Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!mnetor!seismo!husc6!husc4!hadeishi
From: hadeishi@husc4.HARVARD.EDU (mitsuharu hadeishi)
Newsgroups: comp.sys.amiga
Subject: Re: How fast is AmigaBasic?
Message-ID: <2460@husc6.UUCP>
Date: Sat, 4-Jul-87 12:04:42 EDT
Article-I.D.: husc6.2460
Posted: Sat Jul  4 12:04:42 1987
Date-Received: Sat, 4-Jul-87 20:06:35 EDT
References: <1348@bath63.ux63.bath.ac.uk>
Sender: news@husc6.UUCP
Reply-To: hadeishi@husc4.UUCP (mitsuharu hadeishi)
Distribution: world
Organization: Harvard Univ. Science Center
Lines: 62
Summary: Fast, but there's no comparison

In article <1348@bath63.ux63.bath.ac.uk> pes@ux63.bath.ac.uk (Paul Smee) writes:
. . .
>The reason I ask is that I've seen adverts for a 3rd machine which include
>comparative timings of their own BASIC on their hardware (the machine is
>BASIC-oriented, that is it's intended as the main language) against standard
>Atari ST-Basic and standard Amiga AmigaBasic.  Their comparative timings
>make their machine look like a bat-out-of-hell (well, they would, wouldn't
>they?) and the ST and Amiga look like dead snails.
. . .

	As a weekend programming project about a year ago I decided to
implement a full-blown better-than-Infocom style text adventure parser.
I did this is AmigaBasic because it wasn't really a serious project
and I also wanted to develop in an interactive language (so I could
do major modifications without having to wait for compilations.)  The
program with extensive comments was eventually about 120K long, and
was a very very capable parser (by Zork standards.)  Something you
should know is that the AmigaBasic interpreter does some optimization
AS IT RUNS.  That is, the FIRST time you use some code, it takes some
three to four times longer than it does the NEXT time.  So my parser
would have an initial response time of some twelve seconds or so (clearly
unacceptably slow).  However, after parsing the first sentence response
time would steadily decrease as the parser used more and more of it's
internal routines, and AmigaBasic optimized them.  (The optimization
consists of optimizing GOTO <label> and GOSUB <label> statements.)
After two inputs or so response time would be down to a comfortable
1/2 to three seconds or so.  (Finally what I did was have the parser "go through
the motions" by parsing a dummy sentence so it would optimize before the
program became interactive, so the slow response time did not affect
the user once the interactive session began.)  Now I must emphasize that
this was a pretty complicated parser which would have had an average
of ten to fifteen second response time on one of the old 8-bit machines
(I know because I wrote a simplified parser version of Zork for my old
PET 2001 with 16K when I was a youngster.  It had a response time of
five to six seconds with a much less sophisticated grammar.  Oh, give me
a LISP machine anyday :-)  I considered the final performance of the
machine after optimization to be FAST, based on my experience with
8-bit BASICs.

	It would help if you provided some information as to what
kind of benchmarks were performed.  If they did something sneaky
like run the program once and multiply the results by 100 or something,
then AmigaBasic would perform terribly.  Note that since AmigaBasic
does NOT use line numbers (it treats line numbers as alphanumeric
labels) any program which was directly entered with just a bunch of
line numbers for labels would run very slowly the first time in AmigaBasic.

	So AmigaBasic is fast enough at certain things.  Floating point
is slow, but who would want to do extensive FP in Basic, anyway?
There are no competitors, anyway (except perhaps TrueBASIC, but that's
a compiler if I'm not mistaken.)  But, few really use AmigaBasic,
so there hasn't been much pressure to optimize it further.  It is a
very handy tool for certain limited applications where speed is
not important and interactive development is, and it is essentially
a superset of Microsoft Extended Basic it is quite a large program,
which might also slow it down relative to more simple-minded BASICs.

	Off the cuff, I'd say AmigaBasic is around ten times slower than
the output of a C compiler, plus or minus half an order of magnitude or
so.

				-Mitsu