This is another step towards implementing register classes with parametrized register/spill sizes.
Details
Diff Detail
- Repository
- rL LLVM
Event Timeline
Makes sense.
include/llvm/Target/TargetRegisterInfo.h | ||
---|---|---|
98 | How hard would it be to rename this to getSpillSize() to avoid confusing? |
include/llvm/Target/TargetRegisterInfo.h | ||
---|---|---|
98 | That would be easy. However there are some places where this function is called with the intent of getting the register size. AMDGPU and Hexagon both have places like that. I'd like to add the register size as a separate field here, but I'm getting close to completing the first version of the variable class patch which already adds that. |
How hard would it be to rename this to getSpillSize() to avoid confusing?
We can however do that in a separate patch.