Re: CUDA/OpenCL version of Octave for GPGPU?

[David Bateman]

Octave uses the single precision blas and lapack if the type used is "single". 
For example

A = single([1 2; 3 4]);
y = A \ ones(2, 1, "single");

Will use the single precision Lapack

D.



Le 16 juin 2011 à 07:57, Martijn Brouwer <address@hidden> a écrit :

> Hi,
> After some searching I found ACML-GPU, the AMD Core Math Library with
> GPU accelartion. It contains BLAS and Lapack. For my ATI mobility, only
> single precsision calculations are supported, so I guess octave won't
> work. For a desktop Radeon it might work. Is there anybody with
> experience with ACML(-GPU) and octave? Would be nice to compare with
> nVidia.
> 
> Martijn
> 
> On Wed, 2011-06-15 at 22:36 +0200, Martijn wrote:
>> Hi Tim,
>> If I am correct, you 'just' have to link octave to an opencl
>> implementation of BLAS and Lapack. I enclosed 'just' because I have not
>> been able to find such implementation.
>> Cula, implements Lapack, but octave also uses BLAS directly. Hopefully
>> AMD and nVidia will come up with an implementation.
>> 
>> bye,
>> 
>> Martijn
>> 
>> On Thu, 2011-03-17 at 23:43 -0700, Tim Rueth wrote:
>>> Have there been any thoughts about a CUDA/OpenCL version of Octave (on
>>> Windows) that can run on a GPGPU?
>>> 
>>> 
>>> 
>>> Thanks,
>>> 
>>> 
>>> 
>>> --Tim
>>> 
>>> 
>>> _______________________________________________
>>> Help-octave mailing list
>>> address@hidden
>>> https://mailman.cae.wisc.edu/listinfo/help-octave
>> 
>> 
>> _______________________________________________
>> Help-octave mailing list
>> address@hidden
>> https://mailman.cae.wisc.edu/listinfo/help-octave
> 
> 
> _______________________________________________
> Help-octave mailing list
> address@hidden
> https://mailman.cae.wisc.edu/listinfo/help-octave