hipnosuer
/
Zuma
forked from ShuriZma/suyu
1
0
Fork 0
Code Pull Requests Activity

buffer_cache: Split CreateBuffer in separate functions 

Allow adding functionality to each function without making CreateBuffer
more complex.
This commit is contained in:
ReinUsesLisp 2021-01-18 00:56:59 -03:00
parent a02b4e1df6
commit ec9354d6d9
1 changed files with 52 additions and 29 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

81
src/video_core/buffer_cache/buffer_cache.h
View File

@ -71,6 +71,12 @@ class BufferCache {
struct Empty {}; struct Empty {};
struct OverlapResult {
std::vector<BufferId> ids;
VAddr begin;
VAddr end;
};
struct Binding { struct Binding {
VAddr cpu_addr{}; VAddr cpu_addr{};
u32 size{}; u32 size{};
@ -220,6 +226,10 @@ private:
[[nodiscard]] BufferId FindBuffer(VAddr cpu_addr, u32 size); [[nodiscard]] BufferId FindBuffer(VAddr cpu_addr, u32 size);
[[nodiscard]] OverlapResult ResolveOverlaps(VAddr cpu_addr, u32 wanted_size);
void JoinOverlap(BufferId new_buffer_id, BufferId overlap_id);
[[nodiscard]] BufferId CreateBuffer(VAddr cpu_addr, u32 wanted_size); [[nodiscard]] BufferId CreateBuffer(VAddr cpu_addr, u32 wanted_size);
void Register(BufferId buffer_id); void Register(BufferId buffer_id);
@ -988,12 +998,12 @@ BufferId BufferCache<P>::FindBuffer(VAddr cpu_addr, u32 size) {
} }
template <class P> template <class P>
BufferId BufferCache<P>::CreateBuffer(VAddr cpu_addr, u32 wanted_size) { typename BufferCache<P>::OverlapResult BufferCache<P>::ResolveOverlaps(VAddr cpu_addr,
u32 wanted_size) {
std::vector<BufferId> overlap_ids; std::vector<BufferId> overlap_ids;
VAddr cpu_addr_begin = cpu_addr; VAddr begin = cpu_addr;
VAddr cpu_addr_end = cpu_addr + wanted_size; VAddr end = cpu_addr + wanted_size;
for (; cpu_addr >> PAGE_BITS < Common::DivCeil(cpu_addr_end, PAGE_SIZE); for (; cpu_addr >> PAGE_BITS < Common::DivCeil(end, PAGE_SIZE); cpu_addr += PAGE_SIZE) {
cpu_addr += PAGE_SIZE) {
const BufferId overlap_id = page_table[cpu_addr >> PAGE_BITS]; const BufferId overlap_id = page_table[cpu_addr >> PAGE_BITS];
if (!overlap_id) { if (!overlap_id) {
continue; continue;
@ -1005,35 +1015,48 @@ BufferId BufferCache<P>::CreateBuffer(VAddr cpu_addr, u32 wanted_size) {
overlap.Pick(); overlap.Pick();
overlap_ids.push_back(overlap_id); overlap_ids.push_back(overlap_id);
const VAddr overlap_cpu_addr = overlap.CpuAddr(); const VAddr overlap_cpu_addr = overlap.CpuAddr();
if (overlap_cpu_addr < cpu_addr_begin) { if (overlap_cpu_addr < begin) {
cpu_addr = cpu_addr_begin = overlap_cpu_addr; cpu_addr = begin = overlap_cpu_addr;
} }
cpu_addr_end = std::max(cpu_addr_end, overlap_cpu_addr + overlap.SizeBytes()); end = std::max(end, overlap_cpu_addr + overlap.SizeBytes());
} }
const u32 size = static_cast<u32>(cpu_addr_end - cpu_addr_begin); return OverlapResult{
const BufferId new_buffer_id = slot_buffers.insert(runtime, rasterizer, cpu_addr_begin, size); .ids = std::move(overlap_ids),
.begin = begin,
.end = end,
};
}
template <class P>
void BufferCache<P>::JoinOverlap(BufferId new_buffer_id, BufferId overlap_id) {
Buffer& new_buffer = slot_buffers[new_buffer_id]; Buffer& new_buffer = slot_buffers[new_buffer_id];
Buffer& overlap = slot_buffers[overlap_id];
for (const BufferId overlap_id : overlap_ids) { std::vector<BufferCopy> copies;
Buffer& overlap = slot_buffers[overlap_id]; const size_t dst_base_offset = overlap.CpuAddr() - new_buffer.CpuAddr();
overlap.Unpick(); overlap.ForEachDownloadRange([&](u64 begin, u64 range_size) {
copies.push_back(BufferCopy{
std::vector<BufferCopy> copies; .src_offset = begin,
const size_t dst_base_offset = overlap.CpuAddr() - new_buffer.CpuAddr(); .dst_offset = dst_base_offset + begin,
overlap.ForEachDownloadRange([&](u64 begin, u64 range_size) { .size = range_size,
copies.push_back(BufferCopy{
.src_offset = begin,
.dst_offset = dst_base_offset + begin,
.size = range_size,
});
new_buffer.UnmarkRegionAsCpuModified(begin, range_size);
new_buffer.MarkRegionAsGpuModified(begin, range_size);
}); });
if (!copies.empty()) { new_buffer.UnmarkRegionAsCpuModified(begin, range_size);
runtime.CopyBuffer(slot_buffers[new_buffer_id], overlap, copies); new_buffer.MarkRegionAsGpuModified(begin, range_size);
} });
ReplaceBufferDownloads(overlap_id, new_buffer_id); if (!copies.empty()) {
DeleteBuffer(overlap_id); runtime.CopyBuffer(slot_buffers[new_buffer_id], overlap, copies);
}
ReplaceBufferDownloads(overlap_id, new_buffer_id);
DeleteBuffer(overlap_id);
}
template <class P>
BufferId BufferCache<P>::CreateBuffer(VAddr cpu_addr, u32 wanted_size) {
const OverlapResult overlap = ResolveOverlaps(cpu_addr, 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);
for (const BufferId overlap_id : overlap.ids) {
JoinOverlap(new_buffer_id, overlap_id);
} }
Register(new_buffer_id); Register(new_buffer_id);
return new_buffer_id; return new_buffer_id;