Experimental Physics and
Industrial Control System

"Jeff Hill" <[email protected]> · Tue, 22 Apr 2003 10:54:00 -0600

> Because  you mention this,  I wonder  what is the  finest clock
> resolution of  epicsTimerStartDelay() that I can get.  Can I  
> employ this technique to get  1 ms delay  out of 
> epicsTimerStartDelay() ?  If so, could you
> please explain further  about the how-to ?

There are two types of timer queues - active and passive. 

The active queues are general purpose. They are scheduled by an
independent asynchronous thread with instantaneous accuracy
related to epicsThreadSleepQuantum() and average accuracy closer
to, but not, zero. The code epicsTimerTest can be used to
determine how well we are doing on a particular system. On
vxWorks epicsThreadSleepQuantum() typically returns 1/60th of a
second (as you no doubt expect).

The passive queues are application specific. They can be
scheduled any way that you choose. If you have specialized
hardware with time delay based interrupts then that, and the
interrupt latency in the OS, might determine the accuracy of your
implementation of a timer queue.

> when  will R3.14.2  be available ?
> 

RSN (real soon now)

There are only a few remaining issues, but as soon as I knock one
off another one seems to get added to the list.

Jeff

> -----Original Message-----
> From: [email protected] [mailto:[email protected]] On
> Behalf Of Kate Feng
> Sent: Tuesday, April 22, 2003 10:06 AM
> To: Jeff Hill; [email protected]
> Cc: [email protected]
> Subject: Re: epicsTicksPerSecond for OSI ?
> 
> Hi Jeff,
> 
> Jeff Hill wrote:
> 
> > Hi Kate,
> >
> > There are certainly easy ways to compute the difference
> between
> > two EPICS time stamps in seconds and you can also convert an
> > EPICS time stamp to an enhanced struct tm (ANSI C) which
> includes
> > the nanoseconds within a second in addition to the
> conventional
> > date, hour, and second fields. For example, the following
> code
> > measures the delay required to do some computationally
> intensive
> > task and also computes the current date and time.
> >
> > epicsTime begin = epicsTime::getCurrent();
> > .
> > . something computationally intensive
> > .
> > // compute the delay
> > double delayInSeconds = epicsTime::getCurrent() - begin;
> >
> > // compute the date
> > local_tm_nano_sec localDAteAndTime = epicsTime::getCurrent();
> >
> 
> Because  you mention this,  I wonder  what is the  finest clock
> resolution
> of  epicsTimerStartDelay() that I can get.  Can I  employ this
> technique
> to
> get  1 ms delay  out of   epicsTimerStartDelay() ?  If so,
> could you
> please
> explain further  about the how-to ?
> 
> >
> > However, perhaps what you are after is thread sleep
> scheduling
> > accuracy? If so, then you may be pleased with the addition of
> > "double epicsThreadSleepQuantum(void)" to the epicsThread API
> in
> > the next release of R3.14. This function returns the minimum
> > sleep interval obtainable from the local OS and has been very
> > useful for improving the accuracy of timers on average.
> >
> 
> Yes,  epicsThreadSleepQuantum  seems to be exactly what I
> wanted.
> I guess I can get around without it for a while.
> However, when  will R3.14.2  be available ?
> 
> 
> Many thanks,
> Kate
> 
> 
> 

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System