segmentation violation for allocatable arrays

[cpmaster]

program test
  real(8), dimension(:,:),allocatable :: a,b,c
  allocate(a(1024,1024))
  allocate(b(1024,1024))
  allocate(c(1024,1024))
  c = 2.*matmul(a,b)
end program test

This code gives a SEGV at the assignment of c using the Intel
Fortran Compiler for Linux 8.0, but works under g95. It works for
both if the assignment line is replaced by c=matmul(a,b), or if the
array dimensions are any smaller than 1024. Am I
missing something basic here? Thanks in advance.

[TimP]

For a WAG, the compiler may be allocating a temporary result, before multiplying by 2. You may not have asked for additional stack space to allow for that temporary. Unfortunately, this seems a little inefficient, but not enough so to try to avoid it.

[cpmaster]

program test
  interface subtest
     subroutine subtest(a)
       real(8),dimension(:,:) :: a
     end subroutine subtest
  end interface  
  real(8), dimension(1024,1024) :: a
  call subtest(a)
end program test

subroutine subtest(a)
  real(8), dimension(:,:) :: a
  OPEN(1,FILE='test.txt',FORM='unformatted')
  WRITE(1) a 
  CLOSE(1)
end subroutine subtest

Fair enough. However, I also get a segv when I try to pass an array to a subroutine that writes the data to a file. This also only happens when the array dimensions are 1024 or more.
Is there a problem in passing "a" as an assumed-shape array?

Thanks, Cyrus

[TimP]

I guess the question here also is why the compiler has to allocate a new copy of the array. When I look at the code generated by -S, the new copy is generated in subtest, and vectorization has to be turned on to make the copy code efficient.
There's more than meets the eye in this assumed-shape thing. People I work with who write production code avoid it, and compiler vendors aren't challenged to figure out whether it can be optimized.

[jtao]

I am not sure if you already tried to set user resource limit to a bigger number. If not, you can trytosetthe stack size to unlimited to test things

out.

for bash: ulimit -s unlimitied

for (t)csh:limit -s unlimited

my 2c,

Jian

[cpmaster]

Thanks, Jian, that remedied the problem.