From mboxrd@z Thu Jan  1 00:00:00 1970
Date: Wed, 18 Nov 1998 22:38:25 -0600
From: Paul Borman prb@bsdi.com
Subject: [9fans] map and unmap pages of memory
Topicbox-Message-UUID: 8af35c40-eac8-11e9-9e20-41e7f4b1d025
Message-ID: <19981119043825.7SSkPDG0w5ZiLoMPmkmUiJXu_n0mR_5_skGZcmWyD9w@z>

Please be careful on how you read the numbers from hbench-OS.  At least
the networking numbers are probably not measuring what you think.  The
benchmarks are flawed in several aspects and actually measure different
things on different OS's.  I have not examined their benchmarks of map
and unmap, however, if they are as simplistic as the networking benchmarks
they they are probably measuring something that is related, but not quite
what you expect.

				-Paul Borman
				 prb@bsdi.com

PS:  A good benchmark is often hard to write


> From: "Russ Cox" <rsc@plan9.bell-labs.com>
> Sender: owner-9fans@cse.psu.edu
> Date: Wed, 18 Nov 1998 18:28:45 -0500
> Subject: Re: [9fans] map and unmap pages of memory
> To: 9fans@cse.psu.edu
> 
> That test from hBench-OS is measuring the speed of the 
> memory-mapped file calls, which don't exist in Plan 9.
...
> I'ld be interested to hear what numbers you got
> out of hBench under Plan 9 on a particular machine as compared
> with that same exact machine running say Linux or one of
> the BSDs.
> 
> Russ