Hi,
I am using an open-source application on ix86 to perform a task which
is cache-intensive. When I run consecutive iterations of the task on
a fixed input, the variance in timing of each iteration is extemely
high. Needless to say, the test machine is always non-occupied.
On every other OS I tried, Solaris, HPUX, FreeBSD and Tru64, the
timing is very consistent between the iterations. My question is, are
there known issues with L2 cache reuse in Linux kernel?
I can provide any necessary information for anyone interested in
addressing this issue, but I purposely skipped most technical details
in this post to keep it simple.
Thanks in advance.
Igor
This kind of problem could be solved by cache coloring. There should be some
patch available for Linux. I don't know if the other OSes you tried
implement cache coloring, though.
Jun
> -----Original Message-----
> From: Igor Schein [mailto:[email protected]]
> Sent: Tuesday, December 03, 2002 6:09 PM
> To: [email protected]
> Subject: performance of cache-intensive applications
>
> Hi,
>
> I am using an open-source application on ix86 to perform a task which
> is cache-intensive. When I run consecutive iterations of the task on
> a fixed input, the variance in timing of each iteration is extemely
> high. Needless to say, the test machine is always non-occupied.
>
> On every other OS I tried, Solaris, HPUX, FreeBSD and Tru64, the
> timing is very consistent between the iterations. My question is, are
> there known issues with L2 cache reuse in Linux kernel?
>
> I can provide any necessary information for anyone interested in
> addressing this issue, but I purposely skipped most technical details
> in this post to keep it simple.
>
> Thanks in advance.
>
> Igor
>
> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to [email protected]
> More majordomo info at http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at http://www.tux.org/lkml/
On 04-Dec-2002 Igor Schein wrote:
> Hi,
>
> I am using an open-source application on ix86 to perform a task which
> is cache-intensive. When I run consecutive iterations of the task on
> a fixed input, the variance in timing of each iteration is extemely
> high. Needless to say, the test machine is always non-occupied.
CPU bound or I/O bound ? What kernel version ? IDE or
SCSI ? vmstat ? Other details ?
Bye.