This is another instance of the lo->lo copy problem. Unfortunately, I don't have a testcase.
Details
Details
- Reviewers
t.p.northover jmolloy mroth
Diff Detail
Diff Detail
Event Timeline
Comment Actions
So why couldn't you create a test? It doesn't seem like a terribly difficult issue to trigger: just a generic load/store merge operation on Thumb1.
llvm/lib/Target/ARM/ARMLoadStoreOptimizer.cpp | ||
---|---|---|
574 | Isn't the whole block already isThumb1? |
Comment Actions
Hi Jon,
Thanks for adding the test. Just one debugging line that crept in. Feel free to commit with that removed.
Tim.
lib/Target/ARM/ARMLoadStoreOptimizer.cpp | ||
---|---|---|
564 ↗ | (On Diff #18554) | Probably best without this. |
Isn't the whole block already isThumb1?