topic The original code, with in Software Archive

Finding elementwise and conditional matrix multiplication implementation with MKL

aketh_t_ — Tue, 22 Sep 2015 14:47:54 GMT

Hi all,

I have been looking for an MKL version of elementwise matrix multiplication that works based on a condional approach.While Vmult can be used it is for only a 1D vector rather than a matrix.

Below is the code I would like to rewrite with MKL version if possible.

logical(log_kind) check(2000,2000)

do i=1,2000

   do j=1,2000

   if ( check(i,j) )

   c(i,j) = a(i,j) * b(i,j) 

   enddo

enddo

I know Vmult helps, but it has no conditional operation.

Is there is a conditional Vector library or elementwise matrix library.

Can this be done by a combination of MKL library operations?

If the check array has a

jimdempseyatthecove — Wed, 23 Sep 2015 12:54:28 GMT

If the check array has a large number of .TRUE., then consider replacing the logical check array with an array of values 0.0 for false and 1.0 for true. Then use

real :: use_this(2000,2000)
...
do j=1,2000 ! outer loop to be rightmost index
   do i=1,2000 ! inner loop to be the leftmost index
       c(i,j) = c(i,j) + a(i,j) * b(i,j) * use_this(i, j)
   enddo
enddo

Jim Dempsey

The original code, with

TimP — Wed, 23 Sep 2015 15:14:25 GMT

The original code, with corrected syntax (either f66 one-line if or inserting endif) should work well, at least since SSE4, possibly with the addition of a directive such as !dir$ vector always in order to relax the constraint about exceptions. At -O3, ifort should fix the backward loop nesting automatically.

The (different) case which Jim quotes can be handled efficiently by e.g.

c(i,j) = c(i,j) + a(i,j) * merge(b(i,j), 0., check(i,j))

(again, SSE4 or later would help)