llvm AArch64 omit frame pointer optimization

Hi All,

I’m looking into enabling frame pointer elimination for non-leaf functions in the AArch64 target. I’d like to get it working when requested by –fomit-frame-pointer first, then consider whether some target triples cause it to be enable it by default when optimizing. I’m aware of issues around debugger support and unwinding, but was wondering if anyone else had any thoughts/guidance that might be helpful. Specifically, I’d like to hear from any of the implementers of the current AArch64 frame lowering if they know of any pitfalls to be aware of.

Thanks,

-Geoff