[Oisf-users] CUDA Licensing Grunts... (a little off-topic)

[Victor Julien]

Our first attempts at GPU acceleration were with OpenCL. Sadly the linux
OpenCL framework we used was buggy as hell. Drivers crashing and such. I
suspect the OpenCL frameworks out there to be more stable now, so it may
be interesting to revisit that at some point in the future.

On 04/20/2011 02:00 PM, Robert Vineyard wrote:
> There's also OpenCL, which is supposed to be a vendor-agnostic GPGPU
> development kit, but it hasn't achieved the same success as Nvidia's
> (currently) proprietary CUDA yet:
> 
> https://secure.wikimedia.org/wikipedia/en/wiki/OpenCL
> http://www.khronos.org/opencl/
> http://www.nvidia.com/object/cuda_opencl_new.html
> http://www.amd.com/streamopencl
> 
> Here's to hoping this type of thing will eventually take off... although
> CUDA already has quite a head start.
> 
> --
> Robert Vineyard, CISSP, RHCE
> Senior Information Security Engineer
> Georgia Tech Office of Information Technology
> 404.385.6900 (office/cell) / 404.894.9548 (fax)
> 
> On 4/20/2011 7:53 AM, Robert Vineyard wrote:
>> Actually a friend of mine recently got hired by Nvidia to create just that -
>> an open source implementation of the CUDA libraries - as an extension of his
>> PhD thesis, assuming I've been correctly informed.
>>
>> Hopefully Nvidia will see fit to foster further work on this and other projects:
>>
>> http://dank.qemfd.net/dankwiki/index.php/CUDA
>> http://dank.qemfd.net/dankwiki/index.php/CUBAR
>> http://dank.qemfd.net/dankwiki/index.php/Libcudest
>> http://dank.qemfd.net/dankwiki/images/d/d2/Cubar2010.pdf
>>
>> There are also others working on such things:
>>
>> https://github.com/pathscale/pscnv/wiki/nvidia_compute
>>
>> --
>> Robert Vineyard, CISSP, RHCE
>> Senior Information Security Engineer
>> Georgia Tech Office of Information Technology
>> 404.385.6900 (office/cell) / 404.894.9548 (fax)
>>
>>
>> On 4/19/2011 11:54 PM, Randal T. Rioux wrote:
>>> I went to setup a server with a couple GeForce 9600GT cards in it (even
>>> though they only support the 1.1 compute capabilities) to test the CUDA
>>> aspects of Suricata.
>>>
>>> I like Arch Linux, so I wanted to use that for a change.
>>>
>>> It wasn't until I tried to install the CUDA toolkit that I realized it
>>> is closed source. I'm not doing RPM mangling for Arch and would never
>>> touch Ubuntu (I have my reasons), so this makes me sad. Sadder is that I
>>> never noticed this before (always used RHEL or SL when CUDA dabbling).
>>>
>>> Does anyone know of an open-source challenger to the CUDA feature set /
>>> general premise? Something not as vendor-specific (of course). I can
>>> understand why NVidia doesn't want us to know how the sausage is made,
>>> but it doesn't mean I have to eat it!
>>>
>>> Thanks,
>>> Randy
>>> _______________________________________________
>>> Oisf-users mailing list
>>> Oisf-users at openinfosecfoundation.org
>>> http://lists.openinfosecfoundation.org/mailman/listinfo/oisf-users
>> _______________________________________________
>> Oisf-users mailing list
>> Oisf-users at openinfosecfoundation.org
>> http://lists.openinfosecfoundation.org/mailman/listinfo/oisf-users
> _______________________________________________
> Oisf-users mailing list
> Oisf-users at openinfosecfoundation.org
> http://lists.openinfosecfoundation.org/mailman/listinfo/oisf-users
> 


-- 
---------------------------------------------
Victor Julien
http://www.inliniac.net/
PGP: http://www.inliniac.net/victorjulien.asc
---------------------------------------------