crash in ndDynamicsUpdate::SortJoints

Report any bugs here and we'll post fixes

Moderators: Sascha Willems, Thomas

Re: crash in ndDynamicsUpdate::SortJoints

Postby Julio Jerez » Mon Dec 28, 2020 5:15 pm

I like that you use those ferocious convex hull shapes for your test. it exposes weaknesses in the engine.

here is a suggestion for your own production test, the convex hull can take a tolerance parameter. I am not sure if you pass it as 0.0
passing 0.0 makes exact convex hulls, but they force the the engine to do lots of work for very little reward.
this is because exact convex hulls can have many points that can be so closed that they do not make any contribution to the collision system since they generate contacts that are so closed that they produce rows in the matrix that are linear combination of other rows. that's bad.
The engine has to run expensive eigen values decomposition algorithm to prune these contacts.
if the contacts are not there then those algorithm take not effect.

if you pass a larger than zero tolerance, something like 1.0e-3, you will see the result is such that you could not tell the differences. but it can be much faster.
Julio Jerez
Moderator
Moderator
 
Posts: 12249
Joined: Sun Sep 14, 2003 2:18 pm
Location: Los Angeles

Re: crash in ndDynamicsUpdate::SortJoints

Postby Bird » Mon Dec 28, 2020 6:32 pm

Thanks for the tip. I did have the tolerance set to 0.0f.

I thought I posted about everything working now but I don't see the post ??? Anyway, the crashing has stopped after the latest update. Thanks again for the fix.
Bird
 
Posts: 623
Joined: Tue Nov 22, 2011 1:27 am

Previous

Return to Bugs and Fixes

Who is online

Users browsing this forum: No registered users and 5 guests

cron