This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: PR target/23303 (worse instruction choice)
> >
> > Is this whole huge complicated peephole really worth it? I'd expect
> > the pattern to match quite a few times, but all the additional checks
> > probably make it not trigger very often. Honza, do you have numbers
And the numbers bellow are triggers in your terminology, just for sure
;)
There are many checks but they should pass most of time...
Honza
> > about the effects of this patch??
>
> The pattern match about every access to array where both base and index
> are load from memory for the address computation on pentium and i686, so
> I can imagine there are testcases where this accounts for important
> portion of code. It matches 315 times on tramp3d, 35 times on Gerald's
> testcase and 20 times in combine.c. Every match is roughly 4 bytes of
> code size.
>
> So I think it sort of worth it just because the costs of extra peep2 are
> low despite it's spagettiness...
>
> Honza
> >
> > Gr.
> > Steven
> >