Mirrors/shadPS4
1
0
Fork 0
mirror of https://github.com/shadps4-emu/shadPS4.git synced 2025-07-12 04:35:56 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 3

Single pass on mapped ranges (no barrier batching)

Browse source
This commit is contained in:
Lander Gallastegi 2025-07-07 21:59:03 +02:00
parent 5d2598c505
commit 17815ad439
1 changed files with 34 additions and 35 deletions
Download patch file Download diff file Expand all files Collapse all files

43
src/video_core/buffer_cache/buffer_cache.cpp
View file

@ -997,53 +997,52 @@ void BufferCache::SynchronizeBuffersInRange(VAddr device_addr, u64 size) {
void BufferCache::SynchronizeBuffersForDma() {
RENDERER_TRACE;
LOG_WARNING(Render_Vulkan, "SYNC RANGES FOR DMA");
boost::container::small_vector<Buffer*, 64> buffers;
boost::container::small_vector<vk::BufferMemoryBarrier2, 64> barriers;
boost::container::small_vector<vk::BufferCopy, 4> copies;
const auto& mapped_ranges = rasterizer.GetMappedRanges();
bool barrier_recorded = false;
memory_tracker->Lock();
scheduler.EndRendering();
const auto cmdbuf = scheduler.CommandBuffer();
mapped_ranges.ForEach([&](VAddr device_addr, u64 size) {
ForEachBufferInRange(device_addr, size, [&](BufferId buffer_id, Buffer& buffer) {
if (memory_tracker->IsRegionCpuModified<false>(device_addr, size)) {
barriers.push_back(vk::BufferMemoryBarrier2{
memory_tracker->ForEachUploadRange<true, false>(
buffer.CpuAddr(), buffer.SizeBytes(), false,
[&](u64 device_addr_out, u64 range_size) {
if (!barrier_recorded) {
barrier_recorded = true;
const vk::BufferMemoryBarrier2 barrier = {
.srcStageMask = vk::PipelineStageFlagBits2::eAllCommands,
.srcAccessMask =
vk::AccessFlagBits2::eMemoryRead | vk::AccessFlagBits2::eMemoryWrite |
vk::AccessFlagBits2::eTransferRead | vk::AccessFlagBits2::eTransferWrite,
.srcAccessMask = vk::AccessFlagBits2::eMemoryRead |
vk::AccessFlagBits2::eMemoryWrite |
vk::AccessFlagBits2::eTransferRead |
vk::AccessFlagBits2::eTransferWrite,
.dstStageMask = vk::PipelineStageFlagBits2::eTransfer,
.dstAccessMask = vk::AccessFlagBits2::eTransferWrite,
.buffer = buffer.Handle(),
.offset = 0,
.size = buffer.SizeBytes(),
});
buffers.push_back(&buffer);
}
});
});
};
cmdbuf.pipelineBarrier2(vk::DependencyInfo{
.dependencyFlags = vk::DependencyFlagBits::eByRegion,
.bufferMemoryBarrierCount = static_cast<u32>(barriers.size()),
.pBufferMemoryBarriers = barriers.data(),
.bufferMemoryBarrierCount = 1,
.pBufferMemoryBarriers = &barrier,
});
for (auto* buffer : buffers) {
memory_tracker->ForEachUploadRange<true, false>(
buffer->CpuAddr(), buffer->SizeBytes(), false,
[&](u64 device_addr_out, u64 range_size) {
}
const u64 offset = staging_buffer.Copy(device_addr_out, range_size);
copies.push_back(vk::BufferCopy{
.srcOffset = offset,
.dstOffset = device_addr_out - buffer->CpuAddr(),
.dstOffset = device_addr_out - buffer.CpuAddr(),
.size = range_size,
});
});
cmdbuf.copyBuffer(staging_buffer.Handle(), buffer->Handle(), copies);
cmdbuf.copyBuffer(staging_buffer.Handle(), buffer.Handle(), copies);
copies.clear();
}
MemoryBarrier();
barrier_recorded = false;
});
});
memory_tracker->PerformDeferredProtections<Type::CPU, false, false>();
MemoryBarrier();
memory_tracker->Unlock();
}