Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!mnetor!seismo!rutgers!ames!ucbcad!ucbvax!decvax!decwrl!pyramid!prls!mips!earl
From: earl@mips.UUCP (Earl Killian)
Newsgroups: comp.arch,comp.sys.nsc.32k
Subject: Re: "Unoptimizing" Dhrystone
Message-ID: <332@gumby.UUCP>
Date: Wed, 22-Apr-87 18:29:51 EST
Article-I.D.: gumby.332
Posted: Wed Apr 22 18:29:51 1987
Date-Received: Fri, 24-Apr-87 23:42:44 EST
References: <4190@nsc.nsc.com> <951@moscom.UUCP> <2577@intelca.UUCP> <1329@ames.UUCP>
Lines: 30
Xref: mnetor comp.arch:1062 comp.sys.nsc.32k:107

Ehud Reiter and Eugene Miya both have good points.  I recently saw a
paper on the VAX 8700 from DEC and they presented their benchmark
results in a variety of ways, including something like:

TOP CAVEAT: I just typed in imaginary data points for illustration
-- THESE ARE NOT REAL RESULTS.

	4.0	ID
	4.5	IICFFDD
	5.0	IIICFFFDDD
	5.5	IIICCCFFFFDDDD
	6.0	IIIIICCCFFFFFDDD
	6.5	IIIIICCFFDD
	7.0	IIIICFFDD
	7.5	IIICFD
	8.0	I

BOTTOM CAVEAT: I just typed in imaginary data points for illustration
-- THESE ARE NOT REAL RESULTS.

They had something like 120 benchmarks, and presented these in the
above form, where "I" represented an integer benchmark, "C" Cobol,
"F" single-precision floating point, and "D" double-precision floating
point, etc.

I liked this presentation.  It really gives you a feel for how wide
the performance variation can be, even within members of the same
family.

Perhaps someone else can post a reference to this paper.