LunaX64/suyu
1
0
Fork 0
forked from suyu/suyu
Code Pull requests Activity

fermi_2d: Implement simple copies with AccelerateSurfaceCopy.

Browse source
This commit is contained in:
bunnei 2018-10-05 23:46:40 -04:00
parent 011cf77796
commit 9aec85d39c
3 changed files with 35 additions and 23 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

48
src/video_core/engines/fermi_2d.cpp
View file

@ -4,11 +4,13 @@
#include "core/memory.h" #include "core/memory.h"
#include "video_core/engines/fermi_2d.h" #include "video_core/engines/fermi_2d.h"
#include "video_core/rasterizer_interface.h"
#include "video_core/textures/decoders.h" #include "video_core/textures/decoders.h"
namespace Tegra::Engines { namespace Tegra::Engines {
Fermi2D::Fermi2D(MemoryManager& memory_manager) : memory_manager(memory_manager) {} Fermi2D::Fermi2D(VideoCore::RasterizerInterface& rasterizer, MemoryManager& memory_manager)
: memory_manager(memory_manager), rasterizer{rasterizer} {}
void Fermi2D::WriteReg(u32 method, u32 value) { void Fermi2D::WriteReg(u32 method, u32 value) {
ASSERT_MSG(method < Regs::NUM_REGS, ASSERT_MSG(method < Regs::NUM_REGS,
@ -44,27 +46,31 @@ void Fermi2D::HandleSurfaceCopy() {
u32 src_bytes_per_pixel = RenderTargetBytesPerPixel(regs.src.format); u32 src_bytes_per_pixel = RenderTargetBytesPerPixel(regs.src.format);
u32 dst_bytes_per_pixel = RenderTargetBytesPerPixel(regs.dst.format); u32 dst_bytes_per_pixel = RenderTargetBytesPerPixel(regs.dst.format);
if (regs.src.linear == regs.dst.linear) { if (!rasterizer.AccelerateSurfaceCopy(regs.src, regs.dst)) {
// If the input layout and the output layout are the same, just perform a raw copy. // TODO(bunnei): The below implementation currently will not get hit, as
ASSERT(regs.src.BlockHeight() == regs.dst.BlockHeight()); // AccelerateSurfaceCopy tries to always copy and will always return success. This should be
Memory::CopyBlock(dest_cpu, source_cpu, // changed once we properly support flushing.
src_bytes_per_pixel * regs.dst.width * regs.dst.height);
return;
}
u8* src_buffer = Memory::GetPointer(source_cpu); if (regs.src.linear == regs.dst.linear) {
u8* dst_buffer = Memory::GetPointer(dest_cpu); // If the input layout and the output layout are the same, just perform a raw copy.
ASSERT(regs.src.BlockHeight() == regs.dst.BlockHeight());
if (!regs.src.linear && regs.dst.linear) { Memory::CopyBlock(dest_cpu, source_cpu,
// If the input is tiled and the output is linear, deswizzle the input and copy it over. src_bytes_per_pixel * regs.dst.width * regs.dst.height);
Texture::CopySwizzledData(regs.src.width, regs.src.height, src_bytes_per_pixel, return;
dst_bytes_per_pixel, src_buffer, dst_buffer, true, }
regs.src.BlockHeight()); u8* src_buffer = Memory::GetPointer(source_cpu);
} else { u8* dst_buffer = Memory::GetPointer(dest_cpu);
// If the input is linear and the output is tiled, swizzle the input and copy it over. if (!regs.src.linear && regs.dst.linear) {
Texture::CopySwizzledData(regs.src.width, regs.src.height, src_bytes_per_pixel, // If the input is tiled and the output is linear, deswizzle the input and copy it over.
dst_bytes_per_pixel, dst_buffer, src_buffer, false, Texture::CopySwizzledData(regs.src.width, regs.src.height, src_bytes_per_pixel,
regs.dst.BlockHeight()); dst_bytes_per_pixel, src_buffer, dst_buffer, true,
regs.src.BlockHeight());
} else {
// If the input is linear and the output is tiled, swizzle the input and copy it over.
Texture::CopySwizzledData(regs.src.width, regs.src.height, src_bytes_per_pixel,
dst_bytes_per_pixel, dst_buffer, src_buffer, false,
regs.dst.BlockHeight());
}
} }
} }

8
src/video_core/engines/fermi_2d.h
View file

@ -12,6 +12,10 @@
#include "video_core/gpu.h" #include "video_core/gpu.h"
#include "video_core/memory_manager.h" #include "video_core/memory_manager.h"
namespace VideoCore {
class RasterizerInterface;
}
namespace Tegra::Engines { namespace Tegra::Engines {
#define FERMI2D_REG_INDEX(field_name) \ #define FERMI2D_REG_INDEX(field_name) \
@ -19,7 +23,7 @@ namespace Tegra::Engines {
class Fermi2D final { class Fermi2D final {
public: public:
explicit Fermi2D(MemoryManager& memory_manager); explicit Fermi2D(VideoCore::RasterizerInterface& rasterizer, MemoryManager& memory_manager);
~Fermi2D() = default; ~Fermi2D() = default;
/// Write the value to the register identified by method. /// Write the value to the register identified by method.
@ -94,6 +98,8 @@ public:
MemoryManager& memory_manager; MemoryManager& memory_manager;
private: private:
VideoCore::RasterizerInterface& rasterizer;
/// Performs the copy from the source surface to the destination surface as configured in the /// Performs the copy from the source surface to the destination surface as configured in the
/// registers. /// registers.
void HandleSurfaceCopy(); void HandleSurfaceCopy();

2
src/video_core/gpu.cpp
View file

@ -25,7 +25,7 @@ u32 FramebufferConfig::BytesPerPixel(PixelFormat format) {
GPU::GPU(VideoCore::RasterizerInterface& rasterizer) { GPU::GPU(VideoCore::RasterizerInterface& rasterizer) {
memory_manager = std::make_unique<Tegra::MemoryManager>(); memory_manager = std::make_unique<Tegra::MemoryManager>();
maxwell_3d = std::make_unique<Engines::Maxwell3D>(rasterizer, *memory_manager); maxwell_3d = std::make_unique<Engines::Maxwell3D>(rasterizer, *memory_manager);
fermi_2d = std::make_unique<Engines::Fermi2D>(*memory_manager); fermi_2d = std::make_unique<Engines::Fermi2D>(rasterizer, *memory_manager);
maxwell_compute = std::make_unique<Engines::MaxwellCompute>(); maxwell_compute = std::make_unique<Engines::MaxwellCompute>();
maxwell_dma = std::make_unique<Engines::MaxwellDMA>(*memory_manager); maxwell_dma = std::make_unique<Engines::MaxwellDMA>(*memory_manager);
kepler_memory = std::make_unique<Engines::KeplerMemory>(*memory_manager); kepler_memory = std::make_unique<Engines::KeplerMemory>(*memory_manager);