[lkml]   [2008]   [Nov]   [7]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
Messages in this thread
Subject[RFC patch 00/18] Trace Clock v2

I've cleaned up the LTTng timestamping code, renamed it into "trace clock",
ripped apart the tsc_sync.c x86 code, added documentation (printk to the console
when the tracing clock is used) about what to do when an unsync TSC is detected.
I however kept the cache-line bouncing workaround for now. However, I now
synchronize the counters every jiffy with a per-cpu timer so it gives an upper
bound to the time imprecision.

The trace clock works with a trace_clock_get()/put(), so all the mechanic and
overhead that might be required to provide correct timestamps on weird systems
is *only* enabled when tracing is active.

I plan to stick to this simple solution for now so we can get reliable tracing
for 95ish% of systems out there, and to keep room for improvement (nice NTP-like
schemes) for a later version. (sadly, given this is actually v2, I cannot say
"let's keep that for v2") ;)

This patchset applies on top of 2.6.28-rc3.


Mathieu Desnoyers
OpenPGP key fingerprint: 8CD5 52C3 8E3C 4140 715F BA06 3F25 A8FE 3BAE 9A68

 \ /
  Last update: 2008-11-07 06:49    [W:0.173 / U:7.740 seconds]
©2003-2018 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site