Parameterized derived type: problem with ASSOCIATE construct.

[FortranFan]

module m

   implicit none

   private

   type, public :: t(n)
      private
      integer, len :: n = 1
      integer :: m_a(n)
   end type t

   public :: init 
   
contains

   subroutine init(this)

      type(t(*)), intent(inout) :: this

      asc: associate ( a => this%m_a )

         a = 1

      end associate asc

      print *, " with associate: t%a = ", this%m_a

      return

   end subroutine init

end module m

program p

   use m, only : t, init

   type(t(n=2)) :: foo

   call init(foo)

   stop

end program p

Upon execution with either IA-32 or Intel(R) 64 compiler version:

  with associate: t%a =  0 0
Press any key to continue . . .

[FortranFan]

Same issue exists with a parameterized derived type with a type-bound procedure:

module m

   implicit none

   private

   type, public :: t(n)
      private
      integer, len :: n = 1
      integer :: m_a(n)
   contains
      private
      procedure, pass(this), public :: init => init_t
   end type t

contains

   subroutine init_t(this)

      class(t(*)), intent(inout) :: this

      asc: associate ( a => this%m_a )

         a = 1

      end associate asc

      print *, " with associate: t%a = ", this%m_a

      return

   end subroutine init_t

end module m

program p

   use m, only : t

   type(t(n=2)) :: foo

   call foo%init()

   stop

end program p

  with associate: t%a =  0 0
Press any key to continue . . .

[Steven_L_Intel1]

This may be one we've already seen, but I'll check it out on Monday.

[andrew_4619]

I must say I do not use PDTs at the moment but can see that have some useful function. I have to say FortranFan you are doing a fantastic job in giving PDTs (and other new features) in Ifort a severe beasting and perhaps in a release or two's time I might risk dipping my toes in the water when the sharks have been cleared out.  :-)

[FortranFan]

app4619 wrote:

I must say I do not use PDTs at the moment but can see that have some useful function. I have to say FortranFan you are doing a fantastic job in giving PDTs (and other new features) in Ifort a severe beasting and perhaps in a release or two's time I might risk dipping my toes in the water when the sharks have been cleared out.  :-)

Thanks, I surely hope Intel doesn't mind all the posts that lead to Steve et al. having to submit tracking incidents! :-)

Re: your plan to wait until the bugs are ironed out, with PDTs yes there are a lot of issues in the current compiler version, so one has no option but to wait for a robust implementation.  But with the other features in Fortran 2003 and 2008 standards, if you do intent to use them in the near future in new code or in upgrade of your existing code, it may makes sense to start dabbling now and put together small cases that mimic your code design patterns.  I've found this firms up my understanding while also identifying any compiler issues earlier on in the code development cycle.

Now with PDTs, I do think they hold a lot of value and Intel can no sooner improve their implementation for users to start appreciating the benefits.

[Steven_L_Intel1]

The developers do suggest that I go on vacation more often, though this just bunches up the reports.... We do appreciate the bug reports and are interested in strengthening the PDT support as soon as we can. We have already made great strides, thanks to past reports from you and others.

[Steven_L_Intel1]

Issue ID is DPD200367383.

[FortranFan]

Thanks Steve, you deserve all the vacation you desire!  But our hope is you only "hang up your boots" after all the computing is done and the world is in nirvana!! :-))

[Steven_L_Intel1]

Well, this year you'll be without me for quite a bit, so get those bugs in early. (My colleagues will try to fill in...)