Re: [PATCH] af_unix: unix_write_space() use keyed wakeups

[Alban Crequy <alban.crequy@collabora.co.uk>]

Le Sat, 30 Oct 2010 08:44:44 +0200,
Eric Dumazet <eric.dumazet@gmail.com> a écrit :

> We still loop on 800 items, on each wake_up_interruptible_sync_poll()
> call, so maybe we want to optimize this later, adding a global key,
> ORing all items keys. I dont think its worth the added complexity,
> given the biased usage of your program (800 'listeners' to one
> event). Is it a real life scenario ?

Pauli Nieminen told me about his performance problem in select() so I
wrote the test program but I don't know what exactly is the real life
scenario.

> [PATCH] af_unix: use keyed wakeups
> [PATCH] af_unix: optimize unix_dgram_poll()

Your 2 patches are good for me. In my opinion the improved performances
are good enough with your 2 patches, so no need to add more complexity
unless we discover new problems.

I am preparing patches to implement multicast features on Unix
datagram+seqpacket sockets and my patches could potentially make things
worse in unix_dgram_poll() because it would need to check the receiving
queues of all multicast members. So I want unix_dgram_poll() to be fast
in the first place before proposing other changes for multicast.