EPICS Controls Argonne National Laboratory

Experimental Physics and
Industrial Control System

1994  1995  1996  <19971998  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  Index 1994  1995  1996  <19971998  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 
<== Date ==> <== Thread ==>

Subject: EPICS printing speed
From: Tom Himel <[email protected]>
To: [email protected]
Date: Mon, 11 Aug 1997 17:46:57 -0700 (PDT)
At SLAC, we have found the printing from EPICS to be annoyingly slow (12
minutes to print a small DM screen). We have made some measurements with
various hardware and software configurations and believe we understand the
cause of the problem. With this knowledge and a little bit of time and money we
believe we can get the printing time down to 30 seconds for a small DM display
and to 50 seconds for a larger DM display. We still consider these times to be
rather long (our old control system prints any display in 20 seconds) so we
would like to know if anyone has any advice for us before we proceed.

Based on measurements given below (and reports about what the EPICS quickprint
option does) our conclusion is that most of the time is spent by the printer's
CPU processing the bitmap sent to it in postscript format. Hence simpler
postscript files and faster printer CPUs improve the performance. In
particular, having the printer convert a color bitmap to a grayscale bitmap
causes very long printing times. File sizes were small enough that network
transfer time was calculated to be insignificant. We also see that the printer
starts processing a file almost immediately after the print job is submitted.
The other contributor to the total print time is the time taken by the host CPU
to do the screen capture and save it to a disk file in postscript format. This
time was measured and is given in the table below. It is typically less than
(but sometimes 50% of) the time taken by the printer itself.

In the data shown below, the time to print is measured from when carriage
return is hit to send a file to the printer until the page is completely out of
the printer. Typically the printing was done in two manual steps. First a
screen capture was done to disk (this time is in the column labelled "CPU").
Then the file was sent to various printers. All times are given in seconds.
Five printers were tested. LW 12/640 is an Apple Laser Writer. An older laser
writer (M5890) was also tested for some files and was similar to the Imagen.
The HP printers are HP Laser Jets. 


Screen grabber and switches            CPU   file  Imagen   LW   HP4SI HP5SI
                                             size   3320  12/640
-------------------------------------------------------------------------------
         The following are for a small DM display (about 2x6 inches printed)
xwd as implemented in DM                      54k    720            90
DECwindows black+white screen capture         38k     30
DECwindows grayscale screen capture          301k     60            25
xgrabsc (deflt is grayscale,compress)         54k    160            45
xgrabsc -nocompress                          303k                   20
xgrabsc -nocompress -bin                 5   165k     40    50      20    25
xgrabsc -nocompress -bin -nolimit        5   164k     55            20   
xgrabsc             -bin -nolimit             27k                 >180

   The following are for a larger DM display (about 8x10 inches printed)
xgrabsc -nocompress                     30                                45
xgrabsc -nocompress -cps                23                              >150
xgrabsc             -bin                15                              >120
xgrabsc -nocompress -bin                15   740k    110    65      45    35

   The following are for printing a half page ASCII file. They give an idea
   of the time a printer takes for paper handling etc.
                                                            32      17    15

As you can see, sending a color postscript file with a postscript program that
the printer uses to translate it to grayscale (done by xwd and xgrabsc -cps) is
very slow. Compressing makes the file much smaller but slows down the
printer a lot. Sending binary instead of ASCII for the bitmap (-bin) with no
compression is the fastest. The HP printers are faster than the other printers
tested.

Unless someone has a clever idea, we will buy an HP5SI printer for the
control room and modify DM so that its print is like quickprint and does an
xgrabsc -nocompress -bin. We would have it provide the windowID as part of the
invocation so the operator wouldn't have to click on the window he wants to
print. To allow non-DM windows to be printed (e.g. APS stripcharts or MEDM
displays) we would provide a DM print other window option which would use the 
-click switch so the operator could click on the window he wants printed. The
reason for routing all this through DM is that we have used it to select which
printer the output should go to. It certainly would be very nice though to have
a uniform printer interface for all the EPICS displays.

Any comments/clever ideas?

Thanks, 
  Tom Himel  [email protected]

Navigate by Date:
Prev: RE: acc? Jeff Hill
Next: dbStatic: Escape Sequences Marty Kraimer
Index: 1994  1995  1996  <19971998  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 
Navigate by Thread:
Prev: RE: acc? Jeff Hill
Next: RE: EPICS printing speed Jeff Hill
Index: 1994  1995  1996  <19971998  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 
ANJ, 10 Aug 2010 Valid HTML 4.01! · Home · News · About · Base · Modules · Extensions · Distributions · Download ·
· Search · EPICS V4 · IRMIS · Talk · Bugs · Documents · Links · Licensing ·