[PATCH] D133808: [RISCV][InsertVSETVLI] Operations with no tail elements are tail agnostic
Craig Topper via Phabricator via llvm-commits
llvm-commits at lists.llvm.org
Tue Sep 13 18:52:58 PDT 2022
craig.topper added inline comments.
================
Comment at: llvm/lib/Target/RISCV/RISCVInsertVSETVLI.cpp:833
int64_t Imm = VLOp.getImm();
// Conver the VLMax sentintel to X0 register.
+ if (Imm == RISCV::VLMaxSentinel) {
----------------
Can fix the spelling of "sentinel" while you're here?
================
Comment at: llvm/lib/Target/RISCV/RISCVInsertVSETVLI.cpp:837
+ // No tail elements --> tail agnostic. Note that vmv.s.x and
+ // vmv.s.f define tail elements differently, and thus VL=VLMAX
+ // does not imply tail agnostic for those.
----------------
vmv.s.f -> vfmv.s.f
================
Comment at: llvm/lib/Target/RISCV/RISCVInsertVSETVLI.cpp:840
+ // TODO: move this to e.g. a dag combine?
+ TailAgnostic |= !isScalarMoveInstr(MI);
+ } else
----------------
reductions and vcompress also define tail elements differently.
Repository:
rG LLVM Github Monorepo
CHANGES SINCE LAST ACTION
https://reviews.llvm.org/D133808/new/
https://reviews.llvm.org/D133808
More information about the llvm-commits
mailing list