EPICS Controls Argonne National Laboratory

Experimental Physics and
Industrial Control System

1994  1995  1996  1997  1998  1999  2000  2001  2002  2003  2004  2005  2006  2007  2008  2009  2010  2011  2012  2013  2014  2015  2016  2017  2018  2019  2020  2021  2022  2023  2024  2025  <2026 Index 1994  1995  1996  1997  1998  1999  2000  2001  2002  2003  2004  2005  2006  2007  2008  2009  2010  2011  2012  2013  2014  2015  2016  2017  2018  2019  2020  2021  2022  2023  2024  2025  <2026
<== Date ==> <== Thread ==>

Subject: Re: iocStats CPU time seems wrong
From: Mark Rivers via Tech-talk <tech-talk at aps.anl.gov>
To: tech-talk <tech-talk at aps.anl.gov>
Date: Sun, 31 May 2026 21:14:05 +0000
Folks,

Looking at the code, the IOC_CPU_LOAD is indeed in units of % of all CPUs, not % of a single CPU which is what the label in the OPI screen says.


It creates a new record IOC_SINGLE_CPU_LOAD, and uses this record in the OPI screens to agree with the label.

It also adds new records for memory statistics that are in units of GB, rather than bytes.  These are much easier to read on systems with lots of memory.

Mark



From: Mark Rivers <rivers at cars.uchicago.edu>
Sent: Sunday, May 31, 2026 1:28 PM
To: tech-talk <tech-talk at aps.anl.gov>
Subject: iocStats CPU time seems wrong

Folks,

I have an areaDetector IOC running spinnakerApp that is using ~27% of  CPU on Linux according to "top".  I have attached a screenshot of "top".

The IOC is running iocStats.  I am using the current master branch of iocStats.  The problem is that iocStats reports IOC_CPU_LOAD is 0.8% (see attached screenshot).  The label on the OPI screen says that means % of a single CPU, which is what "top" is also showing.

Why are these so different?  Is IOC_CPU_LOAD really not in units of one CPU, but in units of all CPUs?

Thanks,
Mark


References:
iocStats CPU time seems wrong Mark Rivers via Tech-talk

Navigate by Date:
Prev: iocStats CPU time seems wrong Mark Rivers via Tech-talk
Next: Re: EPICS modulle for Piezosystem Jena's piezo controller NV 40/1 CLE LiangChih Chiang via Tech-talk
Index: 1994  1995  1996  1997  1998  1999  2000  2001  2002  2003  2004  2005  2006  2007  2008  2009  2010  2011  2012  2013  2014  2015  2016  2017  2018  2019  2020  2021  2022  2023  2024  2025  <2026
Navigate by Thread:
Prev: iocStats CPU time seems wrong Mark Rivers via Tech-talk
Next: aiopvxs 0.3.0 Python library release Mike Smith via Tech-talk
Index: 1994  1995  1996  1997  1998  1999  2000  2001  2002  2003  2004  2005  2006  2007  2008  2009  2010  2011  2012  2013  2014  2015  2016  2017  2018  2019  2020  2021  2022  2023  2024  2025  <2026
ANJ, 03 Jun 2026 · Home · News · About · Talk · Base · Modules · Extensions ·
· Distributions · Download · Documents · Links · Licensing ·