diff options
| author | Matt Arsenault <Matthew.Arsenault@amd.com> | 2019-02-04 22:26:21 +0000 |
|---|---|---|
| committer | Matt Arsenault <Matthew.Arsenault@amd.com> | 2019-02-04 22:26:21 +0000 |
| commit | cba0c6d0c9d2c610c566e4b46b8f0315a2cfebbc (patch) | |
| tree | c7c50c770ff0471d43d0088c24ddbe2f0a744f49 /llvm/lib | |
| parent | 29ac3a5b822ba8c097a3ae78d983cdb94da43dd4 (diff) | |
| download | bcm5719-llvm-cba0c6d0c9d2c610c566e4b46b8f0315a2cfebbc.tar.gz bcm5719-llvm-cba0c6d0c9d2c610c566e4b46b8f0315a2cfebbc.zip | |
AMDGPU: Don't rematerialize mov with implicit operands
This was pulling the mov used for register indexing on gfx9 out of the
loop.
llvm-svn: 353101
Diffstat (limited to 'llvm/lib')
| -rw-r--r-- | llvm/lib/Target/AMDGPU/SIInstrInfo.cpp | 3 |
1 files changed, 2 insertions, 1 deletions
diff --git a/llvm/lib/Target/AMDGPU/SIInstrInfo.cpp b/llvm/lib/Target/AMDGPU/SIInstrInfo.cpp index b857d9df823..06c080c8af8 100644 --- a/llvm/lib/Target/AMDGPU/SIInstrInfo.cpp +++ b/llvm/lib/Target/AMDGPU/SIInstrInfo.cpp @@ -140,7 +140,8 @@ bool SIInstrInfo::isReallyTriviallyReMaterializable(const MachineInstr &MI, case AMDGPU::V_MOV_B32_e32: case AMDGPU::V_MOV_B32_e64: case AMDGPU::V_MOV_B64_PSEUDO: - return true; + // No implicit operands. + return MI.getNumOperands() == MI.getDesc().getNumOperands(); default: return false; } |
