5 ms·

How does it's performance compare to Intel's MKL on a CPU?

• pavanky 12y ago
  The current CPU implementation is single core, non-vectorized code. That said, ArrayFire can link with any BLAS / LAPACK library to accelerate the relevant algorithms. EDIT: The CUDA and OpenCL backends will obviously be faster than MKL. We'll be adding SSE / AVX support at some point which'll make the CPU backend faster as well.
  • weitzj 12y ago
    did you look at http://www.yeppp.info http://www.yeppp.info for sse/avx support?
    • pavanky 12y ago
      Ha! We interviewed the author of that library!
  • 14113 12y ago
    In a similar vein, how does this compare with (say) CuBLAS or ClMath?
    • pavanky 12y ago
      We depend on those libraries. We make the API easier to understand while keeping the performance close to the upstream libraries.