J3/01-220

Date:    21 May 2001
To:      J3
From:    John Reid
Subject: Interpretation F90/000204 (Meaning of "same variable"
         description of MVBITS)

Here are two new draft responses. I have altered the section number in
the question to that of the Fortran 95 standard. The old answer was
<possible response in 95-253 item 85>. These are entirely new. There
are two according to whether statements such as

         CALL MVBITS(DATA(1:2),2,2,DATA(2:3),0)

where there is a partial overlap between the FROM and TO arguments
are disallowed or allowed.

The view that I take is that to allow this would be a change to the
standard since DATA(1:2) is certainly not the same variable as
DATA(2:3), but I have tried to faithfully present the alternative view
that the intention was to have a facility similar to that of default
assignment. Note that if the above statement is not allowed, the
user can still get the effect of assignment with the statement

         CALL MVBITS((DATA(1:2)),2,2,DATA(2:3),0)

Finally, being more restrictive gives the compiler more scope for
optimizing the call.

I would like to thank Henry and Kurt for their help.

NUMBER: F90/000204
TITLE: Meaning of "same variable" description of MVBITS
KEYWORDS: MVBITS
DEFECT TYPE: Erratum
STATUS: J3 consideration in progress

QUESTION: Section 13.14.75 states "TO ... may be the same
variable as FROM".  Given the following statements, which
pairs of variables are the same?

              INTEGER :: I(10), J(10)
              EQUIVALENCE (I,J)
              INTEGER, TARGET :: T(2:11)
              INTEGER, POINTER :: P1(:), P2(:)
              P1 => T
              P2 => T(2:11)

              I and I                P1 and T
              I(1) and I(1)          P1 and T(2:11)
              I(1:10) and I(1:10)    P2 and T
              I(1:1) and I(1:1:-1)   P2 and T(2:11)
              I and I(1:10)          P1 and P2
              I and J
              I(1) and J(1)
              I(1:10) and J(1:10)
              I(1:1) and J(1:1:-1)
              I and J(1:10)

ANSWER:

The quoted wording makes it clear that calls with the same name for
FROM and TO, such as
   CALL MVBITS(DATA,2,2,DATA,0) ! Overwrite bits 0,1 by bits 2,3
are permitted. It was also intended that associated variables be
permitted. The edits correct this.

EDITS:

In the second line of the final paragraph of subclause 12.7.3
[215,7],  change 'may be the same variables' to  'and may be
associated scalar variables or array variables all of whose
corresponding elements are associated'.

In subclause 13.14.75, in the second line of the paragraph that defines
the effect of TO [258:21] change 'and may be the same variable as FROM'
to  'and may be associated with FROM'.

.......................................................................

Alternative response

ANSWER:

The paragraph on TO contains this sentence [258:24-26]: 'On return,
the LEN bits of TO starting at TOPOS are equal to the value that the
LEN bits of FROM starting at FROMPOS had on entry.'

This was intended to indicate that the semantics of MVBITS would model
those of intrinsic assignment; if any elements of the TO and FROM
arguments are associated, the effect is as if the FROM were an
expression. For example, the statement

         CALL MVBITS(DATA(1:2),2,2,DATA(2:3),0)

is interpreted in the same way as the statement

         CALL MVBITS((DATA(1:2)),2,2,DATA(2:3),0)

The edits make this clear.

EDITS:

In the first sentence of the final paragraph of subclause 12.7.3
[215,6-7], change 'the actual arguments. . . may be the same variable'
to  'the value of the actual argument corresponding to the FROM dummy
argument is associated with the corresponding dummy argument'.

In subclause 13.14.75, in the second line of the paragraph that defines
the effect of TO [258:21] change 'and may be the same variable as FROM'
to  'and may be associated with FROM'.

.......................................................................

SUBMITTED BY: /jor in response to IBM public comments
HISTORY: 95-299 m135 submitted