From owner-nwchem-users@emsl.pnl.gov Mon Mar 13 06:30:19 2006 Received: from odyssey.emsl.pnl.gov (localhost [127.0.0.1]) by odyssey.emsl.pnl.gov (8.12.10/8.12.10) with ESMTP id k2DEUJ07016451 for ; Mon, 13 Mar 2006 06:30:19 -0800 (PST) Received: (from majordom@localhost) by odyssey.emsl.pnl.gov (8.12.10/8.12.10/Submit) id k2DEUJrM016450 for nwchem-users-outgoing-0915; Mon, 13 Mar 2006 06:30:19 -0800 (PST) Date: Mon, 13 Mar 2006 15:30:06 +0100 From: =?ISO-8859-1?Q?Lat=E9vi_Max_LAWSON_DAKU?= Subject: Re: [NWCHEM] TCE module + CCSD calculations : problem with memory allocation In-reply-to: <42C7E766869C42408F0360B7BF0CBD9B0256D449@pnlmse27.pnl.gov> To: "Kowalski, Karol" Cc: nwchem-users@emsl.pnl.gov Message-id: <441581EE.9010401@unige.ch> MIME-version: 1.0 Content-type: multipart/alternative; boundary="Boundary_(ID_XDEwailJZetS0QG+ImogDA)" X-Accept-Language: en-us, en User-Agent: Debian Thunderbird 1.0.7 (X11/20051017) References: <42C7E766869C42408F0360B7BF0CBD9B0256D449@pnlmse27.pnl.gov> X-Comment: This message was scanned against viruses by mbox.unige.ch. Sender: owner-nwchem-users@emsl.pnl.gov Precedence: bulk This is a multi-part message in MIME format. --Boundary_(ID_XDEwailJZetS0QG+ImogDA) Content-type: text/plain; charset=ISO-8859-1 Content-transfer-encoding: QUOTED-PRINTABLE Dear Karol, Many thanks for your valuable hints. I will proceed according to them and will let you how this works. All the best, Max Kowalski, Karol wrote: >Dear Max, >I guess the source of your problem is twofold: > >(1) you are storing large matrix with two-electron integrals on glob= al=20 > array using in-core algorithm. If you have 12 CPUs it's better f= or=20 > this calculation to switch to other I/O algorithm - for example = the one=20 > which uses the replicated files. In order to do so please add to= your tce=20 > group in the input file the following directive: > io replicated > >(2) From your output I have noticed that the tilesize is exceedingly= large=20 > (on the order of 100). Please use tilesize equal 15 or 20 instea= d. You > can define the size of tile by putting: > tilesize 15 > >Good luck, >Karol > > >-----Original Message----- >From: Lat=E9vi Max LAWSON DAKU [mailto:max.lawson@unige.ch]=20 >Sent: Friday, March 10, 2006 1:26 AM >To: Kowalski, Karol >Subject: Re: [NWCHEM] TCE module + CCSD calculations : problem with = memory allocation > >Dear Karol, > >Good morning. > >Here is the output file; it also includes the input file. >Please feel free to contact for any further question. > >Best regards, >Max > >Kowalski, Karol wrote: > > =20 > >>Hello, >>Would you please send me your input/output files. >>Karol >> >> >> >>-----Original Message----- >>From: owner-nwchem-developers@emsl.pnl.gov=20 >>[mailto:owner-nwchem-developers@emsl.pnl.gov] On Behalf Of Lat=E9vi= Max=20 >>LAWSON DAKU >>Sent: Thursday, March 09, 2006 1:28 AM >>To: nwchem-users@emsl.pnl.gov >>Subject: [NWCHEM] TCE module + CCSD calculations : problem with mem= ory=20 >>allocation >> >>Dear NWChem users, >> >>While trying to carry out CCSD calculations with the TCE module, I = noticed that the dimension of the memory (RAM or HD) being requested = is negative. See example below. >> >> ... >> Fock matrix recomputed >> 1-e file size =3D 56169 >> 1-e file name =3D /user/scratch/lawson >> Cpu & wall time / sec 47.9 57.6 >> >> 2-e (intermediate) file size =3D -141563493 >> 2-e (intermediate) file name =3D /user/scratch/lawson >> ... >> >>I guess this has to do with the size of the requested memory being = larger than the maximum value allowed for integers of the integer typ= e used ?! Anyway, with the default IO scheme (GA), this makes the cal= culations abort at the very begining. >>Using the SF schem things seem to run smoothly, (is it safe to do s= o=20 >>?), but after some time, the calculations stop with >> >> ... >> put_block: sf problem 1 >> >> =20 >>-------------------------------------------------------------------= --- >>-- >> >> ----------------------------------------------------------------= -------- >> current input line : >> 0: >> >> =20 >>-------------------------------------------------------------------= --- >>-- >> >> ----------------------------------------------------------------= -------- >> An error occured while trying to read or write to disk space >> >> ----------------------------------------------------------------= -------- >> For more information see the NWChem manual at >> http://www.emsl.pnl.gov/docs/nwchem/nwchem.html >> >> >> For further details see manual section: >> >> >> >> 6:6:put_block: sf problem:: 1 >> >> Killed >> Killed >> Killed >> Killed >> ... >> >>Is there a way to perform such calculations ? Thanks. >> >>Best regards, >>LMLD >> >> >>-- >>*********************************************** >>Latevi Max LAWSON DAKU >>Departement de chimie physique >>Universite de Geneve - Sciences II >>30, quai Ernest-Ansermet >>CH-1211 Geneve 4 >>Switzerland >> >>Tel: (41) 22/379 6548 ++ Fax: (41) 22/379 6103 >>*********************************************** >> >> >> >> >>=20 >> >> =20 >> > >-- >*********************************************** >Latevi Max LAWSON DAKU >Departement de chimie physique >Universite de Geneve - Sciences II >30, quai Ernest-Ansermet >CH-1211 Geneve 4 >Switzerland=20 > >Tel: (41) 22/379 6548 ++ Fax: (41) 22/379 6103 >*********************************************** > > > > =20 > --=20 *********************************************** Latevi Max LAWSON DAKU Departement de chimie physique=20 Universite de Geneve - Sciences II 30, quai Ernest-Ansermet =20 CH-1211 Geneve 4=20 Switzerland=20 Tel: (41) 22/379 6548 ++ Fax: (41) 22/379 6103 *********************************************** --Boundary_(ID_XDEwailJZetS0QG+ImogDA) Content-type: text/html; charset=ISO-8859-1 Content-transfer-encoding: QUOTED-PRINTABLE Dear Karol,

Many thanks for your valuable hints. I will proceed according
to them and will let you how this works.

All the best,
Max


Kowalski, Karol wrote: 
Dear Max,
I guess the source of your problem is twofold:

(1) you are storing large matrix with two-electron integrals on globa=
l=20
    array using in-core algorithm. If you have 12 CPUs it's better fo=
r=20
    this calculation to switch to other I/O algorithm - for example t=
he one=20
    which uses the replicated files. In order to do so please add to =
your tce=20
    group in the input file the following directive:
    io replicated

(2) From your output I have noticed that the tilesize is exceedingly =
large=20
    (on the order of 100). Please use tilesize equal 15 or 20 instead=
. You
    can define the size of tile by putting:
    tilesize 15

Good luck,
Karol


-----Original Message-----
=46rom: Latévi Max LAWSON DAKU [mailto:max.lawson@unige.ch]=20
Sent: Friday, March 10, 2006 1:26 AM
To: Kowalski, Karol
Subject: Re: [NWCHEM] TCE module + CCSD calculations : problem with m=
emory allocation

Dear Karol,

Good morning.

Here is the output file; it also includes the input file.
Please feel free to contact for any further question.

Best regards,
Max

Kowalski, Karol wrote:

  
Hello,
Would you please send me your input/output files.
Karol



-----Original Message-----
=46rom: owner-nwchem-developers@emsl.pnl.gov=20
[mailto:owner-nwchem-developers@emsl.pnl.gov] =
On Behalf Of Latévi Max=20
LAWSON DAKU
Sent: Thursday, March 09, 2006 1:28 AM
To: nwchem-users@emsl.pnl.gov
Subject: [NWCHEM] TCE module + CCSD calculations : problem with memor=
y=20
allocation

Dear NWChem users,

While trying to carry out CCSD calculations with the TCE module, I no=
ticed that the dimension of the memory (RAM or HD) being requested is=
 negative. See example below.

   ...
   Fock matrix recomputed
   1-e file size =3D 56169
   1-e file name =3D /user/scratch/lawson
   Cpu & wall time / sec 47.9 57.6

   2-e (intermediate) file size =3D -141563493
   2-e (intermediate) file name =3D /user/scratch/lawson
   ...

I guess this has to do with the size of the requested memory being la=
rger than the maximum value allowed for integers of the integer type =
used ?! Anyway, with the default IO scheme (GA), this makes the calcu=
lations abort at the very begining.
Using the SF schem things seem to run smoothly, (is it safe to do so=
=20
?), but after some time, the calculations stop with

   ...
   put_block: sf problem 1

  =20
---------------------------------------------------------------------=
-
--

   ------------------------------------------------------------------=
------
   current input line :
   0:

  =20
---------------------------------------------------------------------=
-
--

   ------------------------------------------------------------------=
------
   An error occured while trying to read or write to disk space

   ------------------------------------------------------------------=
------
   For more information see the NWChem manual at
   http://www.emsl.pnl.gov/docs/nwchem/nwchem.=
html


   For further details see manual section:



   6:6:put_block: sf problem:: 1

   Killed
   Killed
   Killed
   Killed
   ...

Is there a way to perform such calculations ? Thanks.

Best regards,
LMLD


--
***********************************************
Latevi Max LAWSON DAKU
Departement de chimie physique
Universite de Geneve - Sciences II
30, quai Ernest-Ansermet
CH-1211 Geneve 4
Switzerland

Tel: (41) 22/379 6548 ++ Fax: (41) 22/379 6103
***********************************************




=20

    

--
***********************************************
Latevi Max LAWSON DAKU
Departement de chimie physique
Universite de Geneve - Sciences II
30, quai Ernest-Ansermet
CH-1211 Geneve 4
Switzerland=20

Tel: (41) 22/379 6548 ++ Fax: (41) 22/379 6103
***********************************************



  

--=20
***********************************************
Latevi Max LAWSON DAKU
Departement de chimie physique=20
Universite de Geneve - Sciences II
30, quai Ernest-Ansermet =20
CH-1211 Geneve 4=20
Switzerland=20

Tel: (41) 22/379 6548 ++ Fax: (41) 22/379 6103
***********************************************
--Boundary_(ID_XDEwailJZetS0QG+ImogDA)--