Re: About result of rt/hrt patch, is it reasonable?

[Mark Hounschell <dmarkh@xxxxxxxxxx>]

Lee David wrote:
> On 11/3/06, Mark Hounschell <dmarkh@xxxxxxxxxx> wrote:
>> Well the reason you don't really see much difference between the hrt
>> and the rt
>> is because your not really stressing your machine at the same time you
>> run this.
>> All you are really seeing is the high-res timer capabilities added to
>> your
>> kernel. As you probably know the rt patch does contain the hrt patch
>> set. For a
>> "maybe" better or more stressing test of your system you might see
>>
>> ftp://ftp.compro.net/public/rt-exec
> 
> Hi, Mark
> 
> Do you mean that when under trivial load, rt/hrt will almost behavior
> the same,
> while rt will get significantly better result under high load because of
> its
> latency optimazations?
> 
> Yes I know that hrt already merged into rt patch set, and I did have
> rt-exec
> on my machine, too. I will do some furthur study and thanks for your
> pointing
> out.
> 
> Thanks,
> D.L.

The hrt patch (which is in the rt patch) allows for higher resolution sleep
periods for calls such as clock_nanosecond/nanosecond etc. A non hrt system
making these calls will sleep a minimum of the system HZ value. An hrt system
can sleep for periods with a much higher resolution (micro seconds) accurately.
The cyclictest is using nanosleep and timing the accuracy of the actual sleep
period. The advantage of the rt over the hrt system is that those sleep periods
will stay more accurate in a broader range of system activity.

Mark

--
Kernelnewbies: Help each other learn about the Linux kernel.
Archive:       http://mail.nl.linux.org/kernelnewbies/
FAQ:           http://kernelnewbies.org/faq/