I tested the performance between Intel OpenCL SDK and avx with multithread. I think the OpenCL sdk also using avx and multithread in the opencl driver. So the performance maybe almost equal.
But the result is the time using OpenCL is 890ms, while the avx with multithread is 220ms. I already read the optimize notice for intel opencl sdk. So I want to ask whether the OpenCL sdk using avx and multithread in the driver. or I guess the driver only using sse and multithread. that can explain the performance difference .The test env. is:
WIN7 x64 sp1
i5-4590 DDR3 8G memory dual channel.
OpenCL sdk 16.1