[PATCH] D118917: [AArch64][NFC] Remove call to useSVEForFixedLengthVectors in useSVEForFixedLengthVectorVT

[David Sherwood via Phabricator via llvm-commits]

david-arm created this revision.
david-arm added reviewers: sdesmalen, paulwalker-arm, peterwaller-arm.
Herald added subscribers: ctetreau, hiraditya, kristof.beyls, tschuett.
david-arm requested review of this revision.
Herald added a project: LLVM.
Herald added a subscriber: llvm-commits.

We want to remove the call to useSVEForFixedLengthVectors in
useSVEForFixedLengthVectorVT and instead call it explicitly at
the callsite of useSVEForFixedLengthVectorVT. This gives us more
flexibility to decide when to use SVE for fixed-length vectors,
since useSVEForFixedLengthVectors only returns true if the minimum
SVE vector length is >= 256 bits. There are times where we actually
do want to use SVE to lower the operation even when the min SVE
length is 128 bits, since SVE can be used for certain operations
that are missing from the NEON architecture, e.g. 64-bit integer
vector multiplies.

This patch was created in response to comments on D118802 <https://reviews.llvm.org/D118802>.


Repository:
  rG LLVM Github Monorepo

https://reviews.llvm.org/D118917

Files:
  llvm/lib/Target/AArch64/AArch64ISelLowering.cpp

-------------- next part --------------
A non-text attachment was scrubbed...
Name: D118917.405671.patch
Type: text/x-patch
Size: 16191 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20220203/2f342478/attachment.bin>