mirror of
https://git.suyu.dev/suyu/suyu.git
synced 2024-11-04 14:02:45 +01:00
Merge pull request #10132 from Kelebek1/fermi_blit2
Allow Fermi blit accelerate to work without images in cache
This commit is contained in:
commit
182221b9ff
3 changed files with 24 additions and 12 deletions
|
@ -23,7 +23,10 @@ public:
|
||||||
buffer{Common::make_unique_for_overwrite<T[]>(initial_capacity)} {}
|
buffer{Common::make_unique_for_overwrite<T[]>(initial_capacity)} {}
|
||||||
|
|
||||||
~ScratchBuffer() = default;
|
~ScratchBuffer() = default;
|
||||||
|
ScratchBuffer(const ScratchBuffer&) = delete;
|
||||||
|
ScratchBuffer& operator=(const ScratchBuffer&) = delete;
|
||||||
ScratchBuffer(ScratchBuffer&&) = default;
|
ScratchBuffer(ScratchBuffer&&) = default;
|
||||||
|
ScratchBuffer& operator=(ScratchBuffer&&) = default;
|
||||||
|
|
||||||
/// This will only grow the buffer's capacity if size is greater than the current capacity.
|
/// This will only grow the buffer's capacity if size is greater than the current capacity.
|
||||||
/// The previously held data will remain intact.
|
/// The previously held data will remain intact.
|
||||||
|
@ -87,6 +90,12 @@ public:
|
||||||
return buffer_capacity;
|
return buffer_capacity;
|
||||||
}
|
}
|
||||||
|
|
||||||
|
void swap(ScratchBuffer& other) noexcept {
|
||||||
|
std::swap(last_requested_size, other.last_requested_size);
|
||||||
|
std::swap(buffer_capacity, other.buffer_capacity);
|
||||||
|
std::swap(buffer, other.buffer);
|
||||||
|
}
|
||||||
|
|
||||||
private:
|
private:
|
||||||
size_t last_requested_size{};
|
size_t last_requested_size{};
|
||||||
size_t buffer_capacity{};
|
size_t buffer_capacity{};
|
||||||
|
|
|
@ -5,6 +5,7 @@
|
||||||
#include <cmath>
|
#include <cmath>
|
||||||
#include <vector>
|
#include <vector>
|
||||||
|
|
||||||
|
#include "common/scratch_buffer.h"
|
||||||
#include "video_core/engines/sw_blitter/blitter.h"
|
#include "video_core/engines/sw_blitter/blitter.h"
|
||||||
#include "video_core/engines/sw_blitter/converter.h"
|
#include "video_core/engines/sw_blitter/converter.h"
|
||||||
#include "video_core/memory_manager.h"
|
#include "video_core/memory_manager.h"
|
||||||
|
@ -112,11 +113,11 @@ void Bilinear(std::span<const f32> input, std::span<f32> output, size_t src_widt
|
||||||
} // namespace
|
} // namespace
|
||||||
|
|
||||||
struct SoftwareBlitEngine::BlitEngineImpl {
|
struct SoftwareBlitEngine::BlitEngineImpl {
|
||||||
std::vector<u8> tmp_buffer;
|
Common::ScratchBuffer<u8> tmp_buffer;
|
||||||
std::vector<u8> src_buffer;
|
Common::ScratchBuffer<u8> src_buffer;
|
||||||
std::vector<u8> dst_buffer;
|
Common::ScratchBuffer<u8> dst_buffer;
|
||||||
std::vector<f32> intermediate_src;
|
Common::ScratchBuffer<f32> intermediate_src;
|
||||||
std::vector<f32> intermediate_dst;
|
Common::ScratchBuffer<f32> intermediate_dst;
|
||||||
ConverterFactory converter_factory;
|
ConverterFactory converter_factory;
|
||||||
};
|
};
|
||||||
|
|
||||||
|
@ -158,14 +159,14 @@ bool SoftwareBlitEngine::Blit(Fermi2D::Surface& src, Fermi2D::Surface& dst,
|
||||||
const auto src_bytes_per_pixel = BytesPerBlock(PixelFormatFromRenderTargetFormat(src.format));
|
const auto src_bytes_per_pixel = BytesPerBlock(PixelFormatFromRenderTargetFormat(src.format));
|
||||||
const auto dst_bytes_per_pixel = BytesPerBlock(PixelFormatFromRenderTargetFormat(dst.format));
|
const auto dst_bytes_per_pixel = BytesPerBlock(PixelFormatFromRenderTargetFormat(dst.format));
|
||||||
const size_t src_size = get_surface_size(src, src_bytes_per_pixel);
|
const size_t src_size = get_surface_size(src, src_bytes_per_pixel);
|
||||||
impl->tmp_buffer.resize(src_size);
|
impl->tmp_buffer.resize_destructive(src_size);
|
||||||
memory_manager.ReadBlock(src.Address(), impl->tmp_buffer.data(), src_size);
|
memory_manager.ReadBlock(src.Address(), impl->tmp_buffer.data(), src_size);
|
||||||
|
|
||||||
const size_t src_copy_size = src_extent_x * src_extent_y * src_bytes_per_pixel;
|
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;
|
const size_t dst_copy_size = dst_extent_x * dst_extent_y * dst_bytes_per_pixel;
|
||||||
|
|
||||||
impl->src_buffer.resize(src_copy_size);
|
impl->src_buffer.resize_destructive(src_copy_size);
|
||||||
|
|
||||||
const bool no_passthrough =
|
const bool no_passthrough =
|
||||||
src.format != dst.format || src_extent_x != dst_extent_x || src_extent_y != dst_extent_y;
|
src.format != dst.format || src_extent_x != dst_extent_x || src_extent_y != dst_extent_y;
|
||||||
|
@ -177,8 +178,10 @@ bool SoftwareBlitEngine::Blit(Fermi2D::Surface& src, Fermi2D::Surface& dst,
|
||||||
|
|
||||||
const auto convertion_phase_ir = [&]() {
|
const auto convertion_phase_ir = [&]() {
|
||||||
auto* input_converter = impl->converter_factory.GetFormatConverter(src.format);
|
auto* input_converter = impl->converter_factory.GetFormatConverter(src.format);
|
||||||
impl->intermediate_src.resize((src_copy_size / src_bytes_per_pixel) * ir_components);
|
impl->intermediate_src.resize_destructive((src_copy_size / src_bytes_per_pixel) *
|
||||||
impl->intermediate_dst.resize((dst_copy_size / dst_bytes_per_pixel) * ir_components);
|
ir_components);
|
||||||
|
impl->intermediate_dst.resize_destructive((dst_copy_size / dst_bytes_per_pixel) *
|
||||||
|
ir_components);
|
||||||
input_converter->ConvertTo(impl->src_buffer, impl->intermediate_src);
|
input_converter->ConvertTo(impl->src_buffer, impl->intermediate_src);
|
||||||
|
|
||||||
if (config.filter != Fermi2D::Filter::Bilinear) {
|
if (config.filter != Fermi2D::Filter::Bilinear) {
|
||||||
|
@ -195,7 +198,7 @@ bool SoftwareBlitEngine::Blit(Fermi2D::Surface& src, Fermi2D::Surface& dst,
|
||||||
|
|
||||||
// Do actual Blit
|
// Do actual Blit
|
||||||
|
|
||||||
impl->dst_buffer.resize(dst_copy_size);
|
impl->dst_buffer.resize_destructive(dst_copy_size);
|
||||||
if (src.linear == Fermi2D::MemoryLayout::BlockLinear) {
|
if (src.linear == Fermi2D::MemoryLayout::BlockLinear) {
|
||||||
UnswizzleSubrect(impl->src_buffer, impl->tmp_buffer, src_bytes_per_pixel, src.width,
|
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.height, src.depth, config.src_x0, config.src_y0, src_extent_x,
|
||||||
|
@ -218,7 +221,7 @@ bool SoftwareBlitEngine::Blit(Fermi2D::Surface& src, Fermi2D::Surface& dst,
|
||||||
}
|
}
|
||||||
|
|
||||||
const size_t dst_size = get_surface_size(dst, dst_bytes_per_pixel);
|
const size_t dst_size = get_surface_size(dst, dst_bytes_per_pixel);
|
||||||
impl->tmp_buffer.resize(dst_size);
|
impl->tmp_buffer.resize_destructive(dst_size);
|
||||||
memory_manager.ReadBlock(dst.Address(), impl->tmp_buffer.data(), dst_size);
|
memory_manager.ReadBlock(dst.Address(), impl->tmp_buffer.data(), dst_size);
|
||||||
|
|
||||||
if (dst.linear == Fermi2D::MemoryLayout::BlockLinear) {
|
if (dst.linear == Fermi2D::MemoryLayout::BlockLinear) {
|
||||||
|
|
|
@ -1469,7 +1469,7 @@ std::optional<typename TextureCache<P>::BlitImages> TextureCache<P>::GetBlitImag
|
||||||
if (!copy.must_accelerate) {
|
if (!copy.must_accelerate) {
|
||||||
do {
|
do {
|
||||||
if (!src_id && !dst_id) {
|
if (!src_id && !dst_id) {
|
||||||
return std::nullopt;
|
break;
|
||||||
}
|
}
|
||||||
if (src_id && True(slot_images[src_id].flags & ImageFlagBits::GpuModified)) {
|
if (src_id && True(slot_images[src_id].flags & ImageFlagBits::GpuModified)) {
|
||||||
break;
|
break;
|
||||||
|
|
Loading…
Reference in a new issue