Date: Sun, 31 Jan 2010 23:21:37 +0100 From: Pieter de Goeje <pieter@degoeje.nl> To: James Mansion <james@mansionfamily.plus.com> Cc: freebsd-java@freebsd.org Subject: Re: OpenJDK 6/7 kqueue based NIO provider Message-ID: <201001312321.38447.pieter@degoeje.nl> In-Reply-To: <4B65E6D3.2010809@mansionfamily.plus.com> References: <201001301816.16987.pieter@degoeje.nl> <201001311324.54206.pieter@degoeje.nl> <4B65E6D3.2010809@mansionfamily.plus.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On Sunday 31 January 2010 21:23:47 James Mansion wrote: > Pieter de Goeje wrote: > > That is an interesting idea, such an implementation could potentially > > replace all existing back-ends with the exception of the Windows IOCP > > back-end. However libev doesn't seem to provide a non-blocking poll which > > is required for NIO (ev_loop always blocks AFAIK). > > I'll ask Marc. Would have thought that it wouldn't be too hard to > arrance an immediate timeout or to register a known-available fd too > (/dev/null?) I misread the the documentation, it is in fact possible to do a non-blocking poll using the aptly named EVLOOP_NONBLOCK flag :-) I do think that implementing nio using libev is more work than just using kqueue directly, because the libev API is more complex. Especially considering the fact that I now have an almost correctly working kqueue backend here :D. -- Pieter
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?201001312321.38447.pieter>