[Xcb] [RFC PATCH] New XCB socket handoff mechanism for Xlib/XCB and other libraries

James Cloos cloos at jhcloos.com
Mon Mar 17 06:00:57 PDT 2008

I’m running with this now, and it seems to work well.

When I first started X after recompiling libxcb and libX11 I noticed
considerable temporal ‘graininess’.  With SeaMonkey and icewm’s taksbar
both updating there were frequent pauses.  But I started X just then
while another emerge was running in a vc; without that I do not see
those pauses.  I thought that a vc on an fb would not impact the video
card when it is not fore, but perhaps it does.

This morning everything seems smooth again.  Even with an emerge running
in a pty.

The X process has used 2½% of the cpu since it started; that is a bit
less than it had been using of late.  (The recent EXA changes in the
server and/or the r100 driver have substantially increased the server’s
CPU usage....)

So, subjectively speaking, I’d say that the peformace on a consistant
workload (ie nothing running in a background vc) is at least no worse,
and might be a bit better.

I’ve incorporated the patches to the libxcb and libX11 ebuilds in the
overlay at:


and will update those with any new patch series as they are posted.

James Cloos <cloos at jhcloos.com>         OpenPGP: 1024D/ED7DAEA6

More information about the Xcb mailing list