[Search for users] [Overall Top Noters] [List of all Conferences] [Download this site]

Conference turris::fortran

Title:	Digital Fortran
Notice:	Read notes 1.* for important information
Moderator:	QUARK::LIONEL

Created:	Thu Jun 01 1995
Last Modified:	Fri Jun 06 1997
Last Successful Update:	Fri Jun 06 1997
Number of topics:	1333
Total number of notes:	6734

1325.0. "Dummy arguments" by PRSSOS::DEGAGE () Mon Jun 02 1997 10:46

	Hello,

	I'd like to have a little update on the problem of omitted dummy argu
ments. I know omitted dummy arguments were not supported by VAX FORTRAN,
what about DEC FORTRAN nowadays ?

To say the truth, one customer wanted to use this 'feature' and encountered
the well known ACCVIO problem with an omitted character string argument, but
the accvio only occurs with nooptimized code ... Why ?

Here is a little example :

	implicit none
	character*2 chain /'cc'/
	integer*4 i /1/

	call my_sub(,i)

	end


	subroutine my_sub ( dum_chain, dum_i)
	character*2 dum_chain
	integer*4 dum_i

	type*, dum_i
	return
	end

So compiled /NOOPT/DEBUG it leads to ACCVIO, compiled with default qualifiers,
it runs.

Thanks for your explanations.

Marielle Degage   CSC France

T.R Title User Personal
Name Date Lines

1325.1 QUARK::LIONEL Free advice is worth every cent Mon Jun 02 1997 11:05 30

T.R	Title	User	Personal Name	Date	Lines
1325.1		QUARK::LIONEL	Free advice is worth every cent	`Mon Jun 02 1997 11:05`	30
	Still not supported. Optimization may change where code gets emitted to access the argument list, or the contents of registers/stack, etc. Yes, this program may "run" but if the subroutine actually uses dum_chain, it doesn't. Try this one: implicit none character2 chain /'cc'/ integer4 i /1/ call my_sub(,i) end subroutine my_sub ( dum_chain, dum_i) character2 dum_chain integer4 dum_i type, dum_i if (dum_i .ne. 1) type , dum_chain return end Even though the comparison will fail, so the TYPE of dum_chain doesn't occur, the compiler generates code to fetch the descriptor of dum_chain and fails. In the example you posted, the optimizer was able to tell that dum_chain was never used and thus removed the descriptor access code. Steve

Still not supported.  Optimization may change where code gets emitted to
access the argument list, or the contents of registers/stack, etc.  

Yes, this program may "run" but if the subroutine actually uses dum_chain,
it doesn't.  Try this one:

        implicit none
        character*2 chain /'cc'/
        integer*4 i /1/

        call my_sub(,i)

        end


        subroutine my_sub ( dum_chain, dum_i)
        character*2 dum_chain
        integer*4 dum_i

        type*, dum_i
        if (dum_i .ne. 1) type *, dum_chain
        return
        end

Even though the comparison will fail, so the TYPE of dum_chain doesn't
occur, the compiler generates code to fetch the descriptor of dum_chain and
fails.  In the example you posted, the optimizer was able to tell that
dum_chain was never used and thus removed the descriptor access code.

			Steve