Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!samsung!munnari.oz.au!goanna!ok
From: ok@goanna.cs.rmit.oz.au (Richard A. O'Keefe)
Newsgroups: comp.arch
Subject: Re: int x int -> long for * (or is it 32x32->64)
Keywords: arithmetic,arbitrary precision,benchmark,modular arithmetic
Message-ID: <3759@goanna.cs.rmit.oz.au>
Date: 14 Sep 90 08:06:03 GMT
References: <3984@bingvaxu.cc.binghamton.edu> <41425@mips.mips.COM> <119977@linus.mitre.org>
Organization: Comp Sci, RMIT, Melbourne, Australia
Lines: 19

I was very surprised by the results of Horsell's benchmark.
However, it is heavily biassed towards small numbers.
I decided to focus on calculating factorials.  My test program
calculates n! for n from 0 to 1000.  I modified Horsell's program
to used a fixed buffer, so that malloc() costs could be eliminated.

					User	System	User ratio
Horsell	(16 x 16 -> 16, C code)		311.2	1.9	1.36 slower
Horsell (32 x 32 -> 32, C code)		228.8	2.6
My code (32 x 32 -> 64, assembly)	 52.6	0.5	4.35 faster

These times were obtained on an Encode MultiMax with NS32532 cpus
running Encore 4.3BSD, Encore's C compiler, time from /bin/time.

I expect that multiplying bignums would also show about a 4x speedup
on this machine from using the (32 x 32 -> 64) instruction.

-- 
Heuer's Law:  Any feature is a bug unless it can be turned off.