Just submitted a proposal for a paper at the GCC Summit coming June ...

[jdelia@intec.unl.edu.ar]

1) Co Array Fortran (CAF):

> Is it this actually clear that co-arrays will be in the next  
> standard and not in a TR? I had the feeling that things (e.g. number  
> of co-array ranks etc.) are still in a flux.

   Perhaps, e.g. see the minutes of the WG5 meeting held in London
   in August 2007 at ftp://ftp.nag.co.uk/sc22wg5/N1651-N1700/N1691.txt.
   Also, could be interesting the
   ftp://ftp.nag.co.uk/sc22wg5/N1701-N1750/N1712.txt file, which
   is related to the ISO/IEC JTC1/SC22/WG5 N1712 proposal "Reducing
   the extent of the co-array feature", by J. Xia and J. Reid.
   About the last one, from a cluster manager (and user) perspective,
   the possibility to reduce in some degree the extent of the co-array
   features might be useful if should allow a more quickly availability
   of F2003 and/or F2008 compilers, of both vendors and public-domain
   ones. It is rather amazing that public compilers that implement
   the Unified Parallel C (UPC) extension seem something more
   mature than the CAF ones, e.g. those UPC from Berkeley, Washington
   and Michigan site domains are full operatives on our Linux Beowulf
   cluster using a MPICH2 layer.

2) Fortran codes in distributed memory systems:

> I think one could (should?) generate calls to a MPI library for the  
> distributed-memory version, which the user has then to link.

   Maybe. The UPC and CAF compilers available nowadays use the
   GasNet (http://gasnet.cs.berkeley.edu/) or
   ARMCI (http://www.emsl.pnl.gov/docs/parsoft/armci/index.html)
   libraries, which already include an option of using an internal
   MPI layer.

3) "Pointer lhs" issue:
   On the other hand, the "pointer lhs" feature was agreed by the UK
   panel and went forward as UK-008 in the WG5 repository (N1626).
   It was accepted by WG5 as second priority (at the WG5 meeting in
   May 2005 - minutes in N1631) and it is now included in the draft
   for Fortran 2008.

Regards,
Jorge D'Elia.
---
CIMEC, http://venus.ceride.gov.ar/cimec
INTEC, Guemes 3450, 3000-Santa Fe, ARGENTINA
tel: +54-342-451.15.94, int. 1018
fax: +54-342-455.09.44
Universidad Nacional del Litoral (UNL)
Consejo Nacional de Investigaciones Cientificas y Tecnicas (CONICET)



----------------------------------------------------------------
This message was sent using IMP, the Internet Messaging Program.