- That flag setting should skip spilling stack slot.
Details
Details
Diff Detail
Diff Detail
- Repository
- rL LLVM
Event Timeline
Comment Actions
FlatScratchInit is set by checking whether there are any stack objects in the early stage of codegen. As we could load MIR in any stage, even after RA, it may set that flag incorrectly by counting spilling stack slot.
llvm/test/CodeGen/AMDGPU/undefined-physreg-sgpr-spill.mir | ||
---|---|---|
44 ↗ | (On Diff #207827) | this test case needs revising as the max-sgprs is incorrectly calculated when FlatScratchInit is set wrongly. That results in $sgpr10 and sgpr11 being reserved. After fixing FlatScratchInit setting, we need to add them into livein registers. |