This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: [lno] 50% runtime performance regression since yesterday
- From: Zdenek Dvorak <rakdver at atrey dot karlin dot mff dot cuni dot cz>
- To: Richard Guenther <rguenth at tat dot physik dot uni-tuebingen dot de>
- Cc: gcc at gcc dot gnu dot org, Sebastian Pop <pop at cri dot ensmp dot fr>
- Date: Wed, 23 Jun 2004 16:23:12 +0200
- Subject: Re: [lno] 50% runtime performance regression since yesterday
- References: <Pine.LNX.4.53.0406231024000.2199@bellatrix.tat.physik.uni-tuebingen.de>
Hello,
> Between (not including)
>
> 2004-06-21 Zdenek Dvorak <rakdver@atrey.karlin.mff.cuni.cz>
>
> PR rtl-optimization/16001
> * loop-iv.c (iv_number_of_iterations): Prevent copy propagation in
> niter_expr.
>
> and (including)
>
> 2004-06-22 Dorit Naishlos <dorit@il.ibm.com>
>
> * tree-vectorizer.c (vect_create_data_ref): Fix setting of mem_tag.
> (vect_analyze_data_refs): Make sure we have a mem_tag.
>
> You introduced a 50% runtime performance regression to lno. lno is now
> slightly worse than mainline with -O2 -funroll-loops on the tramp3d-v3
> testcase while it was 50% better before.
>
> May it be all the removal of CHREC nodes? I don't know if this removes
> any functionality, though. But it seems to be the only patch not touching
> exclusively the vectorizer (which I don't use).
>
> Please get me back that performance!
I guess it must have been one of the Sebastian's patches then; I don't
think that removal of the chrec nodes by itself could change
anything (we no longer had any code to do anything sane with them
anyway), but perhaps something got broken by accident during these
changes.
Zdenek