Nabble - GNU Octave Repository

Re: example octave_comm_test updated

2009-11-26T06:16:07Z

Now it should work
I tested the new examples (have a loot at inst folder).
Let me know if it is the same for you.
Bests
Riccardo

--- Gio 26/11/09, Michael Creel <michael.creel@...> ha scritto:

Da: Michael Creel <michael.creel@...>
Oggetto: Re: example octave_comm_test updated
A: "Riccardo Corradini" <riccardocorradini@...>
Cc: "Jaroslav Hajek" <highegg@...>, "octave-forge list" <octave-dev@...>
Data: Giovedì 26 novembre 2009, 10:59

On Thu, Nov 26, 2009 at 10:45 AM, Riccardo Corradini <riccardocorradini@...> wrote:

Hi all,
I have updated the example octave_comm_test.cc ( I have just inserted a check on the input).
Let me know what you think.
Bests
Riccardo

I haven't be able to get things to compile with this new communicator stuff. From a check out right now, "make all" gives me

mkoctfile -c -I/usr/lib/openmpi/include -I/usr/lib/openmpi/include/openmpi comm-util.cc
comm-util.cc: In function ‘ompi_communicator_t* do_get(const octave_value&) [with T = int]’:
comm-util.cc:39: error: invalid conversion from ‘int’ to ‘ompi_communicator_t*’
comm-util.cc: At global scope:
comm-util.cc:34: error: explicit template specialization cannot have a storage class
comm-util.cc: In function ‘ompi_communicator_t* do_get(const octave_value&) [with T = long int]’:
comm-util.cc:51: error: invalid conversion from ‘long int’ to ‘ompi_communicator_t*’
comm-util.cc: At global scope:
comm-util.cc:46: error: explicit template specialization cannot have a storage class
comm-util.cc:72: error: explicit template specialization cannot have a storage class
michael@yosemite:~/Desktop/of/octave/trunk/octave-forge/extra/openmpi_ext/src$

I'm using openmpi 1.3.2 from Ubuntu:
michael@yosemite:~$ dpkg -l openmpi*
Desired=Unknown/Install/Remove/Purge/Hold
| Status=Not/Inst/Cfg-files/Unpacked/Failed-cfg/Half-inst/trig-aWait/Trig-pend
|/ Err?=(none)/Reinst-required (Status,Err: uppercase=bad)
||/ Name                Version             Description
+++-===================-===================-======================================================
ii openmpi-bin         1.3.2-3ubuntu1      high performance message passing library -- binaries
ii openmpi-common      1.3.2-3ubuntu1      high performance message passing library -- common fil
un openmpi-dev         <none>              (no description available)
un openmpi-libs0       <none>              (no description available)
michael@yosemite:~$ dpkg -l libopenm*
Desired=Unknown/Install/Remove/Purge/Hold
| Status=Not/Inst/Cfg-files/Unpacked/Failed-cfg/Half-inst/trig-aWait/Trig-pend
|/ Err?=(none)/Reinst-required (Status,Err: uppercase=bad)
||/ Name                Version             Description
+++-===================-===================-======================================================
ii libopenmpi-dev      1.3.2-3ubuntu1      high performance message passing library -- header fil
un libopenmpi1         <none>              (no description available)
ii libopenmpi1.3       1.3.2-3ubuntu1      high performance message passing library -- shared lib
michael@yosemite:~$

The original checkins of MPI_Probe and MPI_Iprobe do work for me, and I'm making a montecarlo.m that uses them. If you have time to add new functions, I'd like to see MPI_Bcast. I think that would complete the list of things I use for my econometric applications.

Cheers, Michael

------------------------------------------------------------------------------
Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day
trial. Simplify your report design, integration and deployment - and focus on
what you do best, core application coding. Discover what's new with
Crystal Reports now. http://p.sf.net/sfu/bobj-july
_______________________________________________
Octave-dev mailing list
Octave-dev@...
https://lists.sourceforge.net/lists/listinfo/octave-dev

From forum: octave-dev

Re: example octave_comm_test updated

2009-11-26T01:59:41Z

On Thu, Nov 26, 2009 at 10:45 AM, Riccardo Corradini <riccardocorradini@...> wrote:

Hi all,
I have updated the example octave_comm_test.cc ( I have just inserted a check on the input).
Let me know what you think.
Bests
Riccardo

I haven't be able to get things to compile with this new communicator stuff. From a check out right now, "make all" gives me

mkoctfile -c -I/usr/lib/openmpi/include -I/usr/lib/openmpi/include/openmpi comm-util.cc
comm-util.cc: In function ‘ompi_communicator_t* do_get(const octave_value&) [with T = int]’:
comm-util.cc:39: error: invalid conversion from ‘int’ to ‘ompi_communicator_t*’
comm-util.cc: At global scope:
comm-util.cc:34: error: explicit template specialization cannot have a storage class
comm-util.cc: In function ‘ompi_communicator_t* do_get(const octave_value&) [with T = long int]’:
comm-util.cc:51: error: invalid conversion from ‘long int’ to ‘ompi_communicator_t*’
comm-util.cc: At global scope:
comm-util.cc:46: error: explicit template specialization cannot have a storage class
comm-util.cc:72: error: explicit template specialization cannot have a storage class
michael@yosemite:~/Desktop/of/octave/trunk/octave-forge/extra/openmpi_ext/src$

I'm using openmpi 1.3.2 from Ubuntu:
michael@yosemite:~$ dpkg -l openmpi*
Desired=Unknown/Install/Remove/Purge/Hold
| Status=Not/Inst/Cfg-files/Unpacked/Failed-cfg/Half-inst/trig-aWait/Trig-pend
|/ Err?=(none)/Reinst-required (Status,Err: uppercase=bad)
||/ Name                Version             Description
+++-===================-===================-======================================================
ii openmpi-bin         1.3.2-3ubuntu1      high performance message passing library -- binaries
ii openmpi-common      1.3.2-3ubuntu1      high performance message passing library -- common fil
un openmpi-dev         <none>              (no description available)
un openmpi-libs0       <none>              (no description available)
michael@yosemite:~$ dpkg -l libopenm*
Desired=Unknown/Install/Remove/Purge/Hold
| Status=Not/Inst/Cfg-files/Unpacked/Failed-cfg/Half-inst/trig-aWait/Trig-pend
|/ Err?=(none)/Reinst-required (Status,Err: uppercase=bad)
||/ Name                Version             Description
+++-===================-===================-======================================================
ii libopenmpi-dev      1.3.2-3ubuntu1      high performance message passing library -- header fil
un libopenmpi1         <none>              (no description available)
ii libopenmpi1.3       1.3.2-3ubuntu1      high performance message passing library -- shared lib
michael@yosemite:~$

The original checkins of MPI_Probe and MPI_Iprobe do work for me, and I'm making a montecarlo.m that uses them. If you have time to add new functions, I'd like to see MPI_Bcast. I think that would complete the list of things I use for my econometric applications.

Cheers, Michael

From forum: octave-dev

example octave_comm_test updated

2009-11-26T01:45:57Z

Hi all,
I have updated the example octave_comm_test.cc ( I have just inserted a check on the input).
Let me know what you think.
Bests
Riccardo

From forum: octave-dev

New comunicator simple class

2009-11-25T09:33:14Z

Hi all,
I have just uploaded 3 new files.
Please let me know if you like them.
If the answer will be yes I will upgrade MPI_send etc.. accordingly
Bests
Riccardo

From forum: octave-dev

Re: Monte Carlo example for new MPI bindings

2009-11-25T07:44:15Z

Hi Jaroslav,
I have built correctly this cc file
#include <mpi.h>
#include <octave/oct.h>

class octave_comm : public octave_base_value
{
public:

octave_comm (const std::string _name = "", MPI_Comm _comm = MPI_COMM_WORLD )
    : octave_base_value (), name (_name)
    {
     MPI_Comm_set_name(_comm, const_cast<char*>(name.c_str()));
    }

void print (std::ostream& os, bool pr_as_read_syntax = false) const
    {
      os << name << std::endl;
    }
bool is_defined (void) const { return true; }

private:
const std::string name;
DECLARE_OCTAVE_ALLOCATOR
DECLARE_OV_TYPEID_FUNCTIONS_AND_DATA
};

DEFINE_OCTAVE_ALLOCATOR (octave_comm);
DEFINE_OV_TYPEID_FUNCTIONS_AND_DATA (octave_comm, "octave_comm", "octave_comm");

static bool octave_comm_type_loaded = false;

DEFUN_DLD(octave_comm, args, ,"")
{
if (!octave_comm_type_loaded)
    {
      octave_comm::register_type ();
      octave_comm_type_loaded = true;
      mlock ();
    }

octave_value retval;
if (args.length () != 1 || !args (0).is_string ())
    {
      error ("octave_comm: first argument must be a string");
      return retval;
    }

const std::string name = args (0).string_value ();
// For Default MPI_COMM_WORLD
MPI_Comm CW;
octave_comm *b=new octave_comm("",CW);
// For Custom Comunicator
MPI_Comm CCom;
octave_comm *c = new octave_comm("customcom",CCom);
return retval;
}

Now I would like to access the public property _comm .. of the octave_comm class, just from C++ code.
Could you please give me a hint?
Thanks a lot
Riccardo

--- Mar 24/11/09, Michael Creel <michael.creel@...> ha scritto:

Da: Michael Creel <michael.creel@...>
Oggetto: Re: Monte Carlo example for new MPI bindings
A: "Jaroslav Hajek" <highegg@...>
Cc: "Riccardo Corradini" <riccardocorradini@...>, "octave-forge list" <octave-dev@...>
Data: Martedì 24 novembre 2009, 10:36

On Mon, Nov 23, 2009 at 2:31 PM, Jaroslav Hajek <highegg@...> wrote:

Hi Riccardo and Michael,

I have already implemented basic support for the communicators in the SVN tree.
The relevant functions now accept comm as their last argument, defaulting to MPI_COMM_WORLD if not provided.
MPI_COMM_WORLD is a new function.

Note 1: The implementation imposes the requirement that MPI_Comm is a typedef for signed 32-bit or 64-bit integer (int, long, or long long) which is true in all major MPI implementations with the notable exception of OpenMPI. OpenMPI uses raw pointers, which make things a bit difficult. One option is to simply typecast the pointers to integers; but I feel a bit uneasy about working with raw pointers in Octave, so I'm thinking about a solution providing some validation as well.
I would, however, like to avoid creating a user type for each of MPI's types.

Note 2:
There are no functions for creating new communicators yet.

Hi all,
A check out this morning is not compiling for me:

michael@yosemite:~/Desktop/openmpi_ext/src$ make all
mkoctfile -c -I/usr/lib/openmpi/include -I/usr/lib/openmpi/include/openmpi comm-util.cc
comm-util.cc: In function ‘ompi_communicator_t* do_get(const octave_value&) [with T = int]’:
comm-util.cc:39: error: invalid conversion from ‘int’ to ‘ompi_communicator_t*’
comm-util.cc: At global scope:
comm-util.cc:34: error: explicit template specialization cannot have a storage class
comm-util.cc: In function ‘ompi_communicator_t* do_get(const octave_value&) [with T = long int]’:
comm-util.cc:51: error: invalid conversion from ‘long int’ to ‘ompi_communicator_t*’
comm-util.cc: At global scope:
comm-util.cc:46: error: explicit template specialization cannot have a storage class
comm-util.cc:72: error: explicit template specialization cannot have a storage class
comm-util.cc:78: error: explicit template specialization cannot have a storage class
make: *** [comm-util.o] Error 1
michael@yosemite:~/Desktop/openmpi_ext/src$

Is that expected at the moment?

Cheers, Michael

From forum: octave-dev

Re: Release system / web page generation

2009-11-25T05:45:59Z

On 25 Nov 2009, at 14:35, Søren Hauberg wrote:

> I must admit that I quickly give up on the SF documentation, so I just
> might be missing the obvious. To me, it seems like files can be copied
> to the server using
>
> Rsync over SSH
> SFTP
> SCP
>
> I don't see any web tools for uploading html files to the web server,
> but perhaps I'm just not looking hard enough?

you're right the SF documentation is a mess ...
I'll investigate this further.

> Søren
c.
------------------------------------------------------------------------------
Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day
trial. Simplify your report design, integration and deployment - and focus on
what you do best, core application coding. Discover what's new with
Crystal Reports now. http://p.sf.net/sfu/bobj-july
_______________________________________________
Octave-dev mailing list
Octave-dev@...
https://lists.sourceforge.net/lists/listinfo/octave-dev

From forum: octave-dev

Re: Release system / web page generation

2009-11-25T05:44:20Z

On 25 Nov 2009, at 14:38, Søren Hauberg wrote:

> We have quite a lot of packages, so I think we would have to only put
> the most recent announcement on the front page, otherwise there
> would be
> way too much information on the front page.

so the php code would sort by date and then only display the first N
entries.
the NEWS page (which, BTW is currently not working) could include the
same code with a larger value of N.

> But I think your idea of
> just added an 'announce_<date>.html' file to each package makes sense.

if you take care of updating generate_html I could do the php part,
but I doubt I'll be able to do that before Dec 15

> Søren
c.
------------------------------------------------------------------------------
Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day
trial. Simplify your report design, integration and deployment - and focus on
what you do best, core application coding. Discover what's new with
Crystal Reports now. http://p.sf.net/sfu/bobj-july
_______________________________________________
Octave-dev mailing list
Octave-dev@...
https://lists.sourceforge.net/lists/listinfo/octave-dev

From forum: octave-dev

Re: Release system / web page generation

2009-11-25T05:38:14Z

ons, 25 11 2009 kl. 10:02 +0100, skrev Carlo de Falco:

> On 25 Nov 2009, at 09:52, Søren Hauberg wrote:
>
> > Hmm, I guess this is possible. This requires that we can determine
> > which
> > packages have changed most recently. Do you think it is enough to
> > simple
> > check the dates of the directories containing the html for the
> > packages?
>
> more simply, we can have generate_package_html add a file named
>
> announce_<date>.html
>
> containing the announcement to the package directory and the logic in
> the front page would
> sort the announcements by descending date before displaying them.

We have quite a lot of packages, so I think we would have to only put
the most recent announcement on the front page, otherwise there would be
way too much information on the front page. But I think your idea of
just added an 'announce_<date>.html' file to each package makes sense.

Søren

------------------------------------------------------------------------------
Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day
trial. Simplify your report design, integration and deployment - and focus on
what you do best, core application coding. Discover what's new with
Crystal Reports now. http://p.sf.net/sfu/bobj-july
_______________________________________________
Octave-dev mailing list
Octave-dev@...
https://lists.sourceforge.net/lists/listinfo/octave-dev

From forum: octave-dev

Re: Release system / web page generation

2009-11-25T05:35:46Z

ons, 25 11 2009 kl. 13:40 +0100, skrev Carlo de Falco:
> >> But people can't upload HTML this way, right?
> >
> > I actually thought this would be possible, I'm going to check on SF
> > documentation
>
> Actually, https://sourceforge.net/apps/trac/sourceforge/wiki/Project%20web
> says it can be done: "Project web content may be uploaded using our
> File management service or be managed directly using our Shell service."

I must admit that I quickly give up on the SF documentation, so I just
might be missing the obvious. To me, it seems like files can be copied
to the server using

Rsync over SSH
SFTP
SCP

I don't see any web tools for uploading html files to the web server,
but perhaps I'm just not looking hard enough?

Søren

------------------------------------------------------------------------------
Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day
trial. Simplify your report design, integration and deployment - and focus on
what you do best, core application coding. Discover what's new with
Crystal Reports now. http://p.sf.net/sfu/bobj-july
_______________________________________________
Octave-dev mailing list
Octave-dev@...
https://lists.sourceforge.net/lists/listinfo/octave-dev

From forum: octave-dev

Re: Release system / web page generation

2009-11-25T04:40:04Z

On 25 Nov 2009, at 08:58, Carlo de Falco wrote:

>
> On 24 Nov 2009, at 11:48, Søren Hauberg wrote:
>
>> man, 23 11 2009 kl. 10:12 +0100, skrev Carlo de Falco:
>>>> Yeah, I've been thinking a bit about this. I guess we would like to
>>>> have
>>>> a whole bunch of maintainers, meaning we would have many people
>>>> with
>>>> 'admin' status. I don't really have security concerns as I trust
>>>> people
>>>> on this list, but the chance of somebody doing something wrong by
>>>> accident is just very high. So, I don't think we should take this
>>>> route.
>>>
>>> actually I feel that having many different people manually messing
>>> around
>>> with the web pages might be even more error prone, at least the SF
>>> GUI
>>> tools
>>> restrict the kind of operations that one can do on the server and
>>> they
>>> seem
>>> to take care automatically of setting file permissions.
>>
>> But people can't upload HTML this way, right?
>
> I actually thought this would be possible, I'm going to check on SF
> documentation

Actually, https://sourceforge.net/apps/trac/sourceforge/wiki/Project%20web
says it can be done: "Project web content may be uploaded using our
File management service or be managed directly using our Shell service."

> c.

------------------------------------------------------------------------------
Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day
trial. Simplify your report design, integration and deployment - and focus on
what you do best, core application coding. Discover what's new with
Crystal Reports now. http://p.sf.net/sfu/bobj-july
_______________________________________________
Octave-dev mailing list
Octave-dev@...
https://lists.sourceforge.net/lists/listinfo/octave-dev

From forum: octave-dev

Re: Release system / web page generation

2009-11-25T01:02:47Z

On 25 Nov 2009, at 09:52, Søren Hauberg wrote:

> Hmm, I guess this is possible. This requires that we can determine
> which
> packages have changed most recently. Do you think it is enough to
> simple
> check the dates of the directories containing the html for the
> packages?

more simply, we can have generate_package_html add a file named

announce_<date>.html

containing the announcement to the package directory and the logic in
the front page would
sort the announcements by descending date before displaying them.

> Søren
c.

------------------------------------------------------------------------------
Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day
trial. Simplify your report design, integration and deployment - and focus on
what you do best, core application coding. Discover what's new with
Crystal Reports now. http://p.sf.net/sfu/bobj-july
_______________________________________________
Octave-dev mailing list
Octave-dev@...
https://lists.sourceforge.net/lists/listinfo/octave-dev

From forum: octave-dev

Re: Release system / web page generation

2009-11-25T00:52:18Z

ons, 25 11 2009 kl. 09:03 +0100, skrev Carlo de Falco:
> > I can only think of the
> > front page, as it might be nice to announce new package releases
> > there.
>
> we could also think about adding an "announce" input field to
> generate_package_html
> and automate the front page much like the package list page...

Hmm, I guess this is possible. This requires that we can determine which
packages have changed most recently. Do you think it is enough to simple
check the dates of the directories containing the html for the packages?

Søren

------------------------------------------------------------------------------
Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day
trial. Simplify your report design, integration and deployment - and focus on
what you do best, core application coding. Discover what's new with
Crystal Reports now. http://p.sf.net/sfu/bobj-july
_______________________________________________
Octave-dev mailing list
Octave-dev@...
https://lists.sourceforge.net/lists/listinfo/octave-dev

From forum: octave-dev

Re: Gnuplot bug - mingw-octave

2009-11-25T00:31:14Z

I get this too (on Linux Gentoo x86_64, Octave 3.2.3) for large images. I can often trigger this by pressing CTRL-C when Octave is sending data (I guess) to Gnuplot.

/Fredrik

On Tue, Nov 24, 2009 at 10:16 PM, Bernd Heinze <berndheinze69@...> wrote:

Hello,

I'm using the latest octave-3.2.3-2-mingw on Windows XP in order to manipulate and display data taken from a digital oscilloscope. The matrices used are pretty large, mostly about 2 x 1...5 million.

I found out, that if the amount of data is getting too large gnuplot is starting to act strange. After executing plot(x,y); gnuplot starts to pop up its (Windows) helpfile dialog with strange (non-latin) characters in the search field. This repeats several time, afterwards the plot remains empty. However in the working directory several zero-sized files were created, also with strange characters.

Does anybody know the reason for this, or any workaround? My humble guess is some bug in the connection between octave and gnuplot.

Best regards
Bernd

__________________________________________________
Do You Yahoo!?
Sie sind Spam leid? Yahoo! Mail verfügt über einen herausragenden Schutz gegen Massenmails.
http://mail.yahoo.com

------------------------------------------------------------------------------
Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day
trial. Simplify your report design, integration and deployment - and focus on
what you do best, core application coding. Discover what's new with
Crystal Reports now. http://p.sf.net/sfu/bobj-july
_______________________________________________
Octave-dev mailing list
Octave-dev@...
https://lists.sourceforge.net/lists/listinfo/octave-dev

From forum: octave-dev

Re: Release system / web page generation

2009-11-25T00:03:34Z

On 24 Nov 2009, at 11:46, Søren Hauberg wrote:

> man, 23 11 2009 kl. 10:20 +0100, skrev Carlo de Falco:
>> Here I referred to the static part of the web pages, currently to
>> modify them
>> one connects to the SF server and modifies or replaces them in place
>> with a text editor.
>> I see many ways things can go wrong and I'm not sure whether any
>> backups are being made to
>> fix things in case of errors. A wiki usually maintains full history
>> of
>> modifications to the text
>> so roll-back in case of errors is possible much like with a VCS
>> repository.
>> Furthermore, windows users are not likely to be familiar witha a
>> procedure that requires
>> * ssh via a terminal emulator
>> * edit a file in emacs/vi
>> * scp/sftp to transfer local files to the server
>> but I am sure they would be able to edit a wiki page.
>
> Okay, I understand. I don't think package maintainers will be required
> to do much (any?) editing of the static pages.

most likely not, it has happened a few times lately, but probably once
the new version of the site will be in production it won't be needed
any more.

> I can only think of the
> front page, as it might be nice to announce new package releases
> there.

we could also think about adding an "announce" input field to
generate_package_html
and automate the front page much like the package list page...

> If we go for the strategy of having a few selected people upload the
> package+html, then I guess the same selected people could also edit
> the
> front page.

again this is probably the simplest route

> Søren
c.
------------------------------------------------------------------------------
Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day
trial. Simplify your report design, integration and deployment - and focus on
what you do best, core application coding. Discover what's new with
Crystal Reports now. http://p.sf.net/sfu/bobj-july
_______________________________________________
Octave-dev mailing list
Octave-dev@...
https://lists.sourceforge.net/lists/listinfo/octave-dev

From forum: octave-dev

Re: Release system / web page generation

2009-11-24T23:58:29Z

On 24 Nov 2009, at 11:48, Søren Hauberg wrote:

> man, 23 11 2009 kl. 10:12 +0100, skrev Carlo de Falco:
>>> Yeah, I've been thinking a bit about this. I guess we would like to
>>> have
>>> a whole bunch of maintainers, meaning we would have many people with
>>> 'admin' status. I don't really have security concerns as I trust
>>> people
>>> on this list, but the chance of somebody doing something wrong by
>>> accident is just very high. So, I don't think we should take this
>>> route.
>>
>> actually I feel that having many different people manually messing
>> around
>> with the web pages might be even more error prone, at least the SF
>> GUI
>> tools
>> restrict the kind of operations that one can do on the server and
>> they
>> seem
>> to take care automatically of setting file permissions.
>
> But people can't upload HTML this way, right?

I actually thought this would be possible, I'm going to check on SF
documentation
c.
------------------------------------------------------------------------------
Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day
trial. Simplify your report design, integration and deployment - and focus on
what you do best, core application coding. Discover what's new with
Crystal Reports now. http://p.sf.net/sfu/bobj-july
_______________________________________________
Octave-dev mailing list
Octave-dev@...
https://lists.sourceforge.net/lists/listinfo/octave-dev

From forum: octave-dev

Gnuplot bug - mingw-octave

2009-11-24T13:16:54Z

Hello,

I'm using the latest octave-3.2.3-2-mingw on Windows XP in order to manipulate and display data taken from a digital oscilloscope. The matrices used are pretty large, mostly about 2 x 1...5 million.

I found out, that if the amount of data is getting too large gnuplot is starting to act strange. After executing plot(x,y); gnuplot starts to pop up its (Windows) helpfile dialog with strange (non-latin) characters in the search field. This repeats several time, afterwards the plot remains empty. However in the working directory several zero-sized files were created, also with strange characters.

Does anybody know the reason for this, or any workaround? My humble guess is some bug in the connection between octave and gnuplot.

Best regards
Bernd

__________________________________________________
Do You Yahoo!?
Sie sind Spam leid? Yahoo! Mail verfügt über einen herausragenden Schutz gegen Massenmails.
http://mail.yahoo.com

------------------------------------------------------------------------------
Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day
trial. Simplify your report design, integration and deployment - and focus on
what you do best, core application coding. Discover what's new with
Crystal Reports now. http://p.sf.net/sfu/bobj-july
_______________________________________________
Octave-dev mailing list
Octave-dev@...
https://lists.sourceforge.net/lists/listinfo/octave-dev

From forum: octave-dev

Re: Monte Carlo example for new MPI bindings

2009-11-24T06:10:34Z

On Tue, Nov 24, 2009 at 2:41 PM, Riccardo Corradini <riccardocorradini@...> wrote:

Yes, sure.
I would propose this solution

#include "mpi.h"
#include <octave/oct.h>
#include <octave/error.h>

class mpicomm : public octave_base_value
{
public:
// Constructor
mpicomm (const std::string _name = "")
    : octave_base_value (), name (_name)
    {
    }

void print (std::ostream& os, bool pr_as_read_syntax = false) const
    {
      os << name << std::endl;
    }
bool is_defined (void) const { return true; }

private:
const std::string name;

DECLARE_OCTAVE_ALLOCATOR
DECLARE_OV_TYPEID_FUNCTIONS_AND_DATA
};

DEFINE_OCTAVE_ALLOCATOR (mpicomm);
DEFINE_OV_TYPEID_FUNCTIONS_AND_DATA (mpicomm, "mpicomm", "mpicomm");

static bool mpicomm_type_loaded = false;

DEFUN_DLD(mpicomm, args, ,"")
{
if (!mpicomm_type_loaded)
    {
      mpicomm::register_type ();
      mpicomm_type_loaded = true;
      mlock ();
    }

octave_value retval;
if (args.length () != 1 || !args (0).is_string ())
    {
      error ("mpicomm: first argument must be a string");
      return retval;
    }
const std::string name = args (0).string_value ();
      if( name == "MPI_COMM_WORLD" )
      {
      MPI_Comm Mine = MPI_COMM_WORLD;
      }
      else if( name == "MPI_COMM_SELF" )
      {
      MPI_Comm Mine = MPI_COMM_SELF;
      }
      else if( name == "MPI_COMM_NULL" )
      {
    MPI_Comm Mine = MPI_COMM_NULL;
      }
      else
      {
    // Error about invalid type
      error ("MPI: not a valid communicator");
      }


retval = new mpicomm (name);

return retval;
}

At least we get rid of raw pointer and abuse of reinterpret cast like here
#if __ia64__ || __x86_64__                // IA-64 / EM64T/AMD64
#define OpenMPI_OctPtrTyp        octave_uint64
#define OpenMPI_OctIntFcn        uint64_scalar_value
#define OpenMPI_64BIT        1
#define OpenMPI_ST_offset        2

#else    // #elif __i386__                // IA-32
#define OpenMPI_OctPtrTyp        octave_uint32
#define OpenMPI_OctIntFcn        uint32_scalar_value
#define OpenMPI_64BIT        0
#define OpenMPI_ST_offset        1
#endif

#define OpenMPI_intcast(cptr)            OpenMPI_OctPtrTyp        (\
                reinterpret_cast <unsigned long> ( cptr ) )

#include "mpi.h"

#undef       MPI_COMM_WORLD
#define NAME MPI_COMM_WORLD
DEFUN_DLD ( MPI_COMM_WORLD, args, nargout,
"MPI_COMM_WORLD         Openmpi_ext constant for corresponding communicator object\n\
\n\
Current value (openmpi-1.3.3) is:\n\
\n\
typedef struct ompi_communicator_t *MPI_Comm;\n\
#define MPI_COMM_WORLD OMPI_PREDEFINED_GLOBAL( MPI_Comm, ompi_mpi_comm_world)\n\
\n\
SEE ALSO: comms\
")
{

    octave_value result;

//    var = MPI_COMM_WORLD
//    Definition stolen from mpi.h
// #define MPI_COMM_WORLD OMPI_PREDEFINED_GLOBAL( MPI_Comm, ompi_mpi_comm_world)

    result = (OpenMPI_intcast( NAME )    // defined -> expanded
}

that JWE would definitely rule out.
We could consider the first as a temporary solution.
What do you think Jarolaslav?

Respectfully
Riccardo

Yes, something like that would be nice. I don't think you need to pass the strings as arguments, functions will do.
There doesn't need to be an extra .oct file per constant, we should use autoloads. I was up to code it this week, but feel free to start ahead.

--
RNDr. Jaroslav Hajek
computing expert & GNU Octave developer
Aeronautical Research and Test Institute (VZLU)
Prague, Czech Republic
url: www.highegg.matfyz.cz

------------------------------------------------------------------------------
Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day
trial. Simplify your report design, integration and deployment - and focus on
what you do best, core application coding. Discover what's new with
Crystal Reports now. http://p.sf.net/sfu/bobj-july
_______________________________________________
Octave-dev mailing list
Octave-dev@...
https://lists.sourceforge.net/lists/listinfo/octave-dev

From forum: octave-dev

Re: Monte Carlo example for new MPI bindings

2009-11-24T05:41:46Z

Yes, sure.
I would propose this solution

#include "mpi.h"
#include <octave/oct.h>
#include <octave/error.h>

class mpicomm : public octave_base_value
{
public:
// Constructor
mpicomm (const std::string _name = "")
    : octave_base_value (), name (_name)
    {
    }

void print (std::ostream& os, bool pr_as_read_syntax = false) const
    {
      os << name << std::endl;
    }
bool is_defined (void) const { return true; }

private:
const std::string name;

DECLARE_OCTAVE_ALLOCATOR
DECLARE_OV_TYPEID_FUNCTIONS_AND_DATA
};

DEFINE_OCTAVE_ALLOCATOR (mpicomm);
DEFINE_OV_TYPEID_FUNCTIONS_AND_DATA (mpicomm, "mpicomm", "mpicomm");

static bool mpicomm_type_loaded = false;

DEFUN_DLD(mpicomm, args, ,"")
{
if (!mpicomm_type_loaded)
    {
      mpicomm::register_type ();
      mpicomm_type_loaded = true;
      mlock ();
    }

octave_value retval;
if (args.length () != 1 || !args (0).is_string ())
    {
      error ("mpicomm: first argument must be a string");
      return retval;
    }
const std::string name = args (0).string_value ();
      if( name == "MPI_COMM_WORLD" )
      {
      MPI_Comm Mine = MPI_COMM_WORLD;
      }
      else if( name == "MPI_COMM_SELF" )
      {
      MPI_Comm Mine = MPI_COMM_SELF;
      }
      else if( name == "MPI_COMM_NULL" )
      {
    MPI_Comm Mine = MPI_COMM_NULL;
      }
      else
      {
    // Error about invalid type
      error ("MPI: not a valid communicator");
      }


retval = new mpicomm (name);

return retval;
}

At least we get rid of raw pointer and abuse of reinterpret cast like here
#if __ia64__ || __x86_64__                // IA-64 / EM64T/AMD64
#define OpenMPI_OctPtrTyp        octave_uint64
#define OpenMPI_OctIntFcn        uint64_scalar_value
#define OpenMPI_64BIT        1
#define OpenMPI_ST_offset        2

#else    // #elif __i386__                // IA-32
#define OpenMPI_OctPtrTyp        octave_uint32
#define OpenMPI_OctIntFcn        uint32_scalar_value
#define OpenMPI_64BIT        0
#define OpenMPI_ST_offset        1
#endif

#define OpenMPI_intcast(cptr)            OpenMPI_OctPtrTyp        (\
                reinterpret_cast <unsigned long> ( cptr ) )

#include "mpi.h"

#undef       MPI_COMM_WORLD
#define NAME MPI_COMM_WORLD
DEFUN_DLD ( MPI_COMM_WORLD, args, nargout,
"MPI_COMM_WORLD         Openmpi_ext constant for corresponding communicator object\n\
\n\
Current value (openmpi-1.3.3) is:\n\
\n\
typedef struct ompi_communicator_t *MPI_Comm;\n\
#define MPI_COMM_WORLD OMPI_PREDEFINED_GLOBAL( MPI_Comm, ompi_mpi_comm_world)\n\
\n\
SEE ALSO: comms\
")
{

    octave_value result;

//    var = MPI_COMM_WORLD
//    Definition stolen from mpi.h
// #define MPI_COMM_WORLD OMPI_PREDEFINED_GLOBAL( MPI_Comm, ompi_mpi_comm_world)

    result = (OpenMPI_intcast( NAME )    // defined -> expanded
}

that JWE would definitely rule out.
We could consider the first as a temporary solution.
What do you think Jarolaslav?

Respectfully
Riccardo

--- Mar 24/11/09, Michael Creel <michael.creel@...> ha scritto:

Da: Michael Creel <michael.creel@...>
Oggetto: Re: Monte Carlo example for new MPI bindings
A: "Jaroslav Hajek" <highegg@...>
Cc: "Riccardo Corradini" <riccardocorradini@...>, "octave-forge list" <octave-dev@...>
Data: Martedì 24 novembre 2009, 10:36

On Mon, Nov 23, 2009 at 2:31 PM, Jaroslav Hajek <highegg@...> wrote:

Hi Riccardo and Michael,

I have already implemented basic support for the communicators in the SVN tree.
The relevant functions now accept comm as their last argument, defaulting to MPI_COMM_WORLD if not provided.
MPI_COMM_WORLD is a new function.

Note 1: The implementation imposes the requirement that MPI_Comm is a typedef for signed 32-bit or 64-bit integer (int, long, or long long) which is true in all major MPI implementations with the notable exception of OpenMPI. OpenMPI uses raw pointers, which make things a bit difficult. One option is to simply typecast the pointers to integers; but I feel a bit uneasy about working with raw pointers in Octave, so I'm thinking about a solution providing some validation as well.
I would, however, like to avoid creating a user type for each of MPI's types.

Note 2:
There are no functions for creating new communicators yet.

Hi all,
A check out this morning is not compiling for me:

michael@yosemite:~/Desktop/openmpi_ext/src$ make all
mkoctfile -c -I/usr/lib/openmpi/include -I/usr/lib/openmpi/include/openmpi comm-util.cc
comm-util.cc: In function ‘ompi_communicator_t* do_get(const octave_value&) [with T = int]’:
comm-util.cc:39: error: invalid conversion from ‘int’ to ‘ompi_communicator_t*’
comm-util.cc: At global scope:
comm-util.cc:34: error: explicit template specialization cannot have a storage class
comm-util.cc: In function ‘ompi_communicator_t* do_get(const octave_value&) [with T = long int]’:
comm-util.cc:51: error: invalid conversion from ‘long int’ to ‘ompi_communicator_t*’
comm-util.cc: At global scope:
comm-util.cc:46: error: explicit template specialization cannot have a storage class
comm-util.cc:72: error: explicit template specialization cannot have a storage class
comm-util.cc:78: error: explicit template specialization cannot have a storage class
make: *** [comm-util.o] Error 1
michael@yosemite:~/Desktop/openmpi_ext/src$

Is that expected at the moment?

Cheers, Michael

From forum: octave-dev

Re: Release system / web page generation

2009-11-24T02:48:07Z

man, 23 11 2009 kl. 10:12 +0100, skrev Carlo de Falco:

> > Yeah, I've been thinking a bit about this. I guess we would like to
> > have
> > a whole bunch of maintainers, meaning we would have many people with
> > 'admin' status. I don't really have security concerns as I trust
> > people
> > on this list, but the chance of somebody doing something wrong by
> > accident is just very high. So, I don't think we should take this
> > route.
>
> actually I feel that having many different people manually messing
> around
> with the web pages might be even more error prone, at least the SF GUI
> tools
> restrict the kind of operations that one can do on the server and they
> seem
> to take care automatically of setting file permissions.

But people can't upload HTML this way, right?

> > One option would be to have a team of administrators. When a
> > maintainer
> > wants to release a package he/she would then contact this team (on
> > this
> > list perhaps?) and tell them to upload the package and the html. The
> > downside to this approach is that a release depends on this team, but
> > that's why I would like to have more than one person on this team. The
> > current system, where things depend on me, is just terrible as I don't
> > have that much time.
> >
> > Would such a setup be a good idea?
>
> This sounds like the solution requiring the least maintainance and
> development work.
> I would say, as most packages are distributed as source so the total
> file size of package
> tarball + function reference is small, let's set up a "package
> release" forum on the SF website,
> let each package maintainer submit their package to the forum for
> release and a team of admins
> listening to the forum would take care of uploading the files.
> Using the GUI the file upload procedure does not require much time so
> I could volunteer to be one of such admins.

I think this sounds reasonable. Does anybody know how to create a new
forum on SF? If not, then I'll look into it in the not-too-distant
future...

Søren

------------------------------------------------------------------------------
Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day
trial. Simplify your report design, integration and deployment - and focus on
what you do best, core application coding. Discover what's new with
Crystal Reports now. http://p.sf.net/sfu/bobj-july
_______________________________________________
Octave-dev mailing list
Octave-dev@...
https://lists.sourceforge.net/lists/listinfo/octave-dev

From forum: octave-dev

Re: Release system / web page generation

2009-11-24T02:46:14Z

man, 23 11 2009 kl. 10:20 +0100, skrev Carlo de Falco:

> Here I referred to the static part of the web pages, currently to
> modify them
> one connects to the SF server and modifies or replaces them in place
> with a text editor.
> I see many ways things can go wrong and I'm not sure whether any
> backups are being made to
> fix things in case of errors. A wiki usually maintains full history of
> modifications to the text
> so roll-back in case of errors is possible much like with a VCS
> repository.
> Furthermore, windows users are not likely to be familiar witha a
> procedure that requires
> * ssh via a terminal emulator
> * edit a file in emacs/vi
> * scp/sftp to transfer local files to the server
> but I am sure they would be able to edit a wiki page.

Okay, I understand. I don't think package maintainers will be required
to do much (any?) editing of the static pages. I can only think of the
front page, as it might be nice to announce new package releases there.
If we go for the strategy of having a few selected people upload the
package+html, then I guess the same selected people could also edit the
front page.

Søren

------------------------------------------------------------------------------
Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day
trial. Simplify your report design, integration and deployment - and focus on
what you do best, core application coding. Discover what's new with
Crystal Reports now. http://p.sf.net/sfu/bobj-july
_______________________________________________
Octave-dev mailing list
Octave-dev@...
https://lists.sourceforge.net/lists/listinfo/octave-dev

From forum: octave-dev

Re: Monte Carlo example for new MPI bindings

2009-11-24T01:36:20Z

On Mon, Nov 23, 2009 at 2:31 PM, Jaroslav Hajek <highegg@...> wrote:

Hi Riccardo and Michael,

I have already implemented basic support for the communicators in the SVN tree.
The relevant functions now accept comm as their last argument, defaulting to MPI_COMM_WORLD if not provided.
MPI_COMM_WORLD is a new function.

Note 1: The implementation imposes the requirement that MPI_Comm is a typedef for signed 32-bit or 64-bit integer (int, long, or long long) which is true in all major MPI implementations with the notable exception of OpenMPI. OpenMPI uses raw pointers, which make things a bit difficult. One option is to simply typecast the pointers to integers; but I feel a bit uneasy about working with raw pointers in Octave, so I'm thinking about a solution providing some validation as well.
I would, however, like to avoid creating a user type for each of MPI's types.

Note 2:
There are no functions for creating new communicators yet.

Hi all,
A check out this morning is not compiling for me:

michael@yosemite:~/Desktop/openmpi_ext/src$ make all
mkoctfile -c -I/usr/lib/openmpi/include -I/usr/lib/openmpi/include/openmpi comm-util.cc
comm-util.cc: In function ‘ompi_communicator_t* do_get(const octave_value&) [with T = int]’:
comm-util.cc:39: error: invalid conversion from ‘int’ to ‘ompi_communicator_t*’
comm-util.cc: At global scope:
comm-util.cc:34: error: explicit template specialization cannot have a storage class
comm-util.cc: In function ‘ompi_communicator_t* do_get(const octave_value&) [with T = long int]’:
comm-util.cc:51: error: invalid conversion from ‘long int’ to ‘ompi_communicator_t*’
comm-util.cc: At global scope:
comm-util.cc:46: error: explicit template specialization cannot have a storage class
comm-util.cc:72: error: explicit template specialization cannot have a storage class
comm-util.cc:78: error: explicit template specialization cannot have a storage class
make: *** [comm-util.o] Error 1
michael@yosemite:~/Desktop/openmpi_ext/src$

Is that expected at the moment?

Cheers, Michael

------------------------------------------------------------------------------
Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day
trial. Simplify your report design, integration and deployment - and focus on
what you do best, core application coding. Discover what's new with
Crystal Reports now. http://p.sf.net/sfu/bobj-july
_______________________________________________
Octave-dev mailing list
Octave-dev@...
https://lists.sourceforge.net/lists/listinfo/octave-dev

From forum: octave-dev

Re: Monte Carlo example for new MPI bindings

2009-11-24T01:04:22Z

On Mon, Nov 23, 2009 at 10:26 AM, Riccardo Corradini <riccardocorradini@...> wrote:

Dear Michael,
I have posted MPI_Iprobe.cc and MPI_Probe.cc with info as last flag.
Please let me know if it helps in building other examples.
For the comunicator issue I am thinking on passing a simple string as alternative comunicator and if nothing is provided by the user then MPI_COMM_WORLD is the default value.
Thanks and bests
Riccardo

--- Sab 21/11/09, Michael Creel <michael.creel@...> ha scritto:

Da: Michael Creel <michael.creel@...>
Oggetto: Monte Carlo example for new MPI bindings
A: "Riccardo Corradini" <riccardocorradini@...>, "Jaroslav Hajek" <highegg@...>, "octave-forge list" <octave-dev@...>
Data: Sabato 21 novembre 2009, 14:14

Hello all,
Attached are two files, montecarlo.m and mc_example.m that implement parallel Monte Carlo simulation of a user defined function. The function just needs to return a row vector, and montecarlo.m will stack the replications into a matrix. This implementation is basic and illustrative. At http://pareto.uab.es/mcreel/Econometrics/MyOctaveFiles/Econometrics/MonteCarlo/ and http://pareto.uab.es/mcreel/Econometrics/Examples/Parallel/montecarlo/ there is a more sophisticated version that uses MPITB.

The basic version is susceptible to the crash of a node, and will not balance the load between nodes of different speed. To implement that, MPI_Iprobe would be needed.

To use it, doe something like

michael@tuolumne:~/Desktop/openmpi_ext/src$ mpirun -np 2 octave -q --eval mc_example
Sending block to 0
I have sent 4
I have sent 15
I have sent 2
   1.00000   0.96457   1.23784
   1.00000   1.20805   1.19109
   1.00000   1.21014   1.15861
   1.00000   0.93194   0.83522
   1.00000   0.73157   1.36160
   0.00000   1.07361   0.96141
   0.00000   1.32715   1.06895
   0.00000   0.73529   0.94909
   0.00000   1.24376   1.02082
   0.00000   1.05802   1.20701
michael@tuolumne:~/Desktop/openmpi_ext/src$

In that, the first column is the MPI rank of the node that generated the rest of the row, which is the output of the function that is being simulated..

Cheers, Michael

Great. I will adapt my "professional" version of montecarlo.m to use the new wrappers, and when it works I'll check it in to SVN, to serve as an example. Then I will do a little benchmarking to compare to MPITB. Do you know if the wrappers work with Octave 3.0.X?
Cheers, Michael

From forum: octave-dev

Re: a few more questions about MPI

2009-11-24T01:01:50Z

Hi all,

On Mon, Nov 23, 2009 at 3:09 PM, Jaroslav Hajek <highegg@...> wrote:

hi guys,

I have a few more comments/questions regarding the MPI package.

1. Will just OpenMPI be targeted or shall we adhere to the MPI standard to support various implementations?
For instance, the latest probe functions access some non-standard fields of MPI_State and hence require patching to work under Intel MPI.

I think that using the standard would be best, if possible, even if there is a reasonably small performance cost (I don't know if or why that might happen). Open MPI is only one of a number of implementations. MPICH has a lot of users, and LAM/MPI still does, too, I would guess. Intel MPI probably does too. Who knows when another will appear.

Michael

From forum: octave-dev

Re: fzero_compat calls fsolve_options

2009-11-23T22:57:25Z

On Mon, Nov 23, 2009 at 8:20 PM, Thomas Weber <thomas.weber.mail@...> wrote:

Hi,

fzero_compat.m calls fsolve_options, which was removed from Octave some
time ago. Is there much need for fzero_compat? Otherwise, I'd just
remove it.

Thomas

I think the new fzero works relatively well, so I think we can remove it. I just did it.

--
RNDr. Jaroslav Hajek
computing expert & GNU Octave developer
Aeronautical Research and Test Institute (VZLU)
Prague, Czech Republic
url: www.highegg.matfyz.cz

------------------------------------------------------------------------------
Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day
trial. Simplify your report design, integration and deployment - and focus on
what you do best, core application coding. Discover what's new with
Crystal Reports now. http://p.sf.net/sfu/bobj-july
_______________________________________________
Octave-dev mailing list
Octave-dev@...
https://lists.sourceforge.net/lists/listinfo/octave-dev

From forum: octave-dev

fzero_compat calls fsolve_options

2009-11-23T11:20:23Z

Hi,

fzero_compat.m calls fsolve_options, which was removed from Octave some
time ago. Is there much need for fzero_compat? Otherwise, I'd just
remove it.

Thomas

------------------------------------------------------------------------------
Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day
trial. Simplify your report design, integration and deployment - and focus on
what you do best, core application coding. Discover what's new with
Crystal Reports now. http://p.sf.net/sfu/bobj-july
_______________________________________________
Octave-dev mailing list
Octave-dev@...
https://lists.sourceforge.net/lists/listinfo/octave-dev

From forum: octave-dev

I: R: a few more questions about MPI

2009-11-23T07:09:01Z

--- Lun 23/11/09, Riccardo Corradini <riccardocorradini@...> ha scritto:

Da: Riccardo Corradini <riccardocorradini@...>
Oggetto: R: a few more questions about MPI
A: "Jaroslav Hajek" <highegg@...>
Data: Lunedì 23 novembre 2009, 16:08

Dear Jaroslav,
for question number 3 I just wanted to show how everything could be easy with MPI_Derived Datatypes (they are not so scary for beginners) because you pass a lot of pieces of information with just one call.
As a matter of the fact I continue to send vectors, so I handle contiguous blocks of memory.
At the beginning I had the idea of sending a whole octave_value all at the once so I thought
MPI_Derived Datatypes has to be the best tool.
Please have also have a look at this in
http://www.osl.iu.edu/research/oompi/
There is a nice example with MPI derived datatypes with the classes.
4) No problems with octave standard of documentation, but I have to learn form you and others.
Also have a look at this
http://en.wikipedia.org/wiki/Message_Passing_Interface#Derived_Datatypes
They don't say is better. I just preferred it.
I think more experiments on this should be carried out.
1) No problems renaming the package but I would like to run it on open-mpi because it seems very spread and I like (it sounds simple).
2) I don't know enough well octave but solution C will be very interesting for me for learn more about octave. On the other hand I guess you would like to use solution B.

Respectfully
Riccardo Corradini

--- Lun 23/11/09, Jaroslav Hajek <highegg@...> ha scritto:

Da: Jaroslav Hajek <highegg@...>
Oggetto: a few more questions about MPI
A: "octave-forge list" <octave-dev@...>
Cc: "Riccardo Corradini" <riccardocorradini@...>, "Michael Creel" <michael.creel@...>
Data: Lunedì 23 novembre 2009, 15:09

hi guys,

I have a few more comments/questions regarding the MPI package.

1. Will just OpenMPI be targeted or shall we adhere to the MPI standard to support various implementations?
For instance, the latest probe functions access some non-standard fields of MPI_State and hence require patching to work under Intel MPI.
1a. The answer to 1 is MPI Standard, what about renaming the package to not suggest it is OpenMPI specific?
2. My current aim for communicators is just tip of the iceberg; there are a number of MPI types that may need to be wrapped. In the future.
I see three different approaches here:
A: reinterpret-based
reinterpret the MPI type as some of Octave's types (an integer, or character array) and vice versa on the function boundary
B: handle-based
store global maps of objects as created by MPI and return integer handles.
C: object-based
derive new Octave objects for the MPI types.

A is completely unsafe, so I think we should avoid that.
B has the advantage that various manipulation of the handles (such as storing in arrays or structs) would be automatically supported (just like the graphics handles).
C is the most OO approach, however, there are currently some gotchas in defining new types from oct files; further, Octave is not really good when a lot of new types are dynamically registered.

3. this is a question for Riccardo: what are the advantages of using MPI_Datatype for sending, say, real Matrices, over just providing the "count" argument to MPI_Send?
4. are you OK with using Octave style coding and documentation?

--
RNDr. Jaroslav Hajek
computing expert & GNU Octave developer
Aeronautical Research and Test Institute (VZLU)
Prague, Czech Republic
url: www.highegg.matfyz.cz

From forum: octave-dev

a few more questions about MPI

2009-11-23T06:09:22Z

hi guys,

I have a few more comments/questions regarding the MPI package.

1. Will just OpenMPI be targeted or shall we adhere to the MPI standard to support various implementations?
For instance, the latest probe functions access some non-standard fields of MPI_State and hence require patching to work under Intel MPI.
1a. The answer to 1 is MPI Standard, what about renaming the package to not suggest it is OpenMPI specific?
2. My current aim for communicators is just tip of the iceberg; there are a number of MPI types that may need to be wrapped. In the future.
I see three different approaches here:
A: reinterpret-based
reinterpret the MPI type as some of Octave's types (an integer, or character array) and vice versa on the function boundary
B: handle-based
store global maps of objects as created by MPI and return integer handles.
C: object-based
derive new Octave objects for the MPI types.

A is completely unsafe, so I think we should avoid that.
B has the advantage that various manipulation of the handles (such as storing in arrays or structs) would be automatically supported (just like the graphics handles).
C is the most OO approach, however, there are currently some gotchas in defining new types from oct files; further, Octave is not really good when a lot of new types are dynamically registered.

3. this is a question for Riccardo: what are the advantages of using MPI_Datatype for sending, say, real Matrices, over just providing the "count" argument to MPI_Send?
4. are you OK with using Octave style coding and documentation?

--
RNDr. Jaroslav Hajek
computing expert & GNU Octave developer
Aeronautical Research and Test Institute (VZLU)
Prague, Czech Republic
url: www.highegg.matfyz.cz

------------------------------------------------------------------------------
Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day
trial. Simplify your report design, integration and deployment - and focus on
what you do best, core application coding. Discover what's new with
Crystal Reports now. http://p.sf.net/sfu/bobj-july
_______________________________________________
Octave-dev mailing list
Octave-dev@...
https://lists.sourceforge.net/lists/listinfo/octave-dev

From forum: octave-dev

Re: Monte Carlo example for new MPI bindings

2009-11-23T05:31:23Z

On Mon, Nov 23, 2009 at 10:26 AM, Riccardo Corradini <riccardocorradini@...> wrote:

Dear Michael,
I have posted MPI_Iprobe.cc and MPI_Probe.cc with info as last flag.
Please let me know if it helps in building other examples.
For the comunicator issue I am thinking on passing a simple string as alternative comunicator and if nothing is provided by the user then MPI_COMM_WORLD is the default value.
Thanks and bests
Riccardo

Hi Riccardo and Michael,

I have already implemented basic support for the communicators in the SVN tree.
The relevant functions now accept comm as their last argument, defaulting to MPI_COMM_WORLD if not provided.
MPI_COMM_WORLD is a new function.

Note 1: The implementation imposes the requirement that MPI_Comm is a typedef for signed 32-bit or 64-bit integer (int, long, or long long) which is true in all major MPI implementations with the notable exception of OpenMPI. OpenMPI uses raw pointers, which make things a bit difficult. One option is to simply typecast the pointers to integers; but I feel a bit uneasy about working with raw pointers in Octave, so I'm thinking about a solution providing some validation as well.
I would, however, like to avoid creating a user type for each of MPI's types.

Note 2:
There are no functions for creating new communicators yet.

--
RNDr. Jaroslav Hajek
computing expert & GNU Octave developer
Aeronautical Research and Test Institute (VZLU)
Prague, Czech Republic
url: www.highegg.matfyz.cz

------------------------------------------------------------------------------
Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day
trial. Simplify your report design, integration and deployment - and focus on
what you do best, core application coding. Discover what's new with
Crystal Reports now. http://p.sf.net/sfu/bobj-july
_______________________________________________
Octave-dev mailing list
Octave-dev@...
https://lists.sourceforge.net/lists/listinfo/octave-dev

From forum: octave-dev

R: Monte Carlo example for new MPI bindings

2009-11-23T01:26:45Z

Dear Michael,
I have posted MPI_Iprobe.cc and MPI_Probe.cc with info as last flag.
Please let me know if it helps in building other examples.
For the comunicator issue I am thinking on passing a simple string as alternative comunicator and if nothing is provided by the user then MPI_COMM_WORLD is the default value.
Thanks and bests
Riccardo

--- Sab 21/11/09, Michael Creel <michael.creel@...> ha scritto:

Da: Michael Creel <michael.creel@...>
Oggetto: Monte Carlo example for new MPI bindings
A: "Riccardo Corradini" <riccardocorradini@...>, "Jaroslav Hajek" <highegg@...>, "octave-forge list" <octave-dev@...>
Data: Sabato 21 novembre 2009, 14:14

Hello all,
Attached are two files, montecarlo.m and mc_example.m that implement parallel Monte Carlo simulation of a user defined function. The function just needs to return a row vector, and montecarlo.m will stack the replications into a matrix. This implementation is basic and illustrative. At http://pareto.uab.es/mcreel/Econometrics/MyOctaveFiles/Econometrics/MonteCarlo/ and http://pareto.uab.es/mcreel/Econometrics/Examples/Parallel/montecarlo/ there is a more sophisticated version that uses MPITB.

The basic version is susceptible to the crash of a node, and will not balance the load between nodes of different speed. To implement that, MPI_Iprobe would be needed.

To use it, doe something like

michael@tuolumne:~/Desktop/openmpi_ext/src$ mpirun -np 2 octave -q --eval mc_example
Sending block to 0
I have sent 4
I have sent 15
I have sent 2
   1.00000   0.96457   1.23784
   1.00000   1.20805   1.19109
   1.00000   1.21014   1.15861
   1.00000   0.93194   0.83522
   1.00000   0.73157   1.36160
   0.00000   1.07361   0.96141
   0.00000   1.32715   1.06895
   0.00000   0.73529   0.94909
   0.00000   1.24376   1.02082
   0.00000   1.05802   1.20701
michael@tuolumne:~/Desktop/openmpi_ext/src$

In that, the first column is the MPI rank of the node that generated the rest of the row, which is the output of the function that is being simulated..

Cheers, Michael

From forum: octave-dev

Re: Release system / web page generation

2009-11-23T01:20:45Z

On 22 Nov 2009, at 22:10, Søren Hauberg wrote:

> Hi,
>
> Sorry about the late reply
>
> man, 16 11 2009 kl. 14:07 +0100, skrev Carlo de Falco:
>> On 16 Nov 2009, at 10:13, Søren Hauberg wrote:
>>
>>>
>>> This is in general a problem. Whenever I upload something I have to
>>> manually make sure that everybody has the rights to change my
>>> uploaded
>>> files. Sometimes I forget to do this. I'm not sure if we can find a
>>> work-around.
>>
>> the only solution I see is to wikify all the static pages but I am
>> afraid that would require quite some work...
>
> I don't see how that solves anything, but perhaps I don't understand.
> Could you elaborate?

Here I referred to the static part of the web pages, currently to
modify them
one connects to the SF server and modifies or replaces them in place
with a text editor.
I see many ways things can go wrong and I'm not sure whether any
backups are being made to
fix things in case of errors. A wiki usually maintains full history of
modifications to the text
so roll-back in case of errors is possible much like with a VCS
repository.
Furthermore, windows users are not likely to be familiar witha a
procedure that requires
* ssh via a terminal emulator
* edit a file in emacs/vi
* scp/sftp to transfer local files to the server
but I am sure they would be able to edit a wiki page.

> Søren
c.
------------------------------------------------------------------------------
Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day
trial. Simplify your report design, integration and deployment - and focus on
what you do best, core application coding. Discover what's new with
Crystal Reports now. http://p.sf.net/sfu/bobj-july
_______________________________________________
Octave-dev mailing list
Octave-dev@...
https://lists.sourceforge.net/lists/listinfo/octave-dev

From forum: octave-dev

Re: Release system / web page generation

2009-11-23T01:12:58Z

On 22 Nov 2009, at 22:20, Søren Hauberg wrote:

> man, 16 11 2009 kl. 14:13 +0100, skrev Carlo de Falco:
>> well, the current procedure looks a bit complicated and un-secure
>> (manually deleting the contents of the directories directly on the
>> web-
>> server could lead to messy mistakes).
>
> Agreed. I guess we could write some script that did all the necessary
> work, such that you would only have to run something like
>
> install_html_files my_package_html.tar.gz
>
> which would then make sure to setup the rights, etc.

This is a possible solution but it would take some work to develop,
test and maintain
the scripts and I feel we do not have enough manpower at this time...
Anyway, if someone wants to give this a try I would be willing to help
with the testing.

>> A great simplification could be obtained by upgrading all package
>> maintainers to "package admin" status so that they can take advantage
>> of the sorceforge web-based GUI tools, but I'm not sure we want to do
>> this...
>
> Yeah, I've been thinking a bit about this. I guess we would like to
> have
> a whole bunch of maintainers, meaning we would have many people with
> 'admin' status. I don't really have security concerns as I trust
> people
> on this list, but the chance of somebody doing something wrong by
> accident is just very high. So, I don't think we should take this
> route.

actually I feel that having many different people manually messing
around
with the web pages might be even more error prone, at least the SF GUI
tools
restrict the kind of operations that one can do on the server and they
seem
to take care automatically of setting file permissions.

> One option would be to have a team of administrators. When a
> maintainer
> wants to release a package he/she would then contact this team (on
> this
> list perhaps?) and tell them to upload the package and the html. The
> downside to this approach is that a release depends on this team, but
> that's why I would like to have more than one person on this team. The
> current system, where things depend on me, is just terrible as I don't
> have that much time.
>
> Would such a setup be a good idea?

This sounds like the solution requiring the least maintainance and
development work.
I would say, as most packages are distributed as source so the total
file size of package
tarball + function reference is small, let's set up a "package
release" forum on the SF website,
let each package maintainer submit their package to the forum for
release and a team of admins
listening to the forum would take care of uploading the files.
Using the GUI the file upload procedure does not require much time so
I could volunteer to be one of such admins.

> Søren
c.
------------------------------------------------------------------------------
Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day
trial. Simplify your report design, integration and deployment - and focus on
what you do best, core application coding. Discover what's new with
Crystal Reports now. http://p.sf.net/sfu/bobj-july
_______________________________________________
Octave-dev mailing list
Octave-dev@...
https://lists.sourceforge.net/lists/listinfo/octave-dev

From forum: octave-dev

Re: Octave-dev Digest, Vol 42, Issue 28

2009-11-22T18:57:40Z

On 11/22/2009 octave-dev-request@... wrote:
> Date: Sun, 22 Nov 2009 21:38:07 -0500 From: "Steven G. Johnson"
<snip>

>
> Well, it depends what you mean by "integrate". Certainly,
> licensewise I think there should be no problem (LGPL), and as I said
> NLopt already comes with an Octave plugin, so you could certainly
> just ship it as-is with the Octave-forge optimization package if you
> wanted to.
>
> It would be easy to modify functions like fminunc etcetera to call
> NLopt under the hood. However, NLopt seems to provide much more
> functionality than the interfaces of your optimization package
> currently provide, because it allows nonlinear constraints and global
> optimization, so I'm not sure you would want to "bury" it inside
> other functions exclusively.
>
> Also, I should say that the philosophy of NLopt seems to be a bit
> different from that of the optimization package in octave-forge. In
> the optimization package, you provide separate interfaces for
> different algorithms like BFGS and Nelder-Mead. NLopt, on the other
> hand, tries to provide a single interface for all the algorithms,
> with the algorithm specified as a parameter. (Of course, some
> algorithms are more limited than others; e.g. if you specify
> nonlinear constraints with an algorithm that only supports bound
> constraints, it will return an error code.)
>
> (NLopt is designe for general nonlinear optimization, so it is
> probably suboptimal for particualr convex cases that have specialized
> algorithms, such linear programming and quadratic programming [LPs
> and QPs].)
>
> Steven

I would suggest making your NLopt functions (and backend) available, as
they currently are, in the optimization package. Then when someone
installs the optim package, all of NLopt would also be available. This
would require you copying over the code to the octave-forge svn file
tree and adapting whatever install process NLopt currently uses to the
octave-forge installation process.

Jonathan

------------------------------------------------------------------------------
Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day
trial. Simplify your report design, integration and deployment - and focus on
what you do best, core application coding. Discover what's new with
Crystal Reports now. http://p.sf.net/sfu/bobj-july
_______________________________________________
Octave-dev mailing list
Octave-dev@...
https://lists.sourceforge.net/lists/listinfo/octave-dev

From forum: octave-dev

Re: Release system / web page generation

2009-11-22T18:49:56Z

On 11/22/2009 octave-dev-request@... wrote:

> Date: Sun, 22 Nov 2009 22:20:38 +0100
> From: S?ren Hauberg <soren@...>
> Subject: Re: [OctDev] Release system / web page generation
> To: Carlo de Falco <carlo.defalco@...>
> Cc: Octave Forge <octave-dev@...>
> Message-ID: <1258924838.2637.78.camel@hauberg-laptop>
> Content-Type: text/plain; charset="UTF-8"
>
>
> Yeah, I've been thinking a bit about this. I guess we would like to
> have
> a whole bunch of maintainers, meaning we would have many people with
> 'admin' status. I don't really have security concerns as I trust
> people
> on this list, but the chance of somebody doing something wrong by
> accident is just very high. So, I don't think we should take this
> route.
>
> One option would be to have a team of administrators. When a
> maintainer
> wants to release a package he/she would then contact this team (on
> this
> list perhaps?) and tell them to upload the package and the html. The
> downside to this approach is that a release depends on this team, but
> that's why I would like to have more than one person on this team. The
> current system, where things depend on me, is just terrible as I don't
> have that much time.
>
> Would such a setup be a good idea?
>

This sounds like a good idea to me.

Jonathan

------------------------------------------------------------------------------
Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day
trial. Simplify your report design, integration and deployment - and focus on
what you do best, core application coding. Discover what's new with
Crystal Reports now. http://p.sf.net/sfu/bobj-july
_______________________________________________
Octave-dev mailing list
Octave-dev@...
https://lists.sourceforge.net/lists/listinfo/octave-dev

From forum: octave-dev

Re: NLopt - free nonlinear-optimization package for Octave

2009-11-22T18:38:07Z

Søren Hauberg wrote:
> I don't think we have some procedure for linking to external code. I
> think the best place is probably the Octave wiki. I'm sure there's a
> page about optimisation on this wiki, where it would be sensible mention
> your work.

Thanks for the suggestion!

http://wiki.octave.org/wiki.pl?CategoryOptimization

seems like the most likely place to put a link, but I don't have a wiki
login. Can someone add the link to NLopt (or give me a login)?

> I didn't look at your code (lack of time, sorry), but how much work
> would it be to integrate your work with the 'optimization' package?

Well, it depends what you mean by "integrate". Certainly, licensewise I
think there should be no problem (LGPL), and as I said NLopt already
comes with an Octave plugin, so you could certainly just ship it as-is
with the Octave-forge optimization package if you wanted to.

It would be easy to modify functions like fminunc etcetera to call NLopt
under the hood. However, NLopt seems to provide much more
functionality than the interfaces of your optimization package currently
provide, because it allows nonlinear constraints and global
optimization, so I'm not sure you would want to "bury" it inside other
functions exclusively.

Also, I should say that the philosophy of NLopt seems to be a bit
different from that of the optimization package in octave-forge. In the
optimization package, you provide separate interfaces for different
algorithms like BFGS and Nelder-Mead. NLopt, on the other hand, tries
to provide a single interface for all the algorithms, with the algorithm
specified as a parameter. (Of course, some algorithms are more limited
than others; e.g. if you specify nonlinear constraints with an algorithm
that only supports bound constraints, it will return an error code.)

(NLopt is designe for general nonlinear optimization, so it is probably
suboptimal for particualr convex cases that have specialized algorithms,
such linear programming and quadratic programming [LPs and QPs].)

Steven

------------------------------------------------------------------------------
Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day
trial. Simplify your report design, integration and deployment - and focus on
what you do best, core application coding. Discover what's new with
Crystal Reports now. http://p.sf.net/sfu/bobj-july
_______________________________________________
Octave-dev mailing list
Octave-dev@...
https://lists.sourceforge.net/lists/listinfo/octave-dev

From forum: octave-dev

Re: NLopt - free nonlinear-optimization package for Octave

2009-11-22T13:31:59Z

This looks interesting. Thanks for sharing.

> tir, 17 11 2009 kl. 18:53 -0500, skrev Steven G. Johnson:
> I'm not sure if there is a good place for you to link to external Octave
> plugins, but if you wanted to link to NLopt somewhere that would be
> great. (If you wanted to use NLopt more directly in Octave I would be
> happy to help with that as well.)

I don't think we have some procedure for linking to external code. I
think the best place is probably the Octave wiki. I'm sure there's a
page about optimisation on this wiki, where it would be sensible mention
your work.

I didn't look at your code (lack of time, sorry), but how much work
would it be to integrate your work with the 'optimization' package?

Søren

------------------------------------------------------------------------------
Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day
trial. Simplify your report design, integration and deployment - and focus on
what you do best, core application coding. Discover what's new with
Crystal Reports now. http://p.sf.net/sfu/bobj-july
_______________________________________________
Octave-dev mailing list
Octave-dev@...
https://lists.sourceforge.net/lists/listinfo/octave-dev

From forum: octave-dev