topic Calling of VML Mathematical Functions [in-place operation] in Intel® oneAPI Math Kernel Library

Calling of VML Mathematical Functions [in-place operation]

Marcus_W_1 — Mon, 01 Apr 2013 22:16:03 GMT

HI,

I have a question concerning the use of Vector Mathematical Functions. The documentation does not contain a hint whether one can apply the same argument twice [in-place operation] without any side effects. For example for two vectors a, b the operation a = a + b should be computed via

vdadd( n, a, b, a ); [in-place operation]

instead of creating a working array y first and copy the result to array a again:

[plain]y = new double;

vdadd( n, a, b, y );

copy(source: y, destination: a);[/plain]

I have tested it and it works for vdadd. My question is is not restricted to vdadd, more general to all other Mathematical functions as for example vdsqr, vzconj, vdpow etc. I can test each function before I will use it but this is rather time consuming and can change from release to release of the MKL Library.

Thanks in advance,

Markus

Hi Markus,

Chao_Y_Intel — Tue, 02 Apr 2013 02:14:31 GMT

Hi Markus,

It is true that all of the VML can work with the in-plance operation. Check the document here:

http://software.intel.com/sites/products/documentation/doclib/mkl_sa/11/mklman/GUID-CF5F14C8-F9B1-4846-87F5-FB1CC0AD5972.htm

Thanks,

Chao

Hi Chao,

Marcus_W_1 — Tue, 02 Apr 2013 21:57:02 GMT

Hi Chao,

thank you very much, I was blind...

Markus

I think the updated place to

Royi — Sat, 26 May 2018 06:50:51 GMT

I think the updated place to look for it is:

Intel MKL Developer Reference - Vector Mathematical Functions.

At least the In Place support holds for Intel MKL 2018 Update 2.