Test OpenCL built-in math functions in saxpy-kernels (Feature #248)
The OpenCL standard provide a function called "fma(a,b,c)" which returns c + a*b.
This is exactly what is done inside the "saxpy" and "saxsbypz" kernels.
Perhaps using this built-in function can improve performance further.
However, this kernel is most probable bandwidth-limited...
|related to CL2QCD - Feature #247: Usage of OpenCL built-in math functions||New||13 Dec 2011|