Re: [Xenomai-help] Negative Latency results

[Gilles Chanteperdrix <gilles.chanteperdrix@xenomai.org>]

Christoph Permes wrote:
> Am Donnerstag, den 27.08.2009, 14:05 +0200 schrieb Gilles Chanteperdrix:
>> Christoph Permes wrote:
>>> Hi,
>>>
>>> I'm working together with Peter on the latency problem.
>>>
>>> I have attached our kernel configuration and the boot log to this mail.
>>>
>>> We have tried out configuring a latency between 0 and 10000 but this
>>> didn't help.
>> Ok. Could you try the following patch:
>>
>> diff --git a/include/asm-generic/bits/timeconv.h
>> b/include/asm-generic/bits/timeconv.h
>> index 9a89fe9..654f17f 100644
>> --- a/include/asm-generic/bits/timeconv.h
>> +++ b/include/asm-generic/bits/timeconv.h
>> @@ -37,6 +37,11 @@ long long xnarch_tsc_to_ns_rounded(long long ticks)
>>         unsigned int shift = tsc_shift - 1;
>>         return (xnarch_llmulshft(ticks, tsc_scale, shift) + 1) / 2;
>>  }
>> +
>> +long long xnarch_ns_to_tsc(long long ns)
>> +{
>> +       return xnarch_llimd(ns, 1 << tsc_shift, tsc_scale);
>> +}
>>  #else  /* !XNARCH_HAVE_LLMULSHFT */
>>  long long xnarch_tsc_to_ns(long long ticks)
>>  {
>> @@ -46,12 +51,12 @@ long long xnarch_tsc_to_ns_rounded(long long ticks)
>>  {
>>         return (xnarch_llimd(ticks, 1000000000, cpufreq/2) + 1) / 2;
>>  }
>> -#endif /* !XNARCH_HAVE_LLMULSHFT */
>>
>>  long long xnarch_ns_to_tsc(long long ns)
>>  {
>>         return xnarch_llimd(ns, cpufreq, 1000000000);
>>  }
>> +#endif /* !XNARCH_HAVE_LLMULSHFT */
>>
>>  static inline void xnarch_init_timeconv(unsigned long long freq)
>>  {
>>
> 
> Now the latency is positive:
> 
> RTT|  00:01:46  (periodic user-mode task, 100 us period, priority 99)
> RTH|-----lat min|-----lat avg|-----lat max|-overrun|----lat best|---lat worst
> RTD|       2.021|       9.626|      28.016|       0|       1.564|      29.206

Much better.

-- 
					    Gilles.