This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: Dropping of old loop optimizer
- From: Pop Sébastian <pop at gauvain dot u-strasbg dot fr>
- To: Daniel Berlin <dberlin at dberlin dot org>
- Cc: Zdenek Dvorak <rakdver at atrey dot karlin dot mff dot cuni dot cz>,tm_gccmail at mail dot kloo dot net, gcc at gcc dot gnu dot org,m dot hayes at elec dot canterbury dot ac dot nz, rth at redhat dot com, law at redhat dot com,dan at dberlin dot org, jh at suse dot cz
- Date: Thu, 27 Feb 2003 17:50:05 +0100
- Subject: Re: Dropping of old loop optimizer
- References: <20030227151807.GA15920@gauvain.u-strasbg.fr> <3779E2EC-4A6A-11D7-B465-000393575BCC@dberlin.org>
On Thu, Feb 27, 2003 at 04:14:48PM +0000, pop wrote:
> On Thu, Feb 27, 2003 at 10:43:28AM -0500, Daniel Berlin wrote:
> >
> > I had started to work on this method, but decided it was too complex,
> > and moved to monotonic evolution.
> >
> In contrast to the CR method, monotonic evolution (ME) gives approximate
> data dependences: you'll have the direction but not the distance vector
> of a dependence. Fact: not all optimizers use approximate vectors.
>
> In this point you cannot compare these two passes, but anyway ...
>
Forgot to mention the conclusion of one of the papers on ME:
"Monotonic Evolution: an Alternative to Induction Variable Substitution
for Dependence Analysis"
Peng Wu, Albert Cohen, Jay Hoeflinger, and David Padua
"Closed form expression computation [...] is also critical for removing
dependences due to computation of induction variables themselves."
In other words if you generate code for IV functions at loop-phi-nodes,
then you can parallelize these loops.
Sebastian