[Search for users] [Overall Top Noters] [List of all Conferences] [Download this site]

Conference msbcs::hpc

Title:Parallel processing through Workstation Farms.
Notice:MSBCS::HPC (renamed from HPCGRP::WORKSTATION_FARMS)
Moderator:MSBCS::SYSTEM
Created:Tue Oct 27 1992
Last Modified:Fri Jun 06 1997
Last Successful Update:Fri Jun 06 1997
Number of topics:507
Total number of notes:1791

500.0. "parallel i/o ?" by TKOVOA::ISHIBASHI_T () Tue Mar 25 1997 20:17

Hi,I got the question of parallel i/o from our customer.

I think the current version of Digital's HPF & PSE dose not have
the parallel i/o capability. Read/Write operations are done on only peer 0.
MasPar's parallel fortran90 or C had those capability. 

How about this capability or plan to introduce to  digital's HPF. 
And are there the parallel i/o in HPF standard specification ?(v1.* or v2.* )

For example, parallel i/o is good performance for read/write some large 
temporary  files using the distributed memory - i/o module -disk systems.

  !hpf$  distribute a(block)
	  :
	real a(1000)
	  :
	read(10,*) a

best regards,
toshihiro
Digital Japan

	

    
T.RTitleUserPersonal
Name
DateLines
500.1Parallel I/O from HPFNETRIX::"[email protected]"Jonathan HarrisThu Mar 27 1997 14:2718
Parallel I/O to temporary files is easy for users to write themselves.  We
have
included a sample application in the /usr/examples/hpf directory that does
exactly that.

Language and runtime support for parallel I/O is being discussed by the HPF
Forum and other standards bodies, but no consensus has been reached on the
right
way to do it.  We are following those efforts, as we know that this is
important
for users.

Can you ask the customer what kind of parallel I/O support they would like to
see?

                                  Jonathan Harris
                                  HPF Project Leader
[Posted by WWW Notes gateway]
500.2Thank you.TKOVOA::ISHIBASHI_TWed Apr 09 1997 13:387
    
    Thank you, Re:1.
    We hope the implementation of language and runtime parallel i/o in
    our fortran for future, if hpf forum take the consensus of them.
    
    toshihiro