Re: A different performance measure

Hi,

Dave Hudson <dave@humbug.demon.co.uk> writes:
>David Koogler wrote:
>>
>> The new numbers look a lot more reasonable! I just could not see how a context
>> switch could take 100uS, except maybe on a 6-Mhz 80286.
>
>As I just mentioned in my reply to Mike there's toom for improvement yet,
>but I think we have to keep a sense of proportion. When Mike tested a
>kernel for me last week on a 16MHz 386sx he saw a 1.4ms switch time drop to
>about 750us, but only improved throughput by 5-10% on file I/O.

Note that this test measured the time to do sequential I/O to a file on
a DOS filesystem. For this test, the time it takes to do an I/O is
dominated by overheads other than the context switch time (eg.,
overheads associated with a relatively small I/O size). Still, it was
nice to see a I/O performance improvement by just dropping in a
new kernel (easy for me to say).

Given a more highly tuned I/O system, however, the context switch time
would likely take up a larger percentage of the total I/O time. In
this case, improvements to context switch time would have a greater
impact on I/O performance.

                                        Mike Larson
Received on Sat Jan 21 12:28:46 1995