Ct, Intel CPU+GPU computing

Robert Jacques sandford at jhu.edu
Fri Dec 12 07:37:46 PST 2008


On Fri, 12 Dec 2008 09:02:32 -0500, davidl <davidl at 126.com> wrote:
> I think we can learn something from it.
>
> http://techresearch.intel.com/UserFiles/en-us/File/terascale/Whitepaper-Ct.pdf
>
Well, Ct (like Microsoft's Accelerator) is fairly limited in its  
capabilities. (and the white-paper also contains some Intel marketing FUD)
Anyways, it looks like a restricted form of Bulk synchronous parallel  
programming (http://en.wikipedia.org/wiki/Bulk_Synchronous_Parallel) and  
something like Bulk-Synchronous GPU Programming  
(www.kunzhou.net/2008/BSGP.pdf) implemented on top of OpenCL  
(http://www.khronos.org/opencl/) so it may run on CPUs and all major GPUs,  
seems like a much better solution.



More information about the Digitalmars-d mailing list