diff --git a/src/video_core/buffer_cache/buffer_cache.h b/src/video_core/buffer_cache/buffer_cache.h index 5f5a59bba..4def8f076 100644 --- a/src/video_core/buffer_cache/buffer_cache.h +++ b/src/video_core/buffer_cache/buffer_cache.h @@ -329,6 +329,8 @@ private: [[nodiscard]] bool HasFastUniformBufferBound(size_t stage, u32 binding_index) const noexcept; + void ClearDownload(IntervalType subtract_interval); + VideoCore::RasterizerInterface& rasterizer; Tegra::Engines::Maxwell3D& maxwell3d; Tegra::Engines::KeplerCompute& kepler_compute; @@ -467,6 +469,14 @@ void BufferCache
::DownloadMemory(VAddr cpu_addr, u64 size) {
});
}
+template ::ClearDownload(IntervalType subtract_interval) {
+ uncommitted_ranges.subtract(subtract_interval);
+ for (auto& interval_set : committed_ranges) {
+ interval_set.subtract(subtract_interval);
+ }
+}
+
template ::DMACopy(GPUVAddr src_address, GPUVAddr dest_address, u64 amount) {
const std::optional ::DMACopy(GPUVAddr src_address, GPUVAddr dest_address, u64 am
}
const IntervalType subtract_interval{*cpu_dest_address, *cpu_dest_address + amount};
- uncommitted_ranges.subtract(subtract_interval);
- for (auto& interval_set : committed_ranges) {
- interval_set.subtract(subtract_interval);
- }
+ ClearDownload(subtract_interval);
BufferId buffer_a;
BufferId buffer_b;
@@ -496,7 +503,6 @@ bool BufferCache ::DMACopy(GPUVAddr src_address, GPUVAddr dest_address, u64 am
auto& src_buffer = slot_buffers[buffer_a];
auto& dest_buffer = slot_buffers[buffer_b];
SynchronizeBuffer(src_buffer, *cpu_src_address, static_cast ::DMACopy(GPUVAddr src_address, GPUVAddr dest_address, u64 am
ForEachWrittenRange(*cpu_src_address, amount, mirror);
// This subtraction in this order is important for overlapping copies.
common_ranges.subtract(subtract_interval);
+ bool atleast_1_download = tmp_intervals.size() != 0;
for (const IntervalType add_interval : tmp_intervals) {
common_ranges.add(add_interval);
}
+ if (dest_buffer.HasCachedWrites()) {
+ dest_buffer.FlushCachedWrites();
+ }
runtime.CopyBuffer(dest_buffer, src_buffer, copies);
- if (IsRegionGpuModified(*cpu_src_address, amount)) {
+ dest_buffer.UnmarkRegionAsCpuModified(*cpu_dest_address, amount);
+ if (atleast_1_download) {
dest_buffer.MarkRegionAsGpuModified(*cpu_dest_address, amount);
}
std::vector ::DMAClear(GPUVAddr dst_address, u64 amount, u32 value) {
}
const IntervalType subtract_interval{*cpu_dst_address, *cpu_dst_address + amount * sizeof(u32)};
- uncommitted_ranges.subtract(subtract_interval);
- for (auto& interval_set : committed_ranges) {
- interval_set.subtract(subtract_interval);
- }
+ ClearDownload(subtract_interval);
common_ranges.subtract(subtract_interval);
const size_t size = amount * sizeof(u32);
@@ -557,6 +565,7 @@ bool BufferCache ::DMAClear(GPUVAddr dst_address, u64 amount, u32 value) {
auto& dest_buffer = slot_buffers[buffer];
const u32 offset = static_cast ::DownloadBufferMemory(Buffer& buffer, VAddr cpu_addr, u64 si
const VAddr end_address = start_address + range_size;
ForEachWrittenRange(start_address, range_size, add_download);
const IntervalType subtract_interval{start_address, end_address};
+ ClearDownload(subtract_interval);
common_ranges.subtract(subtract_interval);
});
if (total_size_bytes == 0) {