xdlutime results

Susan Blackford susan at cs.utk.edu
Thu Sep 14 08:56:42 EDT 2000


Hi,
The value of TOTMEM reflects the amount of physical memory you have on
your machine.  The default value in the distribution is 2M.  We put the
value small because a lot of people tested spawning multiple processes
per processor. 

*  TOTMEM   INTEGER, default = 2000000
*           TOTMEM is a machine-specific parameter indicating the
*           maximum amount of available memory in bytes.
*           The user should customize TOTMEM to his platform.  Remember
*           to leave room in memory for the operating system, the BLACS
*           buffer, etc.  For example, on a system with 8 MB of memory
*           per process (e.g., one processor on an Intel iPSC/860), the
*           parameters we use are TOTMEM=6200000 (leaving 1.8 MB for OS,
*           code, BLACS buffer, etc).  However, for PVM, we usually set
*           TOTMEM = 2000000.  Some experimenting with the maximum value
*           of TOTMEM may be required.
 

However, depending upon the size of matrix you want to time and the
amount of memory you have on your machine, this value will need to be
increased.  The program will test the amount of memory needed for the
size N that you request, and if TOTMEM isn't large enough, it will
tell you.

Hope this helps.
Susan

"Horatio B. Bogbindero" wrote:
> 
> pardon the question but what value should i set it to??? how do i know
> what is the appropriate value for this. thanks.
> 
> On Thu, 14 Sep 2000, Yoon Jae Ho wrote:
> 
> > In the lutimer.f file
> >
> >       PROGRAM DLUTIMER
> > *
> > *     Simple Timing Program for ScaLAPACK routine for LU factiorization
> > *     and solver
> > *
> > *     The program must be driven by a short data file. The name of
> > *     the data file is 'LU.dat'. An annotated example of a data
> > *     file can be obtained by deleting the first 6 characters from
> > *     the following 5 lines: (The number in the first line is the
> > *     number of different problems the user wants to test. If it
> > *     is 'n', the user should input exactly 'n' numbers in each line
> > *     after. Also, this program will use the first column of N, NB,
> > *     P, Q to do the first test, second column of N, NB, P, Q to do
> > *     the second test, etc.)
> > *     2                 number of problems sizes
> > *     500 600           values of N (N x N square matrix)
> > *     64 64             values of NB
> > *     2 2               values of P
> > *     2 2               values of Q
> > *
> > *     .. Parameters ..
> >       INTEGER            CSRC, DBLESZ, INTGSZ, TOTMEM, MEMSIZ, NBRHS,
> >      $                   NOUT, NRHS, NTESTS, RSRC, CSRC_, DLEN_, LLD_,
> >      $                   M_, MB_, N_, NB_, RSRC_
> >       PARAMETER          ( CSRC = 0, DBLESZ = 8, INTGSZ = 4,
> >      $                   TOTMEM = 40000000, MEMSIZ = TOTMEM / DBLESZ,
> >
> >
> > please change the above TOTMEM = 40000000 to big figure.
> >
> > have a nice day
> >
> > from Yoon Jae Ho
> > Seoul,Korea
> >
> > ----- Original Message -----
> > From: Camm Maguire <camm at enhanced.com>
> > To: Fredrik Augustsson <hamlet at cs.umu.se>
> > Cc: <pgeoffra at cs.utk.edu>; <beowulf at beowulf.org>
> > Sent: Wednesday, September 13, 2000 11:59 PM
> > Subject: Re: xdlutime results
> >
> >
> > > Greetings!  the scalapack testers have statically allocated arrays.
> > > TOTMEM is a paramater, and set to a small value by default, if memory
> > > serves.  Just edit and recompile.  The Debian scalapack-test package
> > > has this pre-adjusted to something more reasonable.
> > >
> > > Take care,
> > >
> > > Fredrik Augustsson <hamlet at cs.umu.se> writes:
> > >
> > > > H!
> > > > I've tried to run the xdlutime to get an estimate for how well our
> > > > cluster performs, but I can only test small matrices. I have 8 dual PIII
> > > > with 512 MB of memory so would at least be abel to run tests on
> > > > 20000x20000, or am I wrong here?
> > > >
> > > > Output ...
> > > >
> > > > dumburk [~/pfs]$ mpirun -np 8 -npn 1 xdlutime
> > > >
> > > > Simple Timer for ScaLAPACK routine PDGESV
> > > > Number of processors used:   8
> > > >
> > > > TIME     N  NB   P   Q  LU Time   Sol Time  MFLOP/S Residual  CHECK
> > > > ---- ----- --- --- --- --------- --------- -------- -------- -------
> > > > Unable to perform LU-solve: need TOTMEM of at least  325494408
> > > > Bad MEMORY parameters: going to next test case.
> > > >
> > > >
> > > > + Fredrik
> > > >
> > > >
> > > >
> > > >
> > > > On Tue, Sep 12, 2000 at 10:56:38PM -0400, Patrick GEOFFRAY wrote:
> > > > > Putchong Uthayopas wrote:
> > > > >
> > > > > > Below are our results. The machines is 8 Athlon 550MHz and one 1Ghz Athlon.
> > > > > > 512 Mb memory each. and Myrinet.
> > > > >
> > > > > > WALL 6000 9 1 9 116.80 0.41 1229.04 0.001097 PASSED
> > > > > >
> > > > > > WALL 6000 9 9 1 200.79 3.65 704.66 0.001926 PASSED
> > > > >
> > > > > Hi,
> > > > >
> > > > > it seems that you are far from what you can get from your cluster
> > > > > !
> > > > > * First, 9 nodes is not a good idea, the (1*9) or (9*1) grids are
> > > > > bad. you should use only 8 nodes and try (2*4) and (4*2) grids,
> > > > > much better.
> > > > > * Then 6000 is not a big matrix. If you have 512 MB per node, you
> > > > > should go up to 20000. It will run longuer but you will be closest
> > > > > to the peak.
> > > > > * Which Blas do you use ? Atlas gives a DGEMM peak at 600 MFlops
> > > > > per Athlon, so you can hope about 4 GFlops on your cluster.
> > > > > * Which MPI do you use ? if it's MPICH-GM, do you use the
> > > > > -gm-can-register-memory flag ?
> > > > >
> > > > > I strongly advice you to use the new High Performance Linpack
> > > > > (HPL) benchmark form Antoine Petitet
> > > > > (http://www.netlib.org/benchmark/index.html), it's faster than the
> > > > > old Linpack and much easier to install.
> > > > >
> > > > > Hope it can help.
> > > > > regards.
> > > > >
> > > > > Patrick Geoffray
> > > > > ---
> > > > > Aerospatiale Matra - Sycomore
> > > > > Universite Lyon I - RESAM
> > > > > http://lhpca.univ-lyon1.fr
> > > > >
> > > > > _______________________________________________
> > > > > Beowulf mailing list
> > > > > Beowulf at beowulf.org
> > > > > http://www.beowulf.org/mailman/listinfo/beowulf
> > > >
> > > > _______________________________________________
> > > > Beowulf mailing list
> > > > Beowulf at beowulf.org
> > > > http://www.beowulf.org/mailman/listinfo/beowulf
> > > >
> > > >
> > >
> > > --
> > > Camm Maguire      camm at enhanced.com
> > > ==========================================================================
> > > "The earth is but one country, and mankind its citizens."  --  Baha'u'llah
> > >
> > > _______________________________________________
> > > Beowulf mailing list
> > > Beowulf at beowulf.org
> > > http://www.beowulf.org/mailman/listinfo/beowulf
> >
> > _______________________________________________
> > Beowulf mailing list
> > Beowulf at beowulf.org
> > http://www.beowulf.org/mailman/listinfo/beowulf
> >
> 
> 
> ---------------------
> william.s.yu at ieee.org
> 
> I bought some used paint. It was in the shape of a house.
>                 -- Steven Wright
> 
> 
> _______________________________________________
> Beowulf mailing list
> Beowulf at beowulf.org
> http://www.beowulf.org/mailman/listinfo/beowulf

_______________________________________________
Beowulf mailing list
Beowulf at beowulf.org
http://www.beowulf.org/mailman/listinfo/beowulf



More information about the Beowulf mailing list