topic use mkl95_precision, only: wp => dp in Intel® oneAPI Math Kernel Library

use mkl95_precision, only: wp => dp

marcioy — Fri, 13 May 2011 03:01:17 GMT

Hi everyone,

I am migrating from IMSL to MKL. I am trying to implement matrix-matrix and matrix-vector calculations using the BLAS.

I have a doubt about what the following statements mean:

(...)

use mkl95_precision, only: wp => dp

implicit none

real(wp):: alpha, beta; !just an example of variables;

(...)

It seems that alpha and beta are double precision variables. There is any difference between the above statement and the following declaration? Which one should I use?

(...)

double precision alpha, beta;

(...)

Thank you in advance

Marcio Yamamoto

use mkl95_precision, only: wp => dp

Chao_Y_Intel — Fri, 13 May 2011 07:05:25 GMT

Hi Marcio,

You are right. They are double precision. Both of those two be compiled. If you write alpha, beta as:
real(wp):: alpha, beta;

It make it some easierto changeto single precision code, and only need to change:
wp => DP to wp = SP

Thanks,
Chao

use mkl95_precision, only: wp => dp

marcioy — Fri, 13 May 2011 07:26:39 GMT

Thank you, Chao!

The guys in Intel work up to midnight??!! You guys are my heroes! ;)

I was wondering if the statement "real(wp):: alpha, beta;" creates any special data structures/object which was necessary for the MKL/BLAS calculation. But it is not the case, right? Both statements generate exactly the same data structure: double precision array?

You should get some sleep, Chao.

Thank you again.

Marcio