Is there still any need for the three hints accessor functions
of the june proposal (page 262f: MPI_HINTS_COUNT,
MPI_HINTS_GET_ELEMENT and MPI_HINTS_GET_VALUE) ?
Rolf
--------------------------------
History of my question:
John Wilson wrote:
>>> I was considering sending the following regarding how would
>>> a library writer cache file related information. Leslie
>>> Hart's proposal for info arguements would permit this type
>>> of operation. I do not think Rolf Rabenseifner's proposal
>>> would permit this.
...
>>> Is there a means in the MPI-IO proposal for caching
>>> information in/with a file handle?
Steve answered:
>> The external interfaces chapter has a section where caching on
>> additional opaque handles is proposed. The current draft adds groups
>> and datatypes. The one to be released soon adds persistant requests.
>> There is a note that we need to decide which other handles to add.
>> The only constraint is that we don't want to include handles that are
>> light weight (such as regular requests) where the overhead would be
>> noticable. I think that file handles are ok. Thus, I will add file
>> handles to the list unless someone objects.
and Rolf Rabenseifner asked:
>> is it enough if I add an analogy to the three hints accessor
>> functions of the june proposal (page 262f) to my proposal?
John Wilson answered:
> Steve Huss-Lederman pointed out that the MPI-misc group is
> working on adding caching to "generic" MPI handles. That
> approach, if it goes forward and includes file handles, is
> probaly preferable to using info/hints for what I had
> in mind.
Rolf Rabenseifner (Computer Center )
Rechenzentrum Universitaet Stuttgart (University of Stuttgart)
Allmandring 30 Phone: ++49 711 6855530
D-70550 Stuttgart 80 FAX: ++49 711 6787626
Germany rabenseifner@rus.uni-stuttgart.de