INSANEMODE/yuzu-android
1
0
Fork 0
mirror of https://github.com/yuzu-emu/yuzu-android.git synced 2025-06-11 08:57:57 -05:00
Code Issues Packages Projects Releases Wiki Activity

Use spans over guest memory where possible instead of copying data.

Browse Source
This commit is contained in:
Kelebek1
2023-05-29 00:35:51 +01:00
parent 95ceae40e6
commit 6f7cb69c94
22 changed files with 462 additions and 233 deletions
Download Patch File Download Diff File Expand all files Collapse all files

28
src/video_core/engines/engine_upload.cpp
View File

@ -5,6 +5,7 @@
#include "common/algorithm.h"
#include "common/assert.h"
#include "core/memory.h"
#include "video_core/engines/engine_upload.h"
#include "video_core/memory_manager.h"
#include "video_core/rasterizer_interface.h"
@ -46,15 +47,11 @@ void State::ProcessData(const u32* data, size_t num_data) {
void State::ProcessData(std::span<const u8> read_buffer) {
const GPUVAddr address{regs.dest.Address()};
if (is_linear) {
if (regs.line_count == 1) {
rasterizer->AccelerateInlineToMemory(address, copy_size, read_buffer);
} else {
for (size_t line = 0; line < regs.line_count; ++line) {
const GPUVAddr dest_line = address + line * regs.dest.pitch;
std::span<const u8> buffer(read_buffer.data() + line * regs.line_length_in,
regs.line_length_in);
rasterizer->AccelerateInlineToMemory(dest_line, regs.line_length_in, buffer);
}
for (size_t line = 0; line < regs.line_count; ++line) {
const GPUVAddr dest_line = address + line * regs.dest.pitch;
std::span<const u8> buffer(read_buffer.data() + line * regs.line_length_in,
regs.line_length_in);
rasterizer->AccelerateInlineToMemory(dest_line, regs.line_length_in, buffer);
}
} else {
u32 width = regs.dest.width;
@ -70,13 +67,14 @@ void State::ProcessData(std::span<const u8> read_buffer) {
const std::size_t dst_size = Tegra::Texture::CalculateSize(
true, bytes_per_pixel, width, regs.dest.height, regs.dest.depth,
regs.dest.BlockHeight(), regs.dest.BlockDepth());
tmp_buffer.resize_destructive(dst_size);
memory_manager.ReadBlock(address, tmp_buffer.data(), dst_size);
Tegra::Texture::SwizzleSubrect(tmp_buffer, read_buffer, bytes_per_pixel, width,
regs.dest.height, regs.dest.depth, x_offset, regs.dest.y,
x_elements, regs.line_count, regs.dest.BlockHeight(),
Core::Memory::GpuGuestMemoryScoped<u8, Core::Memory::GuestMemoryFlags::SafeReadCachedWrite>
tmp(memory_manager, address, dst_size, &tmp_buffer);
Tegra::Texture::SwizzleSubrect(tmp, read_buffer, bytes_per_pixel, width, regs.dest.height,
regs.dest.depth, x_offset, regs.dest.y, x_elements,
regs.line_count, regs.dest.BlockHeight(),
regs.dest.BlockDepth(), regs.line_length_in);
memory_manager.WriteBlockCached(address, tmp_buffer.data(), dst_size);
}
}

1
src/video_core/engines/kepler_compute.cpp
View File

@ -84,7 +84,6 @@ Texture::TICEntry KeplerCompute::GetTICEntry(u32 tic_index) const {
Texture::TICEntry tic_entry;
memory_manager.ReadBlockUnsafe(tic_address_gpu, &tic_entry, sizeof(Texture::TICEntry));
return tic_entry;
}

4
src/video_core/engines/maxwell_3d.cpp
View File

@ -9,6 +9,7 @@
#include "common/settings.h"
#include "core/core.h"
#include "core/core_timing.h"
#include "core/memory.h"
#include "video_core/dirty_flags.h"
#include "video_core/engines/draw_manager.h"
#include "video_core/engines/maxwell_3d.h"
@ -679,17 +680,14 @@ void Maxwell3D::ProcessCBData(u32 value) {
Texture::TICEntry Maxwell3D::GetTICEntry(u32 tic_index) const {
const GPUVAddr tic_address_gpu{regs.tex_header.Address() +
tic_index * sizeof(Texture::TICEntry)};
Texture::TICEntry tic_entry;
memory_manager.ReadBlockUnsafe(tic_address_gpu, &tic_entry, sizeof(Texture::TICEntry));
return tic_entry;
}
Texture::TSCEntry Maxwell3D::GetTSCEntry(u32 tsc_index) const {
const GPUVAddr tsc_address_gpu{regs.tex_sampler.Address() +
tsc_index * sizeof(Texture::TSCEntry)};
Texture::TSCEntry tsc_entry;
memory_manager.ReadBlockUnsafe(tsc_address_gpu, &tsc_entry, sizeof(Texture::TSCEntry));
return tsc_entry;

85
src/video_core/engines/maxwell_dma.cpp
View File

@ -7,6 +7,7 @@
#include "common/microprofile.h"
#include "common/settings.h"
#include "core/core.h"
#include "core/memory.h"
#include "video_core/engines/maxwell_3d.h"
#include "video_core/engines/maxwell_dma.h"
#include "video_core/memory_manager.h"
@ -130,11 +131,12 @@ void MaxwellDMA::Launch() {
UNIMPLEMENTED_IF(regs.offset_out % 16 != 0);
read_buffer.resize_destructive(16);
for (u32 offset = 0; offset < regs.line_length_in; offset += 16) {
memory_manager.ReadBlock(
convert_linear_2_blocklinear_addr(regs.offset_in + offset),
read_buffer.data(), read_buffer.size());
memory_manager.WriteBlockCached(regs.offset_out + offset, read_buffer.data(),
read_buffer.size());
Core::Memory::GpuGuestMemoryScoped<
u8, Core::Memory::GuestMemoryFlags::SafeReadCachedWrite>
tmp_write_buffer(memory_manager,
convert_linear_2_blocklinear_addr(regs.offset_in + offset),
16, &read_buffer);
tmp_write_buffer.SetAddressAndSize(regs.offset_out + offset, 16);
}
} else if (is_src_pitch && !is_dst_pitch) {
UNIMPLEMENTED_IF(regs.line_length_in % 16 != 0);
@ -142,20 +144,19 @@ void MaxwellDMA::Launch() {
UNIMPLEMENTED_IF(regs.offset_out % 16 != 0);
read_buffer.resize_destructive(16);
for (u32 offset = 0; offset < regs.line_length_in; offset += 16) {
memory_manager.ReadBlock(regs.offset_in + offset, read_buffer.data(),
read_buffer.size());
memory_manager.WriteBlockCached(
convert_linear_2_blocklinear_addr(regs.offset_out + offset),
read_buffer.data(), read_buffer.size());
Core::Memory::GpuGuestMemoryScoped<
u8, Core::Memory::GuestMemoryFlags::SafeReadCachedWrite>
tmp_write_buffer(memory_manager, regs.offset_in + offset, 16, &read_buffer);
tmp_write_buffer.SetAddressAndSize(
convert_linear_2_blocklinear_addr(regs.offset_out + offset), 16);
}
} else {
if (!accelerate.BufferCopy(regs.offset_in, regs.offset_out, regs.line_length_in)) {
read_buffer.resize_destructive(regs.line_length_in);
memory_manager.ReadBlock(regs.offset_in, read_buffer.data(),
regs.line_length_in,
VideoCommon::CacheType::NoBufferCache);
memory_manager.WriteBlockCached(regs.offset_out, read_buffer.data(),
regs.line_length_in);
Core::Memory::GpuGuestMemoryScoped<
u8, Core::Memory::GuestMemoryFlags::SafeReadCachedWrite>
tmp_write_buffer(memory_manager, regs.offset_in, regs.line_length_in,
&read_buffer);
tmp_write_buffer.SetAddressAndSize(regs.offset_out, regs.line_length_in);
}
}
}
@ -222,17 +223,15 @@ void MaxwellDMA::CopyBlockLinearToPitch() {
CalculateSize(true, bytes_per_pixel, width, height, depth, block_height, block_depth);
const size_t dst_size = dst_operand.pitch * regs.line_count;
read_buffer.resize_destructive(src_size);
write_buffer.resize_destructive(dst_size);
memory_manager.ReadBlock(src_operand.address, read_buffer.data(), src_size);
memory_manager.ReadBlock(dst_operand.address, write_buffer.data(), dst_size);
Core::Memory::GpuGuestMemory<u8, Core::Memory::GuestMemoryFlags::SafeRead> tmp_read_buffer(
memory_manager, src_operand.address, src_size, &read_buffer);
Core::Memory::GpuGuestMemoryScoped<u8, Core::Memory::GuestMemoryFlags::SafeReadCachedWrite>
tmp_write_buffer(memory_manager, dst_operand.address, dst_size, &write_buffer);
UnswizzleSubrect(write_buffer, read_buffer, bytes_per_pixel, width, height, depth, x_offset,
src_params.origin.y, x_elements, regs.line_count, block_height, block_depth,
dst_operand.pitch);
memory_manager.WriteBlockCached(regs.offset_out, write_buffer.data(), dst_size);
UnswizzleSubrect(tmp_write_buffer, tmp_read_buffer, bytes_per_pixel, width, height, depth,
x_offset, src_params.origin.y, x_elements, regs.line_count, block_height,
block_depth, dst_operand.pitch);
}
void MaxwellDMA::CopyPitchToBlockLinear() {
@ -287,18 +286,17 @@ void MaxwellDMA::CopyPitchToBlockLinear() {
CalculateSize(true, bytes_per_pixel, width, height, depth, block_height, block_depth);
const size_t src_size = static_cast<size_t>(regs.pitch_in) * regs.line_count;
read_buffer.resize_destructive(src_size);
write_buffer.resize_destructive(dst_size);
GPUVAddr src_addr = regs.offset_in;
GPUVAddr dst_addr = regs.offset_out;
Core::Memory::GpuGuestMemory<u8, Core::Memory::GuestMemoryFlags::SafeRead> tmp_read_buffer(
memory_manager, src_addr, src_size, &read_buffer);
Core::Memory::GpuGuestMemoryScoped<u8, Core::Memory::GuestMemoryFlags::SafeReadCachedWrite>
tmp_write_buffer(memory_manager, dst_addr, dst_size, &write_buffer);
memory_manager.ReadBlock(regs.offset_in, read_buffer.data(), src_size);
memory_manager.ReadBlockUnsafe(regs.offset_out, write_buffer.data(), dst_size);
// If the input is linear and the output is tiled, swizzle the input and copy it over.
SwizzleSubrect(write_buffer, read_buffer, bytes_per_pixel, width, height, depth, x_offset,
dst_params.origin.y, x_elements, regs.line_count, block_height, block_depth,
regs.pitch_in);
memory_manager.WriteBlockCached(regs.offset_out, write_buffer.data(), dst_size);
// If the input is linear and the output is tiled, swizzle the input and copy it over.
SwizzleSubrect(tmp_write_buffer, tmp_read_buffer, bytes_per_pixel, width, height, depth,
x_offset, dst_params.origin.y, x_elements, regs.line_count, block_height,
block_depth, regs.pitch_in);
}
void MaxwellDMA::CopyBlockLinearToBlockLinear() {
@ -342,23 +340,20 @@ void MaxwellDMA::CopyBlockLinearToBlockLinear() {
const u32 pitch = x_elements * bytes_per_pixel;
const size_t mid_buffer_size = pitch * regs.line_count;
read_buffer.resize_destructive(src_size);
write_buffer.resize_destructive(dst_size);
intermediate_buffer.resize_destructive(mid_buffer_size);
memory_manager.ReadBlock(regs.offset_in, read_buffer.data(), src_size);
memory_manager.ReadBlock(regs.offset_out, write_buffer.data(), dst_size);
Core::Memory::GpuGuestMemory<u8, Core::Memory::GuestMemoryFlags::SafeRead> tmp_read_buffer(
memory_manager, regs.offset_in, src_size, &read_buffer);
Core::Memory::GpuGuestMemoryScoped<u8, Core::Memory::GuestMemoryFlags::SafeReadCachedWrite>
tmp_write_buffer(memory_manager, regs.offset_out, dst_size, &write_buffer);
UnswizzleSubrect(intermediate_buffer, read_buffer, bytes_per_pixel, src_width, src.height,
UnswizzleSubrect(intermediate_buffer, tmp_read_buffer, bytes_per_pixel, src_width, src.height,
src.depth, src_x_offset, src.origin.y, x_elements, regs.line_count,
src.block_size.height, src.block_size.depth, pitch);
SwizzleSubrect(write_buffer, intermediate_buffer, bytes_per_pixel, dst_width, dst.height,
SwizzleSubrect(tmp_write_buffer, intermediate_buffer, bytes_per_pixel, dst_width, dst.height,
dst.depth, dst_x_offset, dst.origin.y, x_elements, regs.line_count,
dst.block_size.height, dst.block_size.depth, pitch);
memory_manager.WriteBlockCached(regs.offset_out, write_buffer.data(), dst_size);
}
void MaxwellDMA::ReleaseSemaphore() {

29
src/video_core/engines/sw_blitter/blitter.cpp
View File

@ -159,11 +159,11 @@ bool SoftwareBlitEngine::Blit(Fermi2D::Surface& src, Fermi2D::Surface& dst,
const auto src_bytes_per_pixel = BytesPerBlock(PixelFormatFromRenderTargetFormat(src.format));
const auto dst_bytes_per_pixel = BytesPerBlock(PixelFormatFromRenderTargetFormat(dst.format));
const size_t src_size = get_surface_size(src, src_bytes_per_pixel);
impl->tmp_buffer.resize_destructive(src_size);
memory_manager.ReadBlock(src.Address(), impl->tmp_buffer.data(), src_size);
Core::Memory::GpuGuestMemory<u8, Core::Memory::GuestMemoryFlags::SafeRead> tmp_buffer(
memory_manager, src.Address(), src_size, &impl->tmp_buffer);
const size_t src_copy_size = src_extent_x * src_extent_y * src_bytes_per_pixel;
const size_t dst_copy_size = dst_extent_x * dst_extent_y * dst_bytes_per_pixel;
impl->src_buffer.resize_destructive(src_copy_size);
@ -200,12 +200,11 @@ bool SoftwareBlitEngine::Blit(Fermi2D::Surface& src, Fermi2D::Surface& dst,
impl->dst_buffer.resize_destructive(dst_copy_size);
if (src.linear == Fermi2D::MemoryLayout::BlockLinear) {
UnswizzleSubrect(impl->src_buffer, impl->tmp_buffer, src_bytes_per_pixel, src.width,
src.height, src.depth, config.src_x0, config.src_y0, src_extent_x,
src_extent_y, src.block_height, src.block_depth,
src_extent_x * src_bytes_per_pixel);
UnswizzleSubrect(impl->src_buffer, tmp_buffer, src_bytes_per_pixel, src.width, src.height,
src.depth, config.src_x0, config.src_y0, src_extent_x, src_extent_y,
src.block_height, src.block_depth, src_extent_x * src_bytes_per_pixel);
} else {
process_pitch_linear(false, impl->tmp_buffer, impl->src_buffer, src_extent_x, src_extent_y,
process_pitch_linear(false, tmp_buffer, impl->src_buffer, src_extent_x, src_extent_y,
src.pitch, config.src_x0, config.src_y0, src_bytes_per_pixel);
}
@ -221,20 +220,18 @@ bool SoftwareBlitEngine::Blit(Fermi2D::Surface& src, Fermi2D::Surface& dst,
}
const size_t dst_size = get_surface_size(dst, dst_bytes_per_pixel);
impl->tmp_buffer.resize_destructive(dst_size);
memory_manager.ReadBlock(dst.Address(), impl->tmp_buffer.data(), dst_size);
Core::Memory::GpuGuestMemoryScoped<u8, Core::Memory::GuestMemoryFlags::SafeReadWrite>
tmp_buffer2(memory_manager, dst.Address(), dst_size, &impl->tmp_buffer);
if (dst.linear == Fermi2D::MemoryLayout::BlockLinear) {
SwizzleSubrect(impl->tmp_buffer, impl->dst_buffer, dst_bytes_per_pixel, dst.width,
dst.height, dst.depth, config.dst_x0, config.dst_y0, dst_extent_x,
dst_extent_y, dst.block_height, dst.block_depth,
dst_extent_x * dst_bytes_per_pixel);
SwizzleSubrect(tmp_buffer2, impl->dst_buffer, dst_bytes_per_pixel, dst.width, dst.height,
dst.depth, config.dst_x0, config.dst_y0, dst_extent_x, dst_extent_y,
dst.block_height, dst.block_depth, dst_extent_x * dst_bytes_per_pixel);
} else {
process_pitch_linear(true, impl->dst_buffer, impl->tmp_buffer, dst_extent_x, dst_extent_y,
process_pitch_linear(true, impl->dst_buffer, tmp_buffer2, dst_extent_x, dst_extent_y,
dst.pitch, config.dst_x0, config.dst_y0,
static_cast<size_t>(dst_bytes_per_pixel));
}
memory_manager.WriteBlock(dst.Address(), impl->tmp_buffer.data(), dst_size);
return true;
}