PyOpenCL computing backend

[Dan Farmer]

Just as a relevant resource for other people trying to get OpenCL
savvy in a hurry, I've been watching a series of talks that Andreas
Klockner gave in Jan. 2011. There are 4 one hour sessions:

Lecture 1  - http://www.youtube.com/watch?v=X9mflbX1NL8 (Intro stuff)
Lecture 2 - http://www.youtube.com/watch?v=MqvfCE_bKOg (PyOpenCL)
Lecture 3 - http://www.youtube.com/watch?v=TAvKmV7CuUw (Code generation)
Lecture 4 - http://www.youtube.com/watch?v=SsuJ0LvZW1Q (Multi-GPU, PyCUDA)

Hope it helps someone :)

-Dan

2011/5/17 James Bergstra <james.bergstra at gmail.com>:
> Exciting work.  Can I draw your attention to related ongoing work with
> Andreas Klochner and Frederic Bastien on the 'gpundarray' also slated for
> compyte?
> The goal is essentially to provide a c-based datatype
> to supersede PyOpenCL's and PyCUDA's GpuArray. This datatype would be used
> as the gpu-side container by PyCUDA, PyOpenCL, Theano, and hopefully other
> projects too. It will extend the existing GpuArray by supporting strided
> memory layouts like numpy's ndarray.
> For ongoing design discussions - feel free to join the list:
> http://lists.tiker.net/listinfo/gpundarray
> James Bergstra
> On Tue, May 17, 2011 at 11:52 AM, Ralf Gommers <ralf.gommers at googlemail.com>
> wrote:
>>
>>
>> 2011/5/17 Stéfan van der Walt <stefan at sun.ac.za>
>>>
>>> Hi all,
>>>
>>> As you may know, we now have a GSoC student working on a new backend
>>> engine for scikits.image:
>>
>> Sounds great! Perhaps Pieter's blog can be added to Planet Scipy for some
>> more exposure?
>>
>> Cheers,
>> Ralf
>>
>>>
>>> http://pythonimaging.blogspot.com/2011/04/gsoc2011.html
>>>
>>> One of these will be a PyOpenCL-backend, which I hope will be able to
>>> leverage this PyOpenCL-subproject:
>>>
>>> https://github.com/inducer/compyte
>>>
>>> Cheers
>>> Stéfan
>>
>
>
>
> --
> http://www-etud.iro.umontreal.ca/~bergstrj
>