[ODE] Non-determistic behaviour between application runs
Keith Johnston
kjohnston at pinolisoftware.com
Mon Sep 1 22:44:02 2003
This is a multi-part message in MIME format.
------=_NextPart_000_0024_01C370EB.104998B0
Content-Type: text/plain;
charset="iso-8859-1"
Content-Transfer-Encoding: 7bit
>From what I've read in the mailing list - ode should be completely
determinstic. I am currently seeing differences between simulation runs
even when restarting the application - so I don't think this is a
problem with resetting the simulation as others have mentioned.
The differences I see start out very small - in the last decimal place
of the position or velocity, for example. But over time, these small
errors accumulate and the end of the simulation can be very different
from one run to the next.
I am using single precision, release mode, global CFM is set to .0001,
ERP is set to .2. I am using the friction cone model for friction. The
simulation involves spheres bouncing off of tri-mesh objects. I am
using single precision. I am using a fixed time step of .01. I am
removing the contact joints between steps.
Any suggestions on what might cause the differences? I also tried
adding the "/Op" option to increase floating point consistency. That
doesn't seem to help either.
Keith
------=_NextPart_000_0024_01C370EB.104998B0
Content-Type: text/html;
charset="iso-8859-1"
Content-Transfer-Encoding: quoted-printable
<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.0 Transitional//EN">
<HTML><HEAD>
<META HTTP-EQUIV=3D"Content-Type" CONTENT=3D"text/html; =
charset=3Diso-8859-1">
<TITLE>Message</TITLE>
<META content=3D"MSHTML 5.00.3513.900" name=3DGENERATOR></HEAD>
<BODY>
<DIV align=3Dleft class=3DOutlookMessageHeader dir=3Dltr =
lang=3Den-us><FONT face=3DArial=20
size=3D2><SPAN class=3D695250805-02092003>From what I've read in the =
mailing list -=20
ode should be completely determinstic. I am currently seeing =
differences=20
between simulation runs even when restarting the application - so I =
don't think=20
this is a problem with resetting the simulation as others have=20
mentioned.</SPAN></FONT></DIV>
<DIV><FONT face=3DArial size=3D2><SPAN=20
class=3D695250805-02092003></SPAN></FONT> </DIV>
<DIV><FONT face=3DArial size=3D2><SPAN class=3D695250805-02092003>The =
differences I=20
see start out very small - in the last decimal place of the position or=20
velocity, for example. But over time, these small errors =
accumulate and=20
the end of the simulation can be very different from one run to the=20
next.</SPAN></FONT></DIV>
<DIV><FONT face=3DArial size=3D2><SPAN=20
class=3D695250805-02092003></SPAN></FONT> </DIV>
<DIV><FONT face=3DArial size=3D2><SPAN class=3D695250805-02092003>I am =
using single=20
precision, release mode, global CFM is set to .0001, ERP is set to =
.2. I=20
am using the friction cone model for friction. The simulation =
involves=20
spheres bouncing off of tri-mesh objects. I am using single=20
precision. I am using a fixed time step of .01. I am =
removing the=20
contact joints between steps.</SPAN></FONT></DIV>
<DIV><FONT face=3DArial size=3D2><SPAN=20
class=3D695250805-02092003></SPAN></FONT> </DIV>
<DIV><FONT face=3DArial size=3D2><SPAN class=3D695250805-02092003>Any =
suggestions on=20
what might cause the differences? I also tried adding the "/Op" =
option to=20
increase floating point consistency. That doesn't seem to help=20
either.</SPAN></FONT></DIV>
<DIV> </DIV>
<DIV><FONT face=3DArial size=3D2><SPAN=20
class=3D695250805-02092003>Keith </SPAN></FONT></DIV></BODY></HTML>
------=_NextPart_000_0024_01C370EB.104998B0--