Jiawei-Shao commented on code in PR #16976:
URL: https://github.com/apache/tvm/pull/16976#discussion_r1596443825
##########
src/target/source/intrin_rule_webgpu.cc:
##########
@@ -113,6 +113,8 @@ TVM_REGISTER_OP("tir.trunc")
// extra dispatch
TVM_REGISTER_OP("tir.erf").set_attr<FLowerIntrinsic>("webgpu.FLowerIntrinsic",
DispatchFastErf);
+TVM_REGISTER_OP("tir.dot4I8Packed").set_attr<FLowerIntrinsic>("webgpu.FLowerIntrinsic",
DispatchPureExtern<Direct>);
Review Comment:
Oh I find there is no `tir.dp4a` in TVM right now, and I see in TVM `dp4a`
is all called through `call_pure_extern()`:
[vulkan](https://github.com/apache/tvm/blob/main/src/target/spirv/codegen_spirv.cc#L382)
[cuda](https://github.com/apache/tvm/blob/main/python/tvm/topi/cuda/tensor_intrin.py#L94)
Do you mean we add `tir.dp4a` in TVM or still support `dp4a` as a pure
external call like what `dp4a` is supported in `codegen_spirv.cc`?
--
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]
