>Some quite interesting results are about SMP-Performnace and Andrea's
>patches.
Unfortunately you was using apache not sleeping in accept(2) so the only
improvement you had from my code cames from my VM/buffer redesign and I
think from my rewrote of reschedule_idle() for SMP. The improvement that
cames from my new VM code is clear from the graph where the working set is
larger than memory.
In 2.2.9_andrea2 I was also releasing the kernel lock too many times and
this bit is now fixed in my current tree (that's probably the reason the
my tree was a bit slower under light load).
I think by telling apache to use my 2.2.9 accept-wake-one code
(-DSINGLE_LISTEN_UNSERIALIZED_ACCEPT) + the fix for the kernel lock
releasing (+ my all other VM/buffer/scheduler stuff) we could beat NT a
_lot_ :)). And except my reschedule_idle() rewrite all my other stuff is
all UP stuff that will apply to everybody out there :).
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/