Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!sun-barr!olivea!decwrl!sgi!shinobu!odin!flobb4!micah
From: micah@flobb4.csd.sgi.com (Micah Altman)
Newsgroups: comp.sys.sgi
Subject: Re: Use of malloc with multiprocessing code
Message-ID: <1991Feb8.201903.27756@odin.corp.sgi.com>
Date: 8 Feb 91 20:19:03 GMT
References: <1991Feb7.192217.11811@portia.Stanford.EDU>
Sender: news@odin.corp.sgi.com (Net News)
Organization: Silicon Graphics, Inc.  Mountain View, CA
Lines: 24

In <1991Feb7.192217.11811@portia.Stanford.EDU> dhinds@portia.Stanford.EDU (David Hinds) writes:

>I want to use the fast malloc, but I don't need the semaphoring stuff
>in the arena calls.  Should I link with -lmalloc and -lmpc?  When I do
>this, I get linker warnings about duplicate definitions of the malloc
>functions.

If you link with both you will only get one. Instead, link with -lmpc,
and use the usconfig(3P) call to turn off semaphoring in your malloc calls.
(use the CONF_STHREADMALLOCOFF ). Do this call before you create your
parallel threads. If you do this, you must make sure that you aren't
making calls to malloc in parallel.

B.T.W. - An additional difference between -lmalloc and -lmpc malloc calls 
is that mallocs made with when linked with -lmpc are quad-word aligned
rather than double word alligned. This can be important for cache line alignment
and graphics DMA transfers.

 Hope this helps.
--
	"Entia non sunt multiplicanda sine necessitate." - William of Ockham

	Micah Altman				micah@csd.sgi.com
	"Computational Juggler"			Phone (415) 335-1866