mirrors/yuzu
2
1
Fork 0
mirror of https://github.com/yuzu-emu/yuzu.git synced 2024-07-04 23:31:19 +01:00
Code Issues Releases Activity

gl_rasterizer_cache: Remove Citra's rasterizer cache, always load/flush surfaces.

Browse source
This commit is contained in:
bunnei 2018-06-21 19:36:01 -04:00
parent 84cadf9918
commit 5f57ab1b2a
4 changed files with 212 additions and 1496 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

109
src/video_core/renderer_opengl/gl_rasterizer.cpp
View file

@ -146,7 +146,6 @@ std::pair<u8*, GLintptr> RasterizerOpenGL::SetupVertexArrays(u8* array_ptr,
u64 size = end - start + 1; u64 size = end - start + 1;
// Copy vertex array data // Copy vertex array data
res_cache.FlushRegion(start, size, nullptr);
Memory::ReadBlock(*memory_manager->GpuToCpuAddress(start), array_ptr, size); Memory::ReadBlock(*memory_manager->GpuToCpuAddress(start), array_ptr, size);
// Bind the vertex array to the buffer at the current offset. // Bind the vertex array to the buffer at the current offset.
@ -325,29 +324,22 @@ void RasterizerOpenGL::DrawArrays() {
std::tie(color_surface, depth_surface, surfaces_rect) = std::tie(color_surface, depth_surface, surfaces_rect) =
res_cache.GetFramebufferSurfaces(using_color_fb, using_depth_fb, viewport_rect); res_cache.GetFramebufferSurfaces(using_color_fb, using_depth_fb, viewport_rect);
const u16 res_scale = color_surface != nullptr
? color_surface->res_scale
: (depth_surface == nullptr ? 1u : depth_surface->res_scale);
MathUtil::Rectangle<u32> draw_rect{ MathUtil::Rectangle<u32> draw_rect{
static_cast<u32>( static_cast<u32>(std::clamp<s32>(static_cast<s32>(surfaces_rect.left) + viewport_rect.left,
std::clamp<s32>(static_cast<s32>(surfaces_rect.left) + viewport_rect.left * res_scale,
surfaces_rect.left, surfaces_rect.right)), // Left surfaces_rect.left, surfaces_rect.right)), // Left
static_cast<u32>( static_cast<u32>(std::clamp<s32>(static_cast<s32>(surfaces_rect.bottom) + viewport_rect.top,
std::clamp<s32>(static_cast<s32>(surfaces_rect.bottom) + viewport_rect.top * res_scale,
surfaces_rect.bottom, surfaces_rect.top)), // Top surfaces_rect.bottom, surfaces_rect.top)), // Top
static_cast<u32>( static_cast<u32>(std::clamp<s32>(static_cast<s32>(surfaces_rect.left) + viewport_rect.right,
std::clamp<s32>(static_cast<s32>(surfaces_rect.left) + viewport_rect.right * res_scale,
surfaces_rect.left, surfaces_rect.right)), // Right surfaces_rect.left, surfaces_rect.right)), // Right
static_cast<u32>(std::clamp<s32>(static_cast<s32>(surfaces_rect.bottom) + static_cast<u32>(
viewport_rect.bottom * res_scale, std::clamp<s32>(static_cast<s32>(surfaces_rect.bottom) + viewport_rect.bottom,
surfaces_rect.bottom, surfaces_rect.top))}; // Bottom surfaces_rect.bottom, surfaces_rect.top))}; // Bottom
// Bind the framebuffer surfaces // Bind the framebuffer surfaces
BindFramebufferSurfaces(color_surface, depth_surface, has_stencil); BindFramebufferSurfaces(color_surface, depth_surface, has_stencil);
// Sync the viewport // Sync the viewport
SyncViewport(surfaces_rect, res_scale); SyncViewport(surfaces_rect);
// Sync the blend state registers // Sync the blend state registers
SyncBlendState(); SyncBlendState();
@ -442,44 +434,23 @@ void RasterizerOpenGL::DrawArrays() {
state.Apply(); state.Apply();
// Mark framebuffer surfaces as dirty // Mark framebuffer surfaces as dirty
MathUtil::Rectangle<u32> draw_rect_unscaled{
draw_rect.left / res_scale, draw_rect.top / res_scale, draw_rect.right / res_scale,
draw_rect.bottom / res_scale};
if (color_surface != nullptr && write_color_fb) { if (color_surface != nullptr && write_color_fb) {
auto interval = color_surface->GetSubRectInterval(draw_rect_unscaled); res_cache.FlushSurface(color_surface);
res_cache.InvalidateRegion(boost::icl::first(interval), boost::icl::length(interval),
color_surface);
} }
if (depth_surface != nullptr && write_depth_fb) { if (depth_surface != nullptr && write_depth_fb) {
auto interval = depth_surface->GetSubRectInterval(draw_rect_unscaled); res_cache.FlushSurface(depth_surface);
res_cache.InvalidateRegion(boost::icl::first(interval), boost::icl::length(interval),
depth_surface);
} }
} }
void RasterizerOpenGL::NotifyMaxwellRegisterChanged(u32 method) {} void RasterizerOpenGL::NotifyMaxwellRegisterChanged(u32 method) {}
void RasterizerOpenGL::FlushAll() { void RasterizerOpenGL::FlushAll() {}
MICROPROFILE_SCOPE(OpenGL_CacheManagement);
res_cache.FlushAll();
}
void RasterizerOpenGL::FlushRegion(Tegra::GPUVAddr addr, u64 size) { void RasterizerOpenGL::FlushRegion(Tegra::GPUVAddr addr, u64 size) {}
MICROPROFILE_SCOPE(OpenGL_CacheManagement);
res_cache.FlushRegion(addr, size);
}
void RasterizerOpenGL::InvalidateRegion(Tegra::GPUVAddr addr, u64 size) { void RasterizerOpenGL::InvalidateRegion(Tegra::GPUVAddr addr, u64 size) {}
MICROPROFILE_SCOPE(OpenGL_CacheManagement);
res_cache.InvalidateRegion(addr, size, nullptr);
}
void RasterizerOpenGL::FlushAndInvalidateRegion(Tegra::GPUVAddr addr, u64 size) { void RasterizerOpenGL::FlushAndInvalidateRegion(Tegra::GPUVAddr addr, u64 size) {}
MICROPROFILE_SCOPE(OpenGL_CacheManagement);
res_cache.FlushRegion(addr, size);
res_cache.InvalidateRegion(addr, size, nullptr);
}
bool RasterizerOpenGL::AccelerateDisplayTransfer(const void* config) { bool RasterizerOpenGL::AccelerateDisplayTransfer(const void* config) {
MICROPROFILE_SCOPE(OpenGL_Blits); MICROPROFILE_SCOPE(OpenGL_Blits);
@ -500,45 +471,9 @@ bool RasterizerOpenGL::AccelerateFill(const void* config) {
bool RasterizerOpenGL::AccelerateDisplay(const Tegra::FramebufferConfig& framebuffer, bool RasterizerOpenGL::AccelerateDisplay(const Tegra::FramebufferConfig& framebuffer,
VAddr framebuffer_addr, u32 pixel_stride, VAddr framebuffer_addr, u32 pixel_stride,
ScreenInfo& screen_info) { ScreenInfo& screen_info) {
if (framebuffer_addr == 0) { // TODO(bunnei): ImplementMe
return false; return false;
} }
MICROPROFILE_SCOPE(OpenGL_CacheManagement);
SurfaceParams src_params;
src_params.cpu_addr = framebuffer_addr;
src_params.addr = res_cache.TryFindFramebufferGpuAddress(framebuffer_addr).get_value_or(0);
src_params.width = std::min(framebuffer.width, pixel_stride);
src_params.height = framebuffer.height;
src_params.stride = pixel_stride;
src_params.is_tiled = true;
src_params.block_height = Tegra::Texture::TICEntry::DefaultBlockHeight;
src_params.pixel_format =
SurfaceParams::PixelFormatFromGPUPixelFormat(framebuffer.pixel_format);
src_params.component_type =
SurfaceParams::ComponentTypeFromGPUPixelFormat(framebuffer.pixel_format);
src_params.UpdateParams();
MathUtil::Rectangle<u32> src_rect;
Surface src_surface;
std::tie(src_surface, src_rect) =
res_cache.GetSurfaceSubRect(src_params, ScaleMatch::Ignore, true);
if (src_surface == nullptr) {
return false;
}
u32 scaled_width = src_surface->GetScaledWidth();
u32 scaled_height = src_surface->GetScaledHeight();
screen_info.display_texcoords = MathUtil::Rectangle<float>(
(float)src_rect.bottom / (float)scaled_height, (float)src_rect.left / (float)scaled_width,
(float)src_rect.top / (float)scaled_height, (float)src_rect.right / (float)scaled_width);
screen_info.display_texture = src_surface->texture.handle;
return true;
}
void RasterizerOpenGL::SamplerInfo::Create() { void RasterizerOpenGL::SamplerInfo::Create() {
sampler.Create(); sampler.Create();
@ -674,7 +609,7 @@ u32 RasterizerOpenGL::SetupTextures(Maxwell::ShaderStage stage, GLuint program,
texture_samplers[current_bindpoint].SyncWithConfig(texture.tsc); texture_samplers[current_bindpoint].SyncWithConfig(texture.tsc);
Surface surface = res_cache.GetTextureSurface(texture); Surface surface = res_cache.GetTextureSurface(texture);
if (surface != nullptr) { if (surface != nullptr) {
state.texture_units[current_bindpoint].texture_2d = surface->texture.handle; state.texture_units[current_bindpoint].texture_2d = surface->Texture().handle;
state.texture_units[current_bindpoint].swizzle.r = state.texture_units[current_bindpoint].swizzle.r =
MaxwellToGL::SwizzleSource(texture.tic.x_source); MaxwellToGL::SwizzleSource(texture.tic.x_source);
state.texture_units[current_bindpoint].swizzle.g = state.texture_units[current_bindpoint].swizzle.g =
@ -700,16 +635,16 @@ void RasterizerOpenGL::BindFramebufferSurfaces(const Surface& color_surface,
state.Apply(); state.Apply();
glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D, glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D,
color_surface != nullptr ? color_surface->texture.handle : 0, 0); color_surface != nullptr ? color_surface->Texture().handle : 0, 0);
if (depth_surface != nullptr) { if (depth_surface != nullptr) {
if (has_stencil) { if (has_stencil) {
// attach both depth and stencil // attach both depth and stencil
glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_DEPTH_STENCIL_ATTACHMENT, GL_TEXTURE_2D, glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_DEPTH_STENCIL_ATTACHMENT, GL_TEXTURE_2D,
depth_surface->texture.handle, 0); depth_surface->Texture().handle, 0);
} else { } else {
// attach depth // attach depth
glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_DEPTH_ATTACHMENT, GL_TEXTURE_2D, glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_DEPTH_ATTACHMENT, GL_TEXTURE_2D,
depth_surface->texture.handle, 0); depth_surface->Texture().handle, 0);
// clear stencil attachment // clear stencil attachment
glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_STENCIL_ATTACHMENT, GL_TEXTURE_2D, 0, 0); glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_STENCIL_ATTACHMENT, GL_TEXTURE_2D, 0, 0);
} }
@ -720,14 +655,14 @@ void RasterizerOpenGL::BindFramebufferSurfaces(const Surface& color_surface,
} }
} }
void RasterizerOpenGL::SyncViewport(const MathUtil::Rectangle<u32>& surfaces_rect, u16 res_scale) { void RasterizerOpenGL::SyncViewport(const MathUtil::Rectangle<u32>& surfaces_rect) {
const auto& regs = Core::System().GetInstance().GPU().Maxwell3D().regs; const auto& regs = Core::System().GetInstance().GPU().Maxwell3D().regs;
const MathUtil::Rectangle<s32> viewport_rect{regs.viewport_transform[0].GetRect()}; const MathUtil::Rectangle<s32> viewport_rect{regs.viewport_transform[0].GetRect()};
state.viewport.x = static_cast<GLint>(surfaces_rect.left) + viewport_rect.left * res_scale; state.viewport.x = static_cast<GLint>(surfaces_rect.left) + viewport_rect.left;
state.viewport.y = static_cast<GLint>(surfaces_rect.bottom) + viewport_rect.bottom * res_scale; state.viewport.y = static_cast<GLint>(surfaces_rect.bottom) + viewport_rect.bottom;
state.viewport.width = static_cast<GLsizei>(viewport_rect.GetWidth() * res_scale); state.viewport.width = static_cast<GLsizei>(viewport_rect.GetWidth());
state.viewport.height = static_cast<GLsizei>(viewport_rect.GetHeight() * res_scale); state.viewport.height = static_cast<GLsizei>(viewport_rect.GetHeight());
} }
void RasterizerOpenGL::SyncClipEnabled() { void RasterizerOpenGL::SyncClipEnabled() {

2
src/video_core/renderer_opengl/gl_rasterizer.h
View file

@ -109,7 +109,7 @@ private:
u32 current_unit, const std::vector<GLShader::SamplerEntry>& entries); u32 current_unit, const std::vector<GLShader::SamplerEntry>& entries);
/// Syncs the viewport to match the guest state /// Syncs the viewport to match the guest state
void SyncViewport(const MathUtil::Rectangle<u32>& surfaces_rect, u16 res_scale); void SyncViewport(const MathUtil::Rectangle<u32>& surfaces_rect);
/// Syncs the clip enabled status to match the guest state /// Syncs the clip enabled status to match the guest state
void SyncClipEnabled(); void SyncClipEnabled();

1230
src/video_core/renderer_opengl/gl_rasterizer_cache.cpp
View file

File diff suppressed because it is too large Load diff

333
src/video_core/renderer_opengl/gl_rasterizer_cache.h
View file

@ -1,57 +1,22 @@
// Copyright 2015 Citra Emulator Project // Copyright 2018 yuzu Emulator Project
// Licensed under GPLv2 or any later version // Licensed under GPLv2 or any later version
// Refer to the license.txt file included. // Refer to the license.txt file included.
#pragma once #pragma once
#include <array> #include <array>
#include <map>
#include <memory> #include <memory>
#include <set>
#include <tuple>
#ifdef __GNUC__
#pragma GCC diagnostic push
#pragma GCC diagnostic ignored "-Wunused-local-typedefs"
#endif
#include <boost/icl/interval_map.hpp>
#include <boost/icl/interval_set.hpp>
#ifdef __GNUC__
#pragma GCC diagnostic pop
#endif
#include <boost/optional.hpp>
#include <glad/glad.h>
#include "common/assert.h"
#include "common/common_funcs.h"
#include "common/common_types.h" #include "common/common_types.h"
#include "common/math_util.h" #include "common/math_util.h"
#include "video_core/gpu.h"
#include "video_core/memory_manager.h"
#include "video_core/renderer_opengl/gl_resource_manager.h" #include "video_core/renderer_opengl/gl_resource_manager.h"
#include "video_core/textures/texture.h" #include "video_core/textures/texture.h"
struct CachedSurface; class CachedSurface;
using Surface = std::shared_ptr<CachedSurface>; using Surface = std::shared_ptr<CachedSurface>;
using SurfaceSet = std::set<Surface>;
using SurfaceRegions = boost::icl::interval_set<Tegra::GPUVAddr>;
using SurfaceMap = boost::icl::interval_map<Tegra::GPUVAddr, Surface>;
using SurfaceCache = boost::icl::interval_map<Tegra::GPUVAddr, SurfaceSet>;
using SurfaceInterval = SurfaceCache::interval_type;
static_assert(std::is_same<SurfaceRegions::interval_type, SurfaceCache::interval_type>() &&
std::is_same<SurfaceMap::interval_type, SurfaceCache::interval_type>(),
"incorrect interval types");
using SurfaceRect_Tuple = std::tuple<Surface, MathUtil::Rectangle<u32>>;
using SurfaceSurfaceRect_Tuple = std::tuple<Surface, Surface, MathUtil::Rectangle<u32>>; using SurfaceSurfaceRect_Tuple = std::tuple<Surface, Surface, MathUtil::Rectangle<u32>>;
using PageMap = boost::icl::interval_map<u64, int>;
enum class ScaleMatch {
Exact, // only accept same res scale
Upscale, // only allow higher scale than params
Ignore // accept every scaled res
};
struct SurfaceParams { struct SurfaceParams {
enum class PixelFormat { enum class PixelFormat {
ABGR8 = 0, ABGR8 = 0,
@ -93,10 +58,10 @@ struct SurfaceParams {
/** /**
* Gets the compression factor for the specified PixelFormat. This applies to just the * Gets the compression factor for the specified PixelFormat. This applies to just the
* "compressed width" and "compressed height", not the overall compression factor of a * "compressed width" and "compressed height", not the overall compression factor of a
* compressed image. This is used for maintaining proper surface sizes for compressed texture * compressed image. This is used for maintaining proper surface sizes for compressed
* formats. * texture formats.
*/ */
static constexpr u32 GetCompresssionFactor(PixelFormat format) { static constexpr u32 GetCompressionFactor(PixelFormat format) {
if (format == PixelFormat::Invalid) if (format == PixelFormat::Invalid)
return 0; return 0;
@ -118,8 +83,8 @@ struct SurfaceParams {
ASSERT(static_cast<size_t>(format) < compression_factor_table.size()); ASSERT(static_cast<size_t>(format) < compression_factor_table.size());
return compression_factor_table[static_cast<size_t>(format)]; return compression_factor_table[static_cast<size_t>(format)];
} }
u32 GetCompresssionFactor() const { u32 GetCompressionFactor() const {
return GetCompresssionFactor(pixel_format); return GetCompressionFactor(pixel_format);
} }
static constexpr u32 GetFormatBpp(PixelFormat format) { static constexpr u32 GetFormatBpp(PixelFormat format) {
@ -165,25 +130,6 @@ struct SurfaceParams {
} }
} }
static bool IsFormatASTC(PixelFormat format) {
switch (format) {
case PixelFormat::ASTC_2D_4X4:
return true;
default:
return false;
}
}
static PixelFormat PixelFormatFromGPUPixelFormat(Tegra::FramebufferConfig::PixelFormat format) {
switch (format) {
case Tegra::FramebufferConfig::PixelFormat::ABGR8:
return PixelFormat::ABGR8;
default:
NGLOG_CRITICAL(HW_GPU, "Unimplemented format={}", static_cast<u32>(format));
UNREACHABLE();
}
}
static PixelFormat PixelFormatFromTextureFormat(Tegra::Texture::TextureFormat format) { static PixelFormat PixelFormatFromTextureFormat(Tegra::Texture::TextureFormat format) {
// TODO(Subv): Properly implement this // TODO(Subv): Properly implement this
switch (format) { switch (format) {
@ -276,36 +222,6 @@ struct SurfaceParams {
} }
} }
static ComponentType ComponentTypeFromGPUPixelFormat(
Tegra::FramebufferConfig::PixelFormat format) {
switch (format) {
case Tegra::FramebufferConfig::PixelFormat::ABGR8:
return ComponentType::UNorm;
default:
NGLOG_CRITICAL(HW_GPU, "Unimplemented format={}", static_cast<u32>(format));
UNREACHABLE();
}
}
static bool CheckFormatsBlittable(PixelFormat pixel_format_a, PixelFormat pixel_format_b) {
SurfaceType a_type = GetFormatType(pixel_format_a);
SurfaceType b_type = GetFormatType(pixel_format_b);
if (a_type == SurfaceType::ColorTexture && b_type == SurfaceType::ColorTexture) {
return true;
}
if (a_type == SurfaceType::Depth && b_type == SurfaceType::Depth) {
return true;
}
if (a_type == SurfaceType::DepthStencil && b_type == SurfaceType::DepthStencil) {
return true;
}
return false;
}
static SurfaceType GetFormatType(PixelFormat pixel_format) { static SurfaceType GetFormatType(PixelFormat pixel_format) {
if (static_cast<size_t>(pixel_format) < MaxPixelFormat) { if (static_cast<size_t>(pixel_format) < MaxPixelFormat) {
return SurfaceType::ColorTexture; return SurfaceType::ColorTexture;
@ -317,220 +233,79 @@ struct SurfaceParams {
return SurfaceType::Invalid; return SurfaceType::Invalid;
} }
/// Update the params "size", "end" and "type" from the already set "addr", "width", "height"
/// and "pixel_format"
void UpdateParams() {
if (stride == 0) {
stride = width;
}
type = GetFormatType(pixel_format);
size = !is_tiled ? BytesInPixels(stride * (height - 1) + width)
: BytesInPixels(stride * 8 * (height / 8 - 1) + width * 8);
end = addr + size;
}
SurfaceInterval GetInterval() const {
return SurfaceInterval::right_open(addr, end);
}
// Returns the outer rectangle containing "interval"
SurfaceParams FromInterval(SurfaceInterval interval) const;
SurfaceInterval GetSubRectInterval(MathUtil::Rectangle<u32> unscaled_rect) const;
// Returns the region of the biggest valid rectange within interval
SurfaceInterval GetCopyableInterval(const Surface& src_surface) const;
/**
* Gets the actual width (in pixels) of the surface. This is provided because `width` is used
* for tracking the surface region in memory, which may be compressed for certain formats. In
* this scenario, `width` is actually the compressed width.
*/
u32 GetActualWidth() const {
return width * GetCompresssionFactor();
}
/**
* Gets the actual height (in pixels) of the surface. This is provided because `height` is used
* for tracking the surface region in memory, which may be compressed for certain formats. In
* this scenario, `height` is actually the compressed height.
*/
u32 GetActualHeight() const {
return height * GetCompresssionFactor();
}
u32 GetScaledWidth() const {
return width * res_scale;
}
u32 GetScaledHeight() const {
return height * res_scale;
}
MathUtil::Rectangle<u32> GetRect() const { MathUtil::Rectangle<u32> GetRect() const {
return {0, height, width, 0}; return {0, height, width, 0};
} }
MathUtil::Rectangle<u32> GetScaledRect() const { size_t SizeInBytes() const {
return {0, GetScaledHeight(), GetScaledWidth(), 0}; const u32 compression_factor{GetCompressionFactor()};
} ASSERT(width % compression_factor == 0);
ASSERT(height % compression_factor == 0);
u64 PixelsInBytes(u64 size) const { return (width / compression_factor) * (height / compression_factor) *
return size * CHAR_BIT / GetFormatBpp(pixel_format); GetFormatBpp(pixel_format) / CHAR_BIT;
}
u64 BytesInPixels(u64 pixels) const {
return pixels * GetFormatBpp(pixel_format) / CHAR_BIT;
} }
VAddr GetCpuAddr() const; VAddr GetCpuAddr() const;
bool ExactMatch(const SurfaceParams& other_surface) const; Tegra::GPUVAddr addr;
bool CanSubRect(const SurfaceParams& sub_surface) const; u32 width;
bool CanExpand(const SurfaceParams& expanded_surface) const; u32 height;
bool CanTexCopy(const SurfaceParams& texcopy_params) const; u32 block_height;
bool is_tiled;
MathUtil::Rectangle<u32> GetSubRect(const SurfaceParams& sub_surface) const; PixelFormat pixel_format;
MathUtil::Rectangle<u32> GetScaledSubRect(const SurfaceParams& sub_surface) const; SurfaceType type;
ComponentType component_type;
Tegra::GPUVAddr addr = 0;
Tegra::GPUVAddr end = 0;
boost::optional<VAddr> cpu_addr;
u64 size = 0;
u32 width = 0;
u32 height = 0;
u32 stride = 0;
u32 block_height = 0;
u16 res_scale = 1;
bool is_tiled = false;
PixelFormat pixel_format = PixelFormat::Invalid;
SurfaceType type = SurfaceType::Invalid;
ComponentType component_type = ComponentType::Invalid;
}; };
static_assert(std::is_pod<SurfaceParams>::value, "SurfaceParams is not POD");
struct CachedSurface : SurfaceParams { class CachedSurface final {
bool CanFill(const SurfaceParams& dest_surface, SurfaceInterval fill_interval) const; public:
bool CanCopy(const SurfaceParams& dest_surface, SurfaceInterval copy_interval) const; CachedSurface(const SurfaceParams& params);
bool IsRegionValid(SurfaceInterval interval) const { const OGLTexture& Texture() const {
return (invalid_regions.find(interval) == invalid_regions.end()); return texture;
} }
bool IsSurfaceFullyInvalid() const { static constexpr unsigned int GetGLBytesPerPixel(SurfaceParams::PixelFormat format) {
return (invalid_regions & GetInterval()) == SurfaceRegions(GetInterval()); if (format == SurfaceParams::PixelFormat::Invalid)
}
bool registered = false;
SurfaceRegions invalid_regions;
u64 fill_size = 0; /// Number of bytes to read from fill_data
std::array<u8, 4> fill_data;
OGLTexture texture;
static constexpr unsigned int GetGLBytesPerPixel(PixelFormat format) {
if (format == PixelFormat::Invalid)
return 0; return 0;
return SurfaceParams::GetFormatBpp(format) / CHAR_BIT; return SurfaceParams::GetFormatBpp(format) / CHAR_BIT;
} }
std::unique_ptr<u8[]> gl_buffer; const SurfaceParams& GetSurfaceParams() const {
size_t gl_buffer_size = 0; return params;
}
// Read/Write data in Switch memory to/from gl_buffer // Read/Write data in Switch memory to/from gl_buffer
void LoadGLBuffer(Tegra::GPUVAddr load_start, Tegra::GPUVAddr load_end); void LoadGLBuffer();
void FlushGLBuffer(Tegra::GPUVAddr flush_start, Tegra::GPUVAddr flush_end); void FlushGLBuffer();
// Upload/Download data in gl_buffer in/to this surface's texture // Upload/Download data in gl_buffer in/to this surface's texture
void UploadGLTexture(const MathUtil::Rectangle<u32>& rect, GLuint read_fb_handle, void UploadGLTexture(GLuint read_fb_handle, GLuint draw_fb_handle);
GLuint draw_fb_handle); void DownloadGLTexture(GLuint read_fb_handle, GLuint draw_fb_handle);
void DownloadGLTexture(const MathUtil::Rectangle<u32>& rect, GLuint read_fb_handle,
GLuint draw_fb_handle);
};
class RasterizerCacheOpenGL : NonCopyable {
public:
RasterizerCacheOpenGL();
~RasterizerCacheOpenGL();
/// Blit one surface's texture to another
bool BlitSurfaces(const Surface& src_surface, const MathUtil::Rectangle<u32>& src_rect,
const Surface& dst_surface, const MathUtil::Rectangle<u32>& dst_rect);
void ConvertD24S8toABGR(GLuint src_tex, const MathUtil::Rectangle<u32>& src_rect,
GLuint dst_tex, const MathUtil::Rectangle<u32>& dst_rect);
/// Copy one surface's region to another
void CopySurface(const Surface& src_surface, const Surface& dst_surface,
SurfaceInterval copy_interval);
/// Load a texture from Switch memory to OpenGL and cache it (if not already cached)
Surface GetSurface(const SurfaceParams& params, ScaleMatch match_res_scale,
bool load_if_create);
/// Tries to find a framebuffer GPU address based on the provided CPU address
boost::optional<Tegra::GPUVAddr> TryFindFramebufferGpuAddress(VAddr cpu_addr) const;
/// Attempt to find a subrect (resolution scaled) of a surface, otherwise loads a texture from
/// Switch memory to OpenGL and caches it (if not already cached)
SurfaceRect_Tuple GetSurfaceSubRect(const SurfaceParams& params, ScaleMatch match_res_scale,
bool load_if_create);
/// Get a surface based on the texture configuration
Surface GetTextureSurface(const Tegra::Texture::FullTextureInfo& config);
/// Get the color and depth surfaces based on the framebuffer configuration
SurfaceSurfaceRect_Tuple GetFramebufferSurfaces(bool using_color_fb, bool using_depth_fb,
const MathUtil::Rectangle<s32>& viewport);
/// Get a surface that matches the fill config
Surface GetFillSurface(const void* config);
/// Get a surface that matches a "texture copy" display transfer config
SurfaceRect_Tuple GetTexCopySurface(const SurfaceParams& params);
/// Write any cached resources overlapping the region back to memory (if dirty)
void FlushRegion(Tegra::GPUVAddr addr, u64 size, Surface flush_surface = nullptr);
/// Mark region as being invalidated by region_owner (nullptr if Switch memory)
void InvalidateRegion(Tegra::GPUVAddr addr, u64 size, const Surface& region_owner);
/// Flush all cached resources tracked by this cache manager
void FlushAll();
private: private:
void DuplicateSurface(const Surface& src_surface, const Surface& dest_surface); OGLTexture texture;
std::unique_ptr<u8[]> gl_buffer;
size_t gl_buffer_size;
SurfaceParams params;
};
/// Update surface's texture for given region when necessary class RasterizerCacheOpenGL final : NonCopyable {
void ValidateSurface(const Surface& surface, Tegra::GPUVAddr addr, u64 size); public:
RasterizerCacheOpenGL();
/// Create a new surface Surface GetTextureSurface(const Tegra::Texture::FullTextureInfo& config);
Surface CreateSurface(const SurfaceParams& params); SurfaceSurfaceRect_Tuple GetFramebufferSurfaces(bool using_color_fb, bool using_depth_fb,
const MathUtil::Rectangle<s32>& viewport);
void LoadSurface(const Surface& surface);
void FlushSurface(const Surface& surface);
/// Register surface into the cache private:
void RegisterSurface(const Surface& surface); Surface GetSurface(const SurfaceParams& params);
/// Remove surface from the cache
void UnregisterSurface(const Surface& surface);
/// Increase/decrease the number of surface in pages touching the specified region
void UpdatePagesCachedCount(Tegra::GPUVAddr addr, u64 size, int delta);
SurfaceCache surface_cache;
PageMap cached_pages;
SurfaceMap dirty_regions;
SurfaceSet remove_surfaces;
std::map<Tegra::GPUVAddr, Surface> surface_cache;
OGLFramebuffer read_framebuffer; OGLFramebuffer read_framebuffer;
OGLFramebuffer draw_framebuffer; OGLFramebuffer draw_framebuffer;
OGLVertexArray attributeless_vao;
OGLBuffer d24s8_abgr_buffer;
GLsizeiptr d24s8_abgr_buffer_size;
OGLProgram d24s8_abgr_shader;
GLint d24s8_abgr_tbo_size_u_id;
GLint d24s8_abgr_viewport_u_id;
}; };