Experimental Physics and
Industrial Control System

"Williams Jr., Ernest L." <[email protected]> · Thu, 12 May 2011 12:08:44 -0700

Sent from my iPhone

On May 12, 2011, at 12:00, "Till Straumann" <[email protected]> wrote:

> On 05/12/2011 01:42 PM, Eric Norum wrote:
>> On May 12, 2011, at 11:38 AM, Ernest L. Williams Jr. wrote:
>> 
>>> Mark Rivers wrote:
>>>>> 1ms period should be fine but I haven't
>>>>> checked the least time period that epicseventwait can use. Maybe Mark has better idea?
>>>>> 
>>>> 
>>>> That depends on the OS and how it is configured.  With vxWorks it is
>>>> determined by the value of sysClkRateGet/Set.  The default clock period
>>>> on vxWorks is normally 50/60Hz, but it can easily be increased with
>>>> sysClkRateSet().  If you are using RTEMS I don't know what the rate of
>>>> the system clock is.  Also, typically if you specify a time of X for
>>>> epicsEventWait the actual time will be X +- epicsThreadSleepQuantum(),
>>>> which is the system clock period.  In other words if I ask for 16.6 ms
>>>> on a standard 60Hz vxWorks system I will get 0 to 16.6 ms, because the
>>>> next system tick can happen immediately (=0) or we may have just missed
>>>> one and it will happen in 16.6ms.
>>>> 
>>> 
>>> Hi Mark,
>>> 
>>> Awesome, thanks for the info.
>>> 
>>> Till and Eric can you weigh in on this question for RTEMS?
> 
> Basically, the behavior is as described by Mark (i.e., the 
> granularity/resolution of a blocking operation or a sleep/delay
> is determined by the OS system clock).
> 
> Our GeSys images at SLAC are configured the following way:
> 
>    4.9.3/rtems_p1/ssrlApps_p2
>    4.9.4/rtems_p0/ssrlApps_p3
>    4.9.4/rtems_p1/ssrlApps_p3
>    4.9.4/rtems_p1/ssrlApps_p4
> 
> (and upcoming, later releases) use a system clock of 1kHz
> by default (so that we can block for less than a 360Hz
> fiducial). Only slower hardware (mvme167, uC5282, virtex,
> rce405) uses a lower clock rate of 100Hz.
> 
> All older GeSys versions (not listed above) use a clock
> rate of 50Hz on all platforms.

Awesome, thanks for the info here :)
Cheers,
Ernest

> 
> HTH
> -- Till
> 
> PS: The increase of clock speed from the old 50Hz
> is documented in the ChangeLogs.
> 
>> 
>>> From<base>/src/RTEMS/base/rtems_config.c
>> 
>> #define CONFIGURE_MICROSECONDS_PER_TICK 20000
>> 
>> If you want a different value you can copy rtems_config.c to your application source directory, add it to the Makefile and make any changes you wish.  Note that this won't work for those of you dynamically loading your EPICS applications.
>> 
> 

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System