LudovicoYIN commented on code in PR #18879:
URL: https://github.com/apache/tvm/pull/18879#discussion_r2893657587
##########
include/tvm/tir/op.h:
##########
@@ -735,11 +741,17 @@ TVM_DLL PrimExpr fast_erf_float_expr(PrimExpr arg, int
bits);
} \
}
+#define TVM_DECLARE_INTRIN_UNARY(OpName) \
+ inline PrimExpr OpName(PrimExpr x, Span span = Span()) { \
+ static const Op& op = Op::Get("tir." #OpName); \
+ return tir::Call(x.dtype(), op, {x}, span); \
+ }
Review Comment:
I fixed this by restoring the existing `bfloat16` handling for the non-trig
unary math ops, and keeping the new dtype rejection only for the trig-style ops
in this PR. I also added a regression test for the preserved `bfloat16` path.
--
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]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
