[m-dev.] benchmarks for finite domain solvers with --generate-trail-ops-inline

[Samrith UONG]

On 12/17/05, Samrith UONG <samuong at gmail.com> wrote:
> On 12/17/05, Julien Fischer <juliensf at cs.mu.oz.au> wrote:
> >
> >
> > On Fri, 16 Dec 2005, Zoltan Somogyi wrote:
> >
> > > >   This is consistent with the
> > > > recent benchmarking that Samrith has been doing on the compiler, where hlc.gc
> > > > and hl.gc have been consistently faster than asm_fast.gc.
> > >
> > > Were those with --use-local-vars enabled? Some benchmarking I have been doing
> > > on my laptop indicates that we should enable --use-local-vars at the default
> > > optimization level; its cost in compilation time is cheap compared to the speed
> > > improvement it generates.
> > >
> >
> > We'll look into --use-local-vars more closely next week.
> >
>
> I've just started building a new batch using --use-local-vars, and
> will have some results soon.
>

Julien and I have done some benchmarking and we've found that, at -O2,
--use-local-vars results in a 3.5% increase in speed and a 7% decrease
in size.  This is at a cost of 1.5% extra compilation time.

This suggests that we should move --use-local-vars to -O2.  I will
have a patch ready for review soon once the power is back on in ICT.

Regards,
Sam.

--------------------------------------------------------------------------
mercury-developers mailing list
Post messages to:       mercury-developers at cs.mu.oz.au
Administrative Queries: owner-mercury-developers at cs.mu.oz.au
Subscriptions:          mercury-developers-request at cs.mu.oz.au
--------------------------------------------------------------------------