vinx13 commented on a change in pull request #9909:
URL: https://github.com/apache/tvm/pull/9909#discussion_r783441046
##########
File path: src/target/source/codegen_cuda.cc
##########
@@ -723,6 +724,41 @@ void CodeGenCUDA::VisitExpr_(const CallNode* op,
std::ostream& os) {
this->PrintExpr(op->args[i * 2 + 1], os);
os << "]" << ((i < 3) ? ", " : ")");
}
+ } else if ((op->op.same_as(builtin::call_extern()) ||
+ op->op.same_as(builtin::call_pure_extern())) &&
+ Downcast<StringImm>(op->args[0])->value == "ptx_mma") {
Review comment:
It would be better to explicitly define it as `builtin::ptx_mma` op
instead of using call extern. This would make other rewrite passes to be aware
of this op and handle it explicitly if needed
--
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.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
