hdkv/yuzu
1
Fork 0
Code Issues Pull Requests Releases 1 Activity

early-access version 3905

Browse Source
This commit is contained in:
pineappleEA
2023-09-29 15:30:08 +02:00
parent 1a7e37e1d5
commit 02452680a7
52 changed files with 1626 additions and 518 deletions
Download Patch File Download Diff File Expand all files Collapse all files

1
src/video_core/CMakeLists.txt
View File

@@ -15,6 +15,7 @@ add_library(video_core STATIC
buffer_cache/buffer_cache.cpp
buffer_cache/buffer_cache.h
buffer_cache/memory_tracker_base.h
buffer_cache/usage_tracker.h
buffer_cache/word_manager.h
cache_types.h
cdma_pusher.cpp

97
src/video_core/buffer_cache/buffer_cache.h
View File

@@ -67,6 +67,7 @@ void BufferCache<P>::TickFrame() {
if (!channel_state) {
return;
}
runtime.TickFrame(slot_buffers);
// Calculate hits and shots and move hit bits to the right
const u32 hits = std::reduce(channel_state->uniform_cache_hits.begin(),
@@ -230,7 +231,10 @@ bool BufferCache<P>::DMACopy(GPUVAddr src_address, GPUVAddr dest_address, u64 am
for (const IntervalType& add_interval : tmp_intervals) {
common_ranges.add(add_interval);
}
runtime.CopyBuffer(dest_buffer, src_buffer, copies);
const auto& copy = copies[0];
src_buffer.MarkUsage(copy.src_offset, copy.size);
dest_buffer.MarkUsage(copy.dst_offset, copy.size);
runtime.CopyBuffer(dest_buffer, src_buffer, copies, true);
if (has_new_downloads) {
memory_tracker.MarkRegionAsGpuModified(*cpu_dest_address, amount);
}
@@ -258,9 +262,10 @@ bool BufferCache<P>::DMAClear(GPUVAddr dst_address, u64 amount, u32 value) {
common_ranges.subtract(subtract_interval);
const BufferId buffer = FindBuffer(*cpu_dst_address, static_cast<u32>(size));
auto& dest_buffer = slot_buffers[buffer];
Buffer& dest_buffer = slot_buffers[buffer];
const u32 offset = dest_buffer.Offset(*cpu_dst_address);
runtime.ClearBuffer(dest_buffer, offset, size, value);
dest_buffer.MarkUsage(offset, size);
return true;
}
@@ -603,6 +608,7 @@ void BufferCache<P>::CommitAsyncFlushesHigh() {
VAddr orig_cpu_addr = static_cast<VAddr>(second_copy.src_offset);
const IntervalType base_interval{orig_cpu_addr, orig_cpu_addr + copy.size};
async_downloads += std::make_pair(base_interval, 1);
buffer.MarkUsage(copy.src_offset, copy.size);
runtime.CopyBuffer(download_staging.buffer, buffer, copies, false);
normalized_copies.push_back(second_copy);
}
@@ -621,8 +627,9 @@ void BufferCache<P>::CommitAsyncFlushesHigh() {
// Have in mind the staging buffer offset for the copy
copy.dst_offset += download_staging.offset;
const std::array copies{copy};
runtime.CopyBuffer(download_staging.buffer, slot_buffers[buffer_id], copies,
false);
Buffer& buffer = slot_buffers[buffer_id];
buffer.MarkUsage(copy.src_offset, copy.size);
runtime.CopyBuffer(download_staging.buffer, buffer, copies, false);
}
runtime.PostCopyBarrier();
runtime.Finish();
@@ -742,7 +749,7 @@ void BufferCache<P>::BindHostIndexBuffer() {
{BufferCopy{.src_offset = upload_staging.offset, .dst_offset = 0, .size = size}}};
std::memcpy(upload_staging.mapped_span.data(),
draw_state.inline_index_draw_indexes.data(), size);
runtime.CopyBuffer(buffer, upload_staging.buffer, copies);
runtime.CopyBuffer(buffer, upload_staging.buffer, copies, true);
} else {
buffer.ImmediateUpload(0, draw_state.inline_index_draw_indexes);
}
@@ -754,6 +761,7 @@ void BufferCache<P>::BindHostIndexBuffer() {
offset + draw_state.index_buffer.first * draw_state.index_buffer.FormatSizeInBytes();
runtime.BindIndexBuffer(buffer, new_offset, size);
} else {
buffer.MarkUsage(offset, size);
runtime.BindIndexBuffer(draw_state.topology, draw_state.index_buffer.format,
draw_state.index_buffer.first, draw_state.index_buffer.count,
buffer, offset, size);
@@ -790,6 +798,7 @@ void BufferCache<P>::BindHostVertexBuffers() {
const u32 stride = maxwell3d->regs.vertex_streams[index].stride;
const u32 offset = buffer.Offset(binding.cpu_addr);
buffer.MarkUsage(offset, binding.size);
host_bindings.buffers.push_back(&buffer);
host_bindings.offsets.push_back(offset);
@@ -895,6 +904,7 @@ void BufferCache<P>::BindHostGraphicsUniformBuffer(size_t stage, u32 index, u32
if constexpr (HAS_PERSISTENT_UNIFORM_BUFFER_BINDINGS) {
channel_state->uniform_buffer_binding_sizes[stage][binding_index] = size;
}
buffer.MarkUsage(offset, size);
if constexpr (NEEDS_BIND_UNIFORM_INDEX) {
runtime.BindUniformBuffer(stage, binding_index, buffer, offset, size);
} else {
@@ -913,6 +923,7 @@ void BufferCache<P>::BindHostGraphicsStorageBuffers(size_t stage) {
SynchronizeBuffer(buffer, binding.cpu_addr, size);
const u32 offset = buffer.Offset(binding.cpu_addr);
buffer.MarkUsage(offset, size);
const bool is_written = ((channel_state->written_storage_buffers[stage] >> index) & 1) != 0;
if constexpr (NEEDS_BIND_STORAGE_INDEX) {
runtime.BindStorageBuffer(stage, binding_index, buffer, offset, size, is_written);
@@ -933,6 +944,7 @@ void BufferCache<P>::BindHostGraphicsTextureBuffers(size_t stage) {
const u32 offset = buffer.Offset(binding.cpu_addr);
const PixelFormat format = binding.format;
buffer.MarkUsage(offset, size);
if constexpr (SEPARATE_IMAGE_BUFFERS_BINDINGS) {
if (((channel_state->image_texture_buffers[stage] >> index) & 1) != 0) {
runtime.BindImageBuffer(buffer, offset, size, format);
@@ -963,9 +975,10 @@ void BufferCache<P>::BindHostTransformFeedbackBuffers() {
SynchronizeBuffer(buffer, binding.cpu_addr, size);
const u32 offset = buffer.Offset(binding.cpu_addr);
buffer.MarkUsage(offset, size);
host_bindings.buffers.push_back(&buffer);
host_bindings.offsets.push_back(offset);
host_bindings.sizes.push_back(binding.size);
host_bindings.sizes.push_back(size);
}
if (host_bindings.buffers.size() > 0) {
runtime.BindTransformFeedbackBuffers(host_bindings);
@@ -989,6 +1002,7 @@ void BufferCache<P>::BindHostComputeUniformBuffers() {
SynchronizeBuffer(buffer, binding.cpu_addr, size);
const u32 offset = buffer.Offset(binding.cpu_addr);
buffer.MarkUsage(offset, size);
if constexpr (NEEDS_BIND_UNIFORM_INDEX) {
runtime.BindComputeUniformBuffer(binding_index, buffer, offset, size);
++binding_index;
@@ -1009,6 +1023,7 @@ void BufferCache<P>::BindHostComputeStorageBuffers() {
SynchronizeBuffer(buffer, binding.cpu_addr, size);
const u32 offset = buffer.Offset(binding.cpu_addr);
buffer.MarkUsage(offset, size);
const bool is_written =
((channel_state->written_compute_storage_buffers >> index) & 1) != 0;
if constexpr (NEEDS_BIND_STORAGE_INDEX) {
@@ -1030,6 +1045,7 @@ void BufferCache<P>::BindHostComputeTextureBuffers() {
const u32 offset = buffer.Offset(binding.cpu_addr);
const PixelFormat format = binding.format;
buffer.MarkUsage(offset, size);
if constexpr (SEPARATE_IMAGE_BUFFERS_BINDINGS) {
if (((channel_state->image_compute_texture_buffers >> index) & 1) != 0) {
runtime.BindImageBuffer(buffer, offset, size, format);
@@ -1148,10 +1164,11 @@ void BufferCache<P>::UpdateVertexBuffer(u32 index) {
if (!gpu_memory->IsWithinGPUAddressRange(gpu_addr_end)) {
size = static_cast<u32>(gpu_memory->MaxContinuousRange(gpu_addr_begin, size));
}
const BufferId buffer_id = FindBuffer(*cpu_addr, size);
channel_state->vertex_buffers[index] = Binding{
.cpu_addr = *cpu_addr,
.size = size,
.buffer_id = FindBuffer(*cpu_addr, size),
.buffer_id = buffer_id,
};
}
@@ -1403,7 +1420,8 @@ void BufferCache<P>::JoinOverlap(BufferId new_buffer_id, BufferId overlap_id,
.dst_offset = dst_base_offset,
.size = overlap.SizeBytes(),
});
runtime.CopyBuffer(new_buffer, overlap, copies);
new_buffer.MarkUsage(copies[0].dst_offset, copies[0].size);
runtime.CopyBuffer(new_buffer, overlap, copies, true);
DeleteBuffer(overlap_id, true);
}
@@ -1416,7 +1434,9 @@ BufferId BufferCache<P>::CreateBuffer(VAddr cpu_addr, u32 wanted_size) {
const u32 size = static_cast<u32>(overlap.end - overlap.begin);
const BufferId new_buffer_id = slot_buffers.insert(runtime, rasterizer, overlap.begin, size);
auto& new_buffer = slot_buffers[new_buffer_id];
runtime.ClearBuffer(new_buffer, 0, new_buffer.SizeBytes(), 0);
const size_t size_bytes = new_buffer.SizeBytes();
runtime.ClearBuffer(new_buffer, 0, size_bytes, 0);
new_buffer.MarkUsage(0, size_bytes);
for (const BufferId overlap_id : overlap.ids) {
JoinOverlap(new_buffer_id, overlap_id, !overlap.has_stream_leap);
}
@@ -1469,11 +1489,6 @@ void BufferCache<P>::TouchBuffer(Buffer& buffer, BufferId buffer_id) noexcept {
template <class P>
bool BufferCache<P>::SynchronizeBuffer(Buffer& buffer, VAddr cpu_addr, u32 size) {
return SynchronizeBufferImpl(buffer, cpu_addr, size);
}
template <class P>
bool BufferCache<P>::SynchronizeBufferImpl(Buffer& buffer, VAddr cpu_addr, u32 size) {
boost::container::small_vector<BufferCopy, 4> copies;
u64 total_size_bytes = 0;
u64 largest_copy = 0;
@@ -1495,51 +1510,6 @@ bool BufferCache<P>::SynchronizeBufferImpl(Buffer& buffer, VAddr cpu_addr, u32 s
return false;
}
template <class P>
bool BufferCache<P>::SynchronizeBufferNoModified(Buffer& buffer, VAddr cpu_addr, u32 size) {
boost::container::small_vector<BufferCopy, 4> copies;
u64 total_size_bytes = 0;
u64 largest_copy = 0;
IntervalSet found_sets{};
auto make_copies = [&] {
for (auto& interval : found_sets) {
const std::size_t sub_size = interval.upper() - interval.lower();
const VAddr cpu_addr_ = interval.lower();
copies.push_back(BufferCopy{
.src_offset = total_size_bytes,
.dst_offset = cpu_addr_ - buffer.CpuAddr(),
.size = sub_size,
});
total_size_bytes += sub_size;
largest_copy = std::max<u64>(largest_copy, sub_size);
}
const std::span<BufferCopy> copies_span(copies.data(), copies.size());
UploadMemory(buffer, total_size_bytes, largest_copy, copies_span);
};
memory_tracker.ForEachUploadRange(cpu_addr, size, [&](u64 cpu_addr_out, u64 range_size) {
const VAddr base_adr = cpu_addr_out;
const VAddr end_adr = base_adr + range_size;
const IntervalType add_interval{base_adr, end_adr};
found_sets.add(add_interval);
});
if (found_sets.empty()) {
return true;
}
const IntervalType search_interval{cpu_addr, cpu_addr + size};
auto it = common_ranges.lower_bound(search_interval);
auto it_end = common_ranges.upper_bound(search_interval);
if (it == common_ranges.end()) {
make_copies();
return false;
}
while (it != it_end) {
found_sets.subtract(*it);
it++;
}
make_copies();
return false;
}
template <class P>
void BufferCache<P>::UploadMemory(Buffer& buffer, u64 total_size_bytes, u64 largest_copy,
std::span<BufferCopy> copies) {
@@ -1588,7 +1558,8 @@ void BufferCache<P>::MappedUploadMemory([[maybe_unused]] Buffer& buffer,
// Apply the staging offset
copy.src_offset += upload_staging.offset;
}
runtime.CopyBuffer(buffer, upload_staging.buffer, copies);
const bool can_reorder = runtime.CanReorderUpload(buffer, copies);
runtime.CopyBuffer(buffer, upload_staging.buffer, copies, true, can_reorder);
}
}
@@ -1630,7 +1601,8 @@ void BufferCache<P>::InlineMemoryImplementation(VAddr dest_address, size_t copy_
}};
u8* const src_pointer = upload_staging.mapped_span.data();
std::memcpy(src_pointer, inlined_buffer.data(), copy_size);
runtime.CopyBuffer(buffer, upload_staging.buffer, copies);
const bool can_reorder = runtime.CanReorderUpload(buffer, copies);
runtime.CopyBuffer(buffer, upload_staging.buffer, copies, true, can_reorder);
} else {
buffer.ImmediateUpload(buffer.Offset(dest_address), inlined_buffer.first(copy_size));
}
@@ -1683,8 +1655,9 @@ void BufferCache<P>::DownloadBufferMemory(Buffer& buffer, VAddr cpu_addr, u64 si
for (BufferCopy& copy : copies) {
// Modify copies to have the staging offset in mind
copy.dst_offset += download_staging.offset;
buffer.MarkUsage(copy.src_offset, copy.size);
}
runtime.CopyBuffer(download_staging.buffer, buffer, copies_span);
runtime.CopyBuffer(download_staging.buffer, buffer, copies_span, true);
runtime.Finish();
for (const BufferCopy& copy : copies) {
const VAddr copy_cpu_addr = buffer.CpuAddr() + copy.src_offset;

4
src/video_core/buffer_cache/buffer_cache_base.h
View File

@@ -525,10 +525,6 @@ private:
bool SynchronizeBuffer(Buffer& buffer, VAddr cpu_addr, u32 size);
bool SynchronizeBufferImpl(Buffer& buffer, VAddr cpu_addr, u32 size);
bool SynchronizeBufferNoModified(Buffer& buffer, VAddr cpu_addr, u32 size);
void UploadMemory(Buffer& buffer, u64 total_size_bytes, u64 largest_copy,
std::span<BufferCopy> copies);

79
src/video_core/buffer_cache/usage_tracker.h Executable file
View File

@@ -0,0 +1,79 @@
// SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project
// SPDX-License-Identifier: GPL-3.0-or-later
#pragma once
#include "common/alignment.h"
#include "common/common_types.h"
namespace VideoCommon {
class UsageTracker {
static constexpr size_t BYTES_PER_BIT_SHIFT = 6;
static constexpr size_t PAGE_SHIFT = 6 + BYTES_PER_BIT_SHIFT;
static constexpr size_t PAGE_BYTES = 1 << PAGE_SHIFT;
public:
explicit UsageTracker(size_t size) {
const size_t num_pages = (size >> PAGE_SHIFT) + 1;
pages.resize(num_pages, 0ULL);
}
void Reset() noexcept {
std::ranges::fill(pages, 0ULL);
}
void Track(u64 offset, u64 size) noexcept {
const size_t page = offset >> PAGE_SHIFT;
const size_t page_end = (offset + size) >> PAGE_SHIFT;
TrackPage(page, offset, size);
if (page == page_end) {
return;
}
for (size_t i = page + 1; i < page_end; i++) {
pages[i] = ~u64{0};
}
const size_t offset_end = offset + size;
const size_t offset_end_page_aligned = Common::AlignDown(offset_end, PAGE_BYTES);
TrackPage(page_end, offset_end_page_aligned, offset_end - offset_end_page_aligned);
}
[[nodiscard]] bool IsUsed(u64 offset, u64 size) const noexcept {
const size_t page = offset >> PAGE_SHIFT;
const size_t page_end = (offset + size) >> PAGE_SHIFT;
if (IsPageUsed(page, offset, size)) {
return true;
}
for (size_t i = page + 1; i < page_end; i++) {
if (pages[i] != 0) {
return true;
}
}
const size_t offset_end = offset + size;
const size_t offset_end_page_aligned = Common::AlignDown(offset_end, PAGE_BYTES);
return IsPageUsed(page_end, offset_end_page_aligned, offset_end - offset_end_page_aligned);
}
private:
void TrackPage(u64 page, u64 offset, u64 size) noexcept {
const size_t offset_in_page = offset % PAGE_BYTES;
const size_t first_bit = offset_in_page >> BYTES_PER_BIT_SHIFT;
const size_t num_bits = std::min(size, PAGE_BYTES) >> BYTES_PER_BIT_SHIFT;
const size_t mask = ~u64{0} >> (64 - num_bits);
pages[page] |= (~u64{0} & mask) << first_bit;
}
bool IsPageUsed(u64 page, u64 offset, u64 size) const noexcept {
const size_t offset_in_page = offset % PAGE_BYTES;
const size_t first_bit = offset_in_page >> BYTES_PER_BIT_SHIFT;
const size_t num_bits = std::min(size, PAGE_BYTES) >> BYTES_PER_BIT_SHIFT;
const size_t mask = ~u64{0} >> (64 - num_bits);
const size_t mask2 = (~u64{0} & mask) << first_bit;
return (pages[page] & mask2) != 0;
}
private:
std::vector<u64> pages;
};
} // namespace VideoCommon

7
src/video_core/renderer_opengl/gl_buffer_cache.cpp
View File

@@ -178,13 +178,14 @@ void BufferCacheRuntime::CopyBuffer(GLuint dst_buffer, Buffer& src_buffer,
}
void BufferCacheRuntime::CopyBuffer(Buffer& dst_buffer, GLuint src_buffer,
std::span<const VideoCommon::BufferCopy> copies, bool barrier) {
std::span<const VideoCommon::BufferCopy> copies, bool barrier,
bool) {
CopyBuffer(dst_buffer.Handle(), src_buffer, copies, barrier);
}
void BufferCacheRuntime::CopyBuffer(Buffer& dst_buffer, Buffer& src_buffer,
std::span<const VideoCommon::BufferCopy> copies) {
CopyBuffer(dst_buffer.Handle(), src_buffer.Handle(), copies);
std::span<const VideoCommon::BufferCopy> copies, bool) {
CopyBuffer(dst_buffer.Handle(), src_buffer.Handle(), copies, true);
}
void BufferCacheRuntime::PreCopyBarrier() {

17
src/video_core/renderer_opengl/gl_buffer_cache.h
View File

@@ -30,6 +30,8 @@ public:
void MakeResident(GLenum access) noexcept;
void MarkUsage(u64 offset, u64 size) {}
[[nodiscard]] GLuint View(u32 offset, u32 size, VideoCore::Surface::PixelFormat format);
[[nodiscard]] GLuint64EXT HostGpuAddr() const noexcept {
@@ -66,22 +68,29 @@ public:
[[nodiscard]] StagingBufferMap DownloadStagingBuffer(size_t size);
bool CanReorderUpload(const Buffer&, std::span<const VideoCommon::BufferCopy>) {
return false;
}
void CopyBuffer(GLuint dst_buffer, GLuint src_buffer,
std::span<const VideoCommon::BufferCopy> copies, bool barrier = true);
std::span<const VideoCommon::BufferCopy> copies, bool barrier);
void CopyBuffer(GLuint dst_buffer, Buffer& src_buffer,
std::span<const VideoCommon::BufferCopy> copies, bool barrier = true);
std::span<const VideoCommon::BufferCopy> copies, bool barrier);
void CopyBuffer(Buffer& dst_buffer, GLuint src_buffer,
std::span<const VideoCommon::BufferCopy> copies, bool barrier = true);
std::span<const VideoCommon::BufferCopy> copies, bool barrier,
bool can_reorder_upload = false);
void CopyBuffer(Buffer& dst_buffer, Buffer& src_buffer,
std::span<const VideoCommon::BufferCopy> copies);
std::span<const VideoCommon::BufferCopy> copies, bool);
void PreCopyBarrier();
void PostCopyBarrier();
void Finish();
void TickFrame(VideoCommon::SlotVector<Buffer>&) noexcept {}
void ClearBuffer(Buffer& dest_buffer, u32 offset, size_t size, u32 value);
void BindIndexBuffer(Buffer& buffer, u32 offset, u32 size);

51
src/video_core/renderer_vulkan/blit_image.cpp
View File

@@ -454,7 +454,7 @@ void BlitImageHelper::BlitColor(const Framebuffer* dst_framebuffer, VkImageView
const VkPipeline pipeline = FindOrEmplaceColorPipeline(key);
scheduler.RequestRenderpass(dst_framebuffer);
scheduler.Record([this, dst_region, src_region, pipeline, layout, sampler,
src_view](vk::CommandBuffer cmdbuf) {
src_view](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
// TODO: Barriers
const VkDescriptorSet descriptor_set = one_texture_descriptor_allocator.Commit();
UpdateOneTextureDescriptorSet(device, descriptor_set, sampler, src_view);
@@ -479,7 +479,8 @@ void BlitImageHelper::BlitColor(const Framebuffer* dst_framebuffer, VkImageView
const VkPipeline pipeline = FindOrEmplaceColorPipeline(key);
scheduler.RequestOutsideRenderPassOperationContext();
scheduler.Record([this, dst_framebuffer, src_image_view, src_image, src_sampler, dst_region,
src_region, src_size, pipeline, layout](vk::CommandBuffer cmdbuf) {
src_region, src_size, pipeline,
layout](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
TransitionImageLayout(cmdbuf, src_image, VK_IMAGE_LAYOUT_READ_ONLY_OPTIMAL);
BeginRenderPass(cmdbuf, dst_framebuffer);
const VkDescriptorSet descriptor_set = one_texture_descriptor_allocator.Commit();
@@ -512,7 +513,7 @@ void BlitImageHelper::BlitDepthStencil(const Framebuffer* dst_framebuffer,
const VkPipeline pipeline = FindOrEmplaceDepthStencilPipeline(key);
scheduler.RequestRenderpass(dst_framebuffer);
scheduler.Record([dst_region, src_region, pipeline, layout, sampler, src_depth_view,
src_stencil_view, this](vk::CommandBuffer cmdbuf) {
src_stencil_view, this](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
// TODO: Barriers
const VkDescriptorSet descriptor_set = two_textures_descriptor_allocator.Commit();
UpdateTwoTexturesDescriptorSet(device, descriptor_set, sampler, src_depth_view,
@@ -581,17 +582,17 @@ void BlitImageHelper::ClearColor(const Framebuffer* dst_framebuffer, u8 color_ma
const VkPipeline pipeline = FindOrEmplaceClearColorPipeline(key);
const VkPipelineLayout layout = *clear_color_pipeline_layout;
scheduler.RequestRenderpass(dst_framebuffer);
scheduler.Record(
[pipeline, layout, color_mask, clear_color, dst_region](vk::CommandBuffer cmdbuf) {
cmdbuf.BindPipeline(VK_PIPELINE_BIND_POINT_GRAPHICS, pipeline);
const std::array blend_color = {
(color_mask & 0x1) ? 1.0f : 0.0f, (color_mask & 0x2) ? 1.0f : 0.0f,
(color_mask & 0x4) ? 1.0f : 0.0f, (color_mask & 0x8) ? 1.0f : 0.0f};
cmdbuf.SetBlendConstants(blend_color.data());
BindBlitState(cmdbuf, dst_region);
cmdbuf.PushConstants(layout, VK_SHADER_STAGE_FRAGMENT_BIT, clear_color);
cmdbuf.Draw(3, 1, 0, 0);
});
scheduler.Record([pipeline, layout, color_mask, clear_color,
dst_region](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
cmdbuf.BindPipeline(VK_PIPELINE_BIND_POINT_GRAPHICS, pipeline);
const std::array blend_color = {
(color_mask & 0x1) ? 1.0f : 0.0f, (color_mask & 0x2) ? 1.0f : 0.0f,
(color_mask & 0x4) ? 1.0f : 0.0f, (color_mask & 0x8) ? 1.0f : 0.0f};
cmdbuf.SetBlendConstants(blend_color.data());
BindBlitState(cmdbuf, dst_region);
cmdbuf.PushConstants(layout, VK_SHADER_STAGE_FRAGMENT_BIT, clear_color);
cmdbuf.Draw(3, 1, 0, 0);
});
scheduler.InvalidateState();
}
@@ -608,14 +609,13 @@ void BlitImageHelper::ClearDepthStencil(const Framebuffer* dst_framebuffer, bool
const VkPipeline pipeline = FindOrEmplaceClearStencilPipeline(key);
const VkPipelineLayout layout = *clear_color_pipeline_layout;
scheduler.RequestRenderpass(dst_framebuffer);
scheduler.Record([pipeline, layout, clear_depth, dst_region](vk::CommandBuffer cmdbuf) {
constexpr std::array blend_constants{0.0f, 0.0f, 0.0f, 0.0f};
cmdbuf.SetBlendConstants(blend_constants.data());
cmdbuf.BindPipeline(VK_PIPELINE_BIND_POINT_GRAPHICS, pipeline);
BindBlitState(cmdbuf, dst_region);
cmdbuf.PushConstants(layout, VK_SHADER_STAGE_FRAGMENT_BIT, clear_depth);
cmdbuf.Draw(3, 1, 0, 0);
});
scheduler.Record(
[pipeline, layout, clear_depth, dst_region](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
cmdbuf.BindPipeline(VK_PIPELINE_BIND_POINT_GRAPHICS, pipeline);
BindBlitState(cmdbuf, dst_region);
cmdbuf.PushConstants(layout, VK_SHADER_STAGE_FRAGMENT_BIT, clear_depth);
cmdbuf.Draw(3, 1, 0, 0);
});
scheduler.InvalidateState();
}
@@ -627,7 +627,8 @@ void BlitImageHelper::Convert(VkPipeline pipeline, const Framebuffer* dst_frameb
const VkExtent2D extent = GetConversionExtent(src_image_view);
scheduler.RequestRenderpass(dst_framebuffer);
scheduler.Record([pipeline, layout, sampler, src_view, extent, this](vk::CommandBuffer cmdbuf) {
scheduler.Record([pipeline, layout, sampler, src_view, extent, this](vk::CommandBuffer cmdbuf,
vk::CommandBuffer) {
const VkOffset2D offset{
.x = 0,
.y = 0,
@@ -673,7 +674,7 @@ void BlitImageHelper::ConvertDepthStencil(VkPipeline pipeline, const Framebuffer
scheduler.RequestRenderpass(dst_framebuffer);
scheduler.Record([pipeline, layout, sampler, src_depth_view, src_stencil_view, extent,
this](vk::CommandBuffer cmdbuf) {
this](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
const VkOffset2D offset{
.x = 0,
.y = 0,
@@ -867,7 +868,7 @@ VkPipeline BlitImageHelper::FindOrEmplaceClearStencilPipeline(
.sType = VK_STRUCTURE_TYPE_PIPELINE_DEPTH_STENCIL_STATE_CREATE_INFO,
.pNext = nullptr,
.flags = 0,
.depthTestEnable = key.depth_clear,
.depthTestEnable = VK_FALSE,
.depthWriteEnable = key.depth_clear,
.depthCompareOp = VK_COMPARE_OP_ALWAYS,
.depthBoundsTestEnable = VK_FALSE,

2
src/video_core/renderer_vulkan/renderer_vulkan.cpp
View File

@@ -239,7 +239,7 @@ void Vulkan::RendererVulkan::RenderScreenshot(const Tegra::FramebufferConfig& fr
memory_allocator.CreateBuffer(dst_buffer_info, MemoryUsage::Download);
scheduler.RequestOutsideRenderPassOperationContext();
scheduler.Record([&](vk::CommandBuffer cmdbuf) {
scheduler.Record([&](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
const VkImageMemoryBarrier read_barrier{
.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,
.pNext = nullptr,

75
src/video_core/renderer_vulkan/vk_blit_screen.cpp
View File

@@ -201,43 +201,44 @@ void BlitScreen::Draw(const Tegra::FramebufferConfig& framebuffer,
.depth = 1,
},
};
scheduler.Record([this, copy, index = image_index](vk::CommandBuffer cmdbuf) {
const VkImage image = *raw_images[index];
const VkImageMemoryBarrier base_barrier{
.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,
.pNext = nullptr,
.srcAccessMask = 0,
.dstAccessMask = 0,
.oldLayout = VK_IMAGE_LAYOUT_GENERAL,
.newLayout = VK_IMAGE_LAYOUT_GENERAL,
.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.image = image,
.subresourceRange{
.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT,
.baseMipLevel = 0,
.levelCount = 1,
.baseArrayLayer = 0,
.layerCount = 1,
},
};
VkImageMemoryBarrier read_barrier = base_barrier;
read_barrier.srcAccessMask = VK_ACCESS_HOST_WRITE_BIT;
read_barrier.dstAccessMask = VK_ACCESS_TRANSFER_WRITE_BIT;
read_barrier.oldLayout = VK_IMAGE_LAYOUT_UNDEFINED;
scheduler.Record(
[this, copy, index = image_index](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
const VkImage image = *raw_images[index];
const VkImageMemoryBarrier base_barrier{
.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,
.pNext = nullptr,
.srcAccessMask = 0,
.dstAccessMask = 0,
.oldLayout = VK_IMAGE_LAYOUT_GENERAL,
.newLayout = VK_IMAGE_LAYOUT_GENERAL,
.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.image = image,
.subresourceRange{
.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT,
.baseMipLevel = 0,
.levelCount = 1,
.baseArrayLayer = 0,
.layerCount = 1,
},
};
VkImageMemoryBarrier read_barrier = base_barrier;
read_barrier.srcAccessMask = VK_ACCESS_HOST_WRITE_BIT;
read_barrier.dstAccessMask = VK_ACCESS_TRANSFER_WRITE_BIT;
read_barrier.oldLayout = VK_IMAGE_LAYOUT_UNDEFINED;
VkImageMemoryBarrier write_barrier = base_barrier;
write_barrier.srcAccessMask = VK_ACCESS_TRANSFER_WRITE_BIT;
write_barrier.dstAccessMask = VK_ACCESS_SHADER_READ_BIT;
VkImageMemoryBarrier write_barrier = base_barrier;
write_barrier.srcAccessMask = VK_ACCESS_TRANSFER_WRITE_BIT;
write_barrier.dstAccessMask = VK_ACCESS_SHADER_READ_BIT;
cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_HOST_BIT, VK_PIPELINE_STAGE_TRANSFER_BIT, 0,
read_barrier);
cmdbuf.CopyBufferToImage(*buffer, image, VK_IMAGE_LAYOUT_GENERAL, copy);
cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_TRANSFER_BIT,
VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT |
VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT,
0, write_barrier);
});
cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_HOST_BIT, VK_PIPELINE_STAGE_TRANSFER_BIT,
0, read_barrier);
cmdbuf.CopyBufferToImage(*buffer, image, VK_IMAGE_LAYOUT_GENERAL, copy);
cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_TRANSFER_BIT,
VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT |
VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT,
0, write_barrier);
});
}
const auto anti_alias_pass = Settings::values.anti_aliasing.GetValue();
@@ -250,7 +251,7 @@ void BlitScreen::Draw(const Tegra::FramebufferConfig& framebuffer,
.height = (up_scale * framebuffer.height) >> down_shift,
};
scheduler.Record([this, index = image_index, size,
anti_alias_pass](vk::CommandBuffer cmdbuf) {
anti_alias_pass](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
const VkImageMemoryBarrier base_barrier{
.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,
.pNext = nullptr,
@@ -375,7 +376,7 @@ void BlitScreen::Draw(const Tegra::FramebufferConfig& framebuffer,
}
scheduler.Record([this, host_framebuffer, index = image_index,
size = render_area](vk::CommandBuffer cmdbuf) {
size = render_area](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
const f32 bg_red = Settings::values.bg_red.GetValue() / 255.0f;
const f32 bg_green = Settings::values.bg_green.GetValue() / 255.0f;
const f32 bg_blue = Settings::values.bg_blue.GetValue() / 255.0f;

126
src/video_core/renderer_vulkan/vk_buffer_cache.cpp
View File

@@ -79,13 +79,13 @@ vk::Buffer CreateBuffer(const Device& device, const MemoryAllocator& memory_allo
} // Anonymous namespace
Buffer::Buffer(BufferCacheRuntime&, VideoCommon::NullBufferParams null_params)
: VideoCommon::BufferBase<VideoCore::RasterizerInterface>(null_params) {}
: VideoCommon::BufferBase<VideoCore::RasterizerInterface>(null_params), tracker{4096} {}
Buffer::Buffer(BufferCacheRuntime& runtime, VideoCore::RasterizerInterface& rasterizer_,
VAddr cpu_addr_, u64 size_bytes_)
: VideoCommon::BufferBase<VideoCore::RasterizerInterface>(rasterizer_, cpu_addr_, size_bytes_),
device{&runtime.device}, buffer{
CreateBuffer(*device, runtime.memory_allocator, SizeBytes())} {
device{&runtime.device}, buffer{CreateBuffer(*device, runtime.memory_allocator, SizeBytes())},
tracker{SizeBytes()} {
if (runtime.device.HasDebuggingToolAttached()) {
buffer.SetObjectNameEXT(fmt::format("Buffer 0x{:x}", CpuAddr()).c_str());
}
@@ -179,7 +179,8 @@ public:
if (!host_visible) {
scheduler.RequestOutsideRenderPassOperationContext();
scheduler.Record([src_buffer = staging.buffer, src_offset = staging.offset,
dst_buffer = *buffer, size_bytes](vk::CommandBuffer cmdbuf) {
dst_buffer = *buffer,
size_bytes](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
const VkBufferCopy copy{
.srcOffset = src_offset,
.dstOffset = 0,
@@ -210,9 +211,10 @@ public:
const size_t sub_first_offset = static_cast<size_t>(first % 4) * GetQuadsNum(num_indices);
const size_t offset =
(sub_first_offset + GetQuadsNum(first)) * 6ULL * BytesPerIndex(index_type);
scheduler.Record([buffer_ = *buffer, index_type_, offset](vk::CommandBuffer cmdbuf) {
cmdbuf.BindIndexBuffer(buffer_, offset, index_type_);
});
scheduler.Record(
[buffer_ = *buffer, index_type_, offset](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
cmdbuf.BindIndexBuffer(buffer_, offset, index_type_);
});
}
protected:
@@ -355,12 +357,31 @@ bool BufferCacheRuntime::CanReportMemoryUsage() const {
return device.CanReportMemoryUsage();
}
void BufferCacheRuntime::TickFrame(VideoCommon::SlotVector<Buffer>& slot_buffers) noexcept {
for (auto it = slot_buffers.begin(); it != slot_buffers.end(); it++) {
it->ResetUsageTracking();
}
}
void BufferCacheRuntime::Finish() {
scheduler.Finish();
}
bool BufferCacheRuntime::CanReorderUpload(const Buffer& buffer,
std::span<const VideoCommon::BufferCopy> copies) {
if (Settings::values.disable_buffer_reorder) {
return false;
}
const bool can_use_upload_cmdbuf =
std::ranges::all_of(copies, [&](const VideoCommon::BufferCopy& copy) {
return !buffer.IsRegionUsed(copy.dst_offset, copy.size);
});
return can_use_upload_cmdbuf;
}
void BufferCacheRuntime::CopyBuffer(VkBuffer dst_buffer, VkBuffer src_buffer,
std::span<const VideoCommon::BufferCopy> copies, bool barrier) {
std::span<const VideoCommon::BufferCopy> copies, bool barrier,
bool can_reorder_upload) {
if (dst_buffer == VK_NULL_HANDLE || src_buffer == VK_NULL_HANDLE) {
return;
}
@@ -376,21 +397,31 @@ void BufferCacheRuntime::CopyBuffer(VkBuffer dst_buffer, VkBuffer src_buffer,
.srcAccessMask = VK_ACCESS_TRANSFER_WRITE_BIT,
.dstAccessMask = VK_ACCESS_MEMORY_READ_BIT | VK_ACCESS_MEMORY_WRITE_BIT,
};
// Measuring a popular game, this number never exceeds the specified size once data is warmed up
boost::container::small_vector<VkBufferCopy, 8> vk_copies(copies.size());
std::ranges::transform(copies, vk_copies.begin(), MakeBufferCopy);
if (src_buffer == staging_pool.StreamBuf() && can_reorder_upload) {
scheduler.Record([src_buffer, dst_buffer, vk_copies](vk::CommandBuffer,
vk::CommandBuffer upload_cmdbuf) {
upload_cmdbuf.CopyBuffer(src_buffer, dst_buffer, vk_copies);
});
return;
}
scheduler.RequestOutsideRenderPassOperationContext();
scheduler.Record([src_buffer, dst_buffer, vk_copies, barrier](vk::CommandBuffer cmdbuf) {
if (barrier) {
cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_ALL_COMMANDS_BIT,
VK_PIPELINE_STAGE_TRANSFER_BIT, 0, READ_BARRIER);
}
cmdbuf.CopyBuffer(src_buffer, dst_buffer, vk_copies);
if (barrier) {
cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_TRANSFER_BIT,
VK_PIPELINE_STAGE_ALL_COMMANDS_BIT, 0, WRITE_BARRIER);
}
});
scheduler.Record(
[src_buffer, dst_buffer, vk_copies, barrier](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
if (barrier) {
cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_ALL_COMMANDS_BIT,
VK_PIPELINE_STAGE_TRANSFER_BIT, 0, READ_BARRIER);
}
cmdbuf.CopyBuffer(src_buffer, dst_buffer, vk_copies);
if (barrier) {
cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_TRANSFER_BIT,
VK_PIPELINE_STAGE_ALL_COMMANDS_BIT, 0, WRITE_BARRIER);
}
});
}
void BufferCacheRuntime::PreCopyBarrier() {
@@ -401,7 +432,7 @@ void BufferCacheRuntime::PreCopyBarrier() {
.dstAccessMask = VK_ACCESS_TRANSFER_READ_BIT | VK_ACCESS_TRANSFER_WRITE_BIT,
};
scheduler.RequestOutsideRenderPassOperationContext();
scheduler.Record([](vk::CommandBuffer cmdbuf) {
scheduler.Record([](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_ALL_COMMANDS_BIT, VK_PIPELINE_STAGE_TRANSFER_BIT,
0, READ_BARRIER);
});
@@ -415,7 +446,7 @@ void BufferCacheRuntime::PostCopyBarrier() {
.dstAccessMask = VK_ACCESS_MEMORY_READ_BIT | VK_ACCESS_MEMORY_WRITE_BIT,
};
scheduler.RequestOutsideRenderPassOperationContext();
scheduler.Record([](vk::CommandBuffer cmdbuf) {
scheduler.Record([](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_TRANSFER_BIT, VK_PIPELINE_STAGE_ALL_COMMANDS_BIT,
0, WRITE_BARRIER);
});
@@ -439,13 +470,14 @@ void BufferCacheRuntime::ClearBuffer(VkBuffer dest_buffer, u32 offset, size_t si
};
scheduler.RequestOutsideRenderPassOperationContext();
scheduler.Record([dest_buffer, offset, size, value](vk::CommandBuffer cmdbuf) {
cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_ALL_COMMANDS_BIT, VK_PIPELINE_STAGE_TRANSFER_BIT,
0, READ_BARRIER);
cmdbuf.FillBuffer(dest_buffer, offset, size, value);
cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_TRANSFER_BIT, VK_PIPELINE_STAGE_ALL_COMMANDS_BIT,
0, WRITE_BARRIER);
});
scheduler.Record(
[dest_buffer, offset, size, value](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_ALL_COMMANDS_BIT,
VK_PIPELINE_STAGE_TRANSFER_BIT, 0, READ_BARRIER);
cmdbuf.FillBuffer(dest_buffer, offset, size, value);
cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_TRANSFER_BIT,
VK_PIPELINE_STAGE_ALL_COMMANDS_BIT, 0, WRITE_BARRIER);
});
}
void BufferCacheRuntime::BindIndexBuffer(PrimitiveTopology topology, IndexFormat index_format,
@@ -470,15 +502,16 @@ void BufferCacheRuntime::BindIndexBuffer(PrimitiveTopology topology, IndexFormat
ReserveNullBuffer();
vk_buffer = *null_buffer;
}
scheduler.Record([vk_buffer, vk_offset, vk_index_type](vk::CommandBuffer cmdbuf) {
cmdbuf.BindIndexBuffer(vk_buffer, vk_offset, vk_index_type);
});
scheduler.Record(
[vk_buffer, vk_offset, vk_index_type](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
cmdbuf.BindIndexBuffer(vk_buffer, vk_offset, vk_index_type);
});
}
void BufferCacheRuntime::BindQuadIndexBuffer(PrimitiveTopology topology, u32 first, u32 count) {
if (count == 0) {
ReserveNullBuffer();
scheduler.Record([this](vk::CommandBuffer cmdbuf) {
scheduler.Record([this](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
cmdbuf.BindIndexBuffer(*null_buffer, 0, VK_INDEX_TYPE_UINT32);
});
return;
@@ -499,19 +532,20 @@ void BufferCacheRuntime::BindVertexBuffer(u32 index, VkBuffer buffer, u32 offset
return;
}
if (device.IsExtExtendedDynamicStateSupported()) {
scheduler.Record([index, buffer, offset, size, stride](vk::CommandBuffer cmdbuf) {
const VkDeviceSize vk_offset = buffer != VK_NULL_HANDLE ? offset : 0;
const VkDeviceSize vk_size = buffer != VK_NULL_HANDLE ? size : VK_WHOLE_SIZE;
const VkDeviceSize vk_stride = stride;
cmdbuf.BindVertexBuffers2EXT(index, 1, &buffer, &vk_offset, &vk_size, &vk_stride);
});
scheduler.Record(
[index, buffer, offset, size, stride](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
const VkDeviceSize vk_offset = buffer != VK_NULL_HANDLE ? offset : 0;
const VkDeviceSize vk_size = buffer != VK_NULL_HANDLE ? size : VK_WHOLE_SIZE;
const VkDeviceSize vk_stride = stride;
cmdbuf.BindVertexBuffers2EXT(index, 1, &buffer, &vk_offset, &vk_size, &vk_stride);
});
} else {
if (!device.HasNullDescriptor() && buffer == VK_NULL_HANDLE) {
ReserveNullBuffer();
buffer = *null_buffer;
offset = 0;
}
scheduler.Record([index, buffer, offset](vk::CommandBuffer cmdbuf) {
scheduler.Record([index, buffer, offset](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
cmdbuf.BindVertexBuffer(index, buffer, offset);
});
}
@@ -533,7 +567,8 @@ void BufferCacheRuntime::BindVertexBuffers(VideoCommon::HostBindings<Buffer>& bi
}
if (device.IsExtExtendedDynamicStateSupported()) {
scheduler.Record([this, bindings_ = std::move(bindings),
buffer_handles_ = std::move(buffer_handles)](vk::CommandBuffer cmdbuf) {
buffer_handles_ = std::move(buffer_handles)](vk::CommandBuffer cmdbuf,
vk::CommandBuffer) {
cmdbuf.BindVertexBuffers2EXT(bindings_.min_index,
std::min(bindings_.max_index - bindings_.min_index,
device.GetMaxVertexInputBindings()),
@@ -542,7 +577,8 @@ void BufferCacheRuntime::BindVertexBuffers(VideoCommon::HostBindings<Buffer>& bi
});
} else {
scheduler.Record([this, bindings_ = std::move(bindings),
buffer_handles_ = std::move(buffer_handles)](vk::CommandBuffer cmdbuf) {
buffer_handles_ = std::move(buffer_handles)](vk::CommandBuffer cmdbuf,
vk::CommandBuffer) {
cmdbuf.BindVertexBuffers(bindings_.min_index,
std::min(bindings_.max_index - bindings_.min_index,
device.GetMaxVertexInputBindings()),
@@ -565,7 +601,7 @@ void BufferCacheRuntime::BindTransformFeedbackBuffer(u32 index, VkBuffer buffer,
offset = 0;
size = 0;
}
scheduler.Record([index, buffer, offset, size](vk::CommandBuffer cmdbuf) {
scheduler.Record([index, buffer, offset, size](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
const VkDeviceSize vk_offset = offset;
const VkDeviceSize vk_size = size;
cmdbuf.BindTransformFeedbackBuffersEXT(index, 1, &buffer, &vk_offset, &vk_size);
@@ -581,8 +617,8 @@ void BufferCacheRuntime::BindTransformFeedbackBuffers(VideoCommon::HostBindings<
for (u32 index = 0; index < bindings.buffers.size(); ++index) {
buffer_handles.push_back(bindings.buffers[index]->Handle());
}
scheduler.Record([bindings_ = std::move(bindings),
buffer_handles_ = std::move(buffer_handles)](vk::CommandBuffer cmdbuf) {
scheduler.Record([bindings_ = std::move(bindings), buffer_handles_ = std::move(buffer_handles)](
vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
cmdbuf.BindTransformFeedbackBuffersEXT(0, static_cast<u32>(buffer_handles_.size()),
buffer_handles_.data(), bindings_.offsets.data(),
bindings_.sizes.data());
@@ -613,7 +649,7 @@ void BufferCacheRuntime::ReserveNullBuffer() {
}
scheduler.RequestOutsideRenderPassOperationContext();
scheduler.Record([buffer = *null_buffer](vk::CommandBuffer cmdbuf) {
scheduler.Record([buffer = *null_buffer](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
cmdbuf.FillBuffer(buffer, 0, VK_WHOLE_SIZE, 0);
});
}

21
src/video_core/renderer_vulkan/vk_buffer_cache.h
View File

@@ -5,6 +5,7 @@
#include "video_core/buffer_cache/buffer_cache_base.h"
#include "video_core/buffer_cache/memory_tracker_base.h"
#include "video_core/buffer_cache/usage_tracker.h"
#include "video_core/engines/maxwell_3d.h"
#include "video_core/renderer_vulkan/vk_compute_pass.h"
#include "video_core/renderer_vulkan/vk_staging_buffer_pool.h"
@@ -34,6 +35,18 @@ public:
return *buffer;
}
[[nodiscard]] bool IsRegionUsed(u64 offset, u64 size) const noexcept {
return tracker.IsUsed(offset, size);
}
void MarkUsage(u64 offset, u64 size) noexcept {
tracker.Track(offset, size);
}
void ResetUsageTracking() noexcept {
tracker.Reset();
}
operator VkBuffer() const noexcept {
return *buffer;
}
@@ -49,6 +62,7 @@ private:
const Device* device{};
vk::Buffer buffer;
std::vector<BufferView> views;
VideoCommon::UsageTracker tracker;
};
class QuadArrayIndexBuffer;
@@ -67,6 +81,8 @@ public:
ComputePassDescriptorQueue& compute_pass_descriptor_queue,
DescriptorPool& descriptor_pool);
void TickFrame(VideoCommon::SlotVector<Buffer>& slot_buffers) noexcept;
void Finish();
u64 GetDeviceLocalMemory() const;
@@ -79,12 +95,15 @@ public:
[[nodiscard]] StagingBufferRef DownloadStagingBuffer(size_t size, bool deferred = false);
bool CanReorderUpload(const Buffer& buffer, std::span<const VideoCommon::BufferCopy> copies);
void FreeDeferredStagingBuffer(StagingBufferRef& ref);
void PreCopyBarrier();
void CopyBuffer(VkBuffer src_buffer, VkBuffer dst_buffer,
std::span<const VideoCommon::BufferCopy> copies, bool barrier = true);
std::span<const VideoCommon::BufferCopy> copies, bool barrier,
bool can_reorder_upload = false);
void PostCopyBarrier();

49
src/video_core/renderer_vulkan/vk_compute_pass.cpp
View File

@@ -324,22 +324,23 @@ std::pair<VkBuffer, VkDeviceSize> Uint8Pass::Assemble(u32 num_vertices, VkBuffer
const void* const descriptor_data{compute_pass_descriptor_queue.UpdateData()};
scheduler.RequestOutsideRenderPassOperationContext();
scheduler.Record([this, descriptor_data, num_vertices](vk::CommandBuffer cmdbuf) {
static constexpr u32 DISPATCH_SIZE = 1024;
static constexpr VkMemoryBarrier WRITE_BARRIER{
.sType = VK_STRUCTURE_TYPE_MEMORY_BARRIER,
.pNext = nullptr,
.srcAccessMask = VK_ACCESS_SHADER_WRITE_BIT,
.dstAccessMask = VK_ACCESS_VERTEX_ATTRIBUTE_READ_BIT,
};
const VkDescriptorSet set = descriptor_allocator.Commit();
device.GetLogical().UpdateDescriptorSet(set, *descriptor_template, descriptor_data);
cmdbuf.BindPipeline(VK_PIPELINE_BIND_POINT_COMPUTE, *pipeline);
cmdbuf.BindDescriptorSets(VK_PIPELINE_BIND_POINT_COMPUTE, *layout, 0, set, {});
cmdbuf.Dispatch(Common::DivCeil(num_vertices, DISPATCH_SIZE), 1, 1);
cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT,
VK_PIPELINE_STAGE_VERTEX_INPUT_BIT, 0, WRITE_BARRIER);
});
scheduler.Record(
[this, descriptor_data, num_vertices](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
static constexpr u32 DISPATCH_SIZE = 1024;
static constexpr VkMemoryBarrier WRITE_BARRIER{
.sType = VK_STRUCTURE_TYPE_MEMORY_BARRIER,
.pNext = nullptr,
.srcAccessMask = VK_ACCESS_SHADER_WRITE_BIT,
.dstAccessMask = VK_ACCESS_VERTEX_ATTRIBUTE_READ_BIT,
};
const VkDescriptorSet set = descriptor_allocator.Commit();
device.GetLogical().UpdateDescriptorSet(set, *descriptor_template, descriptor_data);
cmdbuf.BindPipeline(VK_PIPELINE_BIND_POINT_COMPUTE, *pipeline);
cmdbuf.BindDescriptorSets(VK_PIPELINE_BIND_POINT_COMPUTE, *layout, 0, set, {});
cmdbuf.Dispatch(Common::DivCeil(num_vertices, DISPATCH_SIZE), 1, 1);
cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT,
VK_PIPELINE_STAGE_VERTEX_INPUT_BIT, 0, WRITE_BARRIER);
});
return {staging.buffer, staging.offset};
}
@@ -383,7 +384,7 @@ std::pair<VkBuffer, VkDeviceSize> QuadIndexedPass::Assemble(
scheduler.RequestOutsideRenderPassOperationContext();
scheduler.Record([this, descriptor_data, num_tri_vertices, base_vertex, index_shift,
is_strip](vk::CommandBuffer cmdbuf) {
is_strip](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
static constexpr u32 DISPATCH_SIZE = 1024;
static constexpr VkMemoryBarrier WRITE_BARRIER{
.sType = VK_STRUCTURE_TYPE_MEMORY_BARRIER,
@@ -423,7 +424,7 @@ void ConditionalRenderingResolvePass::Resolve(VkBuffer dst_buffer, VkBuffer src_
const void* const descriptor_data{compute_pass_descriptor_queue.UpdateData()};
scheduler.RequestOutsideRenderPassOperationContext();
scheduler.Record([this, descriptor_data](vk::CommandBuffer cmdbuf) {
scheduler.Record([this, descriptor_data](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
static constexpr VkMemoryBarrier read_barrier{
.sType = VK_STRUCTURE_TYPE_MEMORY_BARRIER,
.pNext = nullptr,
@@ -483,7 +484,7 @@ void QueriesPrefixScanPass::Run(VkBuffer accumulation_buffer, VkBuffer dst_buffe
scheduler.RequestOutsideRenderPassOperationContext();
scheduler.Record([this, descriptor_data, min_accumulation_limit, max_accumulation_limit,
runs_to_do, used_offset](vk::CommandBuffer cmdbuf) {
runs_to_do, used_offset](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
static constexpr VkMemoryBarrier read_barrier{
.sType = VK_STRUCTURE_TYPE_MEMORY_BARRIER,
.pNext = nullptr,
@@ -548,8 +549,8 @@ void ASTCDecoderPass::Assemble(Image& image, const StagingBufferRef& map,
const VkImageAspectFlags aspect_mask = image.AspectMask();
const VkImage vk_image = image.Handle();
const bool is_initialized = image.ExchangeInitialization();
scheduler.Record([vk_pipeline, vk_image, aspect_mask,
is_initialized](vk::CommandBuffer cmdbuf) {
scheduler.Record([vk_pipeline, vk_image, aspect_mask, is_initialized](vk::CommandBuffer cmdbuf,
vk::CommandBuffer) {
const VkImageMemoryBarrier image_barrier{
.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,
.pNext = nullptr,
@@ -592,7 +593,7 @@ void ASTCDecoderPass::Assemble(Image& image, const StagingBufferRef& map,
ASSERT(params.destination == (std::array<s32, 3>{0, 0, 0}));
ASSERT(params.bytes_per_block_log2 == 4);
scheduler.Record([this, num_dispatches_x, num_dispatches_y, num_dispatches_z, block_dims,
params, descriptor_data](vk::CommandBuffer cmdbuf) {
params, descriptor_data](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
const AstcPushConstants uniforms{
.blocks_dims = block_dims,
.layer_stride = params.layer_stride,
@@ -608,7 +609,7 @@ void ASTCDecoderPass::Assemble(Image& image, const StagingBufferRef& map,
cmdbuf.Dispatch(num_dispatches_x, num_dispatches_y, num_dispatches_z);
});
}
scheduler.Record([vk_image, aspect_mask](vk::CommandBuffer cmdbuf) {
scheduler.Record([vk_image, aspect_mask](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
const VkImageMemoryBarrier image_barrier{
.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,
.pNext = nullptr,
@@ -699,7 +700,7 @@ void MSAACopyPass::CopyImage(Image& dst_image, Image& src_image,
};
scheduler.Record([this, dst = dst_image.Handle(), msaa_pipeline, num_dispatches,
descriptor_data](vk::CommandBuffer cmdbuf) {
descriptor_data](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
const VkDescriptorSet set = descriptor_allocator.Commit();
device.GetLogical().UpdateDescriptorSet(set, *descriptor_template, descriptor_data);
cmdbuf.BindPipeline(VK_PIPELINE_BIND_POINT_COMPUTE, msaa_pipeline);

6
src/video_core/renderer_vulkan/vk_compute_pipeline.cpp
View File

@@ -199,15 +199,15 @@ void ComputePipeline::Configure(Tegra::Engines::KeplerCompute& kepler_compute,
if (!is_built.load(std::memory_order::relaxed)) {
// Wait for the pipeline to be built
scheduler.Record([this](vk::CommandBuffer) {
scheduler.Record([this](vk::CommandBuffer, vk::CommandBuffer) {
std::unique_lock lock{build_mutex};
build_condvar.wait(lock, [this] { return is_built.load(std::memory_order::relaxed); });
});
}
const void* const descriptor_data{guest_descriptor_queue.UpdateData()};
const bool is_rescaling = !info.texture_descriptors.empty() || !info.image_descriptors.empty();
scheduler.Record([this, descriptor_data, is_rescaling,
rescaling_data = rescaling.Data()](vk::CommandBuffer cmdbuf) {
scheduler.Record([this, descriptor_data, is_rescaling, rescaling_data = rescaling.Data()](
vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
cmdbuf.BindPipeline(VK_PIPELINE_BIND_POINT_COMPUTE, *pipeline);
if (!descriptor_set_layout) {
return;

3
src/video_core/renderer_vulkan/vk_fsr.cpp
View File

@@ -38,7 +38,8 @@ VkImageView FSR::Draw(Scheduler& scheduler, size_t image_index, VkImageView imag
UpdateDescriptorSet(image_index, image_view);
scheduler.Record([this, image_index, input_image_extent, crop_rect](vk::CommandBuffer cmdbuf) {
scheduler.Record([this, image_index, input_image_extent, crop_rect](vk::CommandBuffer cmdbuf,
vk::CommandBuffer) {
const VkImageMemoryBarrier base_barrier{
.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,
.pNext = nullptr,

73
src/video_core/renderer_vulkan/vk_graphics_pipeline.cpp
View File

@@ -493,7 +493,7 @@ void GraphicsPipeline::ConfigureDraw(const RescalingPushConstant& rescaling,
if (!is_built.load(std::memory_order::relaxed)) {
// Wait for the pipeline to be built
scheduler.Record([this](vk::CommandBuffer) {
scheduler.Record([this](vk::CommandBuffer, vk::CommandBuffer) {
std::unique_lock lock{build_mutex};
build_condvar.wait(lock, [this] { return is_built.load(std::memory_order::relaxed); });
});
@@ -502,42 +502,43 @@ void GraphicsPipeline::ConfigureDraw(const RescalingPushConstant& rescaling,
const bool update_rescaling{scheduler.UpdateRescaling(is_rescaling)};
const bool bind_pipeline{scheduler.UpdateGraphicsPipeline(this)};
const void* const descriptor_data{guest_descriptor_queue.UpdateData()};
scheduler.Record([this, descriptor_data, bind_pipeline, rescaling_data = rescaling.Data(),
is_rescaling, update_rescaling,
uses_render_area = render_area.uses_render_area,
render_area_data = render_area.words](vk::CommandBuffer cmdbuf) {
if (bind_pipeline) {
cmdbuf.BindPipeline(VK_PIPELINE_BIND_POINT_GRAPHICS, *pipeline);
}
cmdbuf.PushConstants(*pipeline_layout, VK_SHADER_STAGE_ALL_GRAPHICS,
RESCALING_LAYOUT_WORDS_OFFSET, sizeof(rescaling_data),
rescaling_data.data());
if (update_rescaling) {
const f32 config_down_factor{Settings::values.resolution_info.down_factor};
const f32 scale_down_factor{is_rescaling ? config_down_factor : 1.0f};
scheduler.Record(
[this, descriptor_data, bind_pipeline, rescaling_data = rescaling.Data(), is_rescaling,
update_rescaling, uses_render_area = render_area.uses_render_area,
render_area_data = render_area.words](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
if (bind_pipeline) {
cmdbuf.BindPipeline(VK_PIPELINE_BIND_POINT_GRAPHICS, *pipeline);
}
cmdbuf.PushConstants(*pipeline_layout, VK_SHADER_STAGE_ALL_GRAPHICS,
RESCALING_LAYOUT_DOWN_FACTOR_OFFSET, sizeof(scale_down_factor),
&scale_down_factor);
}
if (uses_render_area) {
cmdbuf.PushConstants(*pipeline_layout, VK_SHADER_STAGE_ALL_GRAPHICS,
RENDERAREA_LAYOUT_OFFSET, sizeof(render_area_data),
&render_area_data);
}
if (!descriptor_set_layout) {
return;
}
if (uses_push_descriptor) {
cmdbuf.PushDescriptorSetWithTemplateKHR(*descriptor_update_template, *pipeline_layout,
0, descriptor_data);
} else {
const VkDescriptorSet descriptor_set{descriptor_allocator.Commit()};
const vk::Device& dev{device.GetLogical()};
dev.UpdateDescriptorSet(descriptor_set, *descriptor_update_template, descriptor_data);
cmdbuf.BindDescriptorSets(VK_PIPELINE_BIND_POINT_GRAPHICS, *pipeline_layout, 0,
descriptor_set, nullptr);
}
});
RESCALING_LAYOUT_WORDS_OFFSET, sizeof(rescaling_data),
rescaling_data.data());
if (update_rescaling) {
const f32 config_down_factor{Settings::values.resolution_info.down_factor};
const f32 scale_down_factor{is_rescaling ? config_down_factor : 1.0f};
cmdbuf.PushConstants(*pipeline_layout, VK_SHADER_STAGE_ALL_GRAPHICS,
RESCALING_LAYOUT_DOWN_FACTOR_OFFSET, sizeof(scale_down_factor),
&scale_down_factor);
}
if (uses_render_area) {
cmdbuf.PushConstants(*pipeline_layout, VK_SHADER_STAGE_ALL_GRAPHICS,
RENDERAREA_LAYOUT_OFFSET, sizeof(render_area_data),
&render_area_data);
}
if (!descriptor_set_layout) {
return;
}
if (uses_push_descriptor) {
cmdbuf.PushDescriptorSetWithTemplateKHR(*descriptor_update_template,
*pipeline_layout, 0, descriptor_data);
} else {
const VkDescriptorSet descriptor_set{descriptor_allocator.Commit()};
const vk::Device& dev{device.GetLogical()};
dev.UpdateDescriptorSet(descriptor_set, *descriptor_update_template,
descriptor_data);
cmdbuf.BindDescriptorSets(VK_PIPELINE_BIND_POINT_GRAPHICS, *pipeline_layout, 0,
descriptor_set, nullptr);
}
});
}
void GraphicsPipeline::MakePipeline(VkRenderPass render_pass) {

29
src/video_core/renderer_vulkan/vk_master_semaphore.cpp
View File

@@ -100,12 +100,14 @@ void MasterSemaphore::Wait(u64 tick) {
Refresh();
}
VkResult MasterSemaphore::SubmitQueue(vk::CommandBuffer& cmdbuf, VkSemaphore signal_semaphore,
VkSemaphore wait_semaphore, u64 host_tick) {
VkResult MasterSemaphore::SubmitQueue(vk::CommandBuffer& cmdbuf, vk::CommandBuffer& upload_cmdbuf,
VkSemaphore signal_semaphore, VkSemaphore wait_semaphore,
u64 host_tick) {
if (semaphore) {
return SubmitQueueTimeline(cmdbuf, signal_semaphore, wait_semaphore, host_tick);
return SubmitQueueTimeline(cmdbuf, upload_cmdbuf, signal_semaphore, wait_semaphore,
host_tick);
} else {
return SubmitQueueFence(cmdbuf, signal_semaphore, wait_semaphore, host_tick);
return SubmitQueueFence(cmdbuf, upload_cmdbuf, signal_semaphore, wait_semaphore, host_tick);
}
}
@@ -115,6 +117,7 @@ static constexpr std::array<VkPipelineStageFlags, 2> wait_stage_masks{
};
VkResult MasterSemaphore::SubmitQueueTimeline(vk::CommandBuffer& cmdbuf,
vk::CommandBuffer& upload_cmdbuf,
VkSemaphore signal_semaphore,
VkSemaphore wait_semaphore, u64 host_tick) {
const VkSemaphore timeline_semaphore = *semaphore;
@@ -123,6 +126,8 @@ VkResult MasterSemaphore::SubmitQueueTimeline(vk::CommandBuffer& cmdbuf,
const std::array signal_values{host_tick, u64(0)};
const std::array signal_semaphores{timeline_semaphore, signal_semaphore};
const std::array cmdbuffers{*upload_cmdbuf, *cmdbuf};
const u32 num_wait_semaphores = wait_semaphore ? 1 : 0;
const VkTimelineSemaphoreSubmitInfo timeline_si{
.sType = VK_STRUCTURE_TYPE_TIMELINE_SEMAPHORE_SUBMIT_INFO,
@@ -138,8 +143,8 @@ VkResult MasterSemaphore::SubmitQueueTimeline(vk::CommandBuffer& cmdbuf,
.waitSemaphoreCount = num_wait_semaphores,
.pWaitSemaphores = &wait_semaphore,
.pWaitDstStageMask = wait_stage_masks.data(),
.commandBufferCount = 1,
.pCommandBuffers = cmdbuf.address(),
.commandBufferCount = static_cast<u32>(cmdbuffers.size()),
.pCommandBuffers = cmdbuffers.data(),
.signalSemaphoreCount = num_signal_semaphores,
.pSignalSemaphores = signal_semaphores.data(),
};
@@ -147,19 +152,23 @@ VkResult MasterSemaphore::SubmitQueueTimeline(vk::CommandBuffer& cmdbuf,
return device.GetGraphicsQueue().Submit(submit_info);
}
VkResult MasterSemaphore::SubmitQueueFence(vk::CommandBuffer& cmdbuf, VkSemaphore signal_semaphore,
VkSemaphore wait_semaphore, u64 host_tick) {
VkResult MasterSemaphore::SubmitQueueFence(vk::CommandBuffer& cmdbuf,
vk::CommandBuffer& upload_cmdbuf,
VkSemaphore signal_semaphore, VkSemaphore wait_semaphore,
u64 host_tick) {
const u32 num_signal_semaphores = signal_semaphore ? 1 : 0;
const u32 num_wait_semaphores = wait_semaphore ? 1 : 0;
const std::array cmdbuffers{*upload_cmdbuf, *cmdbuf};
const VkSubmitInfo submit_info{
.sType = VK_STRUCTURE_TYPE_SUBMIT_INFO,
.pNext = nullptr,
.waitSemaphoreCount = num_wait_semaphores,
.pWaitSemaphores = &wait_semaphore,
.pWaitDstStageMask = wait_stage_masks.data(),
.commandBufferCount = 1,
.pCommandBuffers = cmdbuf.address(),
.commandBufferCount = static_cast<u32>(cmdbuffers.size()),
.pCommandBuffers = cmdbuffers.data(),
.signalSemaphoreCount = num_signal_semaphores,
.pSignalSemaphores = &signal_semaphore,
};

14
src/video_core/renderer_vulkan/vk_master_semaphore.h
View File

@@ -52,14 +52,16 @@ public:
void Wait(u64 tick);
/// Submits the device graphics queue, updating the tick as necessary
VkResult SubmitQueue(vk::CommandBuffer& cmdbuf, VkSemaphore signal_semaphore,
VkSemaphore wait_semaphore, u64 host_tick);
VkResult SubmitQueue(vk::CommandBuffer& cmdbuf, vk::CommandBuffer& upload_cmdbuf,
VkSemaphore signal_semaphore, VkSemaphore wait_semaphore, u64 host_tick);
private:
VkResult SubmitQueueTimeline(vk::CommandBuffer& cmdbuf, VkSemaphore signal_semaphore,
VkSemaphore wait_semaphore, u64 host_tick);
VkResult SubmitQueueFence(vk::CommandBuffer& cmdbuf, VkSemaphore signal_semaphore,
VkSemaphore wait_semaphore, u64 host_tick);
VkResult SubmitQueueTimeline(vk::CommandBuffer& cmdbuf, vk::CommandBuffer& upload_cmdbuf,
VkSemaphore signal_semaphore, VkSemaphore wait_semaphore,
u64 host_tick);
VkResult SubmitQueueFence(vk::CommandBuffer& cmdbuf, vk::CommandBuffer& upload_cmdbuf,
VkSemaphore signal_semaphore, VkSemaphore wait_semaphore,
u64 host_tick);
void WaitThread(std::stop_token token);

4
src/video_core/renderer_vulkan/vk_present_manager.cpp
View File

@@ -166,7 +166,7 @@ void PresentManager::Present(Frame* frame) {
return;
}
scheduler.Record([this, frame](vk::CommandBuffer) {
scheduler.Record([this, frame](vk::CommandBuffer, vk::CommandBuffer) {
std::unique_lock lock{queue_mutex};
present_queue.push(frame);
frame_cv.notify_one();
@@ -488,4 +488,4 @@ void PresentManager::CopyToSwapchain(Frame* frame) {
swapchain.Present(render_semaphore);
}
} // namespace Vulkan
} // namespace Vulkan

74
src/video_core/renderer_vulkan/vk_query_cache.cpp
View File

@@ -138,9 +138,10 @@ public:
};
accumulation_buffer = memory_allocator.CreateBuffer(buffer_ci, MemoryUsage::DeviceLocal);
scheduler.RequestOutsideRenderPassOperationContext();
scheduler.Record([buffer = *accumulation_buffer](vk::CommandBuffer cmdbuf) {
cmdbuf.FillBuffer(buffer, 0, 8, 0);
});
scheduler.Record(
[buffer = *accumulation_buffer](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
cmdbuf.FillBuffer(buffer, 0, 8, 0);
});
}
~SamplesStreamer() = default;
@@ -150,8 +151,8 @@ public:
return;
}
ReserveHostQuery();
scheduler.Record([query_pool = current_query_pool,
query_index = current_bank_slot](vk::CommandBuffer cmdbuf) {
scheduler.Record([query_pool = current_query_pool, query_index = current_bank_slot](
vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
const bool use_precise = Settings::IsGPULevelHigh();
cmdbuf.BeginQuery(query_pool, static_cast<u32>(query_index),
use_precise ? VK_QUERY_CONTROL_PRECISE_BIT : 0);
@@ -163,8 +164,8 @@ public:
if (!has_started) {
return;
}
scheduler.Record([query_pool = current_query_pool,
query_index = current_bank_slot](vk::CommandBuffer cmdbuf) {
scheduler.Record([query_pool = current_query_pool, query_index = current_bank_slot](
vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
cmdbuf.EndQuery(query_pool, static_cast<u32>(query_index));
});
has_started = false;
@@ -227,8 +228,8 @@ public:
auto& resolve_buffer = buffers[resolve_buffer_index];
VkQueryPool query_pool = bank->GetInnerPool();
scheduler.RequestOutsideRenderPassOperationContext();
scheduler.Record([start, amount, base_offset, query_pool,
buffer = *resolve_buffer](vk::CommandBuffer cmdbuf) {
scheduler.Record([start, amount, base_offset, query_pool, buffer = *resolve_buffer](
vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
const VkBufferMemoryBarrier copy_query_pool_barrier{
.sType = VK_STRUCTURE_TYPE_BUFFER_MEMORY_BARRIER,
.pNext = nullptr,
@@ -291,9 +292,10 @@ public:
} else {
scheduler.RequestOutsideRenderPassOperationContext();
scheduler.Record([buffer = *accumulation_buffer](vk::CommandBuffer cmdbuf) {
cmdbuf.FillBuffer(buffer, 0, 8, 0);
});
scheduler.Record(
[buffer = *accumulation_buffer](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
cmdbuf.FillBuffer(buffer, 0, 8, 0);
});
}
ReplicateCurrentQueryIfNeeded();
@@ -608,7 +610,7 @@ public:
void Sync(StagingBufferRef& stagging_buffer, size_t extra_offset, size_t start, size_t size) {
scheduler.RequestOutsideRenderPassOperationContext();
scheduler.Record([this, dst_buffer = stagging_buffer.buffer, extra_offset, start,
size](vk::CommandBuffer cmdbuf) {
size](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
std::array<VkBufferCopy, 1> copy{VkBufferCopy{
.srcOffset = start * QUERY_SIZE,
.dstOffset = extra_offset,
@@ -793,7 +795,7 @@ public:
.dstAccessMask = VK_ACCESS_MEMORY_READ_BIT | VK_ACCESS_MEMORY_WRITE_BIT,
};
scheduler.RequestOutsideRenderPassOperationContext();
scheduler.Record([](vk::CommandBuffer cmdbuf) {
scheduler.Record([](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_TRANSFER_BIT,
VK_PIPELINE_STAGE_ALL_COMMANDS_BIT, 0, WRITE_BARRIER);
});
@@ -841,13 +843,14 @@ private:
}
has_flushed_end_pending = true;
if (!has_started || buffers_count == 0) {
scheduler.Record([](vk::CommandBuffer cmdbuf) {
scheduler.Record([](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
cmdbuf.BeginTransformFeedbackEXT(0, 0, nullptr, nullptr);
});
UpdateBuffers();
return;
}
scheduler.Record([this, total = static_cast<u32>(buffers_count)](vk::CommandBuffer cmdbuf) {
scheduler.Record([this, total = static_cast<u32>(buffers_count)](vk::CommandBuffer cmdbuf,
vk::CommandBuffer) {
cmdbuf.BeginTransformFeedbackEXT(0, total, counter_buffers.data(), offsets.data());
});
UpdateBuffers();
@@ -861,12 +864,12 @@ private:
has_flushed_end_pending = false;
if (buffers_count == 0) {
scheduler.Record([](vk::CommandBuffer cmdbuf) {
scheduler.Record([](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
cmdbuf.EndTransformFeedbackEXT(0, 0, nullptr, nullptr);
});
} else {
scheduler.Record([this,
total = static_cast<u32>(buffers_count)](vk::CommandBuffer cmdbuf) {
scheduler.Record([this, total = static_cast<u32>(buffers_count)](
vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
cmdbuf.EndTransformFeedbackEXT(0, total, counter_buffers.data(), offsets.data());
});
}
@@ -918,7 +921,7 @@ private:
scheduler.RequestOutsideRenderPassOperationContext();
scheduler.Record([dst_buffer = current_bank->GetBuffer(),
src_buffer = counter_buffers[stream], src_offset = offsets[stream],
slot](vk::CommandBuffer cmdbuf) {
slot](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_TRANSFORM_FEEDBACK_BIT_EXT,
VK_PIPELINE_STAGE_TRANSFER_BIT, 0, READ_BARRIER);
std::array<VkBufferCopy, 1> copy{VkBufferCopy{
@@ -1253,8 +1256,9 @@ void QueryCacheRuntime::PauseHostConditionalRendering() {
return;
}
if (impl->is_hcr_running) {
impl->scheduler.Record(
[](vk::CommandBuffer cmdbuf) { cmdbuf.EndConditionalRenderingEXT(); });
impl->scheduler.Record([](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
cmdbuf.EndConditionalRenderingEXT();
});
}
impl->is_hcr_running = false;
}
@@ -1264,9 +1268,10 @@ void QueryCacheRuntime::ResumeHostConditionalRendering() {
return;
}
if (!impl->is_hcr_running) {
impl->scheduler.Record([hcr_setup = impl->hcr_setup](vk::CommandBuffer cmdbuf) {
cmdbuf.BeginConditionalRenderingEXT(hcr_setup);
});
impl->scheduler.Record(
[hcr_setup = impl->hcr_setup](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
cmdbuf.BeginConditionalRenderingEXT(hcr_setup);
});
}
impl->is_hcr_running = true;
}
@@ -1436,12 +1441,12 @@ void QueryCacheRuntime::Barriers(bool is_prebarrier) {
.dstAccessMask = VK_ACCESS_MEMORY_READ_BIT | VK_ACCESS_MEMORY_WRITE_BIT,
};
if (is_prebarrier) {
impl->scheduler.Record([](vk::CommandBuffer cmdbuf) {
impl->scheduler.Record([](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_ALL_COMMANDS_BIT,
VK_PIPELINE_STAGE_TRANSFER_BIT, 0, READ_BARRIER);
});
} else {
impl->scheduler.Record([](vk::CommandBuffer cmdbuf) {
impl->scheduler.Record([](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_TRANSFER_BIT,
VK_PIPELINE_STAGE_ALL_COMMANDS_BIT, 0, WRITE_BARRIER);
});
@@ -1536,13 +1541,14 @@ void QueryCacheRuntime::SyncValues(std::span<SyncValuesType> values, VkBuffer ba
}
impl->scheduler.RequestOutsideRenderPassOperationContext();
impl->scheduler.Record([src_buffer, dst_buffers = std::move(impl->buffers_to_upload_to),
vk_copies = std::move(impl->copies_setup)](vk::CommandBuffer cmdbuf) {
size_t size = dst_buffers.size();
for (size_t i = 0; i < size; i++) {
cmdbuf.CopyBuffer(src_buffer, dst_buffers[i].first, vk_copies[i]);
}
});
impl->scheduler.Record(
[src_buffer, dst_buffers = std::move(impl->buffers_to_upload_to),
vk_copies = std::move(impl->copies_setup)](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
size_t size = dst_buffers.size();
for (size_t i = 0; i < size; i++) {
cmdbuf.CopyBuffer(src_buffer, dst_buffers[i].first, vk_copies[i]);
}
});
}
} // namespace Vulkan

166
src/video_core/renderer_vulkan/vk_rasterizer.cpp
View File

@@ -216,7 +216,7 @@ void RasterizerVulkan::Draw(bool is_indexed, u32 instance_count) {
const auto& draw_state = maxwell3d->draw_manager->GetDrawState();
const u32 num_instances{instance_count};
const DrawParams draw_params{MakeDrawParams(draw_state, num_instances, is_indexed)};
scheduler.Record([draw_params](vk::CommandBuffer cmdbuf) {
scheduler.Record([draw_params](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
if (draw_params.is_indexed) {
cmdbuf.DrawIndexed(draw_params.num_vertices, draw_params.num_instances,
draw_params.first_index, draw_params.base_vertex,
@@ -238,7 +238,7 @@ void RasterizerVulkan::DrawIndirect() {
const auto& offset = indirect_buffer.second;
if (params.is_byte_count) {
scheduler.Record([buffer_obj = buffer->Handle(), offset,
stride = params.stride](vk::CommandBuffer cmdbuf) {
stride = params.stride](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
cmdbuf.DrawIndirectByteCountEXT(1, 0, buffer_obj, offset, 0,
static_cast<u32>(stride));
});
@@ -250,7 +250,7 @@ void RasterizerVulkan::DrawIndirect() {
const auto& offset_base = count.second;
scheduler.Record([draw_buffer_obj = draw_buffer->Handle(),
buffer_obj = buffer->Handle(), offset_base, offset,
params](vk::CommandBuffer cmdbuf) {
params](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
if (params.is_indexed) {
cmdbuf.DrawIndexedIndirectCount(
buffer_obj, offset, draw_buffer_obj, offset_base,
@@ -263,7 +263,8 @@ void RasterizerVulkan::DrawIndirect() {
});
return;
}
scheduler.Record([buffer_obj = buffer->Handle(), offset, params](vk::CommandBuffer cmdbuf) {
scheduler.Record([buffer_obj = buffer->Handle(), offset, params](vk::CommandBuffer cmdbuf,
vk::CommandBuffer) {
if (params.is_indexed) {
cmdbuf.DrawIndexedIndirect(buffer_obj, offset,
static_cast<u32>(params.max_draw_counts),
@@ -387,7 +388,8 @@ void RasterizerVulkan::Clear(u32 layer_count) {
if (regs.clear_surface.R && regs.clear_surface.G && regs.clear_surface.B &&
regs.clear_surface.A) {
scheduler.Record([color_attachment, clear_value, clear_rect](vk::CommandBuffer cmdbuf) {
scheduler.Record([color_attachment, clear_value, clear_rect](vk::CommandBuffer cmdbuf,
vk::CommandBuffer) {
const VkClearAttachment attachment{
.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT,
.colorAttachment = color_attachment,
@@ -422,8 +424,7 @@ void RasterizerVulkan::Clear(u32 layer_count) {
return;
}
if (use_stencil && framebuffer->HasAspectStencilBit() && regs.stencil_front_mask != 0xFF &&
regs.stencil_front_mask != 0) {
if (use_stencil && regs.stencil_front_mask != 0xFF && regs.stencil_front_mask != 0) {
Region2D dst_region = {
Offset2D{.x = clear_rect.rect.offset.x, .y = clear_rect.rect.offset.y},
Offset2D{.x = clear_rect.rect.offset.x + static_cast<s32>(clear_rect.rect.extent.width),
@@ -434,7 +435,7 @@ void RasterizerVulkan::Clear(u32 layer_count) {
regs.stencil_front_func_mask, dst_region);
} else {
scheduler.Record([clear_depth = regs.clear_depth, clear_stencil = regs.clear_stencil,
clear_rect, aspect_flags](vk::CommandBuffer cmdbuf) {
clear_rect, aspect_flags](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
VkClearAttachment attachment;
attachment.aspectMask = aspect_flags;
attachment.colorAttachment = 0;
@@ -466,14 +467,16 @@ void RasterizerVulkan::DispatchCompute() {
buffer_cache.ObtainBuffer(*indirect_address, 12, sync_info, post_op);
scheduler.RequestOutsideRenderPassOperationContext();
scheduler.Record([indirect_buffer = buffer->Handle(),
indirect_offset = offset](vk::CommandBuffer cmdbuf) {
indirect_offset = offset](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
cmdbuf.DispatchIndirect(indirect_buffer, indirect_offset);
});
return;
}
const std::array<u32, 3> dim{qmd.grid_dim_x, qmd.grid_dim_y, qmd.grid_dim_z};
scheduler.RequestOutsideRenderPassOperationContext();
scheduler.Record([dim](vk::CommandBuffer cmdbuf) { cmdbuf.Dispatch(dim[0], dim[1], dim[2]); });
scheduler.Record([dim](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
cmdbuf.Dispatch(dim[0], dim[1], dim[2]);
});
}
void RasterizerVulkan::ResetCounter(VideoCommon::QueryType type) {
@@ -691,7 +694,7 @@ void RasterizerVulkan::WaitForIdle() {
query_cache.NotifyWFI();
scheduler.RequestOutsideRenderPassOperationContext();
scheduler.Record([event = *wfi_event, flags](vk::CommandBuffer cmdbuf) {
scheduler.Record([event = *wfi_event, flags](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
cmdbuf.SetEvent(event, flags);
cmdbuf.WaitEvents(event, flags, VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT, {}, {}, {});
});
@@ -949,7 +952,9 @@ void RasterizerVulkan::UpdateViewportsState(Tegra::Engines::Maxwell3D::Regs& reg
.minDepth = 0.0f,
.maxDepth = 1.0f,
};
scheduler.Record([viewport](vk::CommandBuffer cmdbuf) { cmdbuf.SetViewport(0, viewport); });
scheduler.Record([viewport](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
cmdbuf.SetViewport(0, viewport);
});
return;
}
const bool is_rescaling{texture_cache.IsRescaling()};
@@ -964,7 +969,7 @@ void RasterizerVulkan::UpdateViewportsState(Tegra::Engines::Maxwell3D::Regs& reg
GetViewportState(device, regs, 12, scale), GetViewportState(device, regs, 13, scale),
GetViewportState(device, regs, 14, scale), GetViewportState(device, regs, 15, scale),
};
scheduler.Record([this, viewport_list](vk::CommandBuffer cmdbuf) {
scheduler.Record([this, viewport_list](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
const u32 num_viewports = std::min<u32>(device.GetMaxViewports(), Maxwell::NumViewports);
const vk::Span<VkViewport> viewports(viewport_list.data(), num_viewports);
cmdbuf.SetViewport(0, viewports);
@@ -999,7 +1004,7 @@ void RasterizerVulkan::UpdateScissorsState(Tegra::Engines::Maxwell3D::Regs& regs
GetScissorState(regs, 14, up_scale, down_shift),
GetScissorState(regs, 15, up_scale, down_shift),
};
scheduler.Record([this, scissor_list](vk::CommandBuffer cmdbuf) {
scheduler.Record([this, scissor_list](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
const u32 num_scissors = std::min<u32>(device.GetMaxViewports(), Maxwell::NumViewports);
const vk::Span<VkRect2D> scissors(scissor_list.data(), num_scissors);
cmdbuf.SetScissor(0, scissors);
@@ -1032,22 +1037,23 @@ void RasterizerVulkan::UpdateDepthBias(Tegra::Engines::Maxwell3D::Regs& regs) {
units = static_cast<float>(static_cast<double>(units) * rescale_factor);
return false;
})();
scheduler.Record([constant = units, clamp = regs.depth_bias_clamp,
factor = regs.slope_scale_depth_bias, force_unorm,
precise = device.HasExactDepthBiasControl()](vk::CommandBuffer cmdbuf) {
if (force_unorm) {
VkDepthBiasRepresentationInfoEXT info{
.sType = VK_STRUCTURE_TYPE_DEPTH_BIAS_REPRESENTATION_INFO_EXT,
.pNext = nullptr,
.depthBiasRepresentation =
VK_DEPTH_BIAS_REPRESENTATION_LEAST_REPRESENTABLE_VALUE_FORCE_UNORM_EXT,
.depthBiasExact = precise ? VK_TRUE : VK_FALSE,
};
cmdbuf.SetDepthBias(constant, clamp, factor, &info);
return;
}
cmdbuf.SetDepthBias(constant, clamp, factor);
});
scheduler.Record(
[constant = units, clamp = regs.depth_bias_clamp, factor = regs.slope_scale_depth_bias,
force_unorm,
precise = device.HasExactDepthBiasControl()](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
if (force_unorm) {
VkDepthBiasRepresentationInfoEXT info{
.sType = VK_STRUCTURE_TYPE_DEPTH_BIAS_REPRESENTATION_INFO_EXT,
.pNext = nullptr,
.depthBiasRepresentation =
VK_DEPTH_BIAS_REPRESENTATION_LEAST_REPRESENTABLE_VALUE_FORCE_UNORM_EXT,
.depthBiasExact = precise ? VK_TRUE : VK_FALSE,
};
cmdbuf.SetDepthBias(constant, clamp, factor, &info);
return;
}
cmdbuf.SetDepthBias(constant, clamp, factor);
});
}
void RasterizerVulkan::UpdateBlendConstants(Tegra::Engines::Maxwell3D::Regs& regs) {
@@ -1056,16 +1062,19 @@ void RasterizerVulkan::UpdateBlendConstants(Tegra::Engines::Maxwell3D::Regs& reg
}
const std::array blend_color = {regs.blend_color.r, regs.blend_color.g, regs.blend_color.b,
regs.blend_color.a};
scheduler.Record(
[blend_color](vk::CommandBuffer cmdbuf) { cmdbuf.SetBlendConstants(blend_color.data()); });
scheduler.Record([blend_color](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
cmdbuf.SetBlendConstants(blend_color.data());
});
}
void RasterizerVulkan::UpdateDepthBounds(Tegra::Engines::Maxwell3D::Regs& regs) {
if (!state_tracker.TouchDepthBounds()) {
return;
}
scheduler.Record([min = regs.depth_bounds[0], max = regs.depth_bounds[1]](
vk::CommandBuffer cmdbuf) { cmdbuf.SetDepthBounds(min, max); });
scheduler.Record([min = regs.depth_bounds[0],
max = regs.depth_bounds[1]](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
cmdbuf.SetDepthBounds(min, max);
});
}
void RasterizerVulkan::UpdateStencilFaces(Tegra::Engines::Maxwell3D::Regs& regs) {
@@ -1093,7 +1102,8 @@ void RasterizerVulkan::UpdateStencilFaces(Tegra::Engines::Maxwell3D::Regs& regs)
}
}
scheduler.Record([front_ref = regs.stencil_front_ref, back_ref = regs.stencil_back_ref,
two_sided = regs.stencil_two_side_enable](vk::CommandBuffer cmdbuf) {
two_sided = regs.stencil_two_side_enable](vk::CommandBuffer cmdbuf,
vk::CommandBuffer) {
const bool set_back = two_sided && front_ref != back_ref;
// Front face
cmdbuf.SetStencilReference(set_back ? VK_STENCIL_FACE_FRONT_BIT
@@ -1119,7 +1129,8 @@ void RasterizerVulkan::UpdateStencilFaces(Tegra::Engines::Maxwell3D::Regs& regs)
}
scheduler.Record([front_write_mask = regs.stencil_front_mask,
back_write_mask = regs.stencil_back_mask,
two_sided = regs.stencil_two_side_enable](vk::CommandBuffer cmdbuf) {
two_sided = regs.stencil_two_side_enable](vk::CommandBuffer cmdbuf,
vk::CommandBuffer) {
const bool set_back = two_sided && front_write_mask != back_write_mask;
// Front face
cmdbuf.SetStencilWriteMask(set_back ? VK_STENCIL_FACE_FRONT_BIT
@@ -1145,7 +1156,8 @@ void RasterizerVulkan::UpdateStencilFaces(Tegra::Engines::Maxwell3D::Regs& regs)
}
scheduler.Record([front_test_mask = regs.stencil_front_func_mask,
back_test_mask = regs.stencil_back_func_mask,
two_sided = regs.stencil_two_side_enable](vk::CommandBuffer cmdbuf) {
two_sided = regs.stencil_two_side_enable](vk::CommandBuffer cmdbuf,
vk::CommandBuffer) {
const bool set_back = two_sided && front_test_mask != back_test_mask;
// Front face
cmdbuf.SetStencilCompareMask(set_back ? VK_STENCIL_FACE_FRONT_BIT
@@ -1166,7 +1178,8 @@ void RasterizerVulkan::UpdateLineWidth(Tegra::Engines::Maxwell3D::Regs& regs) {
}
const float width =
regs.line_anti_alias_enable ? regs.line_width_smooth : regs.line_width_aliased;
scheduler.Record([width](vk::CommandBuffer cmdbuf) { cmdbuf.SetLineWidth(width); });
scheduler.Record(
[width](vk::CommandBuffer cmdbuf, vk::CommandBuffer) { cmdbuf.SetLineWidth(width); });
}
void RasterizerVulkan::UpdateCullMode(Tegra::Engines::Maxwell3D::Regs& regs) {
@@ -1174,7 +1187,7 @@ void RasterizerVulkan::UpdateCullMode(Tegra::Engines::Maxwell3D::Regs& regs) {
return;
}
scheduler.Record([enabled = regs.gl_cull_test_enabled,
cull_face = regs.gl_cull_face](vk::CommandBuffer cmdbuf) {
cull_face = regs.gl_cull_face](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
cmdbuf.SetCullModeEXT(enabled ? MaxwellToVK::CullFace(cull_face) : VK_CULL_MODE_NONE);
});
}
@@ -1188,7 +1201,7 @@ void RasterizerVulkan::UpdateDepthBoundsTestEnable(Tegra::Engines::Maxwell3D::Re
LOG_WARNING(Render_Vulkan, "Depth bounds is enabled but not supported");
enabled = false;
}
scheduler.Record([enable = enabled](vk::CommandBuffer cmdbuf) {
scheduler.Record([enable = enabled](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
cmdbuf.SetDepthBoundsTestEnableEXT(enable);
});
}
@@ -1197,36 +1210,40 @@ void RasterizerVulkan::UpdateDepthTestEnable(Tegra::Engines::Maxwell3D::Regs& re
if (!state_tracker.TouchDepthTestEnable()) {
return;
}
scheduler.Record([enable = regs.depth_test_enable](vk::CommandBuffer cmdbuf) {
cmdbuf.SetDepthTestEnableEXT(enable);
});
scheduler.Record(
[enable = regs.depth_test_enable](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
cmdbuf.SetDepthTestEnableEXT(enable);
});
}
void RasterizerVulkan::UpdateDepthWriteEnable(Tegra::Engines::Maxwell3D::Regs& regs) {
if (!state_tracker.TouchDepthWriteEnable()) {
return;
}
scheduler.Record([enable = regs.depth_write_enabled](vk::CommandBuffer cmdbuf) {
cmdbuf.SetDepthWriteEnableEXT(enable);
});
scheduler.Record(
[enable = regs.depth_write_enabled](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
cmdbuf.SetDepthWriteEnableEXT(enable);
});
}
void RasterizerVulkan::UpdatePrimitiveRestartEnable(Tegra::Engines::Maxwell3D::Regs& regs) {
if (!state_tracker.TouchPrimitiveRestartEnable()) {
return;
}
scheduler.Record([enable = regs.primitive_restart.enabled](vk::CommandBuffer cmdbuf) {
cmdbuf.SetPrimitiveRestartEnableEXT(enable);
});
scheduler.Record(
[enable = regs.primitive_restart.enabled](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
cmdbuf.SetPrimitiveRestartEnableEXT(enable);
});
}
void RasterizerVulkan::UpdateRasterizerDiscardEnable(Tegra::Engines::Maxwell3D::Regs& regs) {
if (!state_tracker.TouchRasterizerDiscardEnable()) {
return;
}
scheduler.Record([disable = regs.rasterize_enable](vk::CommandBuffer cmdbuf) {
cmdbuf.SetRasterizerDiscardEnableEXT(disable == 0);
});
scheduler.Record(
[disable = regs.rasterize_enable](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
cmdbuf.SetRasterizerDiscardEnableEXT(disable == 0);
});
}
void RasterizerVulkan::UpdateDepthBiasEnable(Tegra::Engines::Maxwell3D::Regs& regs) {
@@ -1260,15 +1277,16 @@ void RasterizerVulkan::UpdateDepthBiasEnable(Tegra::Engines::Maxwell3D::Regs& re
};
const u32 topology_index = static_cast<u32>(maxwell3d->draw_manager->GetDrawState().topology);
const u32 enable = enabled_lut[POLYGON_OFFSET_ENABLE_LUT[topology_index]];
scheduler.Record(
[enable](vk::CommandBuffer cmdbuf) { cmdbuf.SetDepthBiasEnableEXT(enable != 0); });
scheduler.Record([enable](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
cmdbuf.SetDepthBiasEnableEXT(enable != 0);
});
}
void RasterizerVulkan::UpdateLogicOpEnable(Tegra::Engines::Maxwell3D::Regs& regs) {
if (!state_tracker.TouchLogicOpEnable()) {
return;
}
scheduler.Record([enable = regs.logic_op.enable](vk::CommandBuffer cmdbuf) {
scheduler.Record([enable = regs.logic_op.enable](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
cmdbuf.SetLogicOpEnableEXT(enable != 0);
});
}
@@ -1283,15 +1301,16 @@ void RasterizerVulkan::UpdateDepthClampEnable(Tegra::Engines::Maxwell3D::Regs& r
Maxwell::ViewportClipControl::GeometryClip::FrustumXYZ ||
regs.viewport_clip_control.geometry_clip ==
Maxwell::ViewportClipControl::GeometryClip::FrustumZ);
scheduler.Record(
[is_enabled](vk::CommandBuffer cmdbuf) { cmdbuf.SetDepthClampEnableEXT(is_enabled); });
scheduler.Record([is_enabled](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
cmdbuf.SetDepthClampEnableEXT(is_enabled);
});
}
void RasterizerVulkan::UpdateDepthCompareOp(Tegra::Engines::Maxwell3D::Regs& regs) {
if (!state_tracker.TouchDepthCompareOp()) {
return;
}
scheduler.Record([func = regs.depth_test_func](vk::CommandBuffer cmdbuf) {
scheduler.Record([func = regs.depth_test_func](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
cmdbuf.SetDepthCompareOpEXT(MaxwellToVK::ComparisonOp(func));
});
}
@@ -1306,8 +1325,9 @@ void RasterizerVulkan::UpdateFrontFace(Tegra::Engines::Maxwell3D::Regs& regs) {
front_face = front_face == VK_FRONT_FACE_CLOCKWISE ? VK_FRONT_FACE_COUNTER_CLOCKWISE
: VK_FRONT_FACE_CLOCKWISE;
}
scheduler.Record(
[front_face](vk::CommandBuffer cmdbuf) { cmdbuf.SetFrontFaceEXT(front_face); });
scheduler.Record([front_face](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
cmdbuf.SetFrontFaceEXT(front_face);
});
}
void RasterizerVulkan::UpdateStencilOp(Tegra::Engines::Maxwell3D::Regs& regs) {
@@ -1325,7 +1345,7 @@ void RasterizerVulkan::UpdateStencilOp(Tegra::Engines::Maxwell3D::Regs& regs) {
const Maxwell::StencilOp::Op back_zpass = regs.stencil_back_op.zpass;
const Maxwell::ComparisonOp back_compare = regs.stencil_back_op.func;
scheduler.Record([fail, zfail, zpass, compare, back_fail, back_zfail, back_zpass,
back_compare](vk::CommandBuffer cmdbuf) {
back_compare](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
cmdbuf.SetStencilOpEXT(VK_STENCIL_FACE_FRONT_BIT, MaxwellToVK::StencilOp(fail),
MaxwellToVK::StencilOp(zpass), MaxwellToVK::StencilOp(zfail),
MaxwellToVK::ComparisonOp(compare));
@@ -1336,11 +1356,12 @@ void RasterizerVulkan::UpdateStencilOp(Tegra::Engines::Maxwell3D::Regs& regs) {
});
} else {
// Front face defines the stencil op of both faces
scheduler.Record([fail, zfail, zpass, compare](vk::CommandBuffer cmdbuf) {
cmdbuf.SetStencilOpEXT(VK_STENCIL_FACE_FRONT_AND_BACK, MaxwellToVK::StencilOp(fail),
MaxwellToVK::StencilOp(zpass), MaxwellToVK::StencilOp(zfail),
MaxwellToVK::ComparisonOp(compare));
});
scheduler.Record(
[fail, zfail, zpass, compare](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
cmdbuf.SetStencilOpEXT(VK_STENCIL_FACE_FRONT_AND_BACK, MaxwellToVK::StencilOp(fail),
MaxwellToVK::StencilOp(zpass), MaxwellToVK::StencilOp(zfail),
MaxwellToVK::ComparisonOp(compare));
});
}
}
@@ -1351,7 +1372,8 @@ void RasterizerVulkan::UpdateLogicOp(Tegra::Engines::Maxwell3D::Regs& regs) {
const auto op_value = static_cast<u32>(regs.logic_op.op);
auto op = op_value >= 0x1500 && op_value < 0x1510 ? static_cast<VkLogicOp>(op_value - 0x1500)
: VK_LOGIC_OP_NO_OP;
scheduler.Record([op](vk::CommandBuffer cmdbuf) { cmdbuf.SetLogicOpEXT(op); });
scheduler.Record(
[op](vk::CommandBuffer cmdbuf, vk::CommandBuffer) { cmdbuf.SetLogicOpEXT(op); });
}
void RasterizerVulkan::UpdateBlending(Tegra::Engines::Maxwell3D::Regs& regs) {
@@ -1377,7 +1399,7 @@ void RasterizerVulkan::UpdateBlending(Tegra::Engines::Maxwell3D::Regs& regs) {
current |= VK_COLOR_COMPONENT_A_BIT;
}
}
scheduler.Record([setup_masks](vk::CommandBuffer cmdbuf) {
scheduler.Record([setup_masks](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
cmdbuf.SetColorWriteMaskEXT(0, setup_masks);
});
}
@@ -1387,7 +1409,7 @@ void RasterizerVulkan::UpdateBlending(Tegra::Engines::Maxwell3D::Regs& regs) {
std::ranges::transform(
regs.blend.enable, setup_enables.begin(),
[&](const auto& is_enabled) { return is_enabled != 0 ? VK_TRUE : VK_FALSE; });
scheduler.Record([setup_enables](vk::CommandBuffer cmdbuf) {
scheduler.Record([setup_enables](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
cmdbuf.SetColorBlendEnableEXT(0, setup_enables);
});
}
@@ -1410,7 +1432,7 @@ void RasterizerVulkan::UpdateBlending(Tegra::Engines::Maxwell3D::Regs& regs) {
}
blend_setup(regs.blend_per_target[index]);
}
scheduler.Record([setup_blends](vk::CommandBuffer cmdbuf) {
scheduler.Record([setup_blends](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
cmdbuf.SetColorBlendEquationEXT(0, setup_blends);
});
}
@@ -1420,7 +1442,7 @@ void RasterizerVulkan::UpdateStencilTestEnable(Tegra::Engines::Maxwell3D::Regs&
if (!state_tracker.TouchStencilTestEnable()) {
return;
}
scheduler.Record([enable = regs.stencil_enable](vk::CommandBuffer cmdbuf) {
scheduler.Record([enable = regs.stencil_enable](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
cmdbuf.SetStencilTestEnableEXT(enable);
});
}
@@ -1478,7 +1500,7 @@ void RasterizerVulkan::UpdateVertexInput(Tegra::Engines::Maxwell3D::Regs& regs)
.divisor = is_instanced ? input_binding.frequency : 1,
});
}
scheduler.Record([bindings, attributes](vk::CommandBuffer cmdbuf) {
scheduler.Record([bindings, attributes](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
cmdbuf.SetVertexInputEXT(bindings, attributes);
});
}

55
src/video_core/renderer_vulkan/vk_scheduler.cpp
View File

@@ -22,11 +22,12 @@ namespace Vulkan {
MICROPROFILE_DECLARE(Vulkan_WaitForWorker);
void Scheduler::CommandChunk::ExecuteAll(vk::CommandBuffer cmdbuf) {
void Scheduler::CommandChunk::ExecuteAll(vk::CommandBuffer cmdbuf,
vk::CommandBuffer upload_cmdbuf) {
auto command = first;
while (command != nullptr) {
auto next = command->GetNext();
command->Execute(cmdbuf);
command->Execute(cmdbuf, upload_cmdbuf);
command->~Command();
command = next;
}
@@ -102,22 +103,23 @@ void Scheduler::RequestRenderpass(const Framebuffer* framebuffer) {
state.framebuffer = framebuffer_handle;
state.render_area = render_area;
Record([renderpass, framebuffer_handle, render_area](vk::CommandBuffer cmdbuf) {
const VkRenderPassBeginInfo renderpass_bi{
.sType = VK_STRUCTURE_TYPE_RENDER_PASS_BEGIN_INFO,
.pNext = nullptr,
.renderPass = renderpass,
.framebuffer = framebuffer_handle,
.renderArea =
{
.offset = {.x = 0, .y = 0},
.extent = render_area,
},
.clearValueCount = 0,
.pClearValues = nullptr,
};
cmdbuf.BeginRenderPass(renderpass_bi, VK_SUBPASS_CONTENTS_INLINE);
});
Record(
[renderpass, framebuffer_handle, render_area](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
const VkRenderPassBeginInfo renderpass_bi{
.sType = VK_STRUCTURE_TYPE_RENDER_PASS_BEGIN_INFO,
.pNext = nullptr,
.renderPass = renderpass,
.framebuffer = framebuffer_handle,
.renderArea =
{
.offset = {.x = 0, .y = 0},
.extent = render_area,
},
.clearValueCount = 0,
.pClearValues = nullptr,
};
cmdbuf.BeginRenderPass(renderpass_bi, VK_SUBPASS_CONTENTS_INLINE);
});
num_renderpass_images = framebuffer->NumImages();
renderpass_images = framebuffer->Images();
renderpass_image_ranges = framebuffer->ImageRanges();
@@ -180,7 +182,7 @@ void Scheduler::WorkerThread(std::stop_token stop_token) {
// Perform the work, tracking whether the chunk was a submission
// before executing.
const bool has_submit = work->HasSubmit();
work->ExecuteAll(current_cmdbuf);
work->ExecuteAll(current_cmdbuf, current_upload_cmdbuf);
// If the chunk was a submission, reallocate the command buffer.
if (has_submit) {
@@ -205,6 +207,13 @@ void Scheduler::AllocateWorkerCommandBuffer() {
.flags = VK_COMMAND_BUFFER_USAGE_ONE_TIME_SUBMIT_BIT,
.pInheritanceInfo = nullptr,
});
current_upload_cmdbuf = vk::CommandBuffer(command_pool->Commit(), device.GetDispatchLoader());
current_upload_cmdbuf.Begin({
.sType = VK_STRUCTURE_TYPE_COMMAND_BUFFER_BEGIN_INFO,
.pNext = nullptr,
.flags = VK_COMMAND_BUFFER_USAGE_ONE_TIME_SUBMIT_BIT,
.pInheritanceInfo = nullptr,
});
}
u64 Scheduler::SubmitExecution(VkSemaphore signal_semaphore, VkSemaphore wait_semaphore) {
@@ -212,7 +221,9 @@ u64 Scheduler::SubmitExecution(VkSemaphore signal_semaphore, VkSemaphore wait_se
InvalidateState();
const u64 signal_value = master_semaphore->NextTick();
Record([signal_semaphore, wait_semaphore, signal_value, this](vk::CommandBuffer cmdbuf) {
Record([signal_semaphore, wait_semaphore, signal_value, this](vk::CommandBuffer cmdbuf,
vk::CommandBuffer upload_cmdbuf) {
upload_cmdbuf.End();
cmdbuf.End();
if (on_submit) {
@@ -221,7 +232,7 @@ u64 Scheduler::SubmitExecution(VkSemaphore signal_semaphore, VkSemaphore wait_se
std::scoped_lock lock{submit_mutex};
switch (const VkResult result = master_semaphore->SubmitQueue(
cmdbuf, signal_semaphore, wait_semaphore, signal_value)) {
cmdbuf, upload_cmdbuf, signal_semaphore, wait_semaphore, signal_value)) {
case VK_SUCCESS:
break;
case VK_ERROR_DEVICE_LOST:
@@ -275,7 +286,7 @@ void Scheduler::EndRenderPass() {
return;
}
Record([num_images = num_renderpass_images, images = renderpass_images,
ranges = renderpass_image_ranges](vk::CommandBuffer cmdbuf) {
ranges = renderpass_image_ranges](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
std::array<VkImageMemoryBarrier, 9> barriers;
for (size_t i = 0; i < num_images; ++i) {
barriers[i] = VkImageMemoryBarrier{

9
src/video_core/renderer_vulkan/vk_scheduler.h
View File

@@ -119,7 +119,7 @@ private:
public:
virtual ~Command() = default;
virtual void Execute(vk::CommandBuffer cmdbuf) const = 0;
virtual void Execute(vk::CommandBuffer cmdbuf, vk::CommandBuffer upload_cmdbuf) const = 0;
Command* GetNext() const {
return next;
@@ -142,8 +142,8 @@ private:
TypedCommand(TypedCommand&&) = delete;
TypedCommand& operator=(TypedCommand&&) = delete;
void Execute(vk::CommandBuffer cmdbuf) const override {
command(cmdbuf);
void Execute(vk::CommandBuffer cmdbuf, vk::CommandBuffer upload_cmdbuf) const override {
command(cmdbuf, upload_cmdbuf);
}
private:
@@ -152,7 +152,7 @@ private:
class CommandChunk final {
public:
void ExecuteAll(vk::CommandBuffer cmdbuf);
void ExecuteAll(vk::CommandBuffer cmdbuf, vk::CommandBuffer upload_cmdbuf);
template <typename T>
bool Record(T& command) {
@@ -228,6 +228,7 @@ private:
VideoCommon::QueryCacheBase<QueryCacheParams>* query_cache = nullptr;
vk::CommandBuffer current_cmdbuf;
vk::CommandBuffer current_upload_cmdbuf;
std::unique_ptr<CommandChunk> chunk;
std::function<void()> on_submit;

6
src/video_core/renderer_vulkan/vk_smaa.cpp
View File

@@ -103,7 +103,7 @@ void UploadImage(const Device& device, MemoryAllocator& allocator, Scheduler& sc
}}};
scheduler.RequestOutsideRenderPassOperationContext();
scheduler.Record([&](vk::CommandBuffer cmdbuf) {
scheduler.Record([&](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
TransitionImageLayout(cmdbuf, *image, VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL,
VK_IMAGE_LAYOUT_UNDEFINED);
cmdbuf.CopyBufferToImage(*upload_buffer, *image, VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL,
@@ -672,7 +672,7 @@ void SMAA::UploadImages(Scheduler& scheduler) {
UploadImage(m_device, m_allocator, scheduler, m_static_images[Search], search_extent,
VK_FORMAT_R8_UNORM, ARRAY_TO_SPAN(searchTexBytes));
scheduler.Record([&](vk::CommandBuffer& cmdbuf) {
scheduler.Record([&](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
for (auto& images : m_dynamic_images) {
for (size_t i = 0; i < MaxDynamicImage; i++) {
ClearColorImage(cmdbuf, *images.images[i]);
@@ -707,7 +707,7 @@ VkImageView SMAA::Draw(Scheduler& scheduler, size_t image_index, VkImage source_
UpdateDescriptorSets(source_image_view, image_index);
scheduler.RequestOutsideRenderPassOperationContext();
scheduler.Record([=, this](vk::CommandBuffer& cmdbuf) {
scheduler.Record([=, this](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
TransitionImageLayout(cmdbuf, source_image, VK_IMAGE_LAYOUT_GENERAL);
TransitionImageLayout(cmdbuf, edges_image, VK_IMAGE_LAYOUT_GENERAL);
BeginRenderPass(cmdbuf, m_renderpasses[EdgeDetection], edge_detection_framebuffer,

4
src/video_core/renderer_vulkan/vk_staging_buffer_pool.h
View File

@@ -36,6 +36,10 @@ public:
StagingBufferRef Request(size_t size, MemoryUsage usage, bool deferred = false);
void FreeDeferred(StagingBufferRef& ref);
[[nodiscard]] VkBuffer StreamBuf() const noexcept {
return *stream_buffer;
}
void TickFrame();
private:

14
src/video_core/renderer_vulkan/vk_texture_cache.cpp
View File

@@ -709,7 +709,7 @@ void BlitScale(Scheduler& scheduler, VkImage src_image, VkImage dst_image, const
scheduler.RequestOutsideRenderPassOperationContext();
scheduler.Record([dst_image, src_image, extent, resources, aspect_mask, resolution, is_2d,
vk_filter, up_scaling](vk::CommandBuffer cmdbuf) {
vk_filter, up_scaling](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
const VkOffset2D src_size{
.x = static_cast<s32>(up_scaling ? extent.width : resolution.ScaleUp(extent.width)),
.y = static_cast<s32>(is_2d && up_scaling ? extent.height
@@ -955,7 +955,7 @@ void TextureCacheRuntime::ReinterpretImage(Image& dst, Image& src,
const VkImage src_image = src.Handle();
scheduler.RequestOutsideRenderPassOperationContext();
scheduler.Record([dst_image, src_image, copy_buffer, src_aspect_mask, dst_aspect_mask,
vk_in_copies, vk_out_copies](vk::CommandBuffer cmdbuf) {
vk_in_copies, vk_out_copies](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
RangedBarrierRange dst_range;
RangedBarrierRange src_range;
for (const VkBufferImageCopy& copy : vk_in_copies) {
@@ -1104,7 +1104,7 @@ void TextureCacheRuntime::BlitImage(Framebuffer* dst_framebuffer, ImageView& dst
const bool is_resolve = is_src_msaa && !is_dst_msaa;
scheduler.RequestOutsideRenderPassOperationContext();
scheduler.Record([filter, dst_region, src_region, dst_image, src_image, dst_layers, src_layers,
aspect_mask, is_resolve](vk::CommandBuffer cmdbuf) {
aspect_mask, is_resolve](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
const std::array read_barriers{
VkImageMemoryBarrier{
.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,
@@ -1240,7 +1240,8 @@ void TextureCacheRuntime::CopyImage(Image& dst, Image& src,
const VkImage dst_image = dst.Handle();
const VkImage src_image = src.Handle();
scheduler.RequestOutsideRenderPassOperationContext();
scheduler.Record([dst_image, src_image, aspect_mask, vk_copies](vk::CommandBuffer cmdbuf) {
scheduler.Record([dst_image, src_image, aspect_mask, vk_copies](vk::CommandBuffer cmdbuf,
vk::CommandBuffer) {
RangedBarrierRange dst_range;
RangedBarrierRange src_range;
for (const VkImageCopy& copy : vk_copies) {
@@ -1402,7 +1403,7 @@ void Image::UploadMemory(VkBuffer buffer, VkDeviceSize offset,
const VkImageAspectFlags vk_aspect_mask = aspect_mask;
const bool is_initialized = std::exchange(initialized, true);
scheduler->Record([src_buffer, vk_image, vk_aspect_mask, is_initialized,
vk_copies](vk::CommandBuffer cmdbuf) {
vk_copies](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
CopyBufferToImage(cmdbuf, src_buffer, vk_image, vk_aspect_mask, is_initialized, vk_copies);
});
if (is_rescaled) {
@@ -1441,7 +1442,8 @@ void Image::DownloadMemory(std::span<VkBuffer> buffers_span, std::span<VkDeviceS
}
scheduler->RequestOutsideRenderPassOperationContext();
scheduler->Record([buffers = std::move(buffers_vector), image = *original_image,
aspect_mask_ = aspect_mask, vk_copies](vk::CommandBuffer cmdbuf) {
aspect_mask_ = aspect_mask,
vk_copies](vk::CommandBuffer cmdbuf, vk::CommandBuffer) {
const VkImageMemoryBarrier read_barrier{
.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,
.pNext = nullptr,

4
src/video_core/texture_cache/slot_vector.h
View File

@@ -138,6 +138,10 @@ public:
return Iterator(this, SlotId{SlotId::INVALID_INDEX});
}
[[nodiscard]] size_t size() const noexcept {
return values_capacity - free_list.size();
}
private:
struct NonTrivialDummy {
NonTrivialDummy() noexcept {}

4
src/video_core/vulkan_common/vulkan_wrapper.h
View File

@@ -1098,6 +1098,10 @@ public:
return &handle;
}
VkCommandBuffer operator*() const noexcept {
return handle;
}
void Begin(const VkCommandBufferBeginInfo& begin_info) const {
Check(dld->vkBeginCommandBuffer(handle, &begin_info));
}