[molpro-user] ARMCI test errors after compiling with GA/MPICH2, gfortran, ATLAS on Ubuntu 11.04

[Andy May]

Greg,

These jobs run with no problem when I compile ATLAS and Molpro with 
exactly the same options on my machine. Of course there are small 
differences because of the different Linux flavours we are using (I'm 
using openSUSE 11.4, gcc 4.5.1), but I think it unlikely this is the issue.

Tracking back through this discussion, I see that:

./configure -batch -gcc -gfortran -mpp -auto-ga-tcgmsg-mpich2 -instroot 
/usr/local/molpro2010.1 -noblas -nolapack

worked, but that:

./configure -batch -gcc -gfortran -mpp -auto-ga-tcgmsg-mpich2 -instroot 
/usr/local/molpro2010.1 -blaspath /opt/ATLAS3.8.4/lib

does not. It therefore looks very much like the ATLAS on your machine is 
not functioning correctly, but I have no idea why since it is built with 
exactly the same options which work for me. It's pretty difficult for me 
to debug without being able to reproduce the failures, can you send me 
one of the .errout files from a failing job? Can you also try with a 
different external library, eg. ACML:

http://developer.amd.com/Downloads/acml-4-4-0-gfortran-64bit-int64.tgz

./configure -batch -gcc -gfortran -mpp -auto-ga-tcgmsg-mpich2 -instroot 
/usr/local/molpro2010.1 -blaspath /opt/acml4.4.0/gfortran64_int64/lib

to see what happens in this case?

Best wishes,

Andy

On 10/06/11 20:26, Gregory Magoon wrote:
> Thanks Andy. Although this looked promising at first, it seems to give
> an error
> on cn_samc_forces.test and hangs on cs2_qcitgrad_c2v.test. (For what it's
> worth, one of my earlier attempts in which I was using the Ubuntu atlas
> package
> encountered a hang on cn_samc_forces.test.) This time, I followed your
> example
> much more closely (no lapack and same flags as you in ATLAS compile, no
> shared
> object libraries for ATLAS, used /opt/ATLAS3.8.4 install location).
>
> Notes:
> #get dependencies
>
> sudo apt-get install mpich2
> sudo apt-get install curl
> sudo apt-get install gfortran
> sudo apt-get install zlib1g-dev
> sudo apt-get install libcr-dev
>
> #disable CPU throttling (http://ubuntuforums.org/showthread.php?t=611084)
> sudo /usr/bin/cpufreq-selector -g performance
>
>
> #from atlas build directory:
> #build8noLAPACK directory:
> ../configure --prefix=/opt/ATLAS3.8.4 -Si cputhrchk 0 -C if `which
> gfortran` -F
> if '-O3 -m64 -fdefault-integer-8'
> sed -i -e 's/F77_INTEGER=long/F77_INTEGER="long long"/' Make.inc
> make
> make check; make ptcheck
> make time
> sudo make install
>
> #(clear old ATLAS .so and LAPACK libraries)
>
> #compile molpro, using patch level 22
>
> ./configure -batch -gcc -gfortran -mpp -auto-ga-tcgmsg-mpich2 -instroot
> /usr/local/molpro2010.1 -blaspath /opt/ATLAS3.8.4/lib
> make
>
> #tuning, testing
> make patch #enter username, password to download token
> make tuning
> make quicktest
> nohup make MOLPRO_OPTIONS=-n8 test > test061011.txt &
> #***cn_samc_forces test fails; compare with attempt 3 where it hung on
> this job;
> timing for big_lattice: 107/108
>
> Test output:
> Running job Cs_DKH10.test
> Running job Cs_DKH2.test
> Running job Cs_DKH2_standard.test
> Running job Cs_DKH3.test
> Running job Cs_DKH4.test
> Running job Cs_DKH7.test
> Running job Cs_DKH8.test
> Running job Cs_nr.test
> Running job allene_opt.test
> Running job allyl_cipt2.test
> Running job allyl_ls.test
> Running job ar2_dk_dummy.test
> Running job au2o_optdftecp1.test
> Running job au2o_optdftecp2.test
> Running job au2o_optecp.test
> Running job aucs4k2.test
> Running job b_cidft.test
> Running job basisinput.test
> Running job bccd_opt.test
> Running job bccd_save.test
> Running job benz_nlmo.test
> Running job benzol_giao.test
> Running job big_lattice.test
> Running job br2_f12_multgem.test
> Running job c2f4_cosmo.test
> Running job c2h2_dfmp2.test
> Running job c2h4_c1_freq.test
> Running job c2h4_ccsd-f12.test
> Running job c2h4_ccsdfreq.test
> Running job c2h4_cosmo.test
> Running job c2h4_cosmo_direct.test
> Running job c2h4_d2.test
> Running job c2h4_d2h.test
> Running job c2h4_d2h_freq.test
> Running job c2h4_ksfreq.test
> Running job c2h4_lccsd.test
> Running job c2h4_lccsd2.test
> Running job c2h4_lccsd3.test
> Running job c2h4_lmp2.test
> Running job c2h4_optnum.test
> Running job c2h4_prop.test
> Running job c2h4o_cosmo.test
> Running job c6h6_freq.test
> Running job c6h6_freq_restart.test
> Running job c6h6_opt.test
> Running job caffeine_opt.test
> Running job casvb.test
> Running job casvb2.test
> Running job casvb3.test
> Running job ccsapt_hebe2_exdi_dcbs.test
> Running job ccsapt_hebe_df_ccdisp_dcbs.test
> Running job ccsapt_hebe_df_ccdisp_mcbs.test
> Running job ccsapt_hebe_df_ccexdi_dcbs.test
> Running job ccsapt_ne2ar_dcbs.test
> Running job ccsapt_ne2ar_mcbs.test
> Running job ccsapt_ne2ar_nosym_dcbs.test
> Running job ccsapt_ne2ar_tdhfdisp_dcbs.test
> Running job ccsapt_ne2ar_tdhfdisp_mcbs.test
> Running job ccsd_f12_dipmom.test
> Running job ch2_rs3.test
> Running job ch2o_cosmo_rad.test
> Running job ch3_cosmo.test
> Running job ch3_cosmo_dft.test
> Running job ch4_b2plyp.test
> Running job ch4_b2plyp_df.test
> Running job ch4_b2plyp_direct.test
> Running job changegeom.test
> Running job cl-_cosmo.test
> Running job cl2_cosmo.test
> Running job cl2_cosmo2.test
> Running job cl2dft.test
> Running job cl_cosmo.test
> Running job cn_optrs2av.test
> Running job cn_proj.test
> Running job cn_samc_forces.test
> 3:3:fehler:: 1
> (rank:3 hostname:kamet pid:32327):ARMCI DASSERT fail.
> src/armci.c:ARMCI_Error():276 cond:0
> 4:4:fehler:: 1
> (rank:4 hostname:kamet pid:32328):ARMCI DASSERT fail.
> src/armci.c:ARMCI_Error():276 cond:0
> 5:5:fehler:: 1
> (rank:5 hostname:kamet pid:32329):ARMCI DASSERT fail.
> src/armci.c:ARMCI_Error():276 cond:0
> 7:7:fehler:: 1
> (rank:7 hostname:kamet pid:32331):ARMCI DASSERT fail.
> src/armci.c:ARMCI_Error():276 cond:0
> 0:0:fehler:: 1
> (rank:0 hostname:kamet pid:32324):ARMCI DASSERT fail.
> src/armci.c:ARMCI_Error():276 cond:0
> 1:1:fehler:: 1
> (rank:1 hostname:kamet pid:32325):ARMCI DASSERT fail.
> src/armci.c:ARMCI_Error():276 cond:0
> 2:2:fehler:: 1
> (rank:2 hostname:kamet pid:32326):ARMCI DASSERT fail.
> src/armci.c:ARMCI_Error():276 cond:0
> 6:6:fehler:: 1
> (rank:6 hostname:kamet pid:32330):ARMCI DASSERT fail.
> src/armci.c:ARMCI_Error():276 cond:0
>
> =====================================================================================
>
> = BAD TERMINATION OF ONE OF YOUR APPLICATION PROCESSES
> = EXIT CODE: 256
> = CLEANING UP REMAINING PROCESSES
> = YOU CAN IGNORE THE BELOW CLEANUP MESSAGES
> =====================================================================================
>
> **** PROBLEMS WITH JOB cn_samc_forces.test
> **** For further information, look in the output file
> **** /home/user/Molpro/testjobs/cn_samc_forces.errout
> Running job cn_tranh.test
> Running job cn_uno.test
> Running job cndft.test
> Running job co2_mp2freq.test
> Running job co2_vci_pmp.test
> Running job co_cordip.test
> Running job co_qcidip.test
> Running job contest_lih2_D0D1.test
> Running job coreocc.test
> Running job cpp.test
> Running job cppopt.test
> Running job cr2_ecp_forces.test
> Running job cs2_qcitgrad_c2v.test
>
> Quoting Andy May <MayAJ1 at cardiff.ac.uk>:
>
>> Greg,
>>
>> I can now only think the problems relate to the MPICH2 binaries or GA
>> install. I have made an updated Molpro tarball,
>> molpro.2010.1.22.tar.gz. Please can you build atlas without the lapack
>> routines, i.e.:
>>
>> # build ATLAS and install in /opt/atlas/3.8.4
>> mkdir -p /tmp/ATLAS/build
>> tar -C /tmp/ATLAS -xjf ~/atlas3.8.4.tar.bz2
>> cd /tmp/ATLAS/build
>> /tmp/ATLAS/ATLAS/configure --prefix=/opt/atlas/3.8.4 -Si cputhrchk 0
>> -C if `which gfortran` -F if '-O3 -m64 -fdefault-integer-8'
>> sed -i -e 's/F77_INTEGER=long/F77_INTEGER="long long"/' Make.inc
>> make
>> make check
>> make ptcheck
>> sudo make install
>>
>> and then for Molpro, use the auto-build options which will build
>> MPICH2 from scratch and automatically build GA. Also, with this
>> tarball the extra lapack routines are all automatically compiled
>> within Molpro if ATLAS is detected.
>>
>> tar -xzf ~/molpro.2010.1.22.tar.gz
>> cd Molpro
>> ./configure -batch -gcc -gfortran -mpp -auto-ga-tcgmsg-mpich2
>> -instroot /usr/local/molpro2010.1 -blaspath /opt/atlas/3.8.4/lib
>>
>> For me, this gave a working build. I haven't run all of the testjobs,
>> but at least the first few of the ones where you saw problems, and
>> they worked here.
>>
>> There is one disadvantage to the above way of building, 'make install'
>> will not then work for Molpro. This is because the MPICH2 install can
>> not simply be copied since some of the install directories are
>> hard-coded into the executable. We have a bug open to fix this
>> eventually. The configure script does print out exactly how it is
>> building MPICH2 and GA, so if everything works you can simply build
>> the packages as per-configure, but change the install directory to
>> some global location, eg. /opt or /usr/local and then build Molpro
>> pointing to these installs instead.
>>
>> Best wishes,
>>
>> Andy
>>
>