[llvm] [LLVM] treat `@llvm.ssub.sat` the same as `@llvm.aarch64.neon.sqsub` (PR #140454)

[David Green via llvm-commits]

================
@@ -1795,6 +1795,14 @@ bool AArch64LegalizerInfo::legalizeIntrinsic(LegalizerHelper &Helper,
     MI.eraseFromParent();
     return true;
   }
+  case Intrinsic::aarch64_neon_sqadd:
+    return LowerBinOp(TargetOpcode::G_SADDSAT);
+  case Intrinsic::aarch64_neon_sqsub:
+    return LowerBinOp(TargetOpcode::G_SSUBSAT);
+  case Intrinsic::aarch64_neon_uqadd:
+    return LowerBinOp(TargetOpcode::G_UADDSAT);
+  case Intrinsic::aarch64_neon_uqsub:
+    return LowerBinOp(TargetOpcode::G_USUBSAT);
----------------
davemgreen wrote:

Yeah I think that sounds OK. Can you make it similar to the SDAG variants where it checks for isVector() before doing the conversion? (I'm not sure what the scalar intrinsics will do at the moment in gisel, but lets get the vector variants done first).

You might need to update the tests too.

https://github.com/llvm/llvm-project/pull/140454