Elemental finalization

IanH — Tue, 03 Sep 2013 21:02:08 GMT

Following on from a recent c.l.f discussion I think there's a glitch with elemental finalization in the recent release that wasn't in previous releases. The following little program seems to take an excessive period of time to run.

[fortran]MODULE m20130902
IMPLICIT NONE
TYPE :: t
    INTEGER :: comp = 1
CONTAINS
    FINAL :: t_final
END TYPE t
CONTAINS
ELEMENTAL SUBROUTINE t_final(arg)
    TYPE(t), INTENT(IN) :: arg
    IF (arg%comp /= 1) THEN
      CONTINUE
    END IF
END SUBROUTINE t_final
END MODULE m20130902

PROGRAM finally
USE m20130902
IMPLICIT NONE
TYPE(t) :: lhs(1)
TYPE(t) :: rhs(0)
lhs(:SIZE(rhs)) = rhs
END PROGRAM finally
[/fortran]

[plain]>ifort /Od /check:all /warn:all /standard-semantics "2013-09-02 finally.f90"
Intel(R) Visual Fortran Intel(R) 64 Compiler XE for applications running on Intel(R) 64, Version 14.0.0.103 Build 20130728
Copyright (C) 1985-2013 Intel Corporation. All rights reserved.

Microsoft (R) Incremental Linker Version 10.00.40219.01
Copyright (C) Microsoft Corporation. All rights reserved.

"-out:2013-09-02 finally.exe"
-subsystem:console
"2013-09-02 finally.obj"

>"2013-09-02 finally.exe"
forrtl: error (200): program aborting due to control-C event
Image              PC                Routine            Line        Source
kernel32.dll       0000000076C94803 Unknown               Unknown Unknown
kernel32.dll       0000000076C5652D Unknown               Unknown Unknown
ntdll.dll          0000000076D8C541 Unknown               Unknown Unknown[/plain]

(Ctrl-C pressed after a minute or so)

If you use the debugger on this similar program, and investigate values of the component of the argument to the finalizer, you see some odd stuff going on - component value of zero, lots of extra finalizer calls associated with the assignment.

[fortran]!! Check that
MODULE m20130902b
IMPLICIT NONE
TYPE :: t
    INTEGER :: component
CONTAINS
    FINAL :: t_final
END TYPE t
CONTAINS
ELEMENTAL SUBROUTINE t_final(arg)
    TYPE(t), INTENT(IN) :: arg
    CONTINUE      ! Put a breakpoint here.
END SUBROUTINE t_final
END MODULE m20130902b

PROGRAM finally_off_by_one
USE m20130902b
IMPLICIT NONE
CALL sub
CONTAINS
SUBROUTINE sub
    TYPE(t) :: array(3), scalar
    array%component = [1, 2, 3]
    scalar%component = 9
    array(2) = scalar
    ! Put another bp here to separate assignment related finalization
    ! from end of scope related.
    PRINT *, array%component
END SUBROUTINE sub
END PROGRAM finally_off_by_one
[/fortran]

Yeah, I had tried one of the

Steven_L_Intel1 — Tue, 03 Sep 2013 21:07:28 GMT

Yeah, I had tried one of the earlier variants of that program and noticed the delay, though the program did eventually finish after about 15 seconds. I will try your version here and see what's going on.

Thanks, Ian. I think

Steven_L_Intel1 — Mon, 09 Sep 2013 20:10:21 GMT

Thanks, Ian. I think something is walking off the end of a list here. I can reproduce problems with both test cases. Escalated as issue DPD200247880.

I expect that this will be

Steven_L_Intel1 — Thu, 19 Sep 2013 15:19:02 GMT

I expect that this will be fixed in Update 1, planned for October.

topic Yeah, I had tried one of the in Intel® Fortran Compiler

Elemental finalization

Yeah, I had tried one of the

Thanks, Ian. I think

I expect that this will be