[clang] [llvm] [AMDGPU] Restrict `@llvm.amdgcn.image.*` intrinsic's dmask argument (PR #179511)
Alexey Sachkov via cfe-commits
cfe-commits at lists.llvm.org
Fri Feb 20 06:49:01 PST 2026
================
@@ -7151,6 +7151,561 @@ void Verifier::visitIntrinsicCall(Intrinsic::ID ID, CallBase &Call) {
&Call, Op);
break;
}
+ case Intrinsic::amdgcn_image_load_1d:
----------------
AlexeySachkov wrote:
This list is enormously huge which I don't like and I'm looking for feedback from reviewers:
- there is a searchable table for those intrinsics which we can use to detect them, but that will mean a runtime lookup instead of `switch...case`
- we could generate some `.def`/`.inc` file to outline all those `case Intrinsic::...` in there
- we could create a separate function (or even file) to outline all `amdgcn` intrinsics check in there so that they do not pollute the existing `switch...case` as much
https://github.com/llvm/llvm-project/pull/179511
More information about the cfe-commits
mailing list