Quantcast
Channel: Intel Developer Zone Articles
Viewing all articles
Browse latest Browse all 3384

Diagnostic 15032: loop was not vectorized: unsupported reduction (Fortran)

$
0
0

Cause:

Reduction idioms can be recognized by the compiler and vectorized in simple cases.The compiler is unable to vectorize more complex constructs and emits a warning message:

Example:

integer function minIndex(x,n)
   implicit none
   real,    intent(in), dimension(n) :: x
   real                              :: minValue
   integer, intent(in )              :: n
   integer                           :: i
   
   minValue = 999.
   minIndex = -1

   do i=1,n
      if(x(i).lt.100. .and. x(i).lt.minValue) then
         minValue = x(i)
         minIndex = i
      endif
   enddo
   
end function minIndex
> ifort -c -vec-report2 d_15032_1.f90
d_15032_1.f90(13): (col. 10) remark: loop was not vectorized: unsupported reduction
 

Resolution:

If a reduction idiom is not supported by the vectorizer, adding a SIMD directive will not help. However, if you are able to simplify your reduction construct, the compiler may vectorize it. In the above example, one of the conditions can be moved outside the loop, which allows the simplified loop to be vectorized:
 
integer function minIndex(x,n)
   implicit none
   real,    intent(in), dimension(n) :: x
   real                              :: minValue
   integer, intent(in )              :: n
   integer                           :: i
   
   minValue = 999.
   minIndex = -1

   do i=1,n
      if(x(i).lt.minValue) then
         minValue = x(i)
         minIndex = i
      endif
   enddo
   
   if(minValue.ge.100.) then
      minValue = 999.
      minIndex=-1
   endif
   
end function minIndex

> ifort -c -vec-report2 d_15032_2.f90
d_15032_2.f90(11): (col. 4) remark: LOOP WAS VECTORIZED

 


Viewing all articles
Browse latest Browse all 3384

Trending Articles



<script src="https://jsc.adskeeper.com/r/s/rssing.com.1596347.js" async> </script>