Path: utzoo!censor!geac!torsqnt!news-server.csri.toronto.edu!cs.utexas.edu!usc!zaphod.mps.ohio-state.edu!sol.ctr.columbia.edu!emory!hubcap!retrac
From: retrac@titan.rice.edu (John Carter)
Newsgroups: comp.parallel
Subject: Re: Data Sharing Statistics
Message-ID: <12344@hubcap.clemson.edu>
Date: 19 Dec 90 14:16:43 GMT
References: <12327@hubcap.clemson.edu>
Sender: fpst@hubcap.clemson.edu
Organization: Rice University, Houston
Lines: 92
Approved: parallel@hubcap.clemson.edu

In article <12327@hubcap.clemson.edu> galles@mips.com (Mike Galles) writes:
>For purposes of protocol and design evaluation, the amount of data sharing
>which occurs between various nodes in a shared memory system is a very
>important statistic.  It is also important to know whether data that is
>shared is more often clean or dirty.  After searching for references on 
>this topic, I found only two articles which directly addressed these issues:
>
>"A Characterization of Sharing in Parallel Programs and Its Application
>to Coherency Protocol Evaluation", Eggers & Katz (UCB), 1988 IEEE
>
>"Memory-Reference Characteristics of Multiprocessor Applications under MACH",
>Agarwal & Gupta, 1988 Proceedings of ACM Sigmetrics

The best conference proceedings to scan are probably ASPLOS, ISCA (Sigarch),
and Sigmetrics.  Here are a few more publications that you might want to
take a look at that address the issue to varying degrees:

@CONFERENCE{munin:sigarch90,
	AUTHOR = {John K. Bennett and John B. Carter and Willy Zwaenepoel},
	TITLE = {Adaptive Software Cache Management for
		Distributed Shared Memory Architectures},
	BOOKTITLE = sigarch90,
	PAGES = {125-134},
	MONTH = may,
	YEAR = 1990}

@PHDTHESIS{eggers:thesis,
	AUTHOR = {Susan J. Eggers},
	TITLE = {Simulation Analysis of Data Sharing
		in Shared Memory Multiprocessors},
	SCHOOL = {University of California at Berkeley},
	YEAR = 1989,
	MONTH = mar}

@CONFERENCE{eggers:performance,
	AUTHOR = {Susan J. Eggers and Randy H. Katz},
	TITLE = {The Effect of Sharing on the Cache and
		 Bus Performance of Parallel Programs},
	BOOKTITLE = asplos-3,
	PAGES = {257-270},
	YEAR = 1989,
	MONTH = apr}

@CONFERENCE{atum,
	AUTHOR = {Anant Agarwal and Richard L. Sites and Mark Horowitz},
	TITLE = {{ATUM}: A New Technique for Capturing
			Address Traces Using Microcode},
	BOOKTITLE = sigarch86,
	PAGES = {119-127},
	YEAR = 1986,
	MONTH = jun}

@CONFERENCE{atum:multiprocessor,
	AUTHOR = {Richard L. Sites and Anant Agarwal},
	TITLE = {Multiprocessor Cache Analysis Using {ATUM}},
	BOOKTITLE = sigarch88,
	PAGES = {186-195},
	YEAR = 1988,
	MONTH = jun}

@CONFERENCE{gupta:cache-invalidation,
	AUTHOR = {Wolf-Dietrich Weber and Anoop Gupta},
	TITLE = {Analysis of Cache Invalidation Patterns in Multiprocessors},
	BOOKTITLE = asplos-3,
	PAGES = {243-256},
	YEAR = 1989,
	MONTH = apr}

@CONFERENCE{ibm:memory-access-patterns,
	AUTHOR = {F. Darema-Rogers and G.F. Pfister and K. So},
	TITLE = {Memory Access Patterns of Parallel Scientific Programs},
	BOOKTITLE = sigmetrics87,
	PAGES = {46-58},
	YEAR = 1987,
	MONTH = may}

@TECHREPORT{illinois:mem-ref-behavior,
	AUTHOR = {David J. Lilja and David M. Marcovitz and Pen-Chung Yew},
	TITLE = {Memory Referencing Behavior and a Cache Performance Metric
		in a Shared Memory Multiprocessor},
	INSTITUTION = {University of Illinois},
	NUMBER = {CSRD Rpt. No. 838},
	YEAR = 1989,
	MONTH = jan}

Hope this helps,
John

--
John Carter (retrac@rice.edu)
Computer Systems Laboratory // Department of Computer Science
Rice University