[PATCH] D108460: [AArch64] Lower fpto*i.sat intrinsics for NEON.

Sun Aug 22 23:50:54 PDT 2021

dmgreen added inline comments.

================
Comment at: llvm/lib/Target/AArch64/AArch64ISelLowering.cpp:3396
+  //
+  // Warning: We maintain cost tables in AArch64TargetTransformInfo.cpp.
+  // Any additional optimization in this function should be recorded
----------------
I don't think this "Warning" adds much. It's OK to remove for now.

================
Comment at: llvm/lib/Target/AArch64/AArch64TargetTransformInfo.cpp:1084
+    // LowerVectorFP_TO_INT_SAT
+    { ISD::FP_TO_SINT_SAT, MVT::v2i32, MVT::v2f32, 1 },
+    { ISD::FP_TO_SINT_SAT, MVT::v4i32, MVT::v4f32, 1 },
----------------
This would need tests, and is probably best left for a separate commit. I think there would need to be more general FP_TO_SINT_SAT costing changes, it won't ever get here at the moment as they are intrinsics not cast instructions.

================
Comment at: llvm/test/CodeGen/AArch64/fptosi-sat-vector.ll:2
 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc -mtriple=aarch64 < %s | FileCheck %s

----------------
Can you add a run line with fp16 too.

Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D108460/new/

https://reviews.llvm.org/D108460