Details
Diff Detail
- Repository
- rG LLVM Github Monorepo
Event Timeline
| llvm/lib/Transforms/Scalar/LICM.cpp | ||
|---|---|---|
| 484 | I took a quick look at the rest of the uses of eraseInstruction, it seems like this is the only other place where a salvage makes sense. As a follow up we might convert this to the 'OrMarkUndef' variant (or fold that into this patch). | |
| llvm/test/DebugInfo/X86/licm-undef-dbg-value.ll | ||
|---|---|---|
| 84 | do we need the TBAA data? | |
| llvm/lib/Transforms/Scalar/LICM.cpp | ||
|---|---|---|
| 484 | What about other uses of the salvageDebugInfo(), such as: llvm/lib/Transforms/Scalar/GVN.cpp llvm/lib/Transforms/Scalar/DeadStoreElimination.cpp llvm/lib/Transforms/Scalar/DCE.cpp should we address them as well? If so, that should be a different patch. | |
Thanks everyone, I'll land this shortly.
| llvm/lib/Transforms/Scalar/LICM.cpp | ||
|---|---|---|
| 484 | I believe @chrisjackson is going to fix up these cases (salvageDebugInfo -> salvageDebugInfoOrMarkUndef) in an upcoming patch. | |
| llvm/test/DebugInfo/X86/licm-undef-dbg-value.ll | ||
| 84 | Nope, I'll remove it before committing. | |
I took a quick look at the rest of the uses of eraseInstruction, it seems like this is the only other place where a salvage makes sense. As a follow up we might convert this to the 'OrMarkUndef' variant (or fold that into this patch).