Segmentation fault

mathieu52 · ‎10-24-2005

Hi,

The following program creates segmentation
fault under ifc 8.1 and ifort 9.0 but not
under ifc 7.0
PROGRAM BUGSEARCH
CALL gls_corstep_tile(1044574)
CONTAINS
SUBROUTINE gls_corstep_tile(Jend)
integer, intent(in) :: Jend
integer, parameter :: r8 = selected_real_kind(12,300) ! 64-bit
real(r8), dimension(Jend) :: var1
! real(r8), dimension(1044574) :: var1
Print *, 'The program has finished'
END SUBROUTINE gls_corstep_tile
END PROGRAM
Replacing 1044574 by a smaller value make it work.

Replacing Jend by its value 1044574 (i.e. uncommenting and commenting the critical declaration lines) make it works also.

This was tested under linux 2.6.8-24.16-bigsmp
and under 2.4.20-8.

Mathieu

TimP · ‎10-24-2005

Under what operating system do you see this problem? You say you have tested under linux, but you posted in the Windows forum. I suppose we can assume you are running a 32-bit system, since ifc 7.1 didn't have EM64T support.

It's not entirely surprising that the threshold where you would run out of space would change somewhat when you change from an automatic array to a fixed size array.
As Steve suggested on the linux forum, an allocatable array, with error checking, may be preferable to the use of such a large automatic array. The space available for an automatic array will depend on several considerations which aren't under your control.

Message Edited by tim18 on 10-24-2005 08:42 AM

Steven_L_Intel1 · ‎10-24-2005

I posted a reply in the Linux forum where this was also asked.

mathieu52 · ‎10-24-2005

Hi,

thank you to both of you, you solved my problem
and sorry for having put the problem first in
windows forum.

Mathieu