It was not my profile, I only forwarded it to Benno.
It was found on the list, as "Excessive System time in MSP kernel."
And it was run on a 512 MB Dual Pentium(!) system.
[Note: maximum memory on a (well, two) slow Pentium, but the timing
for each function should be the same]
When I reported it I was running in Win95 I could not check out
man readprofile (and I can not now either...)
I noted that I was guessing a little but the data did look
like it was number of calls, label, average time (look at
total line)... [I am in a process of moving all my lists
from Win95 to Linux, but I am waiting for my ISP to add
some mail accounts...]
I have a patch of my own, not related to data in the above
thread, that measures the time from setting of
current->need_resched to the actual (re)scheduling takes place.
It also samples instruction pointer at jiffies and prints data in
schedule() if latency is longer than 5 ms (i.e. definable)
- and that happens a lot!
[I have kept it private group since it is my first patch and
I do not think that it is SMP safe... or safe in any other
way...]
And the readprofile from the forwarded thread "matches" the
data I get...
There are stuff that takes several jiffies before it
reschedules... (PPro 180, 96MB)
Linus Torvalds says later in the tread:
> Also, the si_meminfo() etc stuff is just ridiculous. It's not a question
> of latency: it's a question of CPU usage. We need to just get rid of those
> functions instead of hacking around them - regardless of whether you add
> "reschedule" calls in them, they just eat too much CPU, plain and simple.
> Again, please don't treat the symptoms - I will not accept patches that
> just say "oh, this is crap, so let's reschedule a bit here". They need to
> be fixed properly or not at all.
I fully agree!
But Ingos patches are a nice start - they show that it can be done!
/RogerL
Benno Senoner wrote:
>
> Hi folks, again news
>
> as Roger Larrson suspected, there are some parts in the kernel
> which have a too long execution path:
> On his PPro with 512MB RAM, d_lookup takes up to 80ms to execute !
>
> (I put Roger's latency profiling patch on my page)
>
--The Internet interprets Windows as damage, and routes around it.
Roger Larsson Skellefteċ Sweden
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/