SoftGPU: Fix block transfer and self-render overlap tracking

- The block transfer overlap check passed the stride in pixels where bytes
  are expected, so it only covered part of the rectangle.
- A selfrender/selfdepth flush in UpdateState dropped the current draw's
  pending writes and reads, so later transfers didn't wait for it.

Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>

GPU/Software/BinManager.cpp

@@ -246,21 +246,32 @@ void BinManager::UpdateState() {
 		if (newMaxTasks > MAX_POSSIBLE_TASKS)
 			newMaxTasks = MAX_POSSIBLE_TASKS;
 		// We don't want to overlap wrong, so flush any pending.
+		bool flushed = false;
 		if (maxTasks_ != newMaxTasks) {
 			maxTasks_ = newMaxTasks;
 			Flush("selfrender");
+			flushed = true;
 		}
-		pendingOverlap_ = pendingOverlap_ || selfRender;
 
 		// Lastly, we have to check if we're newly writing depth we were texturing before.
 		// This happens in Call of Duty (depth clear after depth texture), for example.
-		if (!hadDepth && state.pixelID.depthWrite) {
+		if (!flushed && !hadDepth && state.pixelID.depthWrite) {
 			for (size_t i = 0; i < states_.Size(); ++i) {
 				if (HasTextureWrite(states_.Peek(i))) {
 					Flush("selfdepth");
+					flushed = true;
+					break;
 				}
 			}
 		}
+
+		if (flushed) {
+			// The flush forgot what this draw writes and reads, so record it again.
+			MarkPendingWrites(state);
+			MarkPendingReads(state);
+			ClearDirty(SoftDirty::BINNER_RANGE);
+		}
+		pendingOverlap_ = pendingOverlap_ || selfRender;
 		ClearDirty(SoftDirty::BINNER_OVERLAP);
 	}
 }

GPU/Software/SoftGpu.cpp

@@ -826,8 +826,8 @@ void SoftGPU::Execute_BlockTransferStart(u32 op, u32 diff) {
 
 	// Need to flush both source and target, so we overwrite properly.
 	if (Memory::IsValidRange(src, srcSize) && Memory::IsValidRange(dst, dstSize)) {
-		drawEngine_->transformUnit.FlushIfOverlap(this, "blockxfer", false, src, srcStride, width * bpp, height);
-		drawEngine_->transformUnit.FlushIfOverlap(this, "blockxfer", true, dst, dstStride, width * bpp, height);
+		drawEngine_->transformUnit.FlushIfOverlap(this, "blockxfer", false, src, srcStride * bpp, width * bpp, height);
+		drawEngine_->transformUnit.FlushIfOverlap(this, "blockxfer", true, dst, dstStride * bpp, width * bpp, height);
 	} else {
 		drawEngine_->transformUnit.Flush(this, "blockxfer_wrap");
 	}