This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [tree-ssa] Speed up constant propagation [patch]
- From: Richard Guenther <rguenth at tat dot physik dot uni-tuebingen dot de>
- To: Diego Novillo <dnovillo at redhat dot com>
- Cc: "gcc-patches at gcc dot gnu dot org" <gcc-patches at gcc dot gnu dot org>, Gerald Pfeifer <pfeifer at dbai dot tuwien dot ac dot at>
- Date: Tue, 17 Jun 2003 13:38:38 +0200 (CEST)
- Subject: Re: [tree-ssa] Speed up constant propagation [patch]
On 17 Jun 2003, Diego Novillo wrote:
> On Tue, 2003-06-17 at 05:08, Richard Guenther wrote:
>
> > The worst case test went down from 38 minutes to 4:15 (mainline is 3:23)!
> > Nice work! Time report follows (-fno-exceptions -O2 -march=athlon
> > -funroll-loops).
> >
> Cool. Glad to hear that.
>
> > tree alias analysis : 22.45 ( 9%) usr 0.11 ( 3%) sys 22.44 ( 9%) wall
> > tree PHI insertion : 83.14 (33%) usr 0.39 ( 9%) sys 83.50 (33%) wall
> > tree SSA rewrite : 8.73 ( 3%) usr 0.07 ( 2%) sys 8.93 ( 3%) wall
> > tree SSA other : 2.77 ( 1%) usr 0.08 ( 2%) sys 2.87 ( 1%) wall
> > tree SSA to normal : 12.39 ( 5%) usr 0.12 ( 3%) sys 12.44 ( 5%) wall
> > tree CCP : 17.76 ( 7%) usr 0.13 ( 3%) sys 17.88 ( 7%) wall
> > tree COPYPROP : 1.07 ( 0%) usr 0.00 ( 0%) sys 1.05 ( 0%) wall
> > tree DCE : 4.41 ( 2%) usr 0.11 ( 3%) sys 4.57 ( 2%) wall
> >
> Which one is it? PHI insertion times are pretty bad. If we manage to
> get that one below 10 secs, we will be beating mainline :) (not that I
> think this will be necessarily trivial, but if the test has heavy
> aliasing, we may consider introducing zero-versions, a-la ORC, or see if
> PTA helps).
Its src/Field/tests/Cartesian.cpp which you dont have, because thats in my
local tree ;) A preprocessed file is at
http://www.tat.physik.uni-tuebingen.de/~rguenth/gcc/Cartesian.ii.gz
Richard.
--
Richard Guenther <richard dot guenther at uni-tuebingen dot de>
WWW: http://www.tat.physik.uni-tuebingen.de/~rguenth/