[Gllug] Writing my own system call

[Adrian McMenamin]

On Fri, June 17, 2011 5:35 pm, John Hearns wrote:
> On 17 June 2011 17:22, Adrian McMenamin <adrian at mcmen.demon.co.uk> wrote:
>>> He also said "I am researching/confirming phases of memory locality in
>>> the kernel."
>>
>> It is an MSc project.
>>
>> By hacking stuff in user space I can see where the virtual (user space)
>> memory is coming from, but I want to examine what is going on with
>> physical addresses too.
>>
>> Essentially operating system/VMM theory is based on the idea that in a
>> given (short) time delta allocations will show locality. But at a bigger
>> range of time that locality may move - perhaps gradually, though more
>> likely (at least this is the theory/past experimental conclusion) in
>> rather clearer phases.
>
>
> Adrian, thankyou for the reply. That's very interesting.
>
> In response to the person who got out of bed on the wrong side this
> morning,
> I happen to work in high performance computing. Arranging processes to
> run on particular CPUs
> and locality of memory to the job is important.

Actually, I am examining it in terms of "low performance comouting" - ie
thrashing. Wondering if a more aggressive reaping of pages - based on the
sort of working set approach used in VMS (and then in Windows NT) might
have benefits.

It's an MSc rather than a PhD so I cannot claim exactly to be about to
push the boundaries of human knowledge, but as you say its interesting.

--
Gllug mailing list  -  Gllug at gllug.org.uk
http://lists.gllug.org.uk/mailman/listinfo/gllug