From: faculaganymede (faculaganymede++at++yahoo.com)
Date: 05/12/2004 18:30:24
I found out the answer for this question. The answer
is: when I use two pipes, one pfChannel for each pipe,
it turns out that the pfFrameStats keeps track of the
frame times for only one pfChannel (before, I thought
it stores the total frame times for the entire system
- both pipes). Thus, due to the order of operation,
the frame times stored by pfFrameStats for pipe 1
overwrite the frame times for pipe 0, and since APP is
only done once, so the APP time for pipe 1 really is
0.0 msec.
One question down, two more to go... Continue from my
previous question (pfFrameStats::mQuery question)...
QUESTION 1)
I ran my Performer application in both single and
multi-processing mode using 1 pipe, and I noticed that
the CULL time for single processing mode is about 30
times smaller than the CULL time for multiprocessing
mode. How is this possible?????
Here is a sample output from running my app in single
proc mode:
-------------------frame=500---------------------
pipe=0 frame rate=100.000000
total=0.000000; app=0.111200; cull=0.027200;
draw=2.174400; isect=0.000000; dbase=0.001600;
lpoint=0.000000; compute=0.000800
-------------------frame=501---------------------
pipe=0 frame rate=100.000000
total=0.000000; app=0.112000; cull=0.028800;
draw=2.163200; isect=0.000000; dbase=0.001600;
lpoint=0.000000; compute=0.000800
Here is a sample output from running the same app in
multi-proc mode:
-------------------frame=500---------------------
pipe=0 frame rate=100.000000
total=12.056000; app=0.130400; cull=0.612800;
draw=2.175200; isect=0.000000; dbase=0.001600;
lpoint=0.000000; compute=0.002400
-------------------frame=501---------------------
pipe=0 frame rate=100.000000
total=12.056800; app=0.132000; cull=0.600000;
draw=2.177600; isect=0.000000; dbase=0.000800;
lpoint=0.000000; compute=0.001600
Notice the CULL time in the two are significantly
different. Please advise.
QUESTION 2)
When looking at the frame times output, I noticed that
in general (regardless of the number of pipes I used,
single or multi-processing), the frame times for the
first ~20 frames have very strange values. For
example:
The APP time here is 1668.293579 msec, how could this
number be so big?????
-------------------frame=1---------------------
pipe=0 frame rate=100.000000
total=0.000000; app=1668.293579; cull=0.260800;
draw=16.888800; isect=0.000000; dbase=0.000000;
lpoint=0.000000; compute=0.000000
The TOTAL latency time here is -4536.204102 msec, how
could time be negative?????
-------------------frame=4---------------------
pipe=0 frame rate=100.000000
total=-4536.204102; app=0.824000; cull=1.295200;
draw=0.000000; isect=0.000000; dbase=0.404800;
lpoint=0.000000; compute=0.009600
It takes about 20 frames before the frame times become
stabilized (i.e. the process times do not change too
much between frames). Does anyone know why??
=====
Thanks,
faculaganymede
__________________________________
Do you Yahoo!?
Yahoo! Movies - Buy advance tickets for 'Shrek 2'
http://movies.yahoo.com/showtimes/movie?mid=1808405861
This archive was generated by hypermail 2b29 : Wed May 12 2004 - 18:32:08 PDT