Re: tcpbe...@libevent

2010-10-27 Thread Christiano F. Haesbaert
On 26 October 2010 17:36, Damien Miller wrote: > Hi Christiano, > > Thanks very much for doing this. Unfortunately I won't have time to look > at your patch for a couple more weeks, but I certainly have no philosophical > objections to using libevent. > Cool ! I'll run some more tests, as soon as

Re: tcpbe...@libevent

2010-10-26 Thread Damien Miller
Hi Christiano, Thanks very much for doing this. Unfortunately I won't have time to look at your patch for a couple more weeks, but I certainly have no philosophical objections to using libevent. -d On Thu, 14 Oct 2010, Christiano F. Haesbaert wrote: > Sorry, wrong diff forgot the bogus SO_RDOMA

Re: tcpbe...@libevent

2010-10-20 Thread Christiano F. Haesbaert
Sorry to bother, any feedback on this ?

Re: tcpbe...@libevent

2010-10-13 Thread Christiano F. Haesbaert
Sorry, wrong diff forgot the bogus SO_RDOMAIN define (old base). I've just remembered I've changed the -r option from miliseconds to seconds I'll fix that and send another patch with miliseconds, the rest should be ok. Index: tcpbench.c

tcpbe...@libevent

2010-10-13 Thread Christiano F. Haesbaert
Hi, I've rewritten tcpbench to use libevent as suggested before. Code got a little smaller but I couldn't notice any performance changes (tested up to 1500 fds). There is still only one O(n) operation each second: the stats accounting. Turned some variables into globals for the sake of clarity, as