RE: array handling

[Joel Fritz <JFritz@xxxxxxxxxxxxxxxx>]

Ok that answered my question.  A few years ago I did some benchmarking on
setll vs lokup vs RPG binary search vs C binary search in a sorted array
with the ascend keyword.  Both binary searches were faster than lokup or
setll.  Can't remember the numbers, but it was around one order of
magnitude.  I haven't been able to write a binary search in RPG that's
faster than the C function, although the difference isn't dramatic.  That
may be my problem. <g>

> -----Original Message-----
> From: Joe Pluta [mailto:joepluta@PlutaBrothers.com]
> Sent: Monday, December 03, 2001 7:06 AM
> To: midrange-l@midrange.com
> Subject: RE: array handling
>
>
> > -----Original Message-----
> > From: jt
> >
> > But you said ASCEND causes a binary search...?!?  Sheesh...
>  When did THAT
> > happen...?!?  (I'd always thought it still used sequential
> search, and
> > ASCEND just allowed for *LT or *GT type lookups.)  That'd
> be two things,
> > today...:-)
>
> I have to eat a substantial bit of crow here, JT.  I did the
> one thing I
> hate most - I typed in an assumption as fact.  I was told
> this way back in
> the early days of my programming career, and I honestly never
> bothered to
> check it, I assumed it was so because it made such good
> sense.  It made
> sense to me that IBM, with all their knowledge in writing
> compilers, would
> indeed be smart enough to use something as fundamentally
> sound as a binary
> search algorithm, but this turns out not to be true, at least
> from empirical
> evidence.
>
> Modifying my test program to compare the times of a lookup on
> an acending
> array and a lookup on a non-ascending array yielded exactly the same
> results, even after 5,000,000 repetitions.  In my copious
> free time, I might
> try to check the generated MI code, but the actual test shows
> me that the
> code is exactly the same for each one.
>
> Just trying to set the record straight.
>
> _______________________________________________