(j3.2006) Liaison to IFIP WG 2.5

Dan Nagle dannagle
Wed Aug 22 16:48:17 EDT 2007


Hello,

David Muxworthy wrote:
> FWIW the most recent production program I wrote
> was to read a file of tens of thousands of records, sort and analyse
> them and pick out the most interesting for further analysis.
> Parallelism was irrelevant, but the other cores could have been
> working on other tasks at the same time.

Nothing about multithreading prevents other cores
from executing other processes.  It's modern scheduling
and all that.  (And if the main memory is entirely taken
by the multithreaded process, getting it done quickly _helps_
the other processes, rather than hurting them.)

Perhaps David would be amused to hear that the System V Unix
spell program was a multiprocessing program.  The multitasked
part divided portions document into words,
and sorted the resulting lists.
The result of each sort was compared against the dictionary.
It was I/O bound, so it benefited from multiprocessing.
And reusing the dictionary buffers benefited other processes
needing system buffers.

It was completely character processing,
and very much parallel processing,
and from the 1970s.

Sorry, the parallelism-is-a-small-niche argument has been untrue
for decades now.  It will continue to be untrue no matter
how it's phrased.

-- 

Dan Nagle
Purple Sage Computing Solutions, Inc.



More information about the J3 mailing list