From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (qmail 16261 invoked by alias); 29 Apr 2009 21:57:51 -0000 Received: (qmail 16246 invoked by uid 22791); 29 Apr 2009 21:57:50 -0000 X-SWARE-Spam-Status: No, hits=-1.8 required=5.0 tests=AWL,BAYES_00,SARE_MSGID_LONG40,SPF_PASS X-Spam-Check-By: sourceware.org Received: from mail-fx0-f213.google.com (HELO mail-fx0-f213.google.com) (209.85.220.213) by sourceware.org (qpsmtpd/0.43rc1) with ESMTP; Wed, 29 Apr 2009 21:57:45 +0000 Received: by fxm9 with SMTP id 9so1470211fxm.8 for ; Wed, 29 Apr 2009 14:57:42 -0700 (PDT) MIME-Version: 1.0 Received: by 10.204.62.135 with SMTP id x7mr755974bkh.124.1241042261360; Wed, 29 Apr 2009 14:57:41 -0700 (PDT) In-Reply-To: <1241040850.91636.381.camel@localhost> References: <1241040850.91636.381.camel@localhost> Date: Wed, 29 Apr 2009 23:29:00 -0000 Message-ID: <84fc9c000904291457w267bac8as357d7632f80f139f@mail.gmail.com> Subject: Re: [graphite] Weekly phone call notes From: Richard Guenther To: Tobias Grosser Cc: GCC Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: quoted-printable X-IsSubscribed: yes Mailing-List: contact gcc-help@gcc.gnu.org; run by ezmlm Precedence: bulk List-Id: List-Archive: List-Post: List-Help: Sender: gcc-owner@gcc.gnu.org X-SW-Source: 2009-04/txt/msg00797.txt.bz2 On Wed, Apr 29, 2009 at 11:34 PM, Tobias Grosser wrote: > Hi gcc developers, hi graphities > > here are some notes from our weekly phone call. Unfortunately I missed > to send out the notes from the last two phone calls, but I hope to get > them out more regulary. Believe in me! ;-) > > http://gcc.gnu.org/wiki/Graphite_Phone_Call/2009_04_29 > > Attendees: Li, Jan, Konrad, Sebastian, Tobias, David, Christophe > > =A0 =A0 =A0* Reductions: Diego OK with going out of SSA. You will loose all points-to information. I think going out of SSA is a very bad idea. Richard. > =A0 =A0 =A0 =A0 =A0 =A0 =A0* Patch not yet ready. The higher code coverag= e exposes an > =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0IVSTACK bug. > =A0 =A0 =A0* IVStack: Sebastian is working on removing it > =A0 =A0 =A0* Data dependence analysis: Committed to graphite branch. > =A0 =A0 =A0 =A0 =A0 =A0 =A0* Not yet enabled, as the data reference still= does not > =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0bootstrap. The lexicographic smaller than = constraint is > =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0not yet added. This is not wrong but very = conservative. > =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0Dependence test should be enabled with > =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0-fgraphite-identity. > =A0 =A0 =A0* Loop transformations: We need some simple transformations to > =A0 =A0 =A0 =A0check data dependency analysis. > =A0 =A0 =A0 =A0 =A0 =A0 =A0* Pranav? We did not hear anything yet. > =A0 =A0 =A0* Data reference: Last bugs have to be fixed. > =A0 =A0 =A0 =A0 =A0 =A0 =A0* Not yet enabled. Blocks data dependence anal= ysis. > =A0 =A0 =A0* PCP: > =A0 =A0 =A0 =A0 =A0 =A0 =A0* Jan Implemented simplification/canonicalizat= ion for > =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0expressions. This will allow a systematic = translation of > =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0expressions to constraints. Jan will work = on a reduced > =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0polyhedral interface that is used in PCP. = First PCP > =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0integration in gcc - later extended by red= uctions. > > =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0No good representation for reductions for = the general > =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0case. Like conditions and PHI nodes (Sebas= tian) Maybe > =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0for a subset of the reductions we can add = commutativaty > =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0of operations. > > =A0 =A0 =A0* Autopar: Li prepared some patches to trigger autopar by > =A0 =A0 =A0 =A0graphite. > =A0 =A0 =A0 =A0 =A0 =A0 =A0* Autopar failed in the graphite branch even w= ithout > =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0graphite enabled. It seems we introduced a= bug while > =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0using some code of it for graphite. The br= eakage happens > =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0in the reductions part. > =A0 =A0 =A0* Merge from trunk: > =A0 =A0 =A0 =A0 =A0 =A0 =A0* Sebastian wants to work on this Changes in v= ectorizer > =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0Last merge 4 months ago Tried mid march bu= t was > =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0difficult. So stopped because of ENOTIME. > =A0 =A0 =A0* Parts of graphite working well?: The polyhedral part of the > =A0 =A0 =A0 =A0graphite architecture will be described > =A0 =A0 =A0 =A0 =A0 =A0 =A0* in the Tobias' paper at the summit and the P= CP part by > =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0Jan. > =A0 =A0 =A0* Worldwide interest: Cristianno Martins from Brazil showed so= me > =A0 =A0 =A0 =A0interest > =A0 =A0 =A0 =A0 =A0 =A0 =A0* http://gcc.gnu.org/ml/gcc/2009-04/msg00711.h= tml > > =A0 =A0 =A0 =A0 =A0 =A0 =A0* We already have people from the US, Europe a= nd China > =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0contributing. > > > >