I’m running into deeper problems than I expected on this solver bug fix. I
need some more time to trouble shoot.

What’s our deadline? I read something about a freeze till March 1st?

Regarding standard gpus for yade: No, suitesparse depends on double
precision compute power and standard graphics cards are generally weaker
than a multi core cpu for double precision compute. That said, I noticed
during my testing today that using cholmod directly for cpu solving does
seem to re enable multi core factorization (something we used to have
before our third party libs stopped playing nicely, right?). So we should
get a bump in cpu performance...after I get this bug worked out.

On Mon, Feb 12, 2018 at 10:51 AM Bruno Chareyre <
bruno.chare...@grenoble-inp.fr> wrote:

> Awesome.
> Years ago we were using cholmod without the eigen interface, it will be a
> revert...
> These cholmod/openblas issues make me nervous (not mentionning other
> attempts with Pardiso and Taucs, which all worked at some point). It still
> did not stabilize after 6+ years.
>
> Do you think the GPU solver has benefits even with a standard graphical
> card, by the way?
> Cheers
>
> Bruno
>
>
>
> On 02/12/2018 06:52 PM, Robert Caulk wrote:
>
> Yes! We should be able to replace the Eigen cholmod interface entirely
> with the existing direct cholmod solver (useSolver=4). Right now the
> #ifdefs are configured for the special GPU build but I will generalize it
> today and commit it so that it can be used without GPU compilation flags.
> Cholmod is smart enough to revert to CPU on its own if GPU is not available.
>
> I will install 18.04, test the solver, and report back with my findings
> today.
>
> Best,
>
> Robert
>
> On Mon, Feb 12, 2018 at 1:22 AM Bruno Chareyre <
> bruno.chare...@grenoble-inp.fr> wrote:
>
>> Thank you very much for help Janek.
>> I confirm that I could compile trunk on ubuntu18 with CGAL, without
>> special steps (not even dpkg, just cmake+make without options).
>> For DEM-PFV-check.py the problem is with the cholmod solver. I'll try and
>> fix that.
>>
>> @Robert, you mentioned that by-passing the eigen interface to use cholmod
>> directly solved issues on your side IIRC.
>> Could you give more hints about that, do you have some candidate code for
>> commit?
>>
>> Cheers
>>
>>
>> Bruno
>>
>>
>> On 02/11/2018 12:33 PM, janek_li...@wp.pl wrote:
>>
>> OK, so I did the initial commit for cgal 4.11, the most importatnt stuff
>> is in the commit message. I will repeat it here in short:
>>
>> 1.  maybe the #ifdefs that I added in
>> lib/triangulation/RegularTriangulation.h are not necessary?
>>
>> 2. the ALPHASHAPES changes inlib/triangulation/Tesselation.h are not
>> related to cgal, but I couldn't compile without this.
>>
>> 3. In checkPolyhedraCrush.py this python line freezes:
>>      O.run(250, True); checkForcesBodies(25.44893, 4)
>>
>> Moreover:
>> 4. with this patch yade is compiling successfully with cgal 4.9 and 4.11.
>>
>> 5. regardless of cgal version (with 4.9 and 4.11) I have this error
>> in DEM-PFV-check.py:
>>
>>   running:  DEM-PFV-check.py
>>   DEM-PFV: unbalanced Qin vs. Qout ( -0.00279199571618  vs.  0.0 )
>>   DEM-PFV: difference in permeability: 27.9199571618  vs. target
>> 0.040399916554
>>   The difference is more, than the critical tolerance!
>>   DEM-PFV: difference in final pressure: 0.0  vs. target  628.314160434
>>   The difference is more, than the critical tolerance!
>>   DEM-PFV: difference in final deformation -0.00430506304298  vs. target
>> -0.00258113045083
>>   The difference is more, than the critical tolerance!
>>   Status: FAILURE!!!
>>
>> Since it does not depend on cgal version, it must be some other library
>> that I have on my workstation. I have here installed devuan ascii, and I
>> only backported cgal 4.11 from devuan ceres. All other libraries are at
>> their default devuan ascii version.
>>
>> 6. Also  yade version is clean, right from github :) Just to make it all
>> clear. However I could not compile yade just by doing the usual
>>
>>   cd .. ; mkdir build ; cd build
>>   cmake -DCMAKE_INSTALL_PREFIX=../install ../trunk -DDEBUG=1 -DCHUNKSIZE=5
>>   time nice -n 20 make install -j 20 ; xmessage "\nFinished\n"
>>
>> I had some boost errors related to the C++11 declytype support, eg. as
>> discussed in:
>> http://www.boost.org/doc/libs/1_66_0/libs/utility/utility.htm
>> At first I was worried that g++ 6.3 is too old. However I found out that
>> I can compile yade 2018-02 if I copy the ./debian/ directory from yade
>> 2017-01 and do:
>>
>>    time dpkg-buildpackage -rfakeroot -b -j30
>>
>> So it must not be related to compiler version, but rather to compiler
>> flags. And Anton fixed this in debian/ build rules, while my command could
>> not make it correctly. I didn't figure this out yet. Last night I was
>> simply building yade 2018-02 with debian/ from yade 2017-01 :)
>>
>> best regards
>> Janek
>>
>>
>> Dnia 9 lutego 2018 18:46 Anton Gladky <gladky.an...@gmail.com>
>> napisał(a):
>>
>> Yes, it should be OK.
>>
>> Anton
>>
>>
>> 2018-02-09 16:06 GMT+01:00 Bruno Chareyre <bruno.chare...@grenoble-inp.fr
>> >:
>>
>> What is an appropriate system to try this? ubuntu 18.04 beta is ok?
>> B
>>
>>
>> On 02/08/2018 07:35 PM, Janek Kozicki (yade-dev) wrote:
>>
>>
>> Hi Anton and Bruno,
>>
>> I am glad that I can help, and use this opportunity to get back in track
>> ;) So I will do this entire Saturday after I get back home on friday
>> night.
>>
>> Should I work on latest yade trunk?
>>
>> BTW: I'm writing this from an airport ;)
>>
>> Best regards,
>> Janek
>>
>> On 8 Feb 2018, 19:10 +0100, Anton Gladky <gladky.an...@gmail.com>, wrote:
>>
>>
>> Hi Bruno,
>>
>> CGAL_ 4.11 is the only version now for Debian (testing) [1]
>> and upcoming Ubuntu 18.04 [2]. The shipped Yade does
>> not support CGAL due to compilation problems.
>>
>> I am preparing the new Yade upload, but we have a chance
>> to patch the Yade within the next two weeks or prepare
>> the 2018.02b release with CGAL-support and upload it.
>>
>> [1] https://tracker.debian.org/pkg/cgal
>> [2] https://launchpad.net/ubuntu/+source/cgal
>>
>> Anton
>>
>>
>> 2018-02-08 18:45 GMT+01:00 Bruno Chareyre
>> <bruno.chare...@grenoble-inp.fr>:
>>
>>
>> Hi Anton,
>> Thank you very much.
>> I'm not so sure what we are speaking about here.
>> Yade 2018.02a is the candidate source code for producing a binary
>> yade-stable in Ubuntu 18.04, correct (approximately)?
>> If yes, can we build yade-stable with CGAL at the moment?
>> Bruno
>>
>>
>>
>> On 02/08/2018 06:35 PM, Anton Gladky wrote:
>>
>>
>>
>> Well, if we find a way to fix it within the next 2 weeks,
>> I think there is a chance to get it pushed into Debian->Ubuntu.
>>
>> Regards
>>
>> Anton
>>
>>
>> 2018-02-08 15:23 GMT+01:00 Bruno Chareyre
>> <bruno.chare...@grenoble-inp.fr>:
>>
>>
>>
>>
>> On 02/07/2018 05:35 PM, Janek Kozicki (yade-dev) wrote:
>>
>>
>>
>>
>> Regarding compiling yade with CGAL 4.11, I can look into it on
>> Saturday,
>> if my last patches didn't work for you?
>>
>>
>>
>> I did not test it yet I'm afraid. :-/
>> It sounds like a critical issue for a 18.04 release. Is it?
>> Would that mean to skip the whole CGAL related code in the binary
>> release?
>> I'll go back to this asap, thanks for pinging.
>>
>> Bruno
>>
>>
>>
>>
>> _______________________________________________
>> Mailing list: https://launchpad.net/~yade-dev
>> Post to : yade-dev@lists.launchpad.net
>> Unsubscribe : https://launchpad.net/~yade-dev
>> More help : https://help.launchpad.net/ListHelp
>>
>>
>>
>>
>>
>>
>> _______________________________________________
>> Mailing list: https://launchpad.net/~yade-dev
>> Post to : yade-dev@lists.launchpad.net
>> Unsubscribe : https://launchpad.net/~yade-dev
>> More help : https://help.launchpad.net/ListHelp
>>
>>
>>
>> _______________________________________________
>> Mailing list: https://launchpad.net/~yade-dev
>> Post to : yade-dev@lists.launchpad.net
>> Unsubscribe : https://launchpad.net/~yade-dev
>> More help : https://help.launchpad.net/ListHelp
>>
>>
>>
>>
>>
>> _______________________________________________
>> Mailing list: https://launchpad.net/~yade-dev
>> Post to     : yade-dev@lists.launchpad.net
>> Unsubscribe : https://launchpad.net/~yade-dev
>> More help   : https://help.launchpad.net/ListHelp
>>
>>
>> _______________________________________________
>> Mailing list: https://launchpad.net/~yade-dev
>> Post to     : yade-dev@lists.launchpad.net
>> Unsubscribe : https://launchpad.net/~yade-dev
>> More help   : https://help.launchpad.net/ListHelp
>>
>>
>>
>>
>> _______________________________________________
>> Mailing list: https://launchpad.net/~yade-dev
>> Post to     : yade-dev@lists.launchpad.net
>> Unsubscribe : https://launchpad.net/~yade-dev
>> More help   : https://help.launchpad.net/ListHelp
>>
>>
>>
> _______________________________________________
> Mailing list: https://launchpad.net/~yade-dev
> Post to     : yade-dev@lists.launchpad.net
> Unsubscribe : https://launchpad.net/~yade-dev
> More help   : https://help.launchpad.net/ListHelp
>
_______________________________________________
Mailing list: https://launchpad.net/~yade-dev
Post to     : yade-dev@lists.launchpad.net
Unsubscribe : https://launchpad.net/~yade-dev
More help   : https://help.launchpad.net/ListHelp

Reply via email to