I have a vastly improved version I'm about to push out too. Its far more
flexible, is relatively compliant, and gets the fast path serving to
around 6 syscalls or so.
It really wants shared rt queues.
[and david, what workloads have you profiled this stuff under? I'm
definitely interested.
at least under linux I was seeing the kernel spend significant amounts of
time walking the poll lists just to find that a single fd was ready. Given
the tight core of phhttpd it can service an fd quite quickly.. you'd need
gigabit to keep it busy. the trick is to keep _lots_ of 'slow 57.6'
concurrent connections going, not churn through lots of tiny 100mb fetches
over lossless ether. the former is what the real world looks like..]
Zach Brown <zab@redhat.com> -- Software Princess
-
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/