Hzfengsy commented on a change in pull request #4353: [Perf] Enhance cudnn and
cublas backend and enable TensorCore
URL: https://github.com/apache/incubator-tvm/pull/4353#discussion_r347074084
##########
File path: src/runtime/contrib/cublas/cublas.cc
##########
@@ -36,6 +36,36 @@ inline cublasOperation_t BooleanToTranspose(bool item) {
return item ? CUBLAS_OP_T : CUBLAS_OP_N;
}
+inline void TryEnableTensorCore(cublasHandle_t hdl) {
+ // TensorCores are only supported in cublas 9.0 or higher
Review comment:
Cublas and CuDNN will check it. If we try to use TensorCores on
non-TensorCore GPUs, it will fallback to the traditional workload. So I think
there is no need to duplicate this check in tvm.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
With regards,
Apache Git Services
