mirror of
https://github.com/hrydgard/ppsspp.git
synced 2026-10-22 02:39:05 +02:00
SoftGPU: Don't count a block ending at the stride as reaching the next row
PendingWriteIn took a block that ends exactly at the stride for one running into the next row. Ridge Racer keeps its CLUTs in the framebuffer's stride gap (the last 64 bytes of 1024-byte rows, past the 960 drawn), so every CLUT load flushed all queued drawing. Ridge Racer 8% faster. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
This commit is contained in:
1 parent
284934ff23
commit
60be0cde40
1 file changed
+2
-2
@@ -932,8 +932,8 @@ bool BinManager::PendingWriteIn(const BinDirtyRange &range, uint32_t start, uint
|
||||
int32_t rangeY = offset / (int32_t)range.strideBytes;
|
||||
uint32_t rangeX = offset % (int32_t)range.strideBytes;
|
||||
if (rangeY >= 0 && (uint32_t)rangeY < range.height) {
|
||||
// If this row is either within width, or extends beyond stride, overlap.
|
||||
if (rangeX < range.widthBytes || rangeX + w >= range.strideBytes)
|
||||
// If this row is either within width, or extends beyond stride (into the next row), overlap.
|
||||
if (rangeX < range.widthBytes || rangeX + w > range.strideBytes)
|
||||
return true;
|
||||
}
|
||||
|
||||
|
||||
Reference in new issue
Block a user