This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: [tree-ssa] POOMA compile time / memory requirement comparison
On Tue, 4 May 2004, Diego Novillo wrote:
> On Tue, 2004-05-04 at 06:22, Richard Guenther wrote:
>
> > So tree-ssa memory requirement is 127% of mainline, compile time
> > is 134% of mainline (without leafify) and 126% of mainline (with leafify).
> >
> I see two options here: (a) unfreeze the branch and do whatever work is
> necessary to bring POOMA in line with mainline, (b) continue with the
> merge and address these problems in mainline.
>
> IMO option (a) is too costly and may only serve to delay tree-ssa even
> more. I suspect that much of the necessary work to reduce this gap
> involves removing RTL passes and/or doing major surgery on trees, memory
> allocation (the system times are pretty bad).
I would prefer option (b), as POOMA is still uncommon use of C++. Also
fixing the regressions will be easier at the time the leafify patch is
merged upstream - without leafify you really can't compare compile times,
as different expression counting leads to different inlining
characteristics.
I also expect figures to improve once (parts of) the lno loop optimizer is
merged into mainline.
I also don't worry so much about the compile time regression, but the
memory usage needs to improve (and that isn't going to happen with just
removal of RTL passes).
> But I naturally have a warped view of the whole merge process. What do
> others think? If there aren't any major objections, I will continue
> with the merge plans and merge as soon as we address the testsuite
> regressions.
Maybe you are interested in that tree-ssa miscompiles the POOMA testcase
if leafify is enabled - but I'll bug you with a PR as soon as leafify is
available to all developers.
Richard.
--
Richard Guenther <richard dot guenther at uni-tuebingen dot de>
WWW: http://www.tat.physik.uni-tuebingen.de/~rguenth/