[PATCH] D111657: [SVE][CodeGen] Enable reciprocal estimates for scalable fdiv/fsqrt
David Sherwood via Phabricator via llvm-commits
llvm-commits at lists.llvm.org
Thu Oct 21 03:05:15 PDT 2021
david-arm accepted this revision.
david-arm added a comment.
LGTM! It looks like you've addressed @paulwalker-arm's comments. I'm happy for us to look at investigating removing the `fcmeq` and `sel` instructions at a later time.
================
Comment at: llvm/lib/Target/AArch64/AArch64ISelLowering.cpp:8253
+ VT == MVT::f32 || VT == MVT::v1f32 ||
+ VT == MVT::v2f32 || VT == MVT::v4f32)) ||
+ (ST->hasSVE() &&
----------------
nit: Can you fix the formatting issue before merging please?
================
Comment at: llvm/test/CodeGen/AArch64/sve-fp-reciprocal.ll:157
+
+define <vscale x 2 x half> @fsqrt_2f16(<vscale x 2 x half> %a, <vscale x 2 x half> %b) {
+; CHECK-LABEL: fsqrt_2f16:
----------------
nit: For the fsqrt functions can you remove the second `%b` arguments as they seem to be unused?
CHANGES SINCE LAST ACTION
https://reviews.llvm.org/D111657/new/
https://reviews.llvm.org/D111657
More information about the llvm-commits
mailing list