projects / platform / upstream / llvm.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: a4c828a) | patch
[AArch64] More patterns to generate LD1R vector splats
authorSjoerd Meijer <smeijer@nvidia.com>
Wed, 1 Mar 2023 10:39:07 +0000 (10:39 +0000)
committerSjoerd Meijer <smeijer@nvidia.com>
Wed, 1 Mar 2023 10:48:13 +0000 (10:48 +0000)
commit2b462eb98d6777a2fb9385eea74bab9a5a6d3b95
tree8dde062ac5674c82b89444936d121f6678b4fe80tree | snapshot
parenta4c828a9e2b08c3606046aff30ebeecb558f226fcommit | diff
[AArch64] More patterns to generate LD1R vector splats

We are missing patterns to generate vector splats using LD1R. A shuffle vector
with all 0s is a vector splat if the operands are a load and undef for which
we can generate a LD1R.

Differential Revision: https://reviews.llvm.org/D145004
llvm/lib/Target/AArch64/AArch64InstrInfo.td diff | blob | history
llvm/test/CodeGen/AArch64/neon-vector-splat.ll diff | blob | history
Domain: System / Toolchain;