segmentation fault on ifort

[Pfor]

Hallo all,
I just want to tell you about a problem I had with ifort
(probably most of you know about this problem but it took
me a lot of time to solve it).

The code worked fine with gfortran and with old versions of ifort.
I had no clue of what was happening.

In the code there is a subroutine,
and in the subroutine I declare some array e.g.:

If I write something like (being "ndim" a "big"
parameter taken from elsewere in the code):

real (kind = 8), dimension(ndim) :: vett

some recent version of ifort compile but return a segmentation fault when
the code is run,
while something like:

real (kind = 8), dimension), allocatable :: vett
allocate (vett(ndim))

works correctly!

(I found the solution here :

http://jblevins.org/notes/segfault

)

That's all folks,
sorry if you already knew it,
cheers,
Paolo

 

[FJacq]

Yes, this is a well known problem due to a limited stack size : automatic arrays are often allocated on the stack whereas allocatable arrays are always allocated on the heap... But this feature depends on the compiler and even on the compiler version.

I include for years in my .bashrc file the following instruction :

ulimit -s unlimited

This is also the advice proposed for Unix systems in the link you provided.

On Windows, as far as I remembered, the stack size could be provided when linking the executable program. But this solution was not very convenient because the needed stack size might depend on user parameters.

I did not know the option -heap-arrays which is indeed must more convenient. But I don't understand why Intel does not provide a reasonable default value for this parameter...

Thank you for the link

François Jacq