[PATCH] D70253: [AArch64][SVE2] Implement remaining SVE2 floating-point intrinsics

[Eli Friedman via Phabricator via cfe-commits]

efriedma added inline comments.


================
Comment at: llvm/include/llvm/IR/IntrinsicsAArch64.td:898
+                 llvm_i32_ty],
+                [IntrNoMem]>;
+
----------------
kmclaughlin wrote:
> sdesmalen wrote:
> > I'd expect the `llvm_i32_ty` to be an immediate for these instructions, right? If so you'll need to add `ImmArg<OpNo>`  to the list of properties.
> > 
> Thanks for taking a look at this :) I tried your suggestion of adding ImmAr<Op> to the list of properties here but had some problems with it (i.e. Cannot select: intrinsic %llvm.aarch64.sve.fmlalb.lane). I don't think this is too much of an issue here as we have additional checks on the immediate with VectorIndexH32b, which ensures the immediate is in the correct range.
The point of immarg markings isn't to assist the backend; it's to ensure IR optimizations don't break your intrinsic calls.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D70253/new/

https://reviews.llvm.org/D70253