[CPMD-list] [Fwd: [Fwd: Re: Benchmark IBM pSeries]]

Wed Aug 27 11:42:29 CEST 2003

  Dear all,

We are planning to buy a new machine to run large CPMD calculations. 
 Therefore some benchmarks are done for us at the benchmark center of 
IBM in Montpellier.  They have run some examples of our input files and 
encounterd strange error messages.  The output file is attached.  Has 
somebody encountered analogous things?  Maybe it has something to do 
with the 64 bit version of the code.

Has somebody have good experiences or suggestions running CPMD jobs on 
64 bit machines.  The available budget is around 100000 Euros.  

Thanks

Veronique

-------- Original Message --------
Subject: [Fwd: Re: Benchmark IBM pSeries]
Date: Wed, 27 Aug 2003 10:06:37 +0200
From: "Colin Dumontier" <cdumontier at fr.ibm.com>
To: veronique.vanspeybroeck at ugent.be
CC: "Eric Michel" <EMICHEL at fr.ibm.com>, "Sophie Vilarem" 
<Sophie.Vilarem at fr.ibm.com>, Gerd_Vanmechelen/Belgium/IBM%IBMFR 
<Gerd_Vanmechelen/Belgium/IBM at fr.ibm.com>

Dear Veronique,

We are having difficulties running the RUG input decks for CPMD.

We used the provided uspp generation program to create the following atoms
files, and moved them to the run directory, where the CPMD input decks
reside :

-rw-r--r--   1 rug    staff      35916 Aug 26 17:24 001-H-ca--bm.uspp
-rw-r--r--   1 rug    staff      98088 Aug 26 17:26 006-C-ca--bm.uspp
-rw-r--r--   1 rug    staff      99432 Aug 26 17:27 007-N-ca--bm.uspp
-rw-r--r--   1 rug    staff     100776 Aug 26 17:26 008-O-ca--bm.uspp
-rw-r--r--   1 rug    staff     109704 Aug 26 17:27 015-P-ca--vgrp.uspp
-rw-r--r--   1 rug    staff     210612 Aug 26 17:28 019-K-ca-sp-vgrp.uspp
-rw-r--r--   1 rug    staff     232196 Aug 26 17:28
045-Rh-gpw-n-campos.uspp

After succesfully configuring and compiling the MPI code for an IBM 64bit
platform, we submitted the Rh.wf.inp input deck (wavefunction
optimization).

The job crashed after several hours, and didn't generate a RESTART file
that I assume would have been used for the Rh.damp.inp submission.

Here is the log file :

(See attached file: Rh.wf.out)

Are we doing something wrong here ?

Thanks a lot.

Best regards.
__________________________________________
Colin DUMONTIER (cdumontier at fr.ibm.com)
IBM Scientific & Technical Computing EMEA
ATS - P.S.S.C   /  Dpt 1459 / Montpellier FRANCE
Tel: +33 4 67 34 42 37          Fax: +33 4 67 34 64 77
http://w3.pssc.mop.ibm.com

----- Forwarded by Colin Dumontier/France/IBM on 27-08-03 10:03 -----

                      Eric Michel                                                                                                      
                                               To:       Colin Dumontier/France/IBM at IBMFR                                              
                      27-08-03 10:04           cc:                                                                                     
                                               From:     Eric Michel/France/IBM at IBMFR                                                  
                                               Subject:  [Fwd: Re: Benchmark IBM pSeries]                                              

Eric Michel
IBM Scientific & Technical Computing EMEA
ATS-P.S.S.C/ Monptellier FRANCE
emichel at fr.ibm.com  - +33-467344731
----- Forwarded by Eric Michel/France/IBM on 27-08-03 10:04 -----

                      Veronique Van                                                                                                          
                      Speybroeck                     To:       Gerd Vanmechelen/Belgium/IBM at IBMBE, Eric Michel/France/IBM at IBMFR              
                                          Subject:  [Fwd: Re: Benchmark IBM pSeries]                                              

                      25-08-03 12:07                                                                                                         

Dear Gerd, Eric Michel,

concerning the benchmarks following arrangements were made.

We have set up an http server : moldyn10.ugent.be

http://moldyn10.ugent.be

there you can find three links refering to the CPMD code the code for the
pseudopotentials and the input files.
To compile the CPMD code, you must go to the directory SOURCE en let run
the file Configure with the correct option .  More details are given in the
file itself.
Bv. Configure IBM-SP4-64 > Makefile.
then the code can be compiled using the standard make commandos.

To run the CPMD code one also needs pseudopotential files, which have a
binary format and thus have to be generated on the machine the benchmarks
are run.  To compile this code you go to the directory source and again run
the makefile.

I will give more details about how the input files have to be run once you
give me a sign that you compiled the codes succesfully.

Best regards, thanks in advance

veronique

-------- Original Message --------

    Subject: Re: Benchmark IBM pSeries                                                  

       Date: Mon, 25 Aug 2003 10:22:03 +0200                                            

       From: Veronique Van Speybroeck                 

         To: Gerd Vanmechelen , emichel at fr.ibm.com         

 References:      

Beste Gerd,

we zullen de code beschikbaar stellen via een http server nl. de
moldyn10.ugent.be:

http://moldyn10.ugent.be

Dit zal vanmiddag beschikbaar zijn.  Gezien de codes normaal niet mogen
gedistribueerd worden zullen wij vanavond de site weer dicht zetten.
 Hopelijk kunnen jullie de code afhalen deze middag.

Het betreft twee tar files :

cpmdv37_1.tar : Dit is de CPMD code waarvan de benchmarks moeten gelopen
worden.  Om te compileren ga naar de directory SOURCE.  Daar staat een
file Configure .  Daar kan je een makefile aan maken voor het programma
te compileren.  Bv. Configure IBM-SP4-64 > Makefile.
Een je dat hebt kan je de code compileren.

en uspp-734-3.tgz :  Dit is een programma op pseudopotentiaalfiles aan
te maken die nodig zijn om het voorgaande programma te laten lopen.
 Eens het untard is ga je naar de directory Source en compileer je het
programma.

De inputfiles die we wensen te laten lopen zullen ook op de site staan.
 Eens jullie de code goed afgehaald hebben, kan je mij een seintje geven
en dan geef ik verdere instructies over wat precies moet gebeuren om de
code te doen draaien.

Alvast bedankt

veronique

PS : Als je liever hebt dat ik deze mail in het engels opmaak voor Eric
Michel geef dan een seintje

Gerd Vanmechelen wrote:

>Véronique,
>
>Zoals besproken zullen onze Scientific & Technical computing
>specialisten in Montpellier jullie fortran programmas testen.
>
>Kan je een FTP account aanmaken ?
>
>De persoon die de code zal downloaden is Eric Michel (EMICHEL at fr.ibm.com)
>
>Ter info : http://www-1.ibm.com/servers/eserver/briefingcenter/mbc/
>
>Als je nog opmerkingen hebt bij de manier waarop de testen moeten verlopen
>kan je deze per mail bezorgen, met Eric in copie.
>
>
>Regards
>
>Gerd Vanmechelen, Sales Specialist
>IBM Belgium / Luxembourg Systems Sales
>Tel : 0032-2-2253569    Fax : 0032-2-2252368
>E-mail : Gerd_Vanmechelen at be.ibm.com
>
>
>
>

--
ÐÏà¡±á

--
ÐÏà¡±á

-- 
ÐÏà¡±á

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://cpmd.org/pipermail/cpmd-list/attachments/20030827/0ff6d391/attachment.html 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: Rh.wf.out
Type: application/octet-stream
Size: 55442 bytes
Desc: not available
Url : http://cpmd.org/pipermail/cpmd-list/attachments/20030827/0ff6d391/attachment.obj 