diff options
| author | Fabian Ritter <fabian.ritter@amd.com> | 2025-05-27 11:30:51 +0200 |
|---|---|---|
| committer | GitHub <noreply@github.com> | 2025-05-27 11:30:51 +0200 |
| commit | fb27867bd5a04975516b099681587ea32f38f3d4 (patch) | |
| tree | e9644e1601af44479290d03d0c881e37a27ea851 /llvm/lib/Transforms/Utils/LoopPeel.cpp | |
| parent | bf6cd24aaadd8310408d733c944f0d5f9cc14c0a (diff) | |
| download | llvm-fb27867bd5a04975516b099681587ea32f38f3d4.zip llvm-fb27867bd5a04975516b099681587ea32f38f3d4.tar.gz llvm-fb27867bd5a04975516b099681587ea32f38f3d4.tar.bz2 | |
[AMDGPU] SIFoldOperands: Delay foldCopyToVGPROfScalarAddOfFrameIndex (#141558)
foldCopyToVGPROfScalarAddOfFrameIndex transforms s_adds whose results are copied
to vector registers into v_adds. We don't want to do that if foldInstOperand
(which so far runs later) can fold the sreg->vreg copy away.
This patch therefore delays foldCopyToVGPROfScalarAddOfFrameIndex until after
foldInstOperand.
This avoids unnecessary movs in the flat-scratch-svs.ll test and also avoids
regressions in an upcoming patch to enable ISD::PTRADD nodes.
Diffstat (limited to 'llvm/lib/Transforms/Utils/LoopPeel.cpp')
0 files changed, 0 insertions, 0 deletions