Intel® Fortran Compiler
Build applications that can scale for the future with optimized code designed for Intel® Xeon® and compatible processors.
28446 Discussions

Possible bug in ifort: using polymorphic types

Luckner__Paul
Novice
‎03-25-2020 05:08 AM
444 Views

I get different results depending if I comment specific lines out or not (even as those are not accessed at runtime). The bug does not appear using gfortran.

program main

  implicit none

  type :: parent
    integer              :: i1
    integer, allocatable :: i2
  end type

  type, extends(parent) :: child
    integer              :: i3
    integer, allocatable :: i4
  end type

  type :: parent_alc
    class(parent), allocatable :: p
  end type

  type(parent_alc) :: p_alc

  if (.false.) then
    block
      ! type(child) :: e            ! comment this line in/out for different results
    end block
  else
    block
      type(child)      :: c

      c%i1 = 1
      c%i2 = 1
      c%i3 = 1
      c%i4 = 1

      ! ====================
      ! method 1: broken in ifort
      ! p_alc%p = c         
      ! ====================

      ! ==================== 
      ! method 2: broken in ifort 
      ! allocate (p_alc%p, source=c)
      ! ====================

      ! ====================
      ! workaround
      allocate (child :: p_alc%p)
      select type (x => p_alc%p)
        type is (child)
          x%i1 = c%i1
          x%i2 = c%i2
          x%i3 = c%i3
          x%i4 = c%i4
      end select
      ! =====================

      c%i1 = 0
      c%i2 = 0
      c%i3 = 0
      c%i4 = 0
    end block
  end if

  select type (x => p_alc%p)
    type is (child)
      print *, 'p_alc%p%i1..4', x%i1, x%i2, x%i3, x%i4
  end select
end program

Running the program using method 1,2 and commenting in/out the highlighted line 23 yields different results with ifort and the same in gfortran.

Line commented out

$ ifort main.f90 && ./a.out
p_alc%p%i1..4           1           1           1           1

$ gfortran main.f90 && ./a.out
p_alc%p%i1..4           1           1           1           1

Line not commented out

$ ifort main.f90 && ./a.out
p_alc%p%i1..4           1           0           1           0

$ gfortran main.f90 && ./a.out
p_alc%p%i1..4           1           1           1           1

Program versions

$ ifort --version
ifort (IFORT) 19.0.1.144 20181018
Copyright (C) 1985-2018 Intel Corporation.  All rights reserved.

$ gfortran --version
GNU Fortran (Arch Linux 9.3.0-1) 9.3.0
Copyright (C) 2019 Free Software Foundation, Inc.
This is free software; see the source for copying conditions.  There is NO
warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.

The "workaround" code actually works under ifort/gfortran with and without line 23.

0 Kudos

Link Copied

4 Replies
FortranFan
Honored Contributor II
‎03-25-2020 05:19 PM
444 Views

An option you may want to consider if possible is to update your compiler version.  Version 19.1 appears not to face the issue:

C:\Temp>type p.f90
program main

   implicit none

   type :: parent
      integer              :: i1
      integer, allocatable :: i2
   end type

   type, extends(parent) :: child
      integer              :: i3
      integer, allocatable :: i4
   end type

   type :: parent_alc
      class(parent), allocatable :: p
   end type

   type(parent_alc) :: p_alc

   if (.false.) then
      block
         type(child) :: e            ! comment this line in/out for different results
      end block
   else
      block
         type(child)      :: c

         c%i1 = 1
         c%i2 = 1
         c%i3 = 1
         c%i4 = 1

         ! ====================
         ! method 1: broken in ifort
         p_alc%p = c

      end block
   end if

   select type (x => p_alc%p)
      type is (child)
         print *, 'p_alc%p%i1..4', x%i1, x%i2, x%i3, x%i4
   end select
end program

C:\Temp>ifort /standard-semantics /warn:all /stand:f18 p.f90
Intel(R) Visual Fortran Intel(R) 64 Compiler for applications running on Intel(R) 64, Version 19.1.0.166 Build 20191121
Copyright (C) 1985-2019 Intel Corporation.  All rights reserved.

p.f90(23): remark #7712: This variable has not been used.   
         type(child) :: e            ! comment this line in/out for different results
------------------------^
Microsoft (R) Incremental Linker Version 14.24.28316.0
Copyright (C) Microsoft Corporation.  All rights reserved.

-out:p.exe
-subsystem:console
p.obj

C:\Temp>p.exe
 p_alc%p%i1..4 1 1 1 1

C:\Temp>
C:\Temp>type p.f90
program main

   implicit none

   type :: parent
      integer              :: i1
      integer, allocatable :: i2
   end type

   type, extends(parent) :: child
      integer              :: i3
      integer, allocatable :: i4
   end type

   type :: parent_alc
      class(parent), allocatable :: p
   end type

   type(parent_alc) :: p_alc

   if (.false.) then
      block
         type(child) :: e            ! comment this line in/out for different results
      end block
   else
      block
         type(child)      :: c

         c%i1 = 1
         c%i2 = 1
         c%i3 = 1
         c%i4 = 1

         ! ====================
         ! method 2: broken in ifort
         allocate (p_alc%p, source=c)

      end block
   end if

   select type (x => p_alc%p)
      type is (child)
         print *, 'p_alc%p%i1..4', x%i1, x%i2, x%i3, x%i4
   end select
end program

C:\Temp>ifort /standard-semantics /warn:all /stand:f18 p.f90
Intel(R) Visual Fortran Intel(R) 64 Compiler for applications running on Intel(R) 64, Version 19.1.0.166 Build 20191121
Copyright (C) 1985-2019 Intel Corporation.  All rights reserved.

p.f90(23): remark #7712: This variable has not been used.   
         type(child) :: e            ! comment this line in/out for different results
------------------------^
Microsoft (R) Incremental Linker Version 14.24.28316.0
Copyright (C) Microsoft Corporation.  All rights reserved.

-out:p.exe
-subsystem:console
p.obj

C:\Temp>p.exe
 p_alc%p%i1..4 1 1 1 1

C:\Temp>

 

0 Kudos
Copy link
Luckner__Paul
Novice
‎03-26-2020 02:17 AM
444 Views

Thanks for the fast reply!

I will try it out with the newer version. Hope it works and not some internal thing caused by the underlying OS (linux/windows).

0 Kudos
Copy link
IanH
Honored Contributor II
‎03-26-2020 04:08 AM
444 Views

The code in #2 is missing the trailing assignment statements necessary to see the bug - 19.1 is still problematic.

I have encountered issues previously with over-enthusiastic deallocation of block local variables (or components of block local variables) with the current compiler.  I suspect that is going on here - move the block local variables up to the enclosing scope and the symptoms disappear.

0 Kudos
Copy link
FortranFan
Honored Contributor II
‎03-26-2020 07:21 AM
444 Views

IanH (Blackbelt) wrote:

The code in #2 is missing the trailing assignment statements necessary to see the bug - 19.1 is still problematic.

I have encountered issues previously with over-enthusiastic deallocation of block local variables (or components of block local variables) with the current compiler.  I suspect that is going on here - move the block local variables up to the enclosing scope and the symptoms disappear.

Bummer!  Support request at Intel OSC appears to be in order: https://supporttickets.intel.com/servicecenter?lang=en-US ;

0 Kudos
Copy link
Reply

Community support is provided during standard business hours (Monday to Friday 7AM - 5PM PST). Other contact methods are available here.

Intel does not verify all solutions, including but not limited to any file transfers that may appear in this community. Accordingly, Intel disclaims all express and implied warranties, including without limitation, the implied warranties of merchantability, fitness for a particular purpose, and non-infringement, as well as any warranty arising from course of performance, course of dealing, or usage in trade.

For more complete information about compiler optimizations, see our Optimization Notice.