From info at asfyon.nl  Sun Jan  4 05:46:51 2009
From: info at asfyon.nl (Dr. R.J.Slagter)
Date: Sun, 04 Jan 2009 12:46:51 +0100
Subject: [Users] Fortran for PDE's
In-Reply-To: <494F335F.7090505@ea.c.u-tokyo.ac.jp>
References: <493801BA.4040807@nasa.gov>
	<6C72C3B5-7C05-4E06-8784-3E59583C240A@cct.lsu.edu>
	<494F335F.7090505@ea.c.u-tokyo.ac.jp>
Message-ID: <CPSMTP-EML1091FFK7d000d7d10@cpsmtp-eml109.kpnxchange.com>

Hello users,
I am looking for a fortran solver for a set of partial differential 
equations in t and 2 space dimensions  F(t,x,y)
Is there a code in Cactus?
Thanks Reinoud


From dprideout at gmail.com  Mon Jan  5 10:54:22 2009
From: dprideout at gmail.com (David Rideout)
Date: Mon, 5 Jan 2009 11:54:22 -0500
Subject: [Users] Fortran for PDE's
In-Reply-To: <CPSMTP-EML1091FFK7d000d7d10@cpsmtp-eml109.kpnxchange.com>
References: <493801BA.4040807@nasa.gov>
	<6C72C3B5-7C05-4E06-8784-3E59583C240A@cct.lsu.edu>
	<494F335F.7090505@ea.c.u-tokyo.ac.jp>
	<CPSMTP-EML1091FFK7d000d7d10@cpsmtp-eml109.kpnxchange.com>
Message-ID: <1ce81abb0901050854lb79943are59d544473ee8b7e@mail.gmail.com>

You can adapt WaveToy to your two dimensional problem.  You will need
to define your own 2d coordinate system.  Everything in Cactus should
be fine with this.  e.g. (in your interface.ccl)

cctk_real F type=GF dim=2 timelevels=3 "my field"

and, if you want, e.g.

cctk_real coords type=GF dim=2
{
 x, y
} "my 2d coordinates"

Cheers,
David

On Sun, Jan 4, 2009 at 6:46 AM, Dr. R.J.Slagter <info at asfyon.nl> wrote:
> Hello users,
> I am looking for a fortran solver for a set of partial differential
> equations in t and 2 space dimensions  F(t,x,y)
> Is there a code in Cactus?
> Thanks Reinoud
>
> _______________________________________________
> Users mailing list
> Users at cactuscode.org
> http://www.cactuscode.org/mailman/listinfo/users
>

From schnetter at cct.lsu.edu  Mon Jan  5 11:46:44 2009
From: schnetter at cct.lsu.edu (Erik Schnetter)
Date: Mon, 5 Jan 2009 11:46:44 -0600
Subject: [Users] Fortran for PDE's
In-Reply-To: <CPSMTP-EML1091FFK7d000d7d10@cpsmtp-eml109.kpnxchange.com>
References: <493801BA.4040807@nasa.gov>
	<6C72C3B5-7C05-4E06-8784-3E59583C240A@cct.lsu.edu>
	<494F335F.7090505@ea.c.u-tokyo.ac.jp>
	<CPSMTP-EML1091FFK7d000d7d10@cpsmtp-eml109.kpnxchange.com>
Message-ID: <4CAE552D-5BA5-46B8-AE98-0279D37555BB@cct.lsu.edu>

On Jan 4, 2009, at 05:46:51, Dr. R.J.Slagter wrote:

> Hello users,
> I am looking for a fortran solver for a set of partial differential
> equations in t and 2 space dimensions  F(t,x,y)
> Is there a code in Cactus?

Reinoud

Cactus is a tool that helps people write solvers for PDEs.  It has  
e.g. well documented examples for the scalar wave equation.

However, whether this helps you depends on what kind of PDEs you have,  
and what kind of spatial discretisation method and what kind of time  
integration would be appropriate for them.  Since there are many very  
different kinds of PDEs this is impossible to tell from the outside.   
In addition, you will need initial and boundary conditions for your  
system.

Can you tell us more about your PDEs and your field of science?

-erik

-- 
Erik Schnetter <schnetter at cct.lsu.edu>   http://www.cct.lsu.edu/~eschnett/

My email is as private as my paper mail.  I therefore support encrypting
and signing email messages.  Get my PGP key from www.keyserver.net.



-------------- next part --------------
A non-text attachment was scrubbed...
Name: PGP.sig
Type: application/pgp-signature
Size: 194 bytes
Desc: This is a digitally signed message part
Url : http://www.cactuscode.org/pipermail/users/attachments/20090105/61268f9c/attachment.bin 

From info at asfyon.nl  Thu Jan  8 02:04:55 2009
From: info at asfyon.nl (Dr. R.J.Slagter)
Date: Thu, 08 Jan 2009 09:04:55 +0100
Subject: [Users] To Erik
Message-ID: <CPSMTP-EML103tRHQq0001001aa@cpsmtp-eml103.kpnxchange.com>

Hello Erik
Thanks for your reply. I want to send you my set of equations, but I 
did not succeed because the keyserver is down.
I can post the article on my website?
Regards reinoud


From tradke at aei.mpg.de  Mon Jan 12 08:20:09 2009
From: tradke at aei.mpg.de (Thomas Radke)
Date: Mon, 12 Jan 2009 15:20:09 +0100
Subject: [Users] per-file tolerance in testsuites
In-Reply-To: <4951E09F.60104@ea.c.u-tokyo.ac.jp>
References: <493801BA.4040807@nasa.gov>	<6C72C3B5-7C05-4E06-8784-3E59583C240A@cct.lsu.edu>	<494F335F.7090505@ea.c.u-tokyo.ac.jp>	<B3780BC1-773D-4E7F-804D-33B1CDBCCD62@cct.lsu.edu>
	<4951E09F.60104@ea.c.u-tokyo.ac.jp>
Message-ID: <496B5199.6020608@aei.mpg.de>

Baiotti Luca wrote:
> Sorry, that information (about NPROCS) is already in the documentation, 
> but not in the documentation on the webpage, which I had looked and 
> which refers to the stable version.
> 
> What about my question / feature request? Is it reasonable?
> 
> I repeat my previous email here:
> 
> "In the Cactus testsuite mechanism it is possible to select the absolute 
> and relative tolerance for a given thorn or for a given testsuite of a 
> thorn. This information must be included in the test.ccl file.
> 
> Would it be possible to specify a tolerance for a specific file of a 
> specific testsuite? It is indeed the case that some output files may 
> have less stringent tolerances (like some norms).
> 
> If this is not possible, could it be implemented?"

Hi Luca,

currently it is not possible to specify a data tolerance value per 
testsuite output file, or for individual colums therein. Implementing 
such a feature is possible but would require careful revision of the 
existing code in the (rather large) testsuite comparison script.

As a simple alternative I suggest you create multiple testsuites, one 
for each tolerance needed.

-- 
Cheers, Thomas.

From woxinfeijian at gmail.com  Tue Jan 13 13:21:38 2009
From: woxinfeijian at gmail.com (Y. Jun Zhang)
Date: Wed, 14 Jan 2009 03:21:38 +0800
Subject: [Users] How do the data exchange between processors?
In-Reply-To: <5881cf820901131120g442fe201vdaeb6f7831db53c4@mail.gmail.com>
References: <5881cf820901131120g442fe201vdaeb6f7831db53c4@mail.gmail.com>
Message-ID: <5881cf820901131121h2851bfe3xd01c0e2815483837@mail.gmail.com>

Dear all:

      How do the data exchange between processors? It seems trouble
in cactus, I try to define the data array in "interface.ccl" and
"type=GF", but Processor A can not accpet the data array of Processor
B. Maybe I should use MPI function to send this array?

     Best regards

--
-----------------------------------------------------------
Y. Jun Zhang
E-Mail:
j.y.zhangcn at gmail.com
woxinfeijian at gmail.com
QQ:3752942
TEL:02763280205 13545299600



-- 
-----------------------------------------------------------
Y. Jun Zhang
E-Mail:
j.y.zhangcn at gmail.com
woxinfeijian at gmail.com
QQ:3752942
TEL:02763280205 13545299600

From jtao at cct.lsu.edu  Tue Jan 13 13:32:38 2009
From: jtao at cct.lsu.edu (Jian Tao)
Date: Tue, 13 Jan 2009 13:32:38 -0600
Subject: [Users] How do the data exchange between processors?
In-Reply-To: <5881cf820901131121h2851bfe3xd01c0e2815483837@mail.gmail.com>
References: <5881cf820901131120g442fe201vdaeb6f7831db53c4@mail.gmail.com>
	<5881cf820901131121h2851bfe3xd01c0e2815483837@mail.gmail.com>
Message-ID: <496CEC56.9030207@cct.lsu.edu>

Hi Jun,

One important reason why people want to use Cactus is to avoid
dealing with MPI functions. If you want to use GFs you probably
need get yourself familiar with the idea of ghostzones which
are explained in the Cactus user's manual.

Anyway, you can do explicit data exchange with MPI functions just
like any other MPI applications but you will lose the benefit of
using Cactus.

Btw, would you please tell us more about the application you
want to develop with Cactus ?

Regards,
Jian

Y. Jun Zhang wrote:
> Dear all:
> 
>       How do the data exchange between processors? It seems trouble
> in cactus, I try to define the data array in "interface.ccl" and
> "type=GF", but Processor A can not accpet the data array of Processor
> B. Maybe I should use MPI function to send this array?
> 
>      Best regards
> 
> --
> -----------------------------------------------------------
> Y. Jun Zhang
> E-Mail:
> j.y.zhangcn at gmail.com
> woxinfeijian at gmail.com
> QQ:3752942
> TEL:02763280205 13545299600
> 
> 
> 


From baiotti at ea.c.u-tokyo.ac.jp  Wed Jan 14 21:21:37 2009
From: baiotti at ea.c.u-tokyo.ac.jp (Baiotti Luca)
Date: Thu, 15 Jan 2009 12:21:37 +0900
Subject: [Users] per-file tolerance in testsuites
In-Reply-To: <496B5199.6020608@aei.mpg.de>
References: <493801BA.4040807@nasa.gov>	<6C72C3B5-7C05-4E06-8784-3E59583C240A@cct.lsu.edu>	<494F335F.7090505@ea.c.u-tokyo.ac.jp>	<B3780BC1-773D-4E7F-804D-33B1CDBCCD62@cct.lsu.edu>	<4951E09F.60104@ea.c.u-tokyo.ac.jp>
	<496B5199.6020608@aei.mpg.de>
Message-ID: <496EABC1.8030603@ea.c.u-tokyo.ac.jp>

Thomas Radke wrote:
> Baiotti Luca wrote:
>> Sorry, that information (about NPROCS) is already in the documentation, 
>> but not in the documentation on the webpage, which I had looked and 
>> which refers to the stable version.
>>
>> What about my question / feature request? Is it reasonable?
>>
>> I repeat my previous email here:
>>
>> "In the Cactus testsuite mechanism it is possible to select the absolute 
>> and relative tolerance for a given thorn or for a given testsuite of a 
>> thorn. This information must be included in the test.ccl file.
>>
>> Would it be possible to specify a tolerance for a specific file of a 
>> specific testsuite? It is indeed the case that some output files may 
>> have less stringent tolerances (like some norms).
>>
>> If this is not possible, could it be implemented?"
> 
> Hi Luca,
> 
> currently it is not possible to specify a data tolerance value per 
> testsuite output file, or for individual colums therein. Implementing 
> such a feature is possible but would require careful revision of the 
> existing code in the (rather large) testsuite comparison script.
> 
> As a simple alternative I suggest you create multiple testsuites, one 
> for each tolerance needed.

Hi Thomas,

I will look into the temporary alternative. I would like this issue to 
be put in the low-priority development plan of Cactus. Maybe a student 
could do that? Is this thread enough or should I open a feature request?

Luca

From tradke at aei.mpg.de  Thu Jan 15 05:39:15 2009
From: tradke at aei.mpg.de (Thomas Radke)
Date: Thu, 15 Jan 2009 12:39:15 +0100
Subject: [Users] per-file tolerance in testsuites
In-Reply-To: <496EABC1.8030603@ea.c.u-tokyo.ac.jp>
References: <493801BA.4040807@nasa.gov>	<6C72C3B5-7C05-4E06-8784-3E59583C240A@cct.lsu.edu>	<494F335F.7090505@ea.c.u-tokyo.ac.jp>	<B3780BC1-773D-4E7F-804D-33B1CDBCCD62@cct.lsu.edu>	<4951E09F.60104@ea.c.u-tokyo.ac.jp>	<496B5199.6020608@aei.mpg.de>
	<496EABC1.8030603@ea.c.u-tokyo.ac.jp>
Message-ID: <496F2063.8030002@aei.mpg.de>

Baiotti Luca wrote:
> Thomas Radke wrote:
>> Baiotti Luca wrote:
>>> Sorry, that information (about NPROCS) is already in the documentation, 
>>> but not in the documentation on the webpage, which I had looked and 
>>> which refers to the stable version.
>>>
>>> What about my question / feature request? Is it reasonable?
>>>
>>> I repeat my previous email here:
>>>
>>> "In the Cactus testsuite mechanism it is possible to select the absolute 
>>> and relative tolerance for a given thorn or for a given testsuite of a 
>>> thorn. This information must be included in the test.ccl file.
>>>
>>> Would it be possible to specify a tolerance for a specific file of a 
>>> specific testsuite? It is indeed the case that some output files may 
>>> have less stringent tolerances (like some norms).
>>>
>>> If this is not possible, could it be implemented?"
>> Hi Luca,
>>
>> currently it is not possible to specify a data tolerance value per 
>> testsuite output file, or for individual colums therein. Implementing 
>> such a feature is possible but would require careful revision of the 
>> existing code in the (rather large) testsuite comparison script.
>>
>> As a simple alternative I suggest you create multiple testsuites, one 
>> for each tolerance needed.
> 
> Hi Thomas,
> 
> I will look into the temporary alternative. I would like this issue to 
> be put in the low-priority development plan of Cactus. Maybe a student 
> could do that? 

Sure. If we only could find one.

> Is this thread enough or should I open a feature request?

Yes, please open a feature request in GNATS 
(http://www.cactuscode.org/cactus_cgi-bin/gnatsweb.pl) so that it's 
recorded at least.

-- 
Cheers, Thomas.

From mohammad.rashti at ece.queensu.ca  Sat Jan 17 16:07:57 2009
From: mohammad.rashti at ece.queensu.ca (Mohammad Javad Rashti)
Date: Sat, 17 Jan 2009 17:07:57 -0500
Subject: [Users] Problem running Bench_BSSN_PUGH with MPI on a cluster
Message-ID: <3aea68870901171407m525b43d5wa59049b4a9cd7644@mail.gmail.com>

Hi,
This is my first time using Cactus.
I am trying to run the
Bench_BSSN_PUGH<http://www.cactuscode.org/Benchmarks/bench_bssn_pugh>with
MPI. I have already compiled the code and the executable is ready.
I just do not know what parameters are to be set in the parameter file to
tell the cactus about the machines and how
to run Cactus on them. I have four DELL machines (using one Xeon processor
per machine) and want to run Cactus using MPI2
(a version of MPICH2),  using the following command line:

mpiexec -machinefile <my_machine_file_path> -np 4 ./exe/cactus_IRIS
./BENCHMARKS/Bench_MY_PUGH.par

I have added these parameters to the existing sample Bench_BSSN_PUGH_80l.par
parameter file, but not sure whether
this is correct:

PUGH::global_nsize = 30

PUGH::periodic = "yes"

PUGH::processor_topology = manual
PUGH::processor_topology_3d_x = 2
PUGH::processor_topology_3d_y = 2
PUGH::processor_topology_3d_z = 1


Unfortunately the application is terminated after printing some output.
The output file and the parameter file are attached to this email.
My MPI works with some microbenchmarks and applications that I have run
before.


Thanks in advance for your help.

Mohammad
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.cactuscode.org/pipermail/users/attachments/20090117/0df1c446/attachment-0001.html 
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: results.txt
Url: http://www.cactuscode.org/pipermail/users/attachments/20090117/0df1c446/attachment-0001.txt 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: Bench_MY_PUGH.par
Type: application/octet-stream
Size: 1766 bytes
Desc: not available
Url : http://www.cactuscode.org/pipermail/users/attachments/20090117/0df1c446/attachment-0001.obj 

From dlleuz at xmission.com  Sun Jan 18 16:54:35 2009
From: dlleuz at xmission.com (rsoares)
Date: Sun, 18 Jan 2009 15:54:35 -0700
Subject: [Users] Mixing 32-bit and 64-bit machines.
Message-ID: <4973B32B.9090206@xmission.com>

Hi,

Anyone know if the computer runs are done on

1. all 32-bit
2. all 64-bit
3.  sometimes a mix of both

If I am running on three 32-bit and one 64-bit what parameters can I set 
to keep things homogeneous.  When does it matter?

Thanks for your time.

R.Soares

From dlleuz at xmission.com  Sun Jan 18 19:13:21 2009
From: dlleuz at xmission.com (rsoares)
Date: Sun, 18 Jan 2009 18:13:21 -0700
Subject: [Users] Problem running Bench_BSSN_PUGH with MPI on a cluster
In-Reply-To: <3aea68870901171407m525b43d5wa59049b4a9cd7644@mail.gmail.com>
References: <3aea68870901171407m525b43d5wa59049b4a9cd7644@mail.gmail.com>
Message-ID: <4973D3B1.5070500@xmission.com>

Mohammad,

I have 4 computers also trying to run parallel.  May I ask what were the 
MPI microbenchmarks you used, also what version MPICH2?  I use 
mpich2-1.0.5p4.

Thanks,

R.Soares

Mohammad Javad Rashti wrote:
> Hi,
> This is my first time using Cactus.
> I am trying to run the Bench_BSSN_PUGH 
> <http://www.cactuscode.org/Benchmarks/bench_bssn_pugh> with MPI. I 
> have already compiled the code and the executable is ready.
> I just do not know what parameters are to be set in the parameter file 
> to tell the cactus about the machines and how
> to run Cactus on them. I have four DELL machines (using one Xeon 
> processor per machine) and want to run Cactus using MPI2
> (a version of MPICH2),  using the following command line:
>
> mpiexec -machinefile <my_machine_file_path> -np 4 ./exe/cactus_IRIS 
> ./BENCHMARKS/Bench_MY_PUGH.par
>
> I have added these parameters to the existing sample 
> Bench_BSSN_PUGH_80l.par parameter file, but not sure whether
> this is correct:
>
> PUGH::global_nsize = 30
>
> PUGH::periodic = "yes"
>
> PUGH::processor_topology = manual
> PUGH::processor_topology_3d_x = 2
> PUGH::processor_topology_3d_y = 2
> PUGH::processor_topology_3d_z = 1
>
>
> Unfortunately the application is terminated after printing some output.
> The output file and the parameter file are attached to this email.
> My MPI works with some microbenchmarks and applications that I have 
> run before.
>
>
> Thanks in advance for your help.
>
> Mohammad
> ------------------------------------------------------------------------
>
> _______________________________________________
> Users mailing list
> Users at cactuscode.org
> http://www.cactuscode.org/mailman/listinfo/users

From tradke at aei.mpg.de  Mon Jan 19 03:35:14 2009
From: tradke at aei.mpg.de (Thomas Radke)
Date: Mon, 19 Jan 2009 10:35:14 +0100
Subject: [Users] Problem running Bench_BSSN_PUGH with MPI on a cluster
In-Reply-To: <3aea68870901171407m525b43d5wa59049b4a9cd7644@mail.gmail.com>
References: <3aea68870901171407m525b43d5wa59049b4a9cd7644@mail.gmail.com>
Message-ID: <49744952.8040901@aei.mpg.de>

Mohammad Javad Rashti wrote:
> Hi,
> This is my first time using Cactus.
> I am trying to run the Bench_BSSN_PUGH 
> <http://www.cactuscode.org/Benchmarks/bench_bssn_pugh> with MPI. I have 
> already compiled the code and the executable is ready.
> I just do not know what parameters are to be set in the parameter file 
> to tell the cactus about the machines and how
> to run Cactus on them.

Hi Mohammad,

the application doesn't need to know anything about how it should be 
launched as a parallel MPI job across multiple machines - this is taken 
care of by the MPI startup mechanism of the actual MPI implementation 
used (in your case: MPICH2's mpiexec script).

> I have four DELL machines (using one Xeon 
> processor per machine) and want to run Cactus using MPI2
> (a version of MPICH2),  using the following command line:
> 
> mpiexec -machinefile <my_machine_file_path> -np 4 ./exe/cactus_IRIS 
> ./BENCHMARKS/Bench_MY_PUGH.par
> 
> I have added these parameters to the existing sample 
> Bench_BSSN_PUGH_80l.par parameter file, but not sure whether
> this is correct:
> 
> PUGH::global_nsize = 30
> 
> PUGH::periodic = "yes"
> 
> PUGH::processor_topology = manual
> PUGH::processor_topology_3d_x = 2
> PUGH::processor_topology_3d_y = 2
> PUGH::processor_topology_3d_z = 1
> 
> 
> Unfortunately the application is terminated after printing some output.
> The output file and the parameter file are attached to this email.
> My MPI works with some microbenchmarks and applications that I have run 
> before.

The MPI error messages at the bottom of your log output indicate that 
task 3 aborted due to a segmentation violation, and task 1 (and probably 
the other two MPI tasks as well) were then explicitly killed.

Are you trying Cactus on a homogeneous set of machines, or is the forth 
machine (where task 3 ran) any special ? Can you run Cactus successfully 
with only 2 or 3 MPI tasks (you'll have to change your topology 
parameters for this), or with 4 tasks on a single machine or across only 
2 or 3 machines ?

-- 
Cheers, Thomas.


From tradke at aei.mpg.de  Mon Jan 19 04:57:15 2009
From: tradke at aei.mpg.de (Thomas Radke)
Date: Mon, 19 Jan 2009 11:57:15 +0100
Subject: [Users] Mixing 32-bit and 64-bit machines.
In-Reply-To: <4973B32B.9090206@xmission.com>
References: <4973B32B.9090206@xmission.com>
Message-ID: <49745C8B.5080304@aei.mpg.de>

rsoares wrote:
> Hi,
> 
> Anyone know if the computer runs are done on
> 
> 1. all 32-bit
> 2. all 64-bit
> 3.  sometimes a mix of both
> 
> If I am running on three 32-bit and one 64-bit what parameters can I set 
> to keep things homogeneous.  When does it matter?

MPI programs should run fine across heterorgeneous machines as long as 
the MPI datatypes in the corresponding communication calls are the same 
(eg. don't mix MPI_INT with MPI_LONG) _and_ they match up with the 
language datatypes of the variables used in those calls (eg. MPI_INT is 
used to communicate native integers).
The MPI routines will take care of necessary datatype conversions due to 
different binary representations of numerical types (eg. high/low 
endianess conversion of integers).

For parallel Cactus simulations in a heterogeneous environment you 
should make sure that the compilers used to generate Cactus executables 
use the same precision for their generic numerical datatypes (typically 
4 bytes for integers, and 8 bytes for double-precision floating point), 
and that you configure Cactus with the same precision for the generic 
Cactus datatypes CCTK_INT and CCTK_REAL.

-- 
Cheers, Thomas.


From mohammad.rashti at ece.queensu.ca  Mon Jan 19 08:46:11 2009
From: mohammad.rashti at ece.queensu.ca (Mohammad Javad Rashti)
Date: Mon, 19 Jan 2009 09:46:11 -0500
Subject: [Users] Problem running Bench_BSSN_PUGH with MPI on a cluster
In-Reply-To: <49744952.8040901@aei.mpg.de>
References: <3aea68870901171407m525b43d5wa59049b4a9cd7644@mail.gmail.com>
	<49744952.8040901@aei.mpg.de>
Message-ID: <3aea68870901190646i18e2749bh6b2273d174943719@mail.gmail.com>

Hi Thomas,
Thanks a lot for your comments.
Yes, the platform is homogeneous and I have tried these other cases of
mapping
(2,3 processors, etc.) and I am receiving the same error. The program runs
perfectly
on 1 processor using mpiexec.

Did you look at the parameter file? Is there any particular error there? I
know some topology information
needs to be added there.

Thanks
Mohammad

On Mon, Jan 19, 2009 at 4:35 AM, Thomas Radke <tradke at aei.mpg.de> wrote:

> Mohammad Javad Rashti wrote:
> > Hi,
> > This is my first time using Cactus.
> > I am trying to run the Bench_BSSN_PUGH
> > <http://www.cactuscode.org/Benchmarks/bench_bssn_pugh> with MPI. I have
> > already compiled the code and the executable is ready.
> > I just do not know what parameters are to be set in the parameter file
> > to tell the cactus about the machines and how
> > to run Cactus on them.
>
> Hi Mohammad,
>
> the application doesn't need to know anything about how it should be
> launched as a parallel MPI job across multiple machines - this is taken
> care of by the MPI startup mechanism of the actual MPI implementation
> used (in your case: MPICH2's mpiexec script).
>
> > I have four DELL machines (using one Xeon
> > processor per machine) and want to run Cactus using MPI2
> > (a version of MPICH2),  using the following command line:
> >
> > mpiexec -machinefile <my_machine_file_path> -np 4 ./exe/cactus_IRIS
> > ./BENCHMARKS/Bench_MY_PUGH.par
> >
> > I have added these parameters to the existing sample
> > Bench_BSSN_PUGH_80l.par parameter file, but not sure whether
> > this is correct:
> >
> > PUGH::global_nsize = 30
> >
> > PUGH::periodic = "yes"
> >
> > PUGH::processor_topology = manual
> > PUGH::processor_topology_3d_x = 2
> > PUGH::processor_topology_3d_y = 2
> > PUGH::processor_topology_3d_z = 1
> >
> >
> > Unfortunately the application is terminated after printing some output.
> > The output file and the parameter file are attached to this email.
> > My MPI works with some microbenchmarks and applications that I have run
> > before.
>
> The MPI error messages at the bottom of your log output indicate that
> task 3 aborted due to a segmentation violation, and task 1 (and probably
> the other two MPI tasks as well) were then explicitly killed.
>
> Are you trying Cactus on a homogeneous set of machines, or is the forth
> machine (where task 3 ran) any special ? Can you run Cactus successfully
> with only 2 or 3 MPI tasks (you'll have to change your topology
> parameters for this), or with 4 tasks on a single machine or across only
> 2 or 3 machines ?
>
> --
> Cheers, Thomas.
>
> _______________________________________________
> Users mailing list
> Users at cactuscode.org
> http://www.cactuscode.org/mailman/listinfo/users
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.cactuscode.org/pipermail/users/attachments/20090119/1fd6ca9b/attachment.html 

From tradke at aei.mpg.de  Mon Jan 19 09:18:50 2009
From: tradke at aei.mpg.de (Thomas Radke)
Date: Mon, 19 Jan 2009 16:18:50 +0100
Subject: [Users] Problem running Bench_BSSN_PUGH with MPI on a cluster
In-Reply-To: <3aea68870901190646i18e2749bh6b2273d174943719@mail.gmail.com>
References: <3aea68870901171407m525b43d5wa59049b4a9cd7644@mail.gmail.com>	<49744952.8040901@aei.mpg.de>
	<3aea68870901190646i18e2749bh6b2273d174943719@mail.gmail.com>
Message-ID: <497499DA.6010401@aei.mpg.de>

Mohammad Javad Rashti wrote:
> Hi Thomas,
> Thanks a lot for your comments.
> Yes, the platform is homogeneous and I have tried these other cases of 
> mapping
> (2,3 processors, etc.) and I am receiving the same error. The program 
> runs perfectly
> on 1 processor using mpiexec.

It also doesn't run with multiple MPI tasks mapped onto the same machine ?
Instead of a BenchBSSN simulation, did you try a simple wavetoy example ?

> Did you look at the parameter file? Is there any particular error there? 
> I know some topology information
> needs to be added there.

No, the topology parameter settings you have should work for a 
simulation with 4 MPI tasks.

For testing you could just leave out those parameters in your parfile, 
and have the PUGH driver choose an automatic topology setting for you. 
This would then work for any number of MPI tasks.

-- 
Cheers, Thomas.

From mohammad.rashti at ece.queensu.ca  Mon Jan 19 09:56:04 2009
From: mohammad.rashti at ece.queensu.ca (Mohammad Javad Rashti)
Date: Mon, 19 Jan 2009 10:56:04 -0500
Subject: [Users] Problem running Bench_BSSN_PUGH with MPI on a cluster
In-Reply-To: <497499DA.6010401@aei.mpg.de>
References: <3aea68870901171407m525b43d5wa59049b4a9cd7644@mail.gmail.com>
	<49744952.8040901@aei.mpg.de>
	<3aea68870901190646i18e2749bh6b2273d174943719@mail.gmail.com>
	<497499DA.6010401@aei.mpg.de>
Message-ID: <3aea68870901190756s21a62357qda18658322626c88@mail.gmail.com>

Thomas,
No it does the same thing on a single machine (I have 2-processor machines,
so 4-procs is not
possible but with 2-procs similar output is printed).
I could not  find the parfile for wavetoy. Can you guide me how to try that.

The website does not have the parfile, neither I could find it in the
distribution.
The file that I have compiled, installed is the tar file from this page:

http://www.cactuscode.org/Benchmarks/bench_bssn_pugh

Should I download another file?

I appreciate your help.
Mohammad

On Mon, Jan 19, 2009 at 10:18 AM, Thomas Radke <tradke at aei.mpg.de> wrote:

> Mohammad Javad Rashti wrote:
> > Hi Thomas,
> > Thanks a lot for your comments.
> > Yes, the platform is homogeneous and I have tried these other cases of
> > mapping
> > (2,3 processors, etc.) and I am receiving the same error. The program
> > runs perfectly
> > on 1 processor using mpiexec.
>
> It also doesn't run with multiple MPI tasks mapped onto the same machine ?
> Instead of a BenchBSSN simulation, did you try a simple wavetoy example ?
>
> > Did you look at the parameter file? Is there any particular error there?
> > I know some topology information
> > needs to be added there.
>
> No, the topology parameter settings you have should work for a
> simulation with 4 MPI tasks.
>
> For testing you could just leave out those parameters in your parfile,
> and have the PUGH driver choose an automatic topology setting for you.
> This would then work for any number of MPI tasks.
>
> --
> Cheers, Thomas.
> _______________________________________________
> Users mailing list
> Users at cactuscode.org
> http://www.cactuscode.org/mailman/listinfo/users
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.cactuscode.org/pipermail/users/attachments/20090119/b43a6fcf/attachment-0001.html 

From tradke at aei.mpg.de  Mon Jan 19 11:55:44 2009
From: tradke at aei.mpg.de (Thomas Radke)
Date: Mon, 19 Jan 2009 18:55:44 +0100
Subject: [Users] Problem running Bench_BSSN_PUGH with MPI on a cluster
In-Reply-To: <3aea68870901190756s21a62357qda18658322626c88@mail.gmail.com>
References: <3aea68870901171407m525b43d5wa59049b4a9cd7644@mail.gmail.com>	<49744952.8040901@aei.mpg.de>	<3aea68870901190646i18e2749bh6b2273d174943719@mail.gmail.com>	<497499DA.6010401@aei.mpg.de>
	<3aea68870901190756s21a62357qda18658322626c88@mail.gmail.com>
Message-ID: <4974BEA0.7040003@aei.mpg.de>

Mohammad Javad Rashti wrote:
> Thomas,
> No it does the same thing on a single machine (I have 2-processor 
> machines, so 4-procs is not
> possible but with 2-procs similar output is printed).

You can run more MPI tasks on a machine than there are physical 
processors. The application would just use multiple processes in such a 
case.
But apparently there is a general problem with Cactus running in 
parallel even on only one of your machines. You could try a less 
complicated Cactus test case (see below), or some alternative MPI 
implementation (eg. LAM, http://www.lam-mpi.org/).

> I could not  find the parfile for wavetoy. Can you guide me how to try 
> that.
> The website does not have the parfile, neither I could find it in the 
> distribution.
> The file that I have compiled, installed is the tar file from this page:
> 
> http://www.cactuscode.org/Benchmarks/bench_bssn_pugh
> 
> Should I download another file?

Instead of the Bench_BSSN_PUGH tarball you can download and install 
BenchIO_HDF5 (http://www.cactuscode.org/Benchmarks/benchio_hdf5). The 
parameter file for this benchmark has no physics at all in it (not even 
a wavetoy example), and you could even take out all the IO stuff to make 
it even simpler for testing.

-- 
Cheers, Thomas.

From mohammad.rashti at ece.queensu.ca  Tue Jan 20 08:38:31 2009
From: mohammad.rashti at ece.queensu.ca (Mohammad Javad Rashti)
Date: Tue, 20 Jan 2009 09:38:31 -0500
Subject: [Users] Problem running Bench_BSSN_PUGH with MPI on a cluster
In-Reply-To: <4974BEA0.7040003@aei.mpg.de>
References: <3aea68870901171407m525b43d5wa59049b4a9cd7644@mail.gmail.com>
	<49744952.8040901@aei.mpg.de>
	<3aea68870901190646i18e2749bh6b2273d174943719@mail.gmail.com>
	<497499DA.6010401@aei.mpg.de>
	<3aea68870901190756s21a62357qda18658322626c88@mail.gmail.com>
	<4974BEA0.7040003@aei.mpg.de>
Message-ID: <3aea68870901200638u1d80e855i3cb1a381bed6ce4b@mail.gmail.com>

Thanks a lot Thomas,
I will try that BenchIO_HDF5, which apparently requires HDF5
library as well, and let you know if I need help.
Mohammad

On Mon, Jan 19, 2009 at 12:55 PM, Thomas Radke <tradke at aei.mpg.de> wrote:

> Mohammad Javad Rashti wrote:
> > Thomas,
> > No it does the same thing on a single machine (I have 2-processor
> > machines, so 4-procs is not
> > possible but with 2-procs similar output is printed).
>
> You can run more MPI tasks on a machine than there are physical
> processors. The application would just use multiple processes in such a
> case.
> But apparently there is a general problem with Cactus running in
> parallel even on only one of your machines. You could try a less
> complicated Cactus test case (see below), or some alternative MPI
> implementation (eg. LAM, http://www.lam-mpi.org/).
>
> > I could not  find the parfile for wavetoy. Can you guide me how to try
> > that.
> > The website does not have the parfile, neither I could find it in the
> > distribution.
> > The file that I have compiled, installed is the tar file from this page:
> >
> > http://www.cactuscode.org/Benchmarks/bench_bssn_pugh
> >
> > Should I download another file?
>
> Instead of the Bench_BSSN_PUGH tarball you can download and install
> BenchIO_HDF5 (http://www.cactuscode.org/Benchmarks/benchio_hdf5). The
> parameter file for this benchmark has no physics at all in it (not even
> a wavetoy example), and you could even take out all the IO stuff to make
> it even simpler for testing.
>
> --
> Cheers, Thomas.
> _______________________________________________
> Users mailing list
> Users at cactuscode.org
> http://www.cactuscode.org/mailman/listinfo/users
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.cactuscode.org/pipermail/users/attachments/20090120/165ddf9b/attachment.html 

From physik at fangwolg.de  Thu Jan 22 08:39:18 2009
From: physik at fangwolg.de (Wolfgang Kastaun)
Date: Thu, 22 Jan 2009 15:39:18 +0100
Subject: [Users] Recent Cactus assumes C99
Message-ID: <200901221539.18461.physik@fangwolg.de>

Hi,

after checking out a recent cactus (12.1.09), it did not compile anymore.
The error was

Preprocessing /net/hydra/home/kastaun/wd/cactus_12.1.09/arrangements/CactusUtils/TimerReport/src/Output.c
Compiling /net/hydra/home/kastaun/wd/cactus_12.1.09/arrangements/CactusUtils/TimerReport/src/Output.c
/net/hydra/home/kastaun/wd/cactus_12.1.09/configs/idfcow_pioneer/build/TimerReport/Output.c: 
In function 'OutputAllTimers':
/net/hydra/home/kastaun/wd/cactus_12.1.09/configs/idfcow_pioneer/build/TimerReport/Output.c:210: 
error: 'for' loop initial declaration used outside C99 mode

in this C-file, there is a C++ style declaration of an int inside the loop 
header. Using the gcc-4.0.1 compiler, I had to add
CFLAGS   = -std=c99
to compile the code again.

Should be easy to fix by declaring the counter outside the loop.

Cheers,
Wolfgang.

From yfeng1 at lsu.edu  Mon Jan 26 23:58:28 2009
From: yfeng1 at lsu.edu (Yin Feng)
Date: Mon, 26 Jan 2009 23:58:28 -0600
Subject: [Users] Errors when Compiling Cactus with PETSC
Message-ID: <20090127055828.7D15611B724@ws4-1.us4.outblaze.com>

I am a beginner for Cactus, when I use gmake xxx-config mpi=openmpi petsc=yes hdf5=yes,
It reports errors like:

 Using HDF5 package in /home/yfeng1/Soft/hdf5-1.8.2
  Found parallel HDF5 library, so Cactus will potentially make use of parallel H
DF5 support.
Configuring with MPI
  OpenMPI selected
  Using "/home/yfeng1/Soft/openmpi-1.3/include /home/yfeng1/Soft/openmpi-1.3/inc
lude/openmpi/" as OpenMPI include path
  Using "/home/yfeng1/Soft/openmpi-1.3/lib" as OpenMPI library path
Configuring with PETSc
  Using PETSc package in /home/yfeng1/Soft/petsc-3.0.0-p2
  Using PETSc architecture 'linux-gnu-c'
  No PETSc support for architecture 'linux-gnu-c' !

Is there any one kown why  Cactus says no PETSc support for architecture 'linux-gnu-c' and how to solve?

Thanks in advance!

Sincerely,
Yin


From schnetter at cct.lsu.edu  Tue Jan 27 00:07:25 2009
From: schnetter at cct.lsu.edu (Erik Schnetter)
Date: Tue, 27 Jan 2009 00:07:25 -0600
Subject: [Users] [CactusMaint] Errors when Compiling Cactus with PETSC
In-Reply-To: <20090127055828.7D15611B724@ws4-1.us4.outblaze.com>
References: <20090127055828.7D15611B724@ws4-1.us4.outblaze.com>
Message-ID: <202AB67D-FD3E-4487-B049-9CFBCA4D7B1C@cct.lsu.edu>

On Jan 26, 2009, at 23:58:28, Yin Feng wrote:

> I am a beginner for Cactus, when I use gmake xxx-config mpi=openmpi  
> petsc=yes hdf5=yes,
> It reports errors like:
>
> Using HDF5 package in /home/yfeng1/Soft/hdf5-1.8.2
>  Found parallel HDF5 library, so Cactus will potentially make use of  
> parallel H
> DF5 support.
> Configuring with MPI
>  OpenMPI selected
>  Using "/home/yfeng1/Soft/openmpi-1.3/include /home/yfeng1/Soft/ 
> openmpi-1.3/inc
> lude/openmpi/" as OpenMPI include path
>  Using "/home/yfeng1/Soft/openmpi-1.3/lib" as OpenMPI library path
> Configuring with PETSc
>  Using PETSc package in /home/yfeng1/Soft/petsc-3.0.0-p2
>  Using PETSc architecture 'linux-gnu-c'
>  No PETSc support for architecture 'linux-gnu-c' !
>
> Is there any one kown why  Cactus says no PETSc support for  
> architecture 'linux-gnu-c' and how to solve?


Yin

When PETSc is installed, it is installed as one (or multiple)  
"architectures".  The names of these architectures are chosen when  
PETSc is installed.  Presumably, your PETSc library as installed under  
a different name than "linux-gnu-c", which is a default that Cactus  
assumes since you didn't specify an architecture name manually.  You  
can set a configuration variable "PETSC_ARCH" to your PETSc  
architecture.

It is advisable to specify configuration options in an options file,  
not on the command line, since this makes it much easier to re- 
configure with slightly different options if this becomes necessary.

-erik

-- 
Erik Schnetter <schnetter at cct.lsu.edu>   http://www.cct.lsu.edu/~eschnett/

My email is as private as my paper mail.  I therefore support encrypting
and signing email messages.  Get my PGP key from <http://pgp.mit.edu>.

-------------- next part --------------
A non-text attachment was scrubbed...
Name: PGP.sig
Type: application/pgp-signature
Size: 194 bytes
Desc: This is a digitally signed message part
Url : http://www.cactuscode.org/pipermail/users/attachments/20090127/d7ab5e07/attachment.bin 

From schnetter at cct.lsu.edu  Tue Jan 27 12:20:33 2009
From: schnetter at cct.lsu.edu (Erik Schnetter)
Date: Tue, 27 Jan 2009 12:20:33 -0600
Subject: [Users] To Erik
In-Reply-To: <CPSMTP-EML103tRHQq0001001aa@cpsmtp-eml103.kpnxchange.com>
References: <CPSMTP-EML103tRHQq0001001aa@cpsmtp-eml103.kpnxchange.com>
Message-ID: <9A25B9CA-C52C-4E76-974A-0AF1A9257983@cct.lsu.edu>

On Jan 8, 2009, at 02:04:55, Dr. R.J.Slagter wrote:

> Hello Erik
> Thanks for your reply. I want to send you my set of equations, but I
> did not succeed because the keyserver is down.
> I can post the article on my website?


Of course.

(You don't have to use a key server to send me email; unencrypted  
email works as well.)

-erik

-- 
Erik Schnetter <schnetter at cct.lsu.edu>   http://www.cct.lsu.edu/~eschnett/

My email is as private as my paper mail.  I therefore support encrypting
and signing email messages.  Get my PGP key from <http://pgp.mit.edu>.

-------------- next part --------------
A non-text attachment was scrubbed...
Name: PGP.sig
Type: application/pgp-signature
Size: 194 bytes
Desc: This is a digitally signed message part
Url : http://www.cactuscode.org/pipermail/users/attachments/20090127/113fd260/attachment.bin 

From info at asfyon.nl  Tue Jan  6 10:12:19 2009
From: info at asfyon.nl (dr.r.j.slagter)
Date: Tue, 06 Jan 2009 16:12:19 -0000
Subject: [Users] PDE's Fortran (Erik Schnetter)
Message-ID: <CAABE5AE9F5F4B7B99D0C975E1CA9A3A@intel>

Hello Erik, Thanks for your reply.I work on GR for some time now ( see http://www.asfyon.nl/webdoc1.htm )
I used for my numerical calculations several routines (Cadsol,PDECOL RNPL) but I want to try for my latest
rsearch ( see attachtment) a different routine in CACTUS. The equations are ( 11) like equation 20.
The independent variables are t (time) r and y.
Hope you can help me. Regards Reinoud
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.cactuscode.org/pipermail/users/attachments/20090106/2801986b/attachment-0001.html 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: text1a.ps
Type: application/postscript
Size: 186222 bytes
Desc: not available
Url : http://www.cactuscode.org/pipermail/users/attachments/20090106/2801986b/attachment-0001.ps