|
About ::
TODO ::
Blog ::
RSS ::
Old blog ::
Projects ::
GIT ::
Gallery ::
Notes
Mon, 08 Jan 2007
New kevent
aio_sendfile() is implemented as described
here,
except destructor callbacks,
which are essentially the same, except that they reschedule AIO processing.Benchmarks of simple sendfile() vs. aio_sendfile() did not show noticeble
win of any approach, but I want to notice, that my receiving side is
not that best in my case (I managed to create a test where usual
sendfile() stuck until signal received, the same hapend with aio_sendfile()
too, but the latter can be buggy).It would be good to use it in lighttpd (I will create a patch after some feedback received and approach will be considered as good. AIO state machine is a base for network AIO (which becomes quite trivial), but I will not start implementation until roadback of kevent as a whole and AIO implementation become more clear. Patches and userspace test application are available in archive from kevent homepage. /devel/kevent :: Link / Comments () |