Re: [AMBER-Developers] Sander parallel build broken by PBSA updates

From: Daniel Roe <daniel.r.roe.gmail.com>
Date: Thu, 18 Feb 2016 10:13:52 -0700

On Wed, Feb 17, 2016 at 6:04 PM, Ray Luo, Ph.D. <ray.luo.uci.edu> wrote:
> Ross,
>
> Please pull again to see whether it fixed the build problem on your
> side. Looks like multi-thread make does not work ... Too many
> dependence issues ...

The multi-threaded build definitely doesn't work for me when I hit pbsa stuff:

$ make -j6 install
...
make[2]: Entering directory '/home/droe/Amber/GIT/amber/AmberTools/src/pbsa'
gcc -c -O3 -mtune=native -fPIC ...
...
Makefile:238: recipe for target 'pb_p3m.LIBPBSA.o' failed
Makefile:238: recipe for target 'pb_fdfrc.LIBPBSA.o' failed

Amber is pretty large nowadays, and not being able to build in
parallel is a real drawback. We can currently get around this by
adding a '.NOTPARALLEL' target to the pbsa Makefile (or forcing -j1),
but ideally I think the multi-threaded build should be fixed.

-Dan

>
> All the best,
> Ray
> --
> Ray Luo, Ph.D.
> Professor
> Biochemistry, Molecular Biophysics, Chemical Physics,
> Chemical and Biomedical Engineering
> University of California, Irvine, CA 92697-3900
>
>
> On Wed, Feb 17, 2016 at 12:59 PM, Ray Luo, Ph.D. <ray.luo.uci.edu> wrote:
>> Okay, I think I've fixed it. Don't know why the same makefile works in
>> amber15 but doesn't work in amber16. I'm testing all the
>> sander/pbsa/nab combs to make sure all can build without interruption
>> and also pass the tests.
>>
>> Ray
>> --
>> Ray Luo, Ph.D.
>> Professor
>> Biochemistry, Molecular Biophysics, Chemical Physics,
>> Chemical and Biomedical Engineering
>> University of California, Irvine, CA 92697-3900
>>
>>
>> On Tue, Feb 16, 2016 at 10:34 PM, Ray Luo, Ph.D. <ray.luo.uci.edu> wrote:
>>> Finally I can reproduce your problem in building the MPI sander ...
>>>
>>> Will get it fixed tomorrow morning ...
>>>
>>> Ray
>>> --
>>> Ray Luo, Ph.D.
>>> Professor
>>> Biochemistry, Molecular Biophysics, Chemical Physics,
>>> Chemical and Biomedical Engineering
>>> University of California, Irvine, CA 92697-3900
>>>
>>>
>>> On Tue, Feb 16, 2016 at 10:17 PM, Ray Luo, Ph.D. <ray.luo.uci.edu> wrote:
>>>> Hi Ross,
>>>>
>>>> Looks like that there is no extra untracked file at least in all the
>>>> src folders, though there were indeed two .swp files in the test
>>>> cases. So I think lack of "git clean -fdx" is not the problem. I've
>>>> finished the serial build and is running "test.sander.BASIC". Looking
>>>> good, so far.
>>>>
>>>> I'll do the MPI build next after another "git clean -fdx" ...
>>>>
>>>> However, I think one possible reason could be the use of multi-thread
>>>> in make. There may be inter-dependent issues involved.
>>>>
>>>> I'll do some experiment with the MPI build to see whether this is the cause.
>>>>
>>>> Ray
>>>> --
>>>> Ray Luo, Ph.D.
>>>> Professor
>>>> Biochemistry, Molecular Biophysics, Chemical Physics,
>>>> Chemical and Biomedical Engineering
>>>> University of California, Irvine, CA 92697-3900
>>>>
>>>>
>>>> On Tue, Feb 16, 2016 at 8:24 PM, Ross Walker <ross.rosswalker.co.uk> wrote:
>>>>> Hi Ray,
>>>>>
>>>>> It's not about whether you compiled in that folder - it is whether you maybe edited files in that folder that you never committed / pushed. Try doing:
>>>>>
>>>>> git clean -f -d -x
>>>>> git status
>>>>>
>>>>> And see what it reports.
>>>>>
>>>>> All the best
>>>>> Ross
>>>>>
>>>>>> On Feb 16, 2016, at 23:04, Ray Luo <rluo.uci.edu> wrote:
>>>>>>
>>>>>> Hi Ross,
>>>>>>
>>>>>> Yes, I think my folder is clean since I have never compiled anything
>>>>>> in my amber git folder. Every time after syncing with the master
>>>>>> branch, I copy the whole master branch to a new folder for compiling
>>>>>> and testing.
>>>>>>
>>>>>> I'm about to test this evening's synced version, but will do "git
>>>>>> clean -fxd" before configuring/compiling. Maybe this is why.
>>>>>>
>>>>>> By the way, could you add my uci email to the mailing list as well,
>>>>>> i.e. "ray.luo.uci.edu"?
>>>>>>
>>>>>> All the best,
>>>>>> Ray
>>>>>> --
>>>>>> Ray Luo, Ph.D.
>>>>>> Professor
>>>>>> Biochemistry, Molecular Biophysics, Chemical Physics,
>>>>>> Chemical and Biomedical Engineering
>>>>>> University of California, Irvine, CA 92697-3900
>>>>>>
>>>>>>
>>>>>> On Tue, Feb 16, 2016 at 7:51 PM, Ross Walker <ross.rosswalker.co.uk> wrote:
>>>>>>> Hi Ray,
>>>>>>>
>>>>>>> If I do this - on my OSX laptop (GCC 4.9.3), MPICH version 3.1.4
>>>>>>>
>>>>>>> git pull
>>>>>>> git clean -f -d -x
>>>>>>> git status
>>>>>>> On branch master
>>>>>>> Your branch is up-to-date with 'origin/master'.
>>>>>>> nothing to commit, working directory clean
>>>>>>>
>>>>>>> ./configure -mpi gnu
>>>>>>> make install
>>>>>>>
>>>>>>> ...
>>>>>>> ...
>>>>>>> dnrm2.F90:113.24:
>>>>>>>
>>>>>>> ASSIGN 110 TO NEXT
>>>>>>> 1
>>>>>>> Warning: Deleted feature: ASSIGN statement at (1)
>>>>>>> mpif90 -DBINTRAJ -DEMIL -DMPI -c -O3 -mtune=native -fPIC -ffree-form -I/Users/rcw/amber/amber/include -I/Users/rcw/amber/amber/include -o dcopy.o dcopy.F90
>>>>>>> mpif90 -DBINTRAJ -DEMIL -DMPI -c -O3 -mtune=native -fPIC -ffree-form -I/Users/rcw/amber/amber/include -I/Users/rcw/amber/amber/include -o pb_force.o pb_force.F90
>>>>>>> pb_force.F90:292.77:
>>>>>>>
>>>>>>> sa_init, sa_driver, sa_free, sa_free_mb, &
>>>>>>> 1
>>>>>>> Error: Symbol 'saslave_init' referenced at (1) not found in module 'solvent_accessibility'
>>>>>>> make[3]: *** [pb_force.o] Error 1
>>>>>>> make[2]: *** [libpbsa] Error 2
>>>>>>> make[1]: *** [parallel] Error 2
>>>>>>> make: *** [install] Error 2
>>>>>>>
>>>>>>> Same problem. Are you sure you have the latest tree and it is clean and doesn't have any modified files locally or files you forgot to add?
>>>>>>>
>>>>>>> All the best
>>>>>>> Ross
>>>>>>>
>>>>>>>
>>>>>>>> On Feb 16, 2016, at 15:56, Ray Luo, Ph.D. <ray.luo.uci.edu> wrote:
>>>>>>>>
>>>>>>>> Hi Ross,
>>>>>>>>
>>>>>>>> So far I can "configure -mpi gnu", "make install", and can also pass
>>>>>>>> "make sander.BASIC.MPI". I'm running "make test.parallel.MM". I'm
>>>>>>>> using "mpirun -np 4" for the mpi jobs. This is on Rocks 6.1/Centos
>>>>>>>> 6.1. Which step was the problem?
>>>>>>>>
>>>>>>>> However I disabled "pbsa_mbfocus" in AmberTools/test/ since I'm in the
>>>>>>>> process of removing this feature to accommodate incoming CUDA. I also
>>>>>>>> need to update the test cases for sander/nab/mmpbsa ... which will be
>>>>>>>> partially in today for sander/nab. I'm trying to get mmpbsa.py to work
>>>>>>>> with the new python environment.
>>>>>>>>
>>>>>>>> All the best,
>>>>>>>> Ray
>>>>>>>> --
>>>>>>>> Ray Luo, Ph.D.
>>>>>>>> Professor
>>>>>>>> Biochemistry, Molecular Biophysics, Chemical Physics,
>>>>>>>> Chemical and Biomedical Engineering
>>>>>>>> University of California, Irvine, CA 92697-3900
>>>>>>>>
>>>>>>>>
>>>>>>>> On Tue, Feb 16, 2016 at 10:21 AM, Ray Luo, Ph.D. <ray.luo.uci.edu> wrote:
>>>>>>>>> Hi Ross,
>>>>>>>>>
>>>>>>>>> I'm looking into this ... There was a major overhaul of the code in
>>>>>>>>> addition to new features in the last check in.
>>>>>>>>>
>>>>>>>>> All the best,
>>>>>>>>>
>>>>>>>>> Ray
>>>>>>>>>
>>>>>>>>> --
>>>>>>>>> Ray Luo, Ph.D.
>>>>>>>>> Professor
>>>>>>>>> Biochemistry, Molecular Biophysics, Chemical Physics,
>>>>>>>>> Chemical and Biomedical Engineering
>>>>>>>>> University of California, Irvine, CA 92697-3900
>>>>>>>
>>>>>
>
> _______________________________________________
> AMBER-Developers mailing list
> AMBER-Developers.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber-developers



-- 
-------------------------
Daniel R. Roe, PhD
Department of Medicinal Chemistry
University of Utah
30 South 2000 East, Room 307
Salt Lake City, UT 84112-5820
http://home.chpc.utah.edu/~cheatham/
(801) 587-9652
(801) 585-6208 (Fax)
_______________________________________________
AMBER-Developers mailing list
AMBER-Developers.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber-developers
Received on Thu Feb 18 2016 - 09:30:04 PST
Custom Search