[petsc-dev] Seeking OLCF users complaining about poor build times

[Barry Smith]

  Satish,

    Thanks,  unfortunately since this number is so good this means you need to run another test with master and all packages on mira. (by my reckoning the Metis, hypre and superlu_dist all support parallel builds so should be much faster with master)

    Maybe mira is not such a bad compile server after all.

  Barry

> On Feb 26, 2015, at 5:14 PM, Satish Balay <balay at mcs.anl.gov> wrote:
> 
> Here is a comparision of PETSc-only build [without any externalpackage builds] on Mira
> 
> logs attached.
> 
> Satish
> 
> --------
> mira.alcf - without externalpackages
> --------
> [balay at miralac1 petsc.maint.35]$ time ./reconfigure-arch-time-test-nopkg.py
> 
> real    6m57.989s
> user    3m3.418s
> sys     3m28.630s
> 
> [balay at miralac1 petsc.maint.35]$ time    make PETSC_DIR=/gpfs/mira-home/balay/petsc.maint.35 PETSC_ARCH=arch-time-test-nopkg all
> 
> real    1m37.266s
> user    22m21.087s
> sys     7m32.159s
> 
> [balay at miralac1 petsc.maint.35]$ uptime
> 22:55:09 up 10 days,  3:19, 82 users,  load average: 3.21, 4.65, 4.65
> 
> --------
> 
> [balay at miralac1 petsc.maint.35]$ cat reconfigure-arch-time-test-nopkg.py 
> #!/usr/bin/python
> if __name__ == '__main__':
>  import sys
>  import os
>  sys.path.insert(0, os.path.abspath('config'))
>  import configure
>  configure_options = [
>    '--with-blas-lapack-lib=-L/soft/libraries/alcf/current/xl/LAPACK/lib -llapack -L/soft/libraries/alcf/current/xl/BLAS/lib -lblas',
>    '--with-c2html=0',
>    '--with-cc=mpixlc_r',
>    '--with-cxx=mpixlcxx_r',
>    '--with-debugging=0',
>    '--with-fc=mpixlf77_r -qnosave',
>    '--with-shared-libraries=0',
>    '--with-x=0',
>    '-COPTFLAGS= -O3 -qhot=level=0 -qsimd=auto -qmaxmem=-1 -qstrict -qstrict_induction',
>    '-CXXOPTFLAGS= -O3 -qhot=level=0 -qsimd=auto -qmaxmem=-1 -qstrict -qstrict_induction',
>    '-FOPTFLAGS= -O3 -qhot=level=0 -qsimd=auto -qmaxmem=-1 -qstrict -qstrict_induction',
>    'PETSC_ARCH=arch-time-test-nopkg',
>  ]
>  configure.petsc_configure(configure_options)
> [balay at miralac1 petsc.maint.35]$ 
> 
> 
> On Thu, 26 Feb 2015, Satish Balay wrote:
> 
>> Attaching logs.
>> 
>> Note: 'uptime' is run a bit after the builds completed..
>> 
>> Satish
>> 
>> --------------
>> mira.alcf build: MAKE_NP = 32
>> --------------
>> [balay at miralac1 petsc.maint.35]$ time ./reconfigure-arch-time-test.py
>> <snip>
>> real    25m57.240s
>> user    17m33.944s
>> sys     6m41.065s
>> 
>> [balay at miralac1 petsc.maint.35]$ time    make PETSC_DIR=/gpfs/mira-home/balay/petsc.maint.35 PETSC_ARCH=arch-time-test all
>> <snip>
>> real    1m39.067s
>> user    22m56.248s
>> sys     7m42.658s
>> 
>> [balay at miralac1 petsc.maint.35]$ uptime
>> 21:50:08 up 10 days,  2:14, 87 users,  load average: 3.86, 5.02, 5.03
>> 
>> --------------
>> Laptop build: MAKE_NP = 4
>> --------------
>> $ time ./configure --download-hypre --download-metis --download-parmetis --download-superlu_dist --download-fblaslapack --with-debugging=0 --with-c2html=0 PETSC_ARCH=arch-test COPTFLAGS="-O3 -fast" FOPTFLAGS="-O3 -fast" CXXOPTFLAGS="-O3 -fast"
>> <snip>
>> real    3m56.755s
>> user    2m47.458s
>> sys     0m57.206s
>> 
>> $ time    make PETSC_DIR=/home/balay/petsc PETSC_ARCH=arch-test all
>> <snip>
>> real    1m13.519s
>> user    3m40.708s
>> sys     0m47.002s
>> 
>> $ uptime
>> 15:50:32 up 3 days,  5:20, 11 users,  load average: 0.05, 0.30, 0.49
>> 
>> --------------
>> 
>> [balay at miralac1 petsc.maint.35]$ cat reconfigure-arch-time-test.py 
>> #!/usr/bin/python
>> if __name__ == '__main__':
>>  import sys
>>  import os
>>  sys.path.insert(0, os.path.abspath('config'))
>>  import configure
>>  configure_options = [
>>    '--download-hypre=1',
>>    '--download-metis=1',
>>    '--download-parmetis=1',
>>    '--download-superlu_dist=1',
>>    #'--with-blas-lapack-lib=-L/soft/libraries/alcf/current/xl/LAPACK/lib -llapack -L/soft/libraries/alcf/current/xl/BLAS/lib -lblas',
>>    '--download-fblaslapack=1',
>>    '--with-c2html=0',
>>    '--with-cc=mpixlc_r',
>>    '--with-cxx=mpixlcxx_r',
>>    '--with-debugging=0',
>>    '--with-fc=mpixlf77_r -qnosave',
>>    '--with-shared-libraries=0',
>>    '--with-x=0',
>>    '-COPTFLAGS= -O3 -qhot=level=0 -qsimd=auto -qmaxmem=-1 -qstrict -qstrict_induction',
>>    '-CXXOPTFLAGS= -O3 -qhot=level=0 -qsimd=auto -qmaxmem=-1 -qstrict -qstrict_induction',
>>    '-FOPTFLAGS= -O3 -qhot=level=0 -qsimd=auto -qmaxmem=-1 -qstrict -qstrict_induction',
>>    'PETSC_ARCH=arch-time-test',
>>  ]
>>  configure.petsc_configure(configure_options)
>> 
>> 
>>>>>> On Feb 26, 2015, at 8:13 AM, Nathan Collier <nathaniel.collier at gmail.com>
>>>>> wrote:
>>>>>> 
>>>>>> Ok, so I built PETSc with metis, parmetis, superlu_dist, and hyper on
>>>>> Titan. The configure time is the second configure--when you run the
>>>>> reconfigure script that the batch submission generates for you.
>>>>>> 
>>>>>> configure:  38m15.488s
>>>>>> make: 15m37.610s
>>>>>> 
>> 
> <configure_mira_nopkg.log.gz><make_mira_nopkg.log.gz>