This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: Tiny speedups..
>
> I think this is a good idea. You shold use NULL_RTX instead of `0' in
> the return values for single_set.
OK. I see - that problem was just copied from 0.
> I actually think it's a mistake to spell function-like macros in
> ALL_CAPS, precisely because that makes changing back-and-forth unduly
> disruptive. In my opinion, we should only use ALL_CAPS for accessors
> and simple predicates. I don't know what others thing about this.
OK. I tought that our convention was to make macros in uppercase, but I agree
that single_set like macro is much prettier lowercase.
Only problem I see that lowercase will not warn you about repeated side
effects.
>
> Is a PARALLEL required to have two instructions? The documentation
> doesn't imply that, but your code will crash if there are zero or one
> instructions in the parallel. Also, the documentation does not
> prescribe the order in the PARALLEL that you are relying upon. I
> think that you should update the TeXinfo documentation to reflect this
> assumption, if we are going to rely on it.
OK. Wehere I should place it? Canonical form or just in the description
of parallel?
I would not expect someone to break first rule (PARALLEL with one member),
but my own string operations in parallels break the second rule, but I do
have patch pending to fix that.
Thats why I've included the checks for ENABLE_CHECKING.
There are similar requirements in other places as well (I only remember
my own function pc_set offhand, but there are more).
>
> I'll let Richard (or someone else) make the final call on all this, as
> I'm less familiar with the exact details in this area.
OK.
Thanks for the comments.
Honza
>
> Thanks,
>
> --
> Mark Mitchell mark@codesourcery.com
> CodeSourcery, LLC http://www.codesourcery.com