It's baby steps in my mind. I left the ability to use select over poll,
currently, was because poll causes a tiny bit higher load for smaller
working sets. I didn't go directly to kqueue and epoll because I didn't
want to have to implement freebsd's version.
I'll rework commit messages a bit
On Fri, 4 Mar 2016, Donald Sharp wrote:
Having the ability to scale beyond 1k fd's by using poll
is a good thing.
Oh, ack.
Also write code to setup for using a library like libev where we can
use a more efficient os agnostic selection mechanism.
On an abstract point, I suspect it'd be
On Fri, 4 Mar 2016, Donald Sharp wrote:
This patch originated w/ Hannes Hofer .
I've taken the patch fixed some bugs and reworked the code
to allow both poll and select to be choosen at compile time.
Couple of comments (with one comment effectively nested inside the
On Fri, 4 Mar 2016, Donald Sharp wrote:
Both the read and write functions used the same code
slightly modified for reading and writing. Combine this
code together.
Ack, with one request:
/* Add new read thread. */
struct thread *
-funcname_thread_add_read (struct thread_master *m,