simoncook planned changes to this revision. simoncook added a comment. Thanks for the feedback. I will improve the test so it more reliably tests what it intends to.
With regards to behaviour surrounding things such as argument registers, before submitting I checked what the riscv port of GCC does, and it matches this behaviour. If a register is used for arg passing/return values then the option is accepted but silently ignored (at least the register is still used for passing arguments, I haven't confirmed for other regalloc purposes). I agree that this has the opportunity for allowing users to think they've reserved a register but it is still used. I will look at something more to what you've described AArch64 does in LLVM, and also check that there isn't also a bug on the GCC side, if so I'll get that fixed too. Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D67185/new/ https://reviews.llvm.org/D67185 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits