>Hi all, (especially Andrea)
Hi Philip ;)
> I've been doing some more SPECWeb96 tests, and with Andrea's
>patch to 2.2.8 (ftp://e-mind.com/pub/andrea/kernel/2.2.8_andrea1.bz)
>
[snipped the cool piece ;)]
>**Please, put the wakeone patch into the 2.2.X kernel if it isn't already. **
Well in my patch I changed many things (starting from reimplementing the
SMP rescheudle_idle, a good rewrite of buffer.c and my VM,
etcc.etc..etc..), there's not only the wake-one thing (that really right
now is a wake-three as pointed out by Dave ;).
I would be interesting if you would repeat the bench with apache compiled
to sleep in flock(2) (not in accept(2)) so the wake-one part of my patch
won't improve performances at all (since only one task at once will sleep
in accept(2); that way all other tasks will be waken-all as before), so
we'll get the figures about how much my wake-one approch (or better
wake-three) made differences in bench results.
The only place where I specified TASK_WAKE_ONE as bit-state was in the
accept(2) sleep path (grep the patch and you'll see).
If my little wake-three patch for accept(2) made the difference, then
merging it in 2.2.x would be trivial since it impacts really a few bits of
code.
Many thanks for the interesting informations!
Andrea Arcangeli
-
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/