[Cplex] Short SIMD vector types

David Keaton dmk at dmk.com
Tue Jun 4 00:57:12 CEST 2013


On 06/03/2013 03:30 PM, Tom Scogland wrote:
> Both OpenCL, as mentioned, and CUDA provide similarly natural short
> vectors through types for specific widths.  They are far easier to
> manage than traditional vector intrinsics, especially for those new to
> SIMD programming.
>
> A big issue with that approach though is that it is naturally limiting
> and verbose when it comes to varying vector widths.  Is it good to have
> int2,int4,int8,int16... ad nauseum? Perhaps it is, but it may be worth
> considering defining a type with more generic semantics as well to allow
> the writing of code to "native SIMD width" rather than specific widths,
> especially given the proliferation of available widths in newer
> processors/accelerators.

      +1.  The fixed-width approach requires writing nonportable code.

      Array slices as first-class objects might be a good candidate to 
use instead.

					David



More information about the Cplex mailing list