Vectorizing HIRLAM 4: complicated access patterns examined.

[Toon Moene]

L.S.,

This code:

      SUBROUTINE S(N)
      DIMENSION A(N), B(N)
      READ*,ISTART,ISTOP,B
      DO I = ISTART, ISTOP
         A(I) = B(I)
      ENDDO
      PRINT*,A
      END

when compiled thusly:

$ gfortran -g -S -O3 -ftree-vectorize -ftree-vectorizer-verbose=2 -msse2 vect4.f

draws the following "not vectorized" message:

vect4.f:4: note: not vectorized: complicated access pattern.
vect4.f:4: note: vectorized 0 loops in function.

This sort of coding is rather prevalent in HIRLAM, especially in the
interesting inner loops.  One needs this when not all of the inner area
is accessible / computable by the code at hand (for instance because there
are inaccessible "halo's" around the subarea that have to be treated
specially).

The funny thing is that replacing ISTART with a constant makes this loop
vectorizable.  So what's the fundamental difference (as far as vectorizing
is concerned) between a constant and a loop invariant :-) ?

Kind regards,

-- 
Toon Moene - e-mail: toon@moene.indiv.nluug.nl - phone: +31 346 214290
Saturnushof 14, 3738 XG  Maartensdijk, The Netherlands
A maintainer of GNU Fortran 95: http://gcc.gnu.org/fortran/