Subject: Re: NEWPIPE: benchmark performance diff microtime(9) vs. =
To: None <eeh@netbsd.org>
From: Alfred Perlstein <bright@mu.org>
List: tech-perform
Date: 10/25/2001 19:47:16
* eeh@netbsd.org <eeh@netbsd.org> [011025 17:36] wrote:
> | Well, the basic idea is that it doesn't seem to make too much sense
> | to loan piece of memory < PAGE_SIZE (Page Loan works by pages). 
> | That's why I used PAGE_SIZE as basic atom. More recent tests
> | show up that the overhead of Page Loan setup seems to still
> | be higher than the gain for i386 page size (4k), so bumping it
> | to higher value is desirable.
> 
> You should check the page size and only use more than one page if
> the page size is 4KB or less.  If you have an 8KB, 16KB, 32KB, 
> or 64KB page size you really don't want to use more pages than you 
> need for a pipe.  It appears arm26 has 32KB pages, so you are using
> 64KB of RAM for each pipe on a small memory machine.

It appears to me that you would also be requiring at least 64k
on ARM to take advantage of the page loaning.  I may be reading
the code wrong though.

-- 
-Alfred Perlstein [alfred@freebsd.org]
'Instead of asking why a piece of software is using "1970s technology,"
 start asking why software is ignoring 30 years of accumulated wisdom.'
                           http://www.morons.org/rants/gpl-harmful.php3