[MPI3 Fortran] (j3.2006) (SC22WG5.3896) [ukfortran] [MPI3 Fortran] MPI non-blocking transfers

N.M. Maclaren nmm1 at cam.ac.uk
Mon Jan 26 17:28:12 CST 2009

On Jan 26 2009, Bill Long wrote:
>Yes, but this is a concept from the days when MPI-like transfers 
>actually used the system's I/O hardware.  Today that is increasing 
>archaic and misleading.  Times change, and we should be willing to 
>change with them.  A good place to start is getting rid of outdated 
>terminology.  It helps people think more clearly about what is going 
>on.  [BTW, when I started computing the only I/O was cards, line 
>printers, paper tape, magnetic tape, and disks.]

MPI still does use the system I/O hardware on most systems.  I/O over
a TCP socket normally uses the same hardware as MPI.  I/O is data
transfer to or from an external location or agent, and a separate MPI
process runs a separate Fortran program.

And terminal connexions to computers date from 1955 - I don't know when
computer-computer connexions started, but I believe that RJE was no later
than about 1960.

Nick Maclaren,
University of Cambridge Computing Service,
New Museums Site, Pembroke Street, Cambridge CB2 3QH, England.
Email:  nmm1 at cam.ac.uk
Tel.:  +44 1223 334761    Fax:  +44 1223 334679

More information about the mpiwg-fortran mailing list