[llvm] [AArch64][SVE] Lower scalar FP converts to SVE when Neon is unavailable (PR #112564)

Paul Walker via llvm-commits llvm-commits at lists.llvm.org
Tue Oct 22 10:07:35 PDT 2024


================
@@ -1454,8 +1454,12 @@ AArch64TargetLowering::AArch64TargetLowering(const TargetMachine &TM,
       setOperationAction(ISD::INSERT_SUBVECTOR, VT, Custom);
       setOperationAction(ISD::UINT_TO_FP, VT, Custom);
       setOperationAction(ISD::SINT_TO_FP, VT, Custom);
+      setOperationAction(ISD::STRICT_UINT_TO_FP, VT, Custom);
+      setOperationAction(ISD::STRICT_SINT_TO_FP, VT, Custom);
----------------
paulwalker-arm wrote:

Is there a requirement to change the behaviour of the `ISD::STRICT_` nodes?  If not then I'd rather drop them from this PR because SVE support for `strict_fp` is not really a thing yet so I'd rather not have partial support.  Plus I'm pretty sure Streaming SVE disables most all trapping math support anyway.

https://github.com/llvm/llvm-project/pull/112564


More information about the llvm-commits mailing list