statistics interpretation

New Message Reply Date view Thread view Subject view Author view

ceti (ceti++at++worldnet.net)
Thu, 13 Jun 1996 08:52:20 +0100

>I am having trouble interpreting some statistics information
>displayed by the Performer stats and would like to get some help in
>understanding the data. I am currently using Performer 2.0, IRIX 5.3
>on a 3 pipe ONYX.
>
>My questions are focused on the average frame time information and
>the missed frame information. I am getting information of the form:
>
>msec: Frame 70.2 app=13.2 cull=2.7 draw=6.9
>Frames 47 misses: total 13 app=0 cull=0 draw=0
>
>The program is running at 30 Hz LIMIT or LOCK, APP_CULL_DRAW mode.
>
>I have the following questions:
>
>1. I know that the frame information is averaged over a 2 sec
>period. However, why is the total frame time so large when the app,
>cull and draw are reasonable? The timing graph does not show any
>jumps in any of the processes. How is the frame time calculated ?
>None of this information is currently described in the performer
>manuals.
>
>2. The sum of the frames and total misses should equal the number of
>frames over 2 seconds (60 frames), which the data does do. If the
>total misses is non-zero, but the app,cull,draw misses are zero, where
>are the frames being blown ? Again, this information is not
>described in the performer manuals.
>
>3. If the program is running in LIMIT or LOCK and the average frame
>time is around 70 msec, why isn't the frame rate being reduced from
>30 Hz to a lower frame rate ?
>
>
>I need this information to try to establish a timing and load
>analysis of adding/removing effects to our scenerio. I am trying to
>establish which effects we are using are cpu intensive. Until i can
>get believable data out of the stats, it is hard to identify which
>code needs attention.
>
>Thanks for any help.

Reading my old 3 days mail, I saw you asked help on a problem I've just
noticed yesterday.
By my side, I discovered this when going from single process to multi process.
I can't understand why frame time stays around 80ms even if app cull draw
goes lower with a sum less than 50.
What surprised me more is the fact that the update if 20 Hz so this mean a
frame at 50 ms ( and not 80 ms as displayed )
So did you get direct answers ??
 /\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\
< _/_/ _/ _/_/_/ _/ _/ _/_/_/ _/_/_/ _/_/_/ >
< _/ _/ _/ _/ _/ _/ _/ _/ _/ _/ >
< _/ _/ _/ _/ _/ _/ _/ _/_/_/ _/ _/ >
< _/ _/ _/ _/ _/ _/ _/ _/ _/_/_/ >
< _/ _/ _/ _/ _/ _/ _/ _/ _/ _/ >
< _/_/ _/_/_/ _/_/_/ _/ _/_/_/ _/_/_/ _/ _/ >
< >
< BILLARD Olivier - Ingeneer R&D ++at++ C&I Software >
< 1 avenue de la mer - 44380 PORNICHET - FRANCE >
< Tel: +33 40 11 68 72 Fax: +33 140 61 68 14 >
< Email: ceti++at++worldnet.net >
 \/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/

=======================================================================
List Archives, FAQ, FTP: http://www.sgi.com/Technology/Performer.html
            Submissions: info-performer++at++sgi.com
        Admin. requests: info-performer-request++at++sgi.com

New Message Reply Date view Thread view Subject view Author view

This archive was generated by hypermail 2.0b2 on Mon Aug 10 1998 - 17:53:00 PDT

This message has been cleansed for anti-spam protection. Replace '++at++' in any mail addresses with the '@' symbol.