[ODE] Re: [ODE-CVS] ode/ode/src quickstep.cpp, NONE, 1.1 quickstep.h, NONE, 1.1

Adam D. Moss adam at gimp.org
Tue May 18 20:51:18 MST 2004


Russell Smith wrote:
> // uncomment the following line to determine a new constraint-solving
> // order for each iteration. however, the qsort per iteration is expensive,
> // and the optimal order is somewhat problem dependent. the unsorted
> // order is quite often the best way to go, especially for low CFM.
> // @@@ try the leaf->root ordering.
> 
> //#define REORDER_CONSTRAINTS 1

Is that analogous to StepFast's 'random' reordering?  (AFAIK
StepFast purposely shuffled the constraint ordering to attempt
to avoid hitting a constantly lousily-convergant ordering, which can
happen if the constraints are left in any sort of natural ordering
[about as likely as accidentally hitting a quickly-convergant
ordering, but it's the consistancy which is really important] --
was just wondering if the ordering in FastStep was of similar
significance.)

Thanks,
--Adam
-- 
Adam D. Moss   . ,,^^   adam at gimp.org   http://www.foxbox.org/   co:3


More information about the ODE mailing list