RE: QCMDEXC errors -- MI400

Leif, you're a great story-teller, with a great story here!  (I should talk
to the moderator(s), and see if more Around Topic stuff like this can be
posted...;-)

At first I was crestfallen with ya, as I wondered how long it would-a taken
you to write the entire APP this fellow was attempting (if I understand it
correctly).  Then I was crestfallen with ya again, as I realized I'm a
daggone YOUNGSTER next to you!!  (So I should, by definition, be able to
disrespect you for no other reason than that, right!?...;-)  But I'm sure
you learned a lesson you never forgot, so I wasn't sorry fer ya a-tall,
about solving the wrong problem efficiently not being all that great as it
might appear at first.

And then I was wondering, not in a rhetorical sense at all, how long WOULD
it take you to code the errant solution this fellow was misapplying.
Wouldn't it seem like a useful algorithm for Grid or BitTorrent (and IBM's
similar software)?  And this may be what these technologies DO do, I dunno.
But if you measured the "distance" in time rather than linear miles, it'd
seem like this would be a useful algorithm, because the distance would vary
from moment to moment between computer and computer or data and processor.

Dunno.


| -----Original Message-----
| From: mi400-bounces@xxxxxxxxxxxx [mailto:mi400-bounces@xxxxxxxxxxxx]On
| Behalf Of Leif Svalgaard
| Sent: Friday, March 19, 2004 5:04 PM
| To: MI Programming on the AS400 / iSeries
| Subject: Re: [MI400] QCMDEXC errors
|
|
| From: Ted Slate <tslateone@xxxxxxxxxxx>
| > Sorry I'm an efficiency guy.
|
| When I started out with computers (circa 1963) I got an
| (unpaid) job to help other users at "the computer center".
| One guy had a real problem: his program ran for hours
| (when computing time cost 1,000 [1963!] dollars per
| hour). And he needed to run once a week every week.
| So he came to me and that he needed a faster square
| root routine. The built-in one calculated 12 significant
| digits and he only needed about 3, so he figured that
| a tailor-made routine would speed up his program.
| Full of youthful enthusiasm I went to work (this was
| done in machine code - not even assembler) and after
| a week that a routine that made his program run 20 times
| as fast. I was mighty proud.
| Then I did something really dumb: I asked him how his
| program worked, what he used my fancy super-optimized
| hyper-efficient square-root routine for. This is what he told
| me: "I make weather maps. Weather stations that supply
| data are located at irregular places and I need to compute
| the data onto a fixed and regular grid. to do this I compute
| the distances from each grid-intersection point and all the
| weather stations that have data for a particular time. Based
| on the distances I pick the five nearest stations and use a
| complicated interpolation formula to obtain interpolated
| data at the regular gridpoints. You routine makes this sooh
| fast, thank you sooh very much". I was crestfallen.