Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!tut.cis.ohio-state.edu!ucbvax!ROBALO.NYU.EDU!mckenney
From: mckenney@ROBALO.NYU.EDU (Alan McKenney)
Newsgroups: comp.lang.fortran
Subject: FORTRAN array aliasing -- potential problems?
Message-ID: <8907311910.AA15243@robalo.nyu.edu>
Date: 31 Jul 89 19:10:13 GMT
Sender: daemon@ucbvax.BERKELEY.EDU
Lines: 49


[USENET posting via mail, since our posting software doesn't work]


Subject: FORTRAN array aliasing -- potential problems?
Newsgroups: comp.lang.fortran,comp.arch
Keywords: FORTRAN argument array aliasing
Followup-To: comp.lang.fortrancomp.lang.fortran
Reply-To: mckenney@robalo.nyu.edu

    I am considering a situation in which I would call a FORTRAN
subroutine which has several dummy argument arrays using the
same actual array.  Example:

	SUBROUTINE SUB1( N, A, LDA, B, LDB, C, LDC, .... )

    .... coding which sets and then uses A, then sets and uses B
	without referencing A again, then sets and uses C without
	referencing A or B again.  ...
	END
   ....
	DIMENSION WORK(LDW,M)
   ....
	CALL SUB1( M, WORK, LDW, WORK, LDW, WORK, LDW, ... )

In other words, if A, B, and C were not arguments, I could EQUIVALENCE
them without breaking things.  As SUB1 stands, if I am interested
in the final values of A, B, and C, I call SUB1 with distinct arrays,
if not, I call it with the same array.

    The call to SUB1 violates the FORTRAN standard, since A, B, and C
actually refer to the same array, and are modified.  In most
implementations that I know of, A, B, and C are effectively
EQUIVALENCEd to WORK, so it is as if I had used A instead of B and C.
The code I have described works on such a system.

    The FORTRAN on the DECsystem-10 used to copy scalar arguments
into local storage on entry and copy them back on exit; I can conceive
of systems which would do that for arrays as well, so WORK would be set
to either A, B, or C (it being undefined which, but it would be exactly
one of them.)  My code would also work on such a system.

    My question then: are there FORTRAN implementations, either
existing or which one could make an argument for using, under which
such code would break?


Alan McKenney        E-mail:  mckenney@robalo.nyu.edu          (INTERNET)
Courant Institute,NYU         ...!uunet!cmcl2!robalo!mckenney  (UUCP)