Octave and multi-processors

classic Classic list List threaded Threaded
7 messages Options
Reply | Threaded
Open this post in threaded view
|

Octave and multi-processors

Clinton Winant-2
I use octave to run fluid mechanics flow solvers.  These codes rely heavily on fft and the related scripts. In the development phase, I don't include parallel coding.  As I keep an eye on a system monitor during execution I notice that for long periods nearly all the processors are busy, when no other processes (that I know of) are running apart from my octave simulation.  Is octave clever enough to know how to use several processors?  Are details described anywhere?

This is a good occasion to thank Octave developers for 4.2.0  I rely heavily on graphics, and have found the OpenGL toolkit to be almost foolproof, with great enhancements.  I have moved to make a contributions

Thanks Octave Team

_______________________________________________
Help-octave mailing list
[hidden email]
https://lists.gnu.org/mailman/listinfo/help-octave
Reply | Threaded
Open this post in threaded view
|

Re: Octave and multi-processors

José Luis García Pallero
2017-01-13 18:32 GMT+01:00 Clinton Winant <[hidden email]>:
> I use octave to run fluid mechanics flow solvers.  These codes rely heavily
> on fft and the related scripts. In the development phase, I don't include
> parallel coding.  As I keep an eye on a system monitor during execution I
> notice that for long periods nearly all the processors are busy, when no
> other processes (that I know of) are running apart from my octave
> simulation.  Is octave clever enough to know how to use several processors?
> Are details described anywhere?

Octave uses FFTW or FFTPACK for FFT computations
(https://www.gnu.org/software/octave/doc/v4.0.3/Signal-Processing.html),
so if these libraries are compiled for multiprocessing environments
the computations will be performed in parallel

>
> This is a good occasion to thank Octave developers for 4.2.0  I rely heavily
> on graphics, and have found the OpenGL toolkit to be almost foolproof, with
> great enhancements.  I have moved to make a contributions
>
> Thanks Octave Team
>
> _______________________________________________
> Help-octave mailing list
> [hidden email]
> https://lists.gnu.org/mailman/listinfo/help-octave
>



--
*****************************************
José Luis García Pallero
[hidden email]
(o<
/ / \
V_/_
Use Debian GNU/Linux and enjoy!
*****************************************

_______________________________________________
Help-octave mailing list
[hidden email]
https://lists.gnu.org/mailman/listinfo/help-octave
Reply | Threaded
Open this post in threaded view
|

Re: Octave and multi-processors

Andreas Weber-6
In reply to this post by Clinton Winant-2
Am 13.01.2017 um 18:32 schrieb Clinton Winant:
> I use octave to run fluid mechanics flow solvers.  These codes rely
> heavily on fft and the related scripts.

GNU Octave uses FFTW (if available at compiletime) and uses as many
threads as "nproc" returns (4 on my system).

You can tweak that setting with
fftw ("threads", NTHREADS)

If you are doing FFT with the same size again and again you should
create "wisdom" (see fftw page and help fftw in Octave)
In a nutshell: This will check what method is the fastest on your CPU
and create plans which then can be used from inside Octave.

If you run into problems we can create a wiki page for that topic, okay?

Hope that helps, Andy

_______________________________________________
Help-octave mailing list
[hidden email]
https://lists.gnu.org/mailman/listinfo/help-octave
Reply | Threaded
Open this post in threaded view
|

Re: Octave and multi-processors

Andreas Weber-6
Am 14.01.2017 um 11:37 schrieb Andreas Weber:
> If you run into problems we can create a wiki page for that topic, okay?

There is already a page for that :-)
http://wiki.octave.org/Optimizing_Fourier_transforms

-- Andy

_______________________________________________
Help-octave mailing list
[hidden email]
https://lists.gnu.org/mailman/listinfo/help-octave
Reply | Threaded
Open this post in threaded view
|

Re: Octave and multi-processors

José Luis García Pallero
2017-01-14 11:40 GMT+01:00 Andreas Weber <[hidden email]>:
> Am 14.01.2017 um 11:37 schrieb Andreas Weber:
>> If you run into problems we can create a wiki page for that topic, okay?
>
> There is already a page for that :-)
> http://wiki.octave.org/Optimizing_Fourier_transforms

Is this page updates. I've installed Octave 4.2.0 compiled by myself
and linked with FFTW and apparently fftw_wisdom does not exist, but
instead there is a fftw() function that, inspecting its help, I think
it does the same task as fftw_wisdom

>
> -- Andy
>
> _______________________________________________
> Help-octave mailing list
> [hidden email]
> https://lists.gnu.org/mailman/listinfo/help-octave



--
*****************************************
José Luis García Pallero
[hidden email]
(o<
/ / \
V_/_
Use Debian GNU/Linux and enjoy!
*****************************************

_______________________________________________
Help-octave mailing list
[hidden email]
https://lists.gnu.org/mailman/listinfo/help-octave
Reply | Threaded
Open this post in threaded view
|

Re: Octave and multi-processors

José Luis García Pallero
2017-01-14 12:04 GMT+01:00 José Luis García Pallero <[hidden email]>:

> 2017-01-14 11:40 GMT+01:00 Andreas Weber <[hidden email]>:
>> Am 14.01.2017 um 11:37 schrieb Andreas Weber:
>>> If you run into problems we can create a wiki page for that topic, okay?
>>
>> There is already a page for that :-)
>> http://wiki.octave.org/Optimizing_Fourier_transforms
>
> Is this page updates. I've installed Octave 4.2.0 compiled by myself
> and linked with FFTW and apparently fftw_wisdom does not exist, but
> instead there is a fftw() function that, inspecting its help, I think
> it does the same task as fftw_wisdom

Excuse me by the mispelling of the first sentence in my previous mail.
It should be:

Is this page updated?

>
>>
>> -- Andy
>>
>> _______________________________________________
>> Help-octave mailing list
>> [hidden email]
>> https://lists.gnu.org/mailman/listinfo/help-octave
>
>
>
> --
> *****************************************
> José Luis García Pallero
> [hidden email]
> (o<
> / / \
> V_/_
> Use Debian GNU/Linux and enjoy!
> *****************************************



--
*****************************************
José Luis García Pallero
[hidden email]
(o<
/ / \
V_/_
Use Debian GNU/Linux and enjoy!
*****************************************

_______________________________________________
Help-octave mailing list
[hidden email]
https://lists.gnu.org/mailman/listinfo/help-octave
Reply | Threaded
Open this post in threaded view
|

Re: Octave and multi-processors

Andreas Weber-6
In reply to this post by José Luis García Pallero
Am 14.01.2017 um 12:04 schrieb José Luis García Pallero:

> 2017-01-14 11:40 GMT+01:00 Andreas Weber <[hidden email]>:
>> Am 14.01.2017 um 11:37 schrieb Andreas Weber:
>>> If you run into problems we can create a wiki page for that topic, okay?
>>
>> There is already a page for that :-)
>> http://wiki.octave.org/Optimizing_Fourier_transforms
>
> Is this page updates. I've installed Octave 4.2.0 compiled by myself
> and linked with FFTW and apparently fftw_wisdom does not exist, but
> instead there is a fftw() function that, inspecting its help, I think
> it does the same task as fftw_wisdom

As far as I can see fftw_wisdom was removed at 2007-01-05
The wiki page http://wiki.octave.org/Optimizing_Fourier_transforms shows
a call to fftw_wisdom to create a plan for a specific size which isn't
possible with fftw.

So I think the wiki page is really outdated...

-- Andy

_______________________________________________
Help-octave mailing list
[hidden email]
https://lists.gnu.org/mailman/listinfo/help-octave