Optimizing an integral calculation for KNCs vector processing unit

[Anders_S_1]

Hi,

I want to calculate an integral on a KNC core in Fortran  double precision by offload. The integrand is built up of intrinsic functions, which I will

vectorize by OpenMP SIMD instructions. Is there any nice code examples around that could assist me?

Best regards

Anders S

[Loc_N_Intel]

Hi Anders,

I saw Tim P posted the C code sample where he mixed OpenMP and intrinsics (SSE4) here https://software.intel.com/en-us/forums/intel-moderncode-for-parallel-architectures/topic/297545 . I hope it is useful for you to write Intel(R) Initial Many Core Instruction (Intel(R) IMCI). Thank you.

[TimP]

The c++ and Fortran compilers optimize those cases well enough now that you may not wish to invest in knc intrinsic coding which will be discarded when avx512 CPU s arrive.  Compilers can take advantage of the 4 times wider lanes than you are restricted to with sse4.