This addresses a bug where vector versions of ctlz are creating false positive reports.
Depends on D136369
Paths
| Differential D136523
[MSAN] Add handleCountZeroes for ctlz and cttz. ClosedPublic Authored by kda on Oct 22 2022, 2:26 AM.
Details Summary This addresses a bug where vector versions of ctlz are creating false positive reports. Depends on D136369
Diff Detail
Event Timelinekda added a parent revision: D136369: [MSAN] Count Zeroes test for covering ctlz and cttz under MSAN. (NFC). Comment Actions LGTM, but I'd merge the two changes. I don't really see the point of submitting a test for the broken behavior first.
Comment Actions
The test does not demonstrate "broken" behavior but rather shows what is generated before my change.
Comment Actions
Actually I rely like splitting stuff like this Comment Actions LGTM with nits
This revision is now accepted and ready to land.Oct 24 2022, 4:24 PM kda retitled this revision from [MSAN] Add handleCountZeroes for vector versions of ctlz and cttz. to [MSAN] Add handleCountZeroes for ctlz and cttz..Oct 24 2022, 5:28 PM This revision was landed with ongoing or failed builds.Oct 24 2022, 5:31 PM Closed by commit rG31bfa4a69b12: [MSAN] Add handleCountZeroes for ctlz and cttz. (authored by kda). · Explain Why This revision was automatically updated to reflect the committed changes.
Revision Contents
Diff 470332 llvm/lib/Transforms/Instrumentation/MemorySanitizer.cpp
llvm/test/Instrumentation/MemorySanitizer/count-zeroes.ll
|
isn't this implementation strictly better than default? It handles the second operand correctly.