ICE with Intel Fortran 16.X and 17.X

[Victor_S_2]

Dear Intel Fortran Compiler Support team,

I'm a member of a research group who is developing its own research software using the lastest features of the Fortran standard. This code succesfully compiles using the GNU Fortran compiler (version 5.X or 6.X), but while compiling with Intel Fortran Compiler (version 16.X or 17.X) we are experiencing an Internal Compiler Error.

We are compiling the code with 2 different sets of compiler flags:

• -r8 -fpp -W1 -fPIC -assume realloc-lhs -O3

• -r8 -fpp -W1 -fPIC -assume realloc-lhs -O0 -traceback -g -debug all -check all,noarg_temp_created -ftrapuv -warn nointerfaces

and with both we get the following message:

"catastrophic error: **Internal compiler error: segmentation violation signal raised** Please report this error along with the circumstances in which it occurred in a Software Problem Report.  Note: File and line given may not be explicit cause of this error."

The point where the ICE occurs contains highly intrincated  module dependencies and we are not able isolate this piece of code or to reproduce the ICE with a minimum sample code.

Currently, our code is not in the public domain yet, could it be possible that you take a look to this issue? If yes we could share with you the source code and compiling instructions using e.g. email

Thank you very much for your attention.
Best regards,
Víctor.

[TimP]

If you don't want to make the source public, you have available the option of filing a problem report on premier.intel.com using your support account.  If you haven't registered your ifort license, you may create your support account at https://registrationcenter.intel.com

[Steven_L_Intel1]

Let us know if you can't use Intel Premier Support for this.

[Victor_S_2]

Dear Tim and Steve,

Currently we are using the Intel Fortran Compilers installed in an HPC cluster. I think that we cannot use Intel Premier Support.

¿is there another possibility to share the code with you?

Best regards,

Víctor

[Steven_L_Intel1]

Where you're using it makes no difference. All that matters is what kind of license you have and if the support term is current. If you were able to download and install 17.0.1 then your support is current. If you want to send to me by email, send to steve.lionel [at] intel dot com. Please send the minimal files needed to reproduce the error. If it's more than one file, put them in a gzipped tar file.

[Victor_S_2]

Thank you Steve,

I've sent to you a private email with the source code.

Let me know if you need any help compiling it.

Víctor.

[Victor_S_2]

Dear Steve,

finally i was able to isolate the ICE reproducer in the code snippet below.

module base_static_triangulation_names
implicit none
private
 
    type cell_accessor_t
    private
        class(base_static_triangulation_t), pointer :: base_static_triangulation
    contains
        procedure, non_overridable :: get_num_vefs => cell_accessor_get_num_vefs
    end type cell_accessor_t
 
    type base_static_triangulation_t
    contains  
        procedure, non_overridable, nopass, private :: cell_unpack => bst_cell_unpack
    end type base_static_triangulation_t

    public :: base_static_triangulation_t
    public :: cell_accessor_t
 
 
contains

    pure function cell_accessor_get_num_vefs (this)
        class(cell_accessor_t), intent(in) :: this
        integer                        :: cell_accessor_get_num_vefs
        cell_accessor_get_num_vefs=0
    end function cell_accessor_get_num_vefs
 
    subroutine bst_cell_unpack (cell, buffer, vef_gids, vef_dimensions, coordinates)
        type(cell_accessor_t), intent(inout) :: cell
        integer,               intent(in)    :: buffer(:)
        integer,               intent(out)   :: vef_gids(cell%get_num_vefs())              ! <- I think that the problem is in the use of this pure function here
        integer, optional, intent(out)   :: vef_dimensions(cell%get_num_vefs())    ! <- and here
        real, optional,     intent(out)   :: coordinates(3,cell%get_num_vefs())        ! <- and here
        vef_gids = 0    
    end subroutine bst_cell_unpack

end module base_static_triangulation_names

module fe_space_names

  use base_static_triangulation_names
 
implicit none
private

  type, extends(cell_accessor_t) :: fe_accessor_t
  end type fe_accessor_t

end module fe_space_names

[Steven_L_Intel1]

Thank you! Escalated as issue DPD200416054. Seems strange that it is necessary to extend fe_accessor_t to trigger the ICE.