From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (qmail 102387 invoked by alias); 3 Apr 2018 16:19:58 -0000 Mailing-List: contact fortran-help@gcc.gnu.org; run by ezmlm Precedence: bulk List-Id: List-Subscribe: List-Post: List-Help: , Sender: fortran-owner@gcc.gnu.org Received: (qmail 102372 invoked by uid 89); 3 Apr 2018 16:19:57 -0000 Authentication-Results: sourceware.org; auth=none X-Virus-Found: No X-Spam-SWARE-Status: No, score=-4.6 required=5.0 tests=AWL,BAYES_00,FREEMAIL_FROM,GIT_PATCH_2,HTML_MESSAGE,KAM_SHORT,RCVD_IN_DNSWL_NONE,SPF_PASS,UNSUBSCRIBE_BODY autolearn=ham version=3.3.2 spammy=becoming, himself, Alpha, H*Ad:D*ie X-HELO: mail-lf0-f41.google.com Received: from mail-lf0-f41.google.com (HELO mail-lf0-f41.google.com) (209.85.215.41) by sourceware.org (qpsmtpd/0.93/v0.84-503-g423c35a) with ESMTP; Tue, 03 Apr 2018 16:19:52 +0000 Received: by mail-lf0-f41.google.com with SMTP id j68-v6so21019498lfg.13 for ; Tue, 03 Apr 2018 09:19:51 -0700 (PDT) X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20161025; h=x-gm-message-state:mime-version:references:in-reply-to:from:date :message-id:subject:to:cc; bh=qJ/wxoGh6FQPrFyqBF2lQM8G+AdVLqSfeyFEqcY7xmQ=; b=pRuZwjfb1qbK64mq0g38QFd54rrQ36thsOv1MNzVjp4DnFkrE7Hm5anuL07Oq+uxVw acMgdtLjdJgtAJTmViM4Rmj2WBgJwapi2DycJkJG/lPO1q6hFAJpaWk8UqBh6vQcgiqu 0LVdABsP0bbH3kmIE4D4ahyGAYoVwGLgruoAhwzvF2zOoMSg3LlaEHeG7ca00Ts+eGZ/ dtnNBjIOCMb4EqkzMkmz4ZJqpvd31qn2hWlTInAdmc1h7E6cvQyyJ4BLMnIqUafns3uC 0MknFX1iGQte9DgqpdEbvAxETexB/hT1xo0LR/vz5KcOCrr064oQI9tsHef8coN7zKj9 YGpg== X-Gm-Message-State: ALQs6tBXgFgtadcnU6+3Z9W9BEyzVF0gG6u/PPYCURvmYQ1g/qi4HedB PoDPJMwnXrrjmd6Q/Eudn4o+thUQqfbfy+HPowuwbw== X-Google-Smtp-Source: AIpwx49xFD4o7qEDbRoyx8ru6a3I/ihjIXkYNcy3nPJbvIG7bWIwF7ZrRrlMOgWRLDJYi0EStdBl/zVMCD+hAxysHG4= X-Received: by 2002:a19:d8ce:: with SMTP id r75-v6mr3525857lfi.144.1522772388483; Tue, 03 Apr 2018 09:19:48 -0700 (PDT) MIME-Version: 1.0 References: <1522766220.22086.ezmlm@gcc.gnu.org> In-Reply-To: <1522766220.22086.ezmlm@gcc.gnu.org> From: Zaak Beekman Date: Tue, 03 Apr 2018 16:19:00 -0000 Message-ID: Subject: Re: Coarrays in Debian Testing. To: fortran@gcc.gnu.org Cc: alastair@sceal.ie Content-Type: text/plain; charset="UTF-8" Content-Transfer-Encoding: quoted-printable X-SW-Source: 2018-04/txt/msg00009.txt.bz2 > I should examine the gcc build process that uses MPICH and extend it to > use OpenMPI as an option, and see whats needed for Failed Images in > OpenMPI: having gfortran depend on mpich while openmpi is the default > MPI in Debian would be .. unhelpful. > > regards > Alastair Oh, fabulous, the package maintainer himself shows up! Thanks for your work! Yes, I agree that care should be taken as many distros/package repositories pick a default MPI implementation, often OpenMPI (same for Homebrew on Mac) Alastair, please do let us know if there's anything we can do upstream in OpenCoarrays to make your life easier and/or reduce the amount of patching etc. you need to do to get things building. As far as failed image support goes, last I checked there was a private fork of OpenMPI that theoretically worked, branched off of the public ULFM OpenMPI: https://bitbucket.org/icldistcomp/ulfm2 but I haven't checked in a very long while. With MPICH, Andre submitted a patch[1] that fixed some issues, but it's been on master for a long long time and is not yet in a release, AFAICT. Without this patch some bugs exist when OpenCoarrays is built with failed image support, both when using failed images and it also triggers a few when using tangentially related or unrelated features. [1]: https://github.com/pmodels/mpich/pull/2730 -Zaak On Tue, Apr 3, 2018 at 10:37 AM wrote: > > fortran Digest 3 Apr 2018 14:37:00 -0000 Issue 4647 > > Topics (messages 50162 through 50164): > > Re: [patch, fortran] Fix PR 85102, take 2 > 50162 by: Thomas K=C3=B6nig > > Re: Patching the GCC build system to build MPICH and OpenCoarrays > 50163 by: Richard Biener > > Re: Coarrays in Debian Testing. > 50164 by: Alastair McKinstry > > Administrivia: > > To subscribe to the digest, e-mail: > fortran-digest-subscribe@gcc.gnu.org > > To unsubscribe from the digest, e-mail: > fortran-digest-unsubscribe@gcc.gnu.org > > To post to the list, e-mail: > fortran@gcc.gnu.org > > > ---------------------------------------------------------------------- > > > > ---------- Forwarded message ---------- > From: "Thomas K=C3=B6nig" > To: sgk@troutmask.apl.washington.edu > Cc: "fortran@gcc.gnu.org" , gcc-patches < > gcc-patches@gcc.gnu.org>, Dominique Dhumieres < > dominiq@tournesol.lps.ens.fr> > Bcc: > Date: Mon, 2 Apr 2018 18:23:11 +0200 > Subject: Re: [patch, fortran] Fix PR 85102, take 2 > Hi Steve, > > > else > > gfc_free_expr (n); > > > > Don't you need the above changes to avoid leaking memory? > > Correct. > > OK with those changes? > > Regards > > Thomas > > > > ---------- Forwarded message ---------- > From: Richard Biener > To: Damian Rouson > Cc: GCC Development , gfortran , > Daniel Celis Garza , Jerry DeLisle < > jvdelisle@charter.net>, "Filippone, Salvatore" < > salvatore.filippone@cranfield.ac.uk> > Bcc: > Date: Tue, 3 Apr 2018 13:36:36 +0200 > Subject: Re: Patching the GCC build system to build MPICH and OpenCoarrays > On Sat, Mar 31, 2018 at 2:44 AM, Damian Rouson > wrote: > > All, > > > > Jerry DeLisle, Daniel Celis Garza, and I would greatly appreciate > feedback on our approach to patching the GCC build system to build MPICH > and OpenCoarrays after it builds gfortran (gfortran requires MPI and > OpenCoarrays to support the parallel features of Fortran 2008 and Fortran > 2018). A diff of our work relative to the GCC trunk is at > http://bit.ly/2pQuzvS. Our steps so far include the following: > > > > 1. Jerry wrote an autotools build system for OpenCoarrays (on the > OpenCoarrays issue-343-gnu-build branch.) > > 2. Jerry got MPICH and OpenCoarrays tar balls onto the GCC ftp server. > > 3. I modified GCC=E2=80=99s contrib/download_prerequisites script to do= wnload > the tar balls. > > 4. Daniel edited GCC=E2=80=99s Makefile.def, config.ac, and Makefile.tp= l to > incorporate MPICH. > > > > Our current issue is that the Makefile produced as follows doesn=E2=80= =99t > create the desired =E2=80=9Cmpi" directory: > > > > git clone -b build-mpich https://github.com/sourceryinstitute/gcc > > mkdir -p gcc/build > > cd gcc/build > > ../configure --disable-bootstrap --enable-languages=3Dfortran > > > > whereupon "grep mpi Makefile=E2=80=9D only returns comments and an echo > statement. Ultimately, our desired sequence is to build gfortran, then > build MPICH, and then build OpenCoarrays. All advice is welcome. > > You probably only want a new target_module for the MPI library. Note > it's name has to match that of the directory containing the sources > which > as far as I see is 'mpich', not 'libmpi'. > > May I ask again why mpi/opencoarrays should be built together with > gfortran rather than separately? Does gfortran or any of its current > target libraries > link against them? > > Richard. > > > Damian > > > > > > > > > > > > > > > > > > ---------- Forwarded message ---------- > From: Alastair McKinstry > To: Zaak Beekman , fortran@gcc.gnu.org > Cc: Toon Moene , "Gerald \"Jerry\" DeLisle" < > jvdelisle@charter.net>, Damian Rouson > Bcc: > Date: Tue, 3 Apr 2018 15:36:44 +0100 > Subject: Re: Coarrays in Debian Testing. > Hi Toon et al, > > I'm the maintainer of the coarrays package in Debian. > > (1) They're not automated, though I typically turn around new releases > within a week. We have a watch mechanism to check for new releases on > the webpage; the process needs to be manual as it checks for things > like new code without copyright and licenses; license changes, etc. > > (2) Its currently tested against openmpi (2.1.1 ; 3.0) but not mpich. I > need to extend the build system to include mpich builds. OpenMPI is the > default MPI on all Debian archs at the moment. > > Its tested against gfortran 7.3.0 ; I have manually tested against 8.0, > but not recently (last 2 months). > > The build procedure runs tests automatically, It fails on Powerpc > (Big-Endian) and Alpha archs, but likely to be due to openmpi issues on > those builds. > > Failed image support is not tested because it is not supported in > OpenMPI yet. > > I should examine the gcc build process that uses MPICH and extend it to > use OpenMPI as an option, and see whats needed for Failed Images in > OpenMPI: having gfortran depend on mpich while openmpi is the default > MPI in Debian would be .. unhelpful. > > regards > Alastair > > > On 31/03/2018 20:12, Zaak Beekman wrote: > > Hi Toon et al., > > > > FYI, I'm not sure how up-to-date thos debian packages are. We have work= ed > > with the Debian package maintainer in the past but I'm not sure if they > > regularly create .debs for our releases through an automated system, or > if > > someone needs to manually update the .debs. > > > > I would be curious to know: 1) which version they have (try running `caf > > --version` assuming they package the caf wrapper script) and 2) how you > > make out running it. > > > > Most tests that Jerry pointed to you are in the form of stand alone > fortran > > programs that will print "Test passed." if there are no test failures. > Stay > > away from the integration tests because those are more complicated to > > build. > > > https://github.com/sourceryinstitute/OpenCoarrays/tree/master/src/tests/u= nit > is > > the best place to start and > > > https://github.com/sourceryinstitute/OpenCoarrays/tree/master/src/tests/r= egression/reported > > but > > be aware that some tests require certain MPI features and some tests are > > only compatible with GFortran >=3D 7 or GFortran >=3D 8. (e.g., failed = images > > needs experimental MPIX routines only implemented in MPICH >=3D 3.2 and > even > > then, there is an MPICH bug that Andre fixed which was merged with mast= er > > but is not in any release yet that can cause some other tests to fail.) > > > > There is also an effort started to get an RPM package into fedora. Anyo= ne > > who wants to take a look or offer advice should check out > > https://bugzilla.redhat.com/show_bug.cgi?id=3D1560874 > > > > Best, > > Zaak > > > > > > > > On Fri, Mar 30, 2018 at 8:45 PM wrote: > > > >> fortran Digest 31 Mar 2018 00:44:45 -0000 Issue 4644 > >> > >> Topics (messages 50150 through 50154): > >> > >> Re: [patch, fortran] Fix PR 85111, some more zero size stuff > >> 50150 by: Jerry DeLisle > >> > >> [patch, fortran, testsuite, committed] Remove illegal code from > >> substr_6.f90 > >> 50151 by: Thomas Koenig > >> > >> Coarrays in Debian Testing. > >> 50152 by: Toon Moene > >> 50153 by: Jerry DeLisle > >> > >> Patching the GCC build system to build MPICH and OpenCoarrays > >> 50154 by: Damian Rouson > >> > >> Administrivia: > >> > >> To subscribe to the digest, e-mail: > >> fortran-digest-subscribe@gcc.gnu.org > >> > >> To unsubscribe from the digest, e-mail: > >> fortran-digest-unsubscribe@gcc.gnu.org > >> > >> To post to the list, e-mail: > >> fortran@gcc.gnu.org > >> > >> > >> ---------------------------------------------------------------------- > >> > >> > >> > >> ---------- Forwarded message ---------- > >> From: Jerry DeLisle > >> To: Thomas Koenig , "fortran@gcc.gnu.org" < > >> fortran@gcc.gnu.org>, gcc-patches > >> Cc: > >> Bcc: > >> Date: Thu, 29 Mar 2018 19:30:47 -0700 > >> Subject: Re: [patch, fortran] Fix PR 85111, some more zero size stuff > >> On 03/29/2018 11:07 AM, Thomas Koenig wrote: > >>> Hello world, > >>> > >>> the attached patch fixes PR 85111, a regression introduced with my > >>> recent simplification patches. Seems like the zero-size saga > >>> has yet another chapter :-) > >>> > >>> Regression-tested. OK for trunk? > >>> > >> Looks OK, > >> > >> Thanks, > >> > >> Jerry > >>> Regards > >>> > >>> Thomas > >>> > >>> 2017-03-29 Thomas Koenig > >>> > >>> PR fortran/85111 > >>> * array.c (gfc_resolve_character_array_constructor): Early > >>> exit for zero-size arrays. > >>> * simplify.c (simplify_transformation_to_array): Exit early > >>> if the result size is zero. > >>> (simplify_minmaxloc_to_array): Likewise. > >>> > >>> 2017-03-29 Thomas Koenig > >>> > >>> PR fortran/85111 > >>> * gfortran.dg/zero_sized_10.f90: New test. > >> > >> > >> > >> ---------- Forwarded message ---------- > >> From: Thomas Koenig > >> To: "fortran@gcc.gnu.org" , gcc-patches < > >> gcc-patches@gcc.gnu.org> > >> Cc: > >> Bcc: > >> Date: Fri, 30 Mar 2018 14:18:42 +0200 > >> Subject: [patch, fortran, testsuite, committed] Remove illegal code fr= om > >> substr_6.f90 > >> Hello world, > >> > >> F2003, 6.1.1. "Substrings" states that > >> > >> Both the starting point and the ending > >> point shall be within the range 1, 2, ..., n unless the starting point > >> exceeds the ending point, in which > >> case the substring has length zero. > >> > >> The attached patch, which corrects the test case, has > >> been committed as obvious. > >> > >> What's left is the accepts-invalid bug. > >> > >> Regards > >> > >> Thomas > >> > >> 2018-03-30 Thomas Koenig > >> > >> PR fortran/85130 > >> * gfortran.dg/substr_6.f90: Remove illegal test for > >> out-of-bounds substring. > >> > >> > >> > >> ---------- Forwarded message ---------- > >> From: Toon Moene > >> To: gfortran > >> Cc: > >> Bcc: > >> Date: Fri, 30 Mar 2018 16:19:29 +0200 > >> Subject: Coarrays in Debian Testing. > >> I just noticed that the following packages were available in Debian > >> Testing: > >> > >> libcaf-mpi2 - Co-Array Fortran libraries for gfortran > >> libcoarrays-dev - Co-Array Fortran libraries for gfortran - development > >> files > >> open-coarrays-bin - Executables for running Co-Array Fortran programs > >> > >> so I installed them. > >> > >> Are there any test programs to test this Debian installation (before I > >> am going proclaim them "useful" ?) > >> > >> Thanks in advance, > >> > >> -- > >> Toon Moene - e-mail: toon@moene.org - phone: +31 346 214290 > <+31%20346%20214%20290> > >> <+31%20346%20214%20290> > >> Saturnushof 14, 3738 XG Maartensdijk, The Netherlands > >> At home: http://moene.org/~toon/; weather: http://moene.org/~hirlam/ > >> Progress of GNU Fortran: http://gcc.gnu.org/wiki/GFortran#news > >> > >> > >> > >> ---------- Forwarded message ---------- > >> From: Jerry DeLisle > >> To: Toon Moene , gfortran > >> Cc: > >> Bcc: > >> Date: Fri, 30 Mar 2018 10:03:32 -0700 > >> Subject: Re: Coarrays in Debian Testing. > >> On 03/30/2018 07:19 AM, Toon Moene wrote: > >>> I just noticed that the following packages were available in Debian > >>> Testing: > >>> > >>> libcaf-mpi2 - Co-Array Fortran libraries for gfortran > >>> libcoarrays-dev - Co-Array Fortran libraries for gfortran - developme= nt > >>> files > >>> open-coarrays-bin - Executables for running Co-Array Fortran programs > >>> > >>> so I installed them. > >>> > >>> Are there any test programs to test this Debian installation (before I > >>> am going proclaim them "useful" ?) > >>> > >>> Thanks in advance, > >>> > >> Go here and look in these test directories: > >> > >> https://github.com/sourceryinstitute/OpenCoarrays/tree/master/src/tests > >> > >> There are examples of pde solvers etc. > >> > >> Jerry > >> > >> > >> > >> ---------- Forwarded message ---------- > >> From: Damian Rouson > >> To: GCC Development , gfortran > >> Cc: Daniel Celis Garza , Jerry DeLisle > < > >> jvdelisle@charter.net>, "Filippone, Salvatore" < > >> salvatore.filippone@cranfield.ac.uk> > >> Bcc: > >> Date: Fri, 30 Mar 2018 17:44:37 -0700 > >> Subject: Patching the GCC build system to build MPICH and OpenCoarrays > >> All, > >> > >> Jerry DeLisle, Daniel Celis Garza, and I would greatly appreciate > feedback > >> on our approach to patching the GCC build system to build MPICH and > >> OpenCoarrays after it builds gfortran (gfortran requires MPI and > >> OpenCoarrays to support the parallel features of Fortran 2008 and > Fortran > >> 2018). A diff of our work relative to the GCC trunk is at > >> http://bit.ly/2pQuzvS. Our steps so far include the following: > >> > >> 1. Jerry wrote an autotools build system for OpenCoarrays (on the > >> OpenCoarrays issue-343-gnu-build branch.) > >> 2. Jerry got MPICH and OpenCoarrays tar balls onto the GCC ftp server. > >> 3. I modified GCC=E2=80=99s contrib/download_prerequisites script to d= ownload > the > >> tar balls. > >> 4. Daniel edited GCC=E2=80=99s Makefile.def, config.ac, and Makefile.t= pl to > >> incorporate MPICH. > >> > >> Our current issue is that the Makefile produced as follows doesn=E2=80= =99t > create > >> the desired =E2=80=9Cmpi" directory: > >> > >> git clone -b build-mpich https://github.com/sourceryinstitute/gcc > >> mkdir -p gcc/build > >> cd gcc/build > >> ../configure --disable-bootstrap --enable-languages=3Dfortran > >> > >> whereupon "grep mpi Makefile=E2=80=9D only returns comments and an echo > >> statement. Ultimately, our desired sequence is to build gfortran, then > >> build MPICH, and then build OpenCoarrays. All advice is welcome. > >> > >> Damian > >> > >> > >> > >> > >> > >> > >> > >> > -- > Alastair McKinstry, , , > https://diaspora.sceal.ie/u/amckinstry > Misentropy: doubting that the Universe is becoming more disordered. > > >