framebuffer;
- OGLFramebuffer draw_framebuffer;
-
- bool texception = false;
-
- /// Use a Pixel Buffer Object to download the previous texture and then upload it to the new one
- /// using the new format.
- OGLBuffer copy_pbo;
-
- std::array<Surface, Maxwell::NumRenderTargets> last_color_buffers;
- std::array<Surface, Maxwell::NumRenderTargets> current_color_buffers;
- Surface last_depth_buffer;
-
- RasterizerTemporaryMemory temporal_memory;
-
- using SurfaceIntervalCache = boost::icl::interval_map<CacheAddr, Surface>;
- using SurfaceInterval = typename SurfaceIntervalCache::interval_type;
-
- static auto GetReinterpretInterval(const Surface& object) {
- return SurfaceInterval::right_open(object->GetCacheAddr() + 1,
- object->GetCacheAddr() + object->GetMemorySize() - 1);
- }
-
- // Reinterpreted surfaces are very fragil as the game may keep rendering into them.
- SurfaceIntervalCache reinterpreted_surfaces;
-
- void RegisterReinterpretSurface(Surface reinterpret_surface) {
- auto interval = GetReinterpretInterval(reinterpret_surface);
- reinterpreted_surfaces.insert({interval, reinterpret_surface});
- reinterpret_surface->MarkReinterpreted();
- }
-
- Surface CollideOnReinterpretedSurface(CacheAddr addr) const {
- const SurfaceInterval interval{addr};
- for (auto& pair :
- boost::make_iterator_range(reinterpreted_surfaces.equal_range(interval))) {
- return pair.second;
- }
- return nullptr;
- }
-
- void Register(const Surface& object) override {
- RasterizerCache<Surface>::Register(object);
- }
-
- /// Unregisters an object from the cache
- void Unregister(const Surface& object) override {
- if (object->IsReinterpreted()) {
- auto interval = GetReinterpretInterval(object);
- reinterpreted_surfaces.erase(interval);
- }
- RasterizerCache<Surface>::Unregister(object);
- }
-};
-
-} // namespace OpenGL