Re: Octave multi threading.

Hi Kai,

I'm using Windows 10 operating system with 2 physical cores and 4 logical
cores. I have a few questions as follows.


   1. Does *Open blas* support the Windows Operating system? and if so, can
   you please provide me with the official link to download it and the
   instructions to install and configure it properly. Please advise.
   2. Can we configure the Open blas with Octave 5.2.0 or any other
   versions of Octave ?
   3. Apart from the above questions, I want to know what is a parallel
   computing or parallel execution package for Octave. Are they similar to
   Open Blas ? and what platform do they support ?

Waiting for your reply.

Best regards
Ilaya bharathi


On Thu, Jun 18, 2020 at 8:11 PM Kai Torben Ohlhus <[email protected]>
wrote:

> On 6/18/20 10:44 PM, Ilaya Bharathi wrote:
> > Hi team,
> >
> > I'm looking for an Octave release which will utilize multiple cores of
> > processors at a same time during heavy operations rather than a single
> > core.
> >
> > If there is any release of octave or any add on libraries that fulfill
> > my expectation, Please let me know. Waiting for your reply.
> >
> > Many thanks
> > Ilaya bharathi
> >
>
>
> What kind of heavy operations do you perform?  For example if you solve
> large linear systems, x = A \ b;, you should make sure that you use a
> (Open-)BLAS version that supports multiple threads with your Octave
> installation.
>
> If you run custom Octave code (m-files), you might be able to
> parallelize tasks by using the Octave Forge (OF) parallel [1] or ocs [2]
> package.
>
> Additionally, it is helpful to know some details about your system to
> give more precise advice.
>
> Personally, I run a custom Singularity Octave image [3] on a cluster of
> a few machines (each system has 28 cores [56 threads]) that really makes
> use of all given resources for linear-algebra operations.  It can be
> combined with the aforementioned OF packages at low installation effort.
>
> HTH,
> Kai
>
> [1] https://octave.sourceforge.io/parallel/
> [2] https://octave.sourceforge.io/ocl/
> [3] https://wiki.octave.org/Octave_for_GNU/Linux#Singularity
>