Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!samsung!dali.cs.montana.edu!uakari.primate.wisc.edu!pikes!aspen.craycos.com!dmk
From: dmk@craycos.com (David Keaton)
Newsgroups: comp.arch
Subject: Re: Page size and linkers (was: Re: SunMMU history)
Message-ID: <1991Jan23.171327.701@craycos.com>
Date: 23 Jan 91 17:13:27 GMT
References: <1991Jan19.133914.23871@bellcore.bellcore.com> <PCG.91Jan20191955@teacho.cs.aber.ac.uk> <3981@skye.ed.ac.uk>
Organization: Cray Computer Corporation
Lines: 18

In article <3981@skye.ed.ac.uk> richard@aiai.UUCP (Richard Tobin) writes:
>  An
>obvious improvement would be to have the linker order procedures so that
>procedures commonly used together were adjacent, and separate from
>rarely used procedures.
>
>Has this been done in any real systems?

     Multiflow's linker used the output of gprof to optimally scatter
frequently cooperating procedures in the I-cache.  Sometimes this
required placing them in different pages.  Typical speedup was about
10%, although I remember a couple of cases that achieved 20%+.

     It was a real system for a while.

					David Keaton
					dmk@craycos.com
					uunet!dmk3b1!dmk