[PATCH] D150958: [SME2/SVE2p1] Change psel intrinsic such that the result/first operand are not overloaded.
Sander de Smalen via Phabricator via llvm-commits
llvm-commits at lists.llvm.org
Mon May 22 06:28:29 PDT 2023
sdesmalen added inline comments.
================
Comment at: llvm/lib/Target/AArch64/SMEInstrFormats.td:1326
let AddedComplexity = 1 in {
- def : Pat<(nxv16i1 (op (nxv16i1 PPRAny:$Pn), (nxv16i1 PPRAny:$Pm),
+ def : Pat<(nxv16i1 (op (nxv16i1 PPRAny:$Pn), (nxv16i1 PPR8:$Pm),
(i32 (tileslice8 MatrixIndexGPR32Op12_15:$idx, sme_elm_idx0_15:$imm)))),
----------------
CarolineConcatto wrote:
> Can you change this to be PPR8?
The instruction itself uses PPRAny (on purpose, because it should print e.g. `p0` instead of `p0.b`). It probably won't make any difference for the pattern, because PPR8 and PPRAny use the same register class, but still I'd rather stay aligned with the instruction definition.
Repository:
rG LLVM Github Monorepo
CHANGES SINCE LAST ACTION
https://reviews.llvm.org/D150958/new/
https://reviews.llvm.org/D150958
More information about the llvm-commits
mailing list