[PATCH] D127576: [RISCV] Teach vsetvli insertion to not insert redundant vsetvli right after VLEFF/VLSEGFF.
Yeting Kuo via Phabricator via llvm-commits
llvm-commits at lists.llvm.org
Tue Jun 14 01:56:01 PDT 2022
fakepaper56 marked 3 inline comments as done.
fakepaper56 added inline comments.
================
Comment at: llvm/lib/Target/RISCV/RISCVInsertVSETVLI.cpp:327
+ // Update to the VSETVLIInfo right after MI.
+ void updateToEndStatus(const MachineInstr &MI) {
+ if (RISCV::isFaultFirstLoad(MI)) {
----------------
reames wrote:
> This should not be a member function on VSETVLIInfo. Its part of the data flow transfer rule, not the abstract state.
Done.
================
Comment at: llvm/lib/Target/RISCV/RISCVInsertVSETVLI.cpp:943
+
+ if (RISCV::isFaultFirstLoad(*DefMI)) {
+ uint64_t TSFlags = MI.getDesc().TSFlags;
----------------
reames wrote:
> This is a separate optimization, move it to its own review.
Done. And sorry the code is redundant.
================
Comment at: llvm/lib/Target/RISCV/RISCVInsertVSETVLI.cpp:1080
+
+ // We need the PHI input to be the output of a VSET(I)VLI/VLEFF/VLSEGFF
+ // and match the predecessor block.
----------------
reames wrote:
> This is a separate optimization, move it to its own review.
Done.
Repository:
rG LLVM Github Monorepo
CHANGES SINCE LAST ACTION
https://reviews.llvm.org/D127576/new/
https://reviews.llvm.org/D127576
More information about the llvm-commits
mailing list