1
0
Fork 0
forked from suyu/suyu

gl_rasterizer: Add rasterizer cache code to handle accerated fermi copies.

This commit is contained in:
bunnei 2018-10-05 23:39:03 -04:00
parent 749aef3dd0
commit 011cf77796
5 changed files with 60 additions and 16 deletions

View file

@ -5,6 +5,7 @@
#pragma once #pragma once
#include "common/common_types.h" #include "common/common_types.h"
#include "video_core/engines/fermi_2d.h"
#include "video_core/gpu.h" #include "video_core/gpu.h"
#include "video_core/memory_manager.h" #include "video_core/memory_manager.h"
@ -33,13 +34,9 @@ public:
/// and invalidated /// and invalidated
virtual void FlushAndInvalidateRegion(VAddr addr, u64 size) = 0; virtual void FlushAndInvalidateRegion(VAddr addr, u64 size) = 0;
/// Attempt to use a faster method to perform a display transfer with is_texture_copy = 0 /// Attempt to use a faster method to perform a surface copy
virtual bool AccelerateDisplayTransfer(const void* config) { virtual bool AccelerateSurfaceCopy(const Tegra::Engines::Fermi2D::Regs::Surface& src,
return false; const Tegra::Engines::Fermi2D::Regs::Surface& dst) {
}
/// Attempt to use a faster method to perform a display transfer with is_texture_copy = 1
virtual bool AccelerateTextureCopy(const void* config) {
return false; return false;
} }

View file

@ -616,14 +616,10 @@ void RasterizerOpenGL::FlushAndInvalidateRegion(VAddr addr, u64 size) {
InvalidateRegion(addr, size); InvalidateRegion(addr, size);
} }
bool RasterizerOpenGL::AccelerateDisplayTransfer(const void* config) { bool RasterizerOpenGL::AccelerateSurfaceCopy(const Tegra::Engines::Fermi2D::Regs::Surface& src,
const Tegra::Engines::Fermi2D::Regs::Surface& dst) {
MICROPROFILE_SCOPE(OpenGL_Blits); MICROPROFILE_SCOPE(OpenGL_Blits);
UNREACHABLE(); res_cache.FermiCopySurface(src, dst);
return true;
}
bool RasterizerOpenGL::AccelerateTextureCopy(const void* config) {
UNREACHABLE();
return true; return true;
} }

View file

@ -52,8 +52,8 @@ public:
void FlushRegion(VAddr addr, u64 size) override; void FlushRegion(VAddr addr, u64 size) override;
void InvalidateRegion(VAddr addr, u64 size) override; void InvalidateRegion(VAddr addr, u64 size) override;
void FlushAndInvalidateRegion(VAddr addr, u64 size) override; void FlushAndInvalidateRegion(VAddr addr, u64 size) override;
bool AccelerateDisplayTransfer(const void* config) override; bool AccelerateSurfaceCopy(const Tegra::Engines::Fermi2D::Regs::Surface& src,
bool AccelerateTextureCopy(const void* config) override; const Tegra::Engines::Fermi2D::Regs::Surface& dst) override;
bool AccelerateFill(const void* config) override; bool AccelerateFill(const void* config) override;
bool AccelerateDisplay(const Tegra::FramebufferConfig& config, VAddr framebuffer_addr, bool AccelerateDisplay(const Tegra::FramebufferConfig& config, VAddr framebuffer_addr,
u32 pixel_stride) override; u32 pixel_stride) override;

View file

@ -143,6 +143,28 @@ static VAddr TryGetCpuAddr(Tegra::GPUVAddr gpu_addr) {
return params; return params;
} }
/*static*/ SurfaceParams SurfaceParams::CreateForFermiCopySurface(
const Tegra::Engines::Fermi2D::Regs::Surface& config) {
SurfaceParams params{};
params.addr = TryGetCpuAddr(config.Address());
params.is_tiled = !config.linear;
params.block_height = params.is_tiled ? config.BlockHeight() : 0,
params.pixel_format = PixelFormatFromRenderTargetFormat(config.format);
params.component_type = ComponentTypeFromRenderTarget(config.format);
params.type = GetFormatType(params.pixel_format);
params.width = config.width;
params.height = config.height;
params.unaligned_height = config.height;
params.target = SurfaceTarget::Texture2D;
params.depth = 1;
params.size_in_bytes_total = params.SizeInBytesTotal();
params.size_in_bytes_2d = params.SizeInBytes2D();
params.max_mip_level = 0;
params.rt = {};
return params;
}
static constexpr std::array<FormatTuple, SurfaceParams::MaxPixelFormat> tex_format_tuples = {{ static constexpr std::array<FormatTuple, SurfaceParams::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_UNSIGNED_INT_8_8_8_8_REV, ComponentType::UNorm, false}, // ABGR8U
{GL_RGBA8, GL_RGBA, GL_BYTE, ComponentType::SNorm, false}, // ABGR8S {GL_RGBA8, GL_RGBA, GL_BYTE, ComponentType::SNorm, false}, // ABGR8S
@ -1045,6 +1067,26 @@ Surface RasterizerCacheOpenGL::GetUncachedSurface(const SurfaceParams& params) {
return surface; return surface;
} }
void RasterizerCacheOpenGL::FermiCopySurface(
const Tegra::Engines::Fermi2D::Regs::Surface& src_config,
const Tegra::Engines::Fermi2D::Regs::Surface& dst_config) {
const auto& src_params = SurfaceParams::CreateForFermiCopySurface(src_config);
const auto& dst_params = SurfaceParams::CreateForFermiCopySurface(dst_config);
ASSERT(src_params.width == dst_params.width);
ASSERT(src_params.height == dst_params.height);
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.depth == 1); // Currently, FastCopySurface only works with 2D surfaces
ASSERT(src_params.target == dst_params.target);
ASSERT(src_params.rt.index == dst_params.rt.index);
FastCopySurface(GetSurface(src_params, true), GetSurface(dst_params, false));
}
Surface RasterizerCacheOpenGL::RecreateSurface(const Surface& old_surface, Surface RasterizerCacheOpenGL::RecreateSurface(const Surface& old_surface,
const SurfaceParams& new_params) { const SurfaceParams& new_params) {
// Verify surface is compatible for blitting // Verify surface is compatible for blitting

View file

@ -13,6 +13,7 @@
#include "common/common_types.h" #include "common/common_types.h"
#include "common/hash.h" #include "common/hash.h"
#include "common/math_util.h" #include "common/math_util.h"
#include "video_core/engines/fermi_2d.h"
#include "video_core/engines/maxwell_3d.h" #include "video_core/engines/maxwell_3d.h"
#include "video_core/rasterizer_cache.h" #include "video_core/rasterizer_cache.h"
#include "video_core/renderer_opengl/gl_resource_manager.h" #include "video_core/renderer_opengl/gl_resource_manager.h"
@ -719,6 +720,10 @@ struct SurfaceParams {
Tegra::GPUVAddr zeta_address, Tegra::GPUVAddr zeta_address,
Tegra::DepthFormat format); Tegra::DepthFormat format);
/// Creates SurfaceParams for a Fermi2D surface copy
static SurfaceParams CreateForFermiCopySurface(
const Tegra::Engines::Fermi2D::Regs::Surface& config);
/// Checks if surfaces are compatible for caching /// Checks if surfaces are compatible for caching
bool IsCompatibleSurface(const SurfaceParams& other) const { bool IsCompatibleSurface(const SurfaceParams& other) const {
return std::tie(pixel_format, type, width, height, target, depth) == return std::tie(pixel_format, type, width, height, target, depth) ==
@ -837,6 +842,10 @@ public:
/// Tries to find a framebuffer using on the provided CPU address /// Tries to find a framebuffer using on the provided CPU address
Surface TryFindFramebufferSurface(VAddr addr) const; Surface TryFindFramebufferSurface(VAddr addr) 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);
private: private:
void LoadSurface(const Surface& surface); void LoadSurface(const Surface& surface);
Surface GetSurface(const SurfaceParams& params, bool preserve_contents = true); Surface GetSurface(const SurfaceParams& params, bool preserve_contents = true);