Path: utzoo!attcan!uunet!aplcen!haven!adm!smoke!gwyn
From: gwyn@smoke.BRL.MIL (Doug Gwyn)
Newsgroups: comp.sys.apple2
Subject: Re: Speed (long), was: Re: A low blow from Apple
Message-ID: <13967@smoke.BRL.MIL>
Date: 30 Sep 90 01:39:19 GMT
References: <MayeQd600Vp015aV50@andrew.cmu.edu> <13961@smoke.BRL.MIL> <7324@darkstar.ucsc.edu>
Organization: U.S. Army Ballistic Research Laboratory, APG, MD.
Lines: 33

In article <7324@darkstar.ucsc.edu> unknown@ucscb.UCSC.EDU (The Unknown User) writes:
>In article <13961@smoke.BRL.MIL> gwyn@smoke.BRL.MIL (Doug Gwyn) writes:
>>Any SANE-using application will thus automatically benefit from the FPE.
>	Just wondering if people knew of any GS programs that -DO- use SANE.

Yes!  In fact, in experimenting with the effect of the FPE, I found that
SANE was sometimes used in applications where you might guess otherwise.
For sure it's used for floating-point applications compiled (for example
by ORCA/C) without enabling the special FPE support.

One of the benchmarks on the version of FPE software that I received was
a Mandelbrot-set image generator.  Because the computational kernel of
the program is, essentially,

	for each pixel(i,j):
		cx = min_x + i * step
		cy = min_y + j * step
		zx = zy = 0
		iter = 0
		repeat:
			x = zx * zx - zy * zy + cx
			zy = 2.0 * zx * zy + cy
			zx = x
			iter = iter + 1
		until zx * zx + zy * zy >= thresh or iter >= max_iters
		assign color[iter] to pixel(i,j)

there is obviously a LOT of floating-point computation involved in
producing the whole image (max_iters is typically 1000).  Thus this
program provides a nice practical demonstration of the effects of the
FPE.  I don't recall the exact timings, but the in-line FPE use
reduced the running time from something like an hour to something like
less than a minute; the improvement was quite dramatic.