------- Additional Comments From tkoenig at gcc dot gnu dot org 2005-04-15
17:24 -------
There's something rotten in the state of Denmark.
I've slightly modified Paul's test program with my patch, and
this is what I got:
program test_spread
implicit none
integer, parameter :: N = 100
integer :: I
integer, dimension(N) :: source
integer, dimension(N,N) :: sink, check,c1,c2
source =(/(i, i=1,N)/)
check = spread( source , 1 , N )
PRINT *,"first 4x4 elements with DIM=1"
write(*,'(1x,4I4)') check(1:4,1:4)
check = spread( source , 2 , N )
PRINT *,"first 4x4 elements with DIM=2"
write(*,'(1x,4I4)') check(1:4,1:4)
c1 = spread(source, 1, N)
c2 = spread(source, 2, N)
sink = spread( source , 1 , N ) * spread( source , 2 , N )
PRINT *,"The product using temporaries"
write(*,'(1x,4I4)') sink(1:4,1:4)
PRINT *,"The product using fixed arrays"
sink = c1 * c2
write(*,'(1x,4I4)') sink(1:4,1:4)
end program test_spread
$ gfortran test_spread.f90
$ ./a.out
first 4x4 elements with DIM=1
1 1 1 1
2 2 2 2
3 3 3 3
4 4 4 4
first 4x4 elements with DIM=2
1 2 3 4
1 2 3 4
1 2 3 4
1 2 3 4
The product using temporaries
100 0 0 0
200 0 0 0
300 0 0 0
400 0 0 0
The product using fixed arrays
1 2 3 4
2 4 6 8
3 6 9 12
4 8 12 16
I am not applying my patch for the moment. A segfault
is better than a silently generated wrong result :-(
Investigating further.
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18495
