Hi Ralph,
I try to update you on the status:
Ralph Lange wrote:
Hello Timo, hello all -
Let me try to summarize the current situation (not meant as an insult,
only to synchronize status):
* The APS event support is old and doesn't match the newer timing
systems.
Correct.
* Most installations are using the "soft slave" fallback flavor of
drvTS to keep the EPICS timestamps in sync across IOCs. This mode
does have a number of weak points.
Correct.
* The SLS has been working on an improved replacement for a long
time. Some versions of this module have been operational for a
long time.
It was never published, though. (Outside of Powerpoint slides.)
Correct. However, not to blame but to protect myself: I made in several
occasions the offer
to publish and re-contribute this code to base, but it was not wanted.
Instead we
agreed to work on a re-write, which has taken longer than expected.
* The SNS has developed an improved replacement ("General Time").
It was never published, though. (Outside of Powerpoint slides.)
Correct; however we have received the sources. More about this below.
* DESY has developed an improved replacement.
It was never published, though. (Outside of Powerpoint slides.)
I heard about this in the last EPICS meeting (or I have forgotten about
it, which is also
remotely possible.)
* Last summer (after the EPICS Meeting at the APS), there were talks
and - as far as I know - an agreement was reached on how the
packages (or at least their ideas) can be combined or merged back.
This is true. We have worked on that at PSI. The new solution is based on
the General Time framework; we developed a replacement for the event timing
as an additional driver to GeneralTime. This is in principle ready for
distribution,
but I have not yet made it publicly available. This requires some small
hooks in IOCcore
(no big changes) and provides a (much) better functionality than drvTS.
I try to get
the source tarball sent to Andrew and others who have been involved, for
review
before it is publicly released. We have doen extensive lab testing and
it looks at the
moment really good - but it is really lab testing so far.
* Between 3.14.7 and 3.14.8 the APS event support was taken out of
base and moved into a separate module.
* Until yesterday, an old, dysfunctional version of that module was
available on the APS web site. The newer version should be there
now. (At the moment, it's only a 404.)
Other people beside us have been working on the (new) event support
and (as far as I remember) the reworked version will be put to the APS site.
* As of today, none of the other packages are available on web
servers.
Seems we have practically disabled the "soft slave" support many
installations were using - without replacement.
I would like to see this changed, possibly for the upcoming 3.14.9
release - are there any ideas which way to go?
I agree - this has to change ;-)
I try to get the tarball out asap, maybe still today; we had a
preliminary review of the code,
thought to take a closer look (first time just looked at the concepts)
but anyway,
the code works in the lab pretty well. Some beautification may still be
needed.
* Making the other packages available and see which survives?
* Separating the "soft slave" mode from the APS hardware stuff
(seems that only drvTS and drvTSconfigure are needed) and putting
this subset back into base?
* Any other ideas?
Separating the 'soft' time from drvTS is not a good idea, and (IMHO) not
necessary anymore. The soft
time support is the cause of about hunderd of problems, and the sync
algorithm is really rudimentary.
The code is a mess, and very much vxWorks-loaded. In the new code we
have got
rid of the watchdog "tasks" (for instance) that needed to run on
interrupt level.
My idea and the procedure that was agreed to in the last EPICS meeting
would be to do whatever is necessary to integrate
General Time and distribute it (with the base, because it is a general
facility that anybody will need.)
The only remaining work is to integrate the sources (not a big deal).
And of course, write documentation...
Another related issue: I realized that epicsTimeShow() and date() are
not available in the iocShell under Linux - why?
I think they have been removed earlier as part of the changes in the
time support in 3.14. I do not remember the details.
best regards,
Timo
Cheers,
Ralph
Korhonen Timo wrote:
Hello Ralph,
Ralph Lange wrote:
If the SLS code was public, we would.
I just can't find it anywhere.
Can you orient me a bit...what code do you mean? drvTS ?
We have a new code that is thought to replace drvTS in the (near?)
future,
but it has not yet been tested, let alone used by anybody else than
us (in the lab.)
We thought we can take it into operation after the winter shutdown
(ends next
week), but we have to postpone it because we did not manage to get
all our drivers
(for other stuff that we need on the timing crate) to run on 3.14.8.
We are very
close but decided to postpone to be safe (and to avoid being called
at night...)
If you want to use drvTS and need some code from us, just let me
know. If you want
to try the new timestamp driver (which is a _major_ improvement over
drvTS, as an
advertisement), let me know too.
Our software distribution is not "optimal", to borrow the words of
our director...please
do not tell me anything about the state of our website ;-)
best regards,
Timo
Ralph
Andrew Johnson wrote:
Hi Ben,
Benjamin Franksen wrote:
According to this web page egDefs.h and erDefs.h are supposed to
be part of the module. But they are not contained in the tar ball.
This causes the build to fail (under 3.14.8).
Oops, we don't appear to have put the latest releases on the
website - I've just done that (I guess Marty's assumption was that
everyone else was now using the SLS code). Hopefully version 3-2
will fix at least one if not both of your timing problems.
- Andrew
--
Timo Korhonen PSI (Paul Scherrer Institut, http://www.psi.ch)
CH-5232 Villigen PSI
tel + 41- 56 3103262 fax + 41 - 56 310 3383
e-mail: [email protected]
- References:
- Re: drvTS Ralph Lange
- Navigate by Date:
- Prev:
Re: drvTS Ralph Lange
- Next:
Re: vxDevWriteProbe does read Andrew Johnson
- Index:
2002
2003
2004
2005
2006
<2007>
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019
2020
2021
2022
2023
2024
- Navigate by Thread:
- Prev:
Re: drvTS Ralph Lange
- Next:
Re: drvTS - where do you stand on this? Andrew Johnson
- Index:
2002
2003
2004
2005
2006
<2007>
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019
2020
2021
2022
2023
2024
|