yade-dev team mailing list archive

[Robert Caulk <rcaulk@xxxxxxxxxxxx>]

Ok, I pushed the fix for the DEM-PFV-check.py failure. I discovered the
issue was compiler related. GCC 5.4 on ubuntu 16.04 initialized
factorizeOnly to false by default, while GCC 7.2 on ubuntu 18.04 did not do
this. Therefore, the only necessary change ended up being the explicit
initialization of factorizeOnly=false.

My experience compiling, installing, and checking Yade master on Ubuntu
18.04:

   - The library installation requires libqglviewer-dev-*qt5 *instead of
   libqglviewer-dev. Otherwise, the libraries install with the same commands
   as on ubuntu 16.04.
   - Clean build is the same as ubuntu 16.04:
   cmake -DCMAKE_INSTALL_PREFIX=../install ../trunk
   - Clean make/install is the same as 16.04: make -jX install
   - Checks (make check) all pass up until checkPolyhedraCrush.py, where it
   hangs. I see others noticed this issue as well.

In addition to the bug fix, I edited the solvers so flow.useSolvers=3 and 4
can now be used with the default CPU build. I can confirm that
flow.useSolver=4 enables multicore CPU factorization, while
flow.useSolver=3 sticks to 1 core factorization :-).

Cheers,

Robert

On Tue, Feb 13, 2018 at 5:20 AM, Bruno Chareyre <
bruno.chareyre@xxxxxxxxxxxxxxx> wrote:

>
>
> On 02/13/2018 10:12 AM, Robert Caulk wrote:
>
>> I’m running into deeper problems than I expected on this solver bug fix.
>> I need some more time to trouble shoot.
>>
> Let me know, I can also try some things. I have been able at least to
> compile and run cholmod on 18.04 with simple c++ programs.
> I wanted to just replicate a hello-cholmod function inside yade to see if
> the problems come from some specific compilation/link options used for yade
> compilation.
>
> What’s our deadline? I read something about a freeze till March 1st?
>>
> Yes beggining of March IIRC.
>
>
>> 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.
>>
> I was expecting that, thanks.
>
> 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?).
>>
> Indeed, it used to work.
>
>
> Bruno
>
>
>
> _______________________________________________
> Mailing list: https://launchpad.net/~yade-dev
> Post to     : yade-dev@xxxxxxxxxxxxxxxxxxx
> Unsubscribe : https://launchpad.net/~yade-dev
> More help   : https://help.launchpad.net/ListHelp
>