[PATCH] GlobalOpt: Apply fastcc to internal x86_thiscallcc functions

Rafael EspĂ­ndola rafael.espindola at gmail.com
Mon Feb 24 18:46:24 PST 2014


> I conceptually agree, but I think that's an issue beyond this optimization.
> I think this patch LGTM, Rafael?

Almost. Please include a corresponding change to the LangRef. Just
state which calling conventions can or cannot cannot be changed when
the target function is internal or private.

Please also add a test showing that a llvm.used function is not optimized.

Changing how we represent llvm.used and llvm.compiler_used is a much
longer project and should not block this patch.

>>
>> > If you just want to flip the naming around here to something like
>> > isProfitableToMakeFastCC() then that works for me.
>
>
> isProfitableToMakeFastCC sounds nice. What really matters is that it needs
> to be a whitelist.
>

Agreed.

Cheers,
Rafael



More information about the llvm-commits mailing list