[ODE] QuickStep solver checked in

[Mohsin Hasan]

Using quickstep with my Ragdolls (9 joints, 7 Hinge and 2 Universal) has
increased the simulation speed by at least 100% if not more. Performance
results follow

note:
Starting FPS = When simulation starts, no collisions
Ending   FPS = When all dolls have fallen and are colliding with each other

StepFast (2 iterations, 2 contacts)

Starting FPS ~ 45
Ending   FPS ~ 27

QuickStep (20 iterations, 2 contacts)

Starting FPS ~ 70
Ending   FPS ~ 67

Since simulation is more stable auto disable kicks in early taking FPS to
around 140

WorldStep (2 Contacts)

Unfortunately the dolls just explode :(

Machine Specs
P4 2.4C GHZ
512 MB ram
Radeon 9700 Pro

Let me post the demo as soon as I clean it up. 

Great job Russ!!

Mohsin
	

-----Original Message-----
From: ode-bounces at q12.org [mailto:ode-bounces at q12.org] On Behalf Of Russ
Smith
Sent: Tuesday, May 18, 2004 11:30 PM
To: ode at q12.org
Subject: [ODE] QuickStep solver checked in


hi all,

i've added the quickstep solver to CVS and also regenerated the source
distribution and documentation. look in the docs under the new "world /
stepping functions". there is a significant amount of tweaking left to
do, but this early version should be good for many things.

i've also added a few functions that are useful for solving some
problems that commonly arise when dealing with contacts, see the
"world / contact parameters" section in the docs.

the boxstack demo has been updated to use quickstep, and to pile up to
100 bodies.

the new solver is of course less accurate than the big matrix solver,
and this inaccuracy is manifested in the boxstack demo mainly by
slipping of the contacts. to fix this problem, some contact coherence
code should be written to make sure that sticking contacts actually
stick.

all those volunteers, please post feedback about how the new solver
works for you...

enjoy,

russ.



-- 
Russell Smith
http://www.q12.org
_______________________________________________
ODE mailing list
ODE at q12.org
http://q12.org/mailman/listinfo/ode