RE: Kernel Single-user Memory usage breakdown results and tools posted in the NetBSD WIKI

["Juha Keski-Saari" <Juha.Keski-Saari%teleca.com@localhost>]

Hello Mindaugas

The purpose of our analysis pass was mainly to profile the actors in the
kernel which reserve memory, I believe what you describe would be more
an analysis of the results of those allocation calls depending on the
allocator used. This is something that could be done by utilizing the
page counter method while performing allocations which would bring out
the characteristics of the specific implementation used. 

It's an interesting idea, I'll see what kind of results I get and report
back to you

Best regards,
JUHA KESKI-SAARI

-----Original Message-----
From: Mindaugas Rasiukevicius [mailto:rmind%NetBSD.org@localhost] 
Sent: 19. marraskuuta 2008 0:03
To: Juha Keski-Saari
Cc: tech-kern%netbsd.org@localhost
Subject: Re: Kernel Single-user Memory usage breakdown results and tools
posted in the NetBSD WIKI

Hello,

"Juha Keski-Saari" <juha.keski-saari%teleca.com@localhost> wrote:
> 
> ...
> 
> But to the point, you can find the results and a brief description of
> our methods along with source code behind the following link
> http://wiki.netbsd.se/NetBSD_Kernel_memory_allocation_breakdown

Interesting data. Have you tried measuring kmem_alloc/kmem_zalloc?
In essence, kmem(9) is a separate allocator based on vmem(9), which
is replacing old malloc(9).

-- 
Best regards,
Mindaugas