Re: [RFC,PATCH] Provide get_attr functions with alternative parameter

[Andreas Krebbel <Andreas.Krebbel@de.ibm.com>]

Hello,

> Interesting!  Is there a reason (in the patch you're working on) why the 
> inline wrappers couldn't just do
> 
>   extract_constrain_insn_cached (insn);
>   return get_alt_attr_blah_blah (insn, which_alternative);
> 
> instead of supporting the special alternative number -1 in 
> get_alt_attr_blah_blah?  (I can see the reason could be code bloat, or 
> GCC deciding not to inline get_attr_blah_blah, but I thought I'd just ask).

Good question. The current code emitted by genattrtab only invokes
extract_constrain_insn if the value of the alternative really depends
on the number of the alternative. So for several attributes it might
not get invoked at all. I was just trying to leave that behaviour as
is in order to avoid performance regressions. It is not required by
the patch I'm working on.

Bye,

-Andreas-