Re: [patch] expmed.c: Fix PR 23971 - slow synth_mult.

[Kazu Hirata <kazu at codesourcery dot com>]

Hi Falk,

> > Here is the timing in seconds with the cross compiler to
> > alpha-linux-gnu from x86_64-pc-linux-gnu.
> >
> > original patched diff(%)
> >   6.939   0.686 -90.114 (wow!)
>
>
> Actually, 0.6 seconds for one line of source still seems quite
> excessive... While this patch certainly seems like a good idea, I'm
> wondering whether we shouldn't just additionally put an upper limit to
> the number of instructions generated, since it seems unlikely that
> expanding a multiply to more than say 12 insns is really such a great
> idea.

Well, we have a limit already although it's not the kind of limit that you are 
talking about.  The number of insns that we generate is limited to 
BITS_PER_WORD.  It's easy to change that number, but we need to respect rtx_cost 
to some extent, too.

Meanwhile, I've got several ideas to further speed up synth_mult.

So I guess we can still improve both features and speed in this area.

Kazu Hirata