diff options
| author | Jakub Chlanda <jakub@codeplay.com> | 2025-01-10 07:49:11 +0100 |
|---|---|---|
| committer | GitHub <noreply@github.com> | 2025-01-10 07:49:11 +0100 |
| commit | 01a7d4e26b9bac27e282b113209f53c4c1d290b2 (patch) | |
| tree | b70f9e6d00f84d4beafa20a919a9746ec522cbd5 /llvm/lib/CodeGen/RegAllocBasic.cpp | |
| parent | 4c0a0f72418b21161b5c1fb9225462bd039121e3 (diff) | |
| download | llvm-01a7d4e26b9bac27e282b113209f53c4c1d290b2.zip llvm-01a7d4e26b9bac27e282b113209f53c4c1d290b2.tar.gz llvm-01a7d4e26b9bac27e282b113209f53c4c1d290b2.tar.bz2 | |
[AMDGPU] Allow selection of BITOP3 for some 2 opcodes and B32 cases (#122267)
This came up in downstream static analysis - as a dead code.
Admittedly, it depends on what the intention was when checking for [`if
(NumOpcodes == 2 &&
IsB32)`](https://github.com/llvm/llvm-project/blob/main/llvm/lib/Target/AMDGPU/AMDGPUInstructionSelector.cpp#L3792C3-L3792C32)
and I took a guess that for certain cases the selection should take
place.
If that's incorrect, that whole if statement can be removed, as it is
after a check for: [`if (NumOpcodes <
4)`](https://github.com/llvm/llvm-project/blob/main/llvm/lib/Target/AMDGPU/AMDGPUInstructionSelector.cpp#L3788)
Diffstat (limited to 'llvm/lib/CodeGen/RegAllocBasic.cpp')
0 files changed, 0 insertions, 0 deletions