Hi again!

Sorry that I can't post replies under the right thread, yet.  The IT people are
 still working on things...

I ran the tests Greg suggested on a simplified network (the same one I posted
to opendx-users at
http://opendx.npaci.edu/mail/opendx-users/1999.11/msg00030.html).  Ran them
under both Irix6.5 and WinNT, for a network with the Image module and one
without Image (using Print instead).  Here are the results I found for "memory
in use" of successive runs. For each "run", the sequencer iterated through 50
steps, so these memory increases are somewhat cumulative.

SGI Irix 6.5
--------------------
w/ image:  1012656, 1392016, 1771424, 2150848, 2530240  (increase of ~379k per
run)
w/ print: 486672, 486112, 486112, 486112, 486112  (stable)

WinNT
-----------------
w/ image: 2885560, 3118632, 3137672, 3156296, 3174920  (increase of ~19k per
run)
w/ print: 1179600, 1276640, 1279472, 1282096, 1284720  (increase of ~3k per
run)

I don't know why the WinNT version w/o image is not running stable... I
could've sworn that yesterday when I was looking at this, both versions without
 image had no memory increases.  Hmmm... Anyway, the SGI version seems to
clearly indicate that something's going on when you add the image module.

Another interesting thing I noted while looking at the original (not this
slimmed-down) network: Using hardware rendering, the memory increase was
~400-500k per time step, but with software rendering the increase was only
~100k per time step.  Perhaps irrelevant, but interesting.  And I did have the
cache=0 option triggered on, although that's only supposed to affect
software-rendering according to the User's Guide.

Thanks!
  Jeremy Zoss
  Southwest Research Institute
  (210)522-3089
  [EMAIL PROTECTED]

Reply via email to