J3/09-108
To:        J3
From:      Van Snyder
Subject:   Alternative and simpler method for atomic accesses, UTI 154
Reference: 09-007
Date: 2009 January 12

1.  The problem

The scheme dreamed up at meeting 186 to specify atomic accesses to a few
types and kinds of objects requires two intrinsic subroutines.  We
already have syntax for reference and definition.  A simpler scheme,
exploiting existing syntactic mechanisms is proposed here.  In addition
to simpler sytax and simpler explanation, atomic references would be
more efficient, because they do not require copying the atomically
referenced object to another variable before using its value.

2.  Edits

[117:8+ R601] ----------------------------------------------------------
Insert "<<or>> <atomic-indicator>" after "<<or>> <array-section>".
Maybe alphabetize all of the right-hand sides.

[121:5-] ---------------------------------------------------------------
Insert a new subclause

"6.4.2a Atomic indicator

R613a <atomic-indicator> <<is>> <designator> % ATOMIC

C618a The <designator> shall be of type integer with kind
      ATOMIC_INT_KIND (13.8.2.2) or of type logical with kind
      ATOMIC_LOGICAL_KIND (13.8.2.3).

C618b The <designator> shall not be a component of an object of a
      sequence type (4.5.2.3) or of a type with the BIND(C) attribute.

{So the processor can arrange convenient alignment:}

C618c If an atomic indicator is an actual argument, the corresponding
      dummy argument shall have the VALUE attribute.

{Require copy-in for atomic actual arguments, to avoid thunking or
hoping for copy in/out; maybe put it at [294:22+ C1236b] with "atomic
indicator" and "actual argument" exchanged:}

If an atomic indicator appears

o in a reference, the entirety of the value of the designated entity is
  produced as if instantaneously, and

o in a variable definition context, the entirety of the value of the
  designated entity is defined as if instantaneously,

without interfering with or interference from other images or processes
defined by means other than Fortran.

NOTE 6.5a
  Suppose a processor represents LOGICAL(ATOMIC_LOGICAL_KIND) using two
  octets, and suppose its values are represented by Z'0000' or Z'FFFF'.
  An atomic reference to or definition of such an object would always
  produce one of those values, and never produce either Z'00FF' or
  Z'FF00', even if another image, or a process defined by means other
  than Fortran, is defining it concurrently."

[190:12-14 8.5.1p6] ----------------------------------------------------
Replace the first sentence of the paragraph:

"An atomic (6.4.2a) reference to a coarray or coindexed object is
permitted during execution of a segment that is unordered relative to
the execution of a segment in which an atomic definition of it occurs."

[193:NOTE 8.38] --------------------------------------------------------
Replace "CALL ATOMIC_DEFINE (LOCKED[Q], FALSE)"
by      "LOCKED[Q]%ATOMIC = .FALSE"
Replace "VAL = .TRUE.
         DO WHILE (VAL)
         CALL ATOMIC_REF (VAL, LOCKED)"
by      "DO WHILE ( LOCKED%ATOMIC )"

[325: Table 13.1] ------------------------------------------------------
Delete "ATOMIC_DEFINE" and "ATOMIC_REF" lines.

[337:17-338:10 13.7.20-21] ---------------------------------------------
Delete subclauses 13.7.20 and 13.7.21 (ATOMIC_DEFINE and ATOMIC_REF).

[402:13, 16] -----------------------------------------------------------
At both places, insert "(6.4.2a)" after the first "atomic", and delete
"specified by atomic subroutines".