[Wien] Compilation in parallel

[Laurence Marks]

I used (beware of line breaks from gmail) in OPTIONS:

FOPT:-FR -mp1 -prec_div -r8 -pc80 -fpconstant -traceback -pad -align
-O3 -ipo1 -DINTEL_VML -i-static -fminshared -xT -thread -DINTEL_VML
FPOPT:-FR -mp1 -prec_div -r8 -pc80 -fpconstant -traceback -pad -align
-O3 -ipo1 -DINTEL_VML -i-static -fminshared -xT -DINTEL_VML
LDFLAGS:-L/opt/intel/mkl/10.0.3.020/lib/em64t
DPARALLEL:'-DParallel'
R_LIBS:$(FOPT) -lmkl_lapack -lmkl_em64t -lmkl_core -liomp5 -pthread -static
RP_LIBS:$(FPOPT) -lmkl_intel_lp64 -lmkl_scalapack_lp64
-lmkl_blacs_lp64 -lmkl_sequential -lmkl_core -liomp5 -pthread
-i-static

In addition:
1) Be aware that you need mvapich compiled with ifort otherwise you
might run into problems. I also used icc and did not compile the c++
part since I could never get it to compile cleanly (and did not need
it).
2) Check how recent your mvapich is. Mine is April 2008 and in one
earlier version they broke scalapack. There is a newer version
released a few weeks ago.
3) While it should not matter for you, there is a recent report on the
mvapich mailing list of problems when using ifort 11.X with a mvapich
compiled with ifort 10.X


On Wed, Nov 26, 2008 at 5:58 AM, Oleg Artamonov <arto at mail.nnz.ru> wrote:
> Dear Wien2k users, dear Wien2k leaders,
>
> We are installing theWien2k-08-3 in the cluster in parallel mode.
> Nodes number - 48, 2*Intel Xeon 5335, 2 G Hz, Infiniband, Linux SLES 10 SP1
> (em64t),
> ifort ver. 10.1.012, (icc ver.10.1.012), MKL ver. 10.0.2.018
> (GotoBLAS-1.19), mvapich-1.0.0-1844 (mpich-1.2.7.).
>
> We have found Gerhard's Fecher options for compilation:
> *****
> current:FOPT:-FR -w -mp1 -prec_div -pad -DINTEL_VML -O3 -xP
> current:LDFLAGS:-L/opt/intel/fce/10.1.017/lib -static -lsvml -lguide
> -lpthread
> current:R_LIBS:-L/opt/intel/mkl/10.0.3.020/lib/em64t -lmkl_lapack
> -lmkl_em64t -lmkl_core -lguide -lpthread
>
> if you have the same pathes like me, check also if you need -xT instead of
> -xP (see quick optimization guide of Intel) use cat /proc/cpuinfo to check
> for the processor.
>
> 8.2 is running without problems on my Core 2 Duo T9500 and 64-bit SUSE 11,
> as well as on a Quad Core Qxxxx and a Quadcore XEON both with 64-bit SUSE
> 10.3,
> all have the options as above with only differing in the -xProcessor
> switches.
> *****
>
> My question is: what do we need to change to compile a parallel mode?
>
> The second question. The names of nodes in our cluster are node-1, node-2
> ....The script lapw1para
> produces names node 1, node 2... , instead of node-1, node-2... . Is it
> possible to change the script?
>
> Many thanks in advance,
>                                         Oleg Artamonov.
>
>
>
> _______________________________________________
> Wien mailing list
> Wien at zeus.theochem.tuwien.ac.at
> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
>
>



-- 
Laurence Marks
Department of Materials Science and Engineering
MSE Rm 2036 Cook Hall
2220 N Campus Drive
Northwestern University
Evanston, IL 60208, USA
Tel: (847) 491-3996 Fax: (847) 491-7820
email: L-marks at northwestern dot edu
Web: www.numis.northwestern.edu
Chair, Commission on Electron Crystallography of IUCR
www.numis.northwestern.edu/
Electron crystallography is the branch of science that uses electron
scattering to study the structure of matter.