eden-emu/eden
24
3
Fork 15
Code Issues 138 Pull requests 56 Projects 1 Releases Packages Wiki Activity Actions

[VK] BufferCache fixes not to be merged yet just for benchmarks
All checks were successful
eden-license / license-header (pull_request) Successful in 26s
Details

Browse source 
DMAClear now treats amount as bytes and marks the region GPU-modified (not subtracting, one of the reasons why need benchmarking].
The page-scanner in IsRegionRegistered() now advances with buf_end_addr so there should be no infinite loop now.
Overlap growth now expands left with expand_begin and right with expand_end (should stop stream buffers thrashing).
Also Skipping zero-fill on new buffers to save bandwidth and is probably safe since we are doing sync in upload paths before GPU reads.
This commit is contained in:
wildcard 2025-08-14 00:18:13 +02:00 • committed by crueter
parent e28b0d2590
commit 4de2faf940
1 changed files with 12 additions and 12 deletions
Download patch file Download diff file Expand all files Collapse all files

24
src/video_core/buffer_cache/buffer_cache.h
View file

@ -249,22 +249,22 @@ bool BufferCache<P>::DMACopy(GPUVAddr src_address, GPUVAddr dest_address, u64 am
template <class P>
bool BufferCache<P>::DMAClear(GPUVAddr dst_address, u64 amount, u32 value) {
const std::optional<DAddr> cpu_dst_address = gpu_memory->GpuToCpuAddress(dst_address);
const auto cpu_dst_address = gpu_memory->GpuToCpuAddress(dst_address);
if (!cpu_dst_address) {
return false;
}
const bool dest_dirty = IsRegionRegistered(*cpu_dst_address, amount);
if (!dest_dirty) {
const size_t size = static_cast<size_t>(amount);// amount should be BYTES (same as DMACopy)
if (!IsRegionRegistered(*cpu_dst_address, size)) {
return false;
}
const size_t size = amount * sizeof(u32);
ClearDownload(*cpu_dst_address, size);
gpu_modified_ranges.Subtract(*cpu_dst_address, size);
ClearDownload(*cpu_dst_address, size);// cancel pending downloads
const BufferId buffer = FindBuffer(*cpu_dst_address, static_cast<u32>(size));
Buffer& dest_buffer = slot_buffers[buffer];
const u32 offset = dest_buffer.Offset(*cpu_dst_address);
// GPU is writing here(dma clear is a gpu write not cpu so dont subtract(ifyify), mark as GPU-modified (same as DMACopy and MarkWrittenBuffer)
MarkWrittenBuffer(buffer, *cpu_dst_address, static_cast<u32>(size)); // adds + tracks uncommitted
runtime.ClearBuffer(dest_buffer, offset, size, value);
dest_buffer.MarkUsage(offset, size);
return true;
@ -671,7 +671,7 @@ bool BufferCache<P>::IsRegionRegistered(DAddr addr, size_t size) {
if (buf_start_addr < end_addr && addr < buf_end_addr) {
return true;
}
page = Common::DivCeil(end_addr, CACHING_PAGESIZE);
page = Common::DivCeil(buf_end_addr, CACHING_PAGESIZE); //fix the scanning
}
return false;
}
@ -1349,10 +1349,10 @@ typename BufferCache<P>::OverlapResult BufferCache<P>::ResolveOverlaps(DAddr dev
// as a stream buffer. Increase the size to skip constantly recreating buffers.
has_stream_leap = true;
if (expands_right) {
expand_begin(CACHING_PAGESIZE * 128);
expand_end(CACHING_PAGESIZE * 128); //previous logic seemed inverse so this fixes it
}
if (expands_left) {
expand_end(CACHING_PAGESIZE * 128);
expand_begin(CACHING_PAGESIZE * 128); //same as above
}
}
}
@ -1393,9 +1393,9 @@ BufferId BufferCache<P>::CreateBuffer(DAddr device_addr, u32 wanted_size) {
const u32 size = static_cast<u32>(overlap.end - overlap.begin);
const BufferId new_buffer_id = slot_buffers.insert(runtime, overlap.begin, size);
auto& new_buffer = slot_buffers[new_buffer_id];
const size_t size_bytes = new_buffer.SizeBytes();
runtime.ClearBuffer(new_buffer, 0, size_bytes, 0);
new_buffer.MarkUsage(0, size_bytes);
//const size_t size_bytes = new_buffer.SizeBytes(); vulkan does not require new buffers be filed with zeros save some bandwidth
//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);
}