@@ -444,7 +444,6 @@ struct GroupedProblemVisitor<ProblemSizeHelper,
|
||||
private:
|
||||
CUTLASS_DEVICE
|
||||
void prefetch_tiles() {
|
||||
// TODO: Consider changing to use async copies from global to shared mem
|
||||
CUTLASS_PRAGMA_UNROLL
|
||||
for (int32_t i = 0; i < kPrefetchTileCount; i += kThreadCount) {
|
||||
int32_t offset = threadIdx.x + i;
|
||||
|
||||
Reference in New Issue
Block a user