topic Internal compiler error when using ASSOCIATE in Intel® Fortran Compiler

Internal compiler error when using ASSOCIATE

sliska — Sun, 05 Dec 2010 02:34:56 GMT

Using ifort12.0.1.107 I get a internal compiler error when trying to compile:----------------------------
program testimplicit nonetype :: block_t integer,allocatable :: arr(:)end type block_ttype(block_t) :: blocks(2) associate(b=>blocks(1)) end associate end program test----------------------------The error occurs when I use the "-O0" or "-debug" options, but does not occur if I use "-O1","-O2", or "-O3".I have notice that this error does not occur if:(a) the array inside block_t is explicit (e.g. replace "integer,allocatable :: arr(:)" with "integer:: arr(-1:1)")(b) the same ASSOCIATE block is used inside a subroutine (see below for example)Another problem I have encountered with ASSOCIATE is illustrated by:----------------------------program testimplicit nonetype :: block_tinteger,allocatable :: arr_alloc(:)integer :: arr_expli(-1:1)end type block_ttype(block_t) :: blocks(2)call getbounds(blocks)containssubroutine getbounds(block)implicit none type(block_t),intent(in) :: block(2) allocate(block(1)%arr_alloc(-1:1)) associate(arr_alloc_assoc=>block(1)%arr_alloc) write(*,*) lbound(arr_alloc_assoc), lbound(block(1)%arr_expli) end associate associate(arr_expli_assoc=>block(1)%arr_expli) write(*,*) lbound(arr_expli_assoc), lbound(block(1)%arr_expli) end associate end subroutine getboundsend program test----------------------------output is:-1 -11 -1----------------------------It seems that the bounds "arr_expli_assoc" are incorrect. I believe this is a compiler problem since the ASSOCIATE documentation says, "The lower bound of each dimension is the result of the intrinsic function LBOUND applied to the corresponding dimension of selector."http://software.intel.com/sites/products/documentation/hpc/composerxe/en-us/fortran/win/lref_for/source_files/statement_associate.htm?wapkw=(ASSOCIATE)
Thanks,
Sebastian

Internal compiler error when using ASSOCIATE

Kevin_D_Intel — Sun, 05 Dec 2010 11:52:03 GMT

Thank you for reporting both errors, Sebastian. The internal error actually only occurs with -O0. -g implies -O0.

I submitted both to Development (internal tracking ids noted below) and will keep the thread updated as I learn more.

(Internal tracking id: DPD200163880 - internal compiler error)
(Internal tracking id: DPD200163881 - associate/lower bound)

(Resolution Update on 02/21/2011): Defect DPD200163880 is fixed in the Intel Fortran Composer XE 2011 Update 2 (2011.2.137 - Linux)

(Resolution Update on 04/27/2011): Defect DPD200163881 is fixed in the Intel Fortran Composer XE 2011 Update 3 (2011.3.174 - Linux)

Internal compiler error when using ASSOCIATE

Kevin_D_Intel — Mon, 21 Feb 2011 19:15:29 GMT

Defect DPD200163880 is fixed in the Intel Fortran Composer XE 2011 Update 2 (2011.2.137 - Linux).

Defect DPD200163881 is expected to be fixed in the Intel Fortran Composer XE 2011 Update 3 due out late next month.

Internal compiler error when using ASSOCIATE

sliska314 — Sun, 02 Oct 2011 23:31:12 GMT

Hi,

I noticed thatDPD200163881 in the l_fcompxe_2011.6.233 'Fixes List' (http://software.intel.com/en-us/articles/intel-composer-xe-2011-compilers-fixes-list/). I tried to run similar test as in the original post (part b):

[fortran]program test

implicit none

type :: block_t
  integer,allocatable :: arr_alloc(:)
  integer :: arr_expli(-1:1)
end type block_t

  type(block_t) :: blocks(2)

  call getbounds(blocks)

  contains

  subroutine getbounds(block)
  implicit none
    type(block_t),intent(inout) :: block(2)

    allocate(block(1)%arr_alloc(-1:1))

    associate(arr_alloc_assoc=>block(1)%arr_alloc)
      write(*,*) lbound(arr_alloc_assoc), lbound(block(1)%arr_alloc)
    end associate

    associate(arr_expli_assoc=>block(1)%arr_expli)
      write(*,*) lbound(arr_expli_assoc), lbound(block(1)%arr_expli)
    end associate

  end subroutine getbounds

end program test[/fortran]

with output:

-1 -1

1 -1

The lower bound for the associated array 'arr_expli_assoc' on line 26 still seems to be incorrect. I am using the latest version of the Composer XE2011Fortran compiler (Update 6).

Thanks,

Sebastian

Internal compiler error when using ASSOCIATE

Kevin_D_Intel — Mon, 03 Oct 2011 08:29:00 GMT

Thank you for your post regarding this earlier reported issue, Sebastian. I confirmed your findings and reopened defect DPD200163881. I will update this thread after I learn more. (Resolution Update on 10/14/2012): This defect is fixed in the Intel® Fortran Composer XE 2013 Initial Release (2013.0.0.079 - Linux)

The additional defect noted

Kevin_D_Intel — Sun, 14 Oct 2012 10:46:28 GMT

The additional defect noted in the similar test posted by sliska314 is fixed in the Intel® Fortran Composer XE 2013 Initial Release (2013.0.0.079 - Linux)