diff options
| author | Matt Arsenault <Matthew.Arsenault@amd.com> | 2017-12-05 18:23:17 +0000 |
|---|---|---|
| committer | Matt Arsenault <Matthew.Arsenault@amd.com> | 2017-12-05 18:23:17 +0000 |
| commit | 7f0a5273001e24b76baaf53e8d94df1e83180479 (patch) | |
| tree | c8e40c7326927d7f840091e6573b7651950eebbe /llvm/lib/Target/AMDGPU/SIOptimizeExecMaskingPreRA.cpp | |
| parent | ea0bafda8a873a1cff8b4e4b3e6a9d0c7326919b (diff) | |
| download | bcm5719-llvm-7f0a5273001e24b76baaf53e8d94df1e83180479.tar.gz bcm5719-llvm-7f0a5273001e24b76baaf53e8d94df1e83180479.zip | |
AMDGPU: Fix infinite loop with dbg_value
Surprisingly SIOptimizeExecMaskingPreRA can infinite loop
in some case with DBG_VALUE. Most tests using dbg_value are
run at -O0, so don't run this pass. This seems to only
happen when the value argument is undef.
llvm-svn: 319808
Diffstat (limited to 'llvm/lib/Target/AMDGPU/SIOptimizeExecMaskingPreRA.cpp')
| -rw-r--r-- | llvm/lib/Target/AMDGPU/SIOptimizeExecMaskingPreRA.cpp | 5 |
1 files changed, 4 insertions, 1 deletions
diff --git a/llvm/lib/Target/AMDGPU/SIOptimizeExecMaskingPreRA.cpp b/llvm/lib/Target/AMDGPU/SIOptimizeExecMaskingPreRA.cpp index 6b2668fe052..647aafe4d2e 100644 --- a/llvm/lib/Target/AMDGPU/SIOptimizeExecMaskingPreRA.cpp +++ b/llvm/lib/Target/AMDGPU/SIOptimizeExecMaskingPreRA.cpp @@ -134,8 +134,11 @@ bool SIOptimizeExecMaskingPreRA::runOnMachineFunction(MachineFunction &MF) { } while (I != E) { - if (I->isDebugValue()) + if (I->isDebugValue()) { + I = std::next(I); continue; + } + if (I->mayStore() || I->isBarrier() || I->isCall() || I->hasUnmodeledSideEffects() || I->hasOrderedMemoryRef()) break; |
