Subject: Re: HEADS UP: timecounters (branch simonb-timecounters) merged into
To: Mike Pumford <firstname.lastname@example.org>
From: Frank Kardel <email@example.com>
Date: 09/10/2006 15:34:16
Mike Pumford wrote:
> Frank Kardel wrote:
>> Mike Pumford wrote:
>> Congrats, pls run t.c as an additional test to test that time does
>> not ever
>> go backwards. You should *not* see any *negative* diff values from t.c
>> positive ones are usually ok due to scheduling.
> Seems to pass the no negative diff values test. The test program
> consistently complains about a positive difference. I've attached a
> couple of lines of output:
> bad time TSA: 0x4504123e.05781845, TSB: 0x4504123e.0610d5ea, diff =
> 10010021 nsec, 20000040 nsec
> bad time TSA: 0x4504123e.06a94761, TSB: 0x4504123e.07420506, diff =
> 10010021 nsec, 20000540 nsec
> Is this within tolerance or have I got some tweaking to do. The system
> is a 200MHz StrongARM system with a fairly slow DRAM memory bus.
That's fine. no worries.
There is a regression test now you can just run to check the
monotonicity of the CLOCK_REALTIME implementation.
The test is found in regress/sys/kern/time - just run make regress.