Re: [MPIWG Fortran] Question about 388 errata

18 Sep 2013

      On 9/18/13 6:03 AM, Rolf Rabenseifner wrote:
...
I propose in
https://svn.mpi-forum.org/trac/mpi-forum-web/attachment/ticket/388/errata-30...
Page 7 lines 14-15 read
For TS 29113 quality, new routines
   are needed only for routines with choice buffer arguments.
but should read
If the Fortran compiler implements TS 29113, then new routines
   are needed only for routines with choice buffer arguments.
This seems to be conflating two concepts.  Whether the Fortran compiler 
supports TS 29113 depends on the compiler vendor.  But this is an Advice 
to the MPI library vendors.  The sentence seems to read that having at 
least one Fortran vendor supporting TS 29113 forces MPI vendors to 
supply the extra routines.  If that is the intent, then the first clause 
of the sentence is unnecessary, since there are already two compilers 
with TS 29113 support.

The version I proposed:

"To take advantage of the features enabled by TS 29113, new routines are 
needed only for routines with choice buffer arguments."

unlinks the requirements and allows the MPI vendor to not implement the 
extra routines and choose to ignore that fact that some of its users 
might have a Fortran compiler with TS support.

I'm not sufficiently informed about earlier MPI discussions to know 
which is intended, but wanted to point out the distinction.   Rolf's 
statement is essentially a requirement that all MPI vendors support the 
new routines.  Mine is just a statement of fact and does not impose a 
requirement on the vendors.

I'm satisfied with either intention, since I use both of the new 
compilers and would always want (as a user, and provide as an MPI 
vendor) the new routines anyway.

Cheers,
Bill
...
(We always use the short-cut TS 29113, the long version is in the bib-list)
Jeff S., Jeff H., Craig, Bill L.,
please give an explicit okay (or a better text).
Best regards
Rolf
----- Original Message -----
...
From: "William Gropp" <[email protected]>
To: [email protected], "MPI-WG Fortran working group" <[email protected]>
Sent: Tuesday, September 17, 2013 1:04:57 AM
Subject: Re: [MPIWG Fortran] Question about 388 errata
What I need is a specific replacement for that sentence to which the
WG agrees. ï¿½ Suggestions?
Bill
Williamï¿½ Gropp
Director, Parallel Computing Institute
Deputy Director for Research
Institute for Advanced Computing Applications and Technologies Thomas
M. Siebel Chair in Computer Science
University of Illinois Urbana-Champaign
On Sep 16, 2013, at 11:49 AM, Bill Long wrote:
I think the intent is "To take advantage of the features enabled by TS
29113, new routines are needed...". ï¿½ The main thrust of the TS, as far
as MPI is concerned, is the capability to pass an argument by (pointer
to a ) descriptor, rather than as a plain address. ï¿½ This is only done
for choice buffer arguments.
The actual side effect of this is that for routines that do not have
choice buffer arguments (MPI_Barrier.... ) ï¿½ the implementation layer
in the C library is the same as before. ï¿½ Only for the ones with choice
buffer arguments is there a need for both the legacy version (with a
(void *) format parameter) and the new one that handles pointer to a
descriptor.
Cheers,
Bill
On 9/16/13 11:39 AM, Jeff Hammond wrote:
I think "the Fortran compiler implements ISO/IEC TS 29113:2012" is
more
accurate.
Jeff
On Mon, Sep 16, 2013 at 10:56 AM, William Gropp < [email protected]
< mailto:[email protected] >> wrote:
...
...
There was one phrase that I found imprecise. ï¿½ Just what does "TS
29113 quality" mean? ï¿½ I can guess, but the standard should not make me
guess.
...
...
Bill
...
...
William Gropp
...
Director, Parallel Computing Institute
...
Deputy Director for Research
...
Institute for Advanced Computing Applications and Technologies
...
Thomas M. Siebel Chair in Computer Science
...
University of Illinois Urbana-Champaign
...
...
...
...
...
...
_______________________________________________
...
mpiwg-fortran mailing list
...
[email protected]
< mailto:[email protected] >
...
http://lists.mpi-forum.org/mailman/listinfo.cgi/mpiwg-fortran
--
Jeff Hammond
[email protected] < mailto:[email protected] >
_______________________________________________
mpiwg-fortran mailing list
[email protected]
http://lists.mpi-forum.org/mailman/listinfo.cgi/mpiwg-fortran
--
Bill Long ï¿½ ï¿½ ï¿½ ï¿½ ï¿½ ï¿½ ï¿½ ï¿½ ï¿½ ï¿½ ï¿½ ï¿½ ï¿½ ï¿½ ï¿½ ï¿½ ï¿½ ï¿½ ï¿½ ï¿½ ï¿½ ï¿½ ï¿½ ï¿½ ï¿½ ï¿½ ï¿½ ï¿½ ï¿½ ï¿½ ï¿½ ï¿½ ï¿½ ï¿½ ï¿½ ï¿½ ï¿½ ï¿½ ï¿½ ï¿½ ï¿½ ï¿½  [email protected]
Fortran Technical Support ï¿½ ï¿½ ï¿½ & ï¿½ ï¿½ ï¿½ ï¿½ ï¿½ ï¿½ ï¿½ ï¿½ ï¿½ ï¿½ ï¿½ ï¿½ ï¿½ ï¿½ ï¿½ ï¿½ voice: 651-605-9024
Bioinformatics Software Development ï¿½ ï¿½ ï¿½ ï¿½ ï¿½ ï¿½ ï¿½ ï¿½ ï¿½ ï¿½ ï¿½ fax: ï¿½ ï¿½ 651-605-9142
Cray Inc./Cray Plaza, Suite 210/380 Jackson St./St. Paul, MN 55101
_______________________________________________
mpiwg-fortran mailing list
[email protected]
http://lists.mpi-forum.org/mailman/listinfo.cgi/mpiwg-fortran
_______________________________________________
mpiwg-fortran mailing list
[email protected]
http://lists.mpi-forum.org/mailman/listinfo.cgi/mpiwg-fortran
-- 
Bill Long                                           [email protected]
Fortran Technical Support    &                 voice: 651-605-9024
Bioinformatics Software Development            fax:   651-605-9142
Cray Inc./Cray Plaza, Suite 210/380 Jackson St./St. Paul, MN 55101

Bill Long

tags

participants (1)