We read every piece of feedback, and take your input very seriously.
To see all available qualifiers, see our documentation.
1 parent 16da670 commit ed6f1b7Copy full SHA for ed6f1b7
1 file changed
lib/Dialect/TritonNvidiaGPU/Transforms/ProxyFenceInsertion.cpp
@@ -32,7 +32,8 @@ namespace {
32
33
bool isAsyncProxyWrite(Operation *op) {
34
return isa<triton::nvidia_gpu::AsyncTMACopyGlobalToLocalOp,
35
- triton::nvidia_gpu::AsyncTMAGatherOp>(op);
+ triton::nvidia_gpu::AsyncTMAGatherOp,
36
+ triton::nvidia_gpu::CLCTryCancelOp>(op);
37
}
38
39
Value getSmemDest(Operation *op) {
@@ -44,6 +45,9 @@ Value getSmemDest(Operation *op) {
44
45
dyn_cast<triton::nvidia_gpu::AsyncTMAGatherOp>(op)) {
46
return asyncTMAGatherOp.getResult();
47
48
+ if (auto clcTryCancelOp = dyn_cast<triton::nvidia_gpu::CLCTryCancelOp>(op)) {
49
+ return clcTryCancelOp.getResult();
50
+ }
51
return Value();
52
53
0 commit comments