Stefan Reinhardt
StefanReinhardt
About Me
Connect
LOCATION
Not Specified
WEBSITE
Houdini Skills
Availability
Not Specified
Recent Forum Posts
Houdini RBD Bullet used mem grows not linear with num Object Feb. 24, 2014, 7:54 a.m.
Hey,
as you can imagine out of the Headline, i'm trying to solve a large number of objects with the bullet solver in houdini.
(The task is to compare bullet with a solver written on my own).
Now i am doing some memory tests and found out that the memory usage which bullet has, is not growing linear with the number of objects.
Here some measurements:
Static object: grid with 6400 Spheres on it.
Tests:
Dynamic Obj. – Mem Usage
12 – 1,1 MB
98 – 110 MB
980 – 811 MB
5600 – 23.299 GB
8000 – >32 GB
Does anyone has a idea why?
I should go into detail about this in my Thesis and i have totally no clue why bullet behaves that way, in this special case…
I attached a example file.
Thanks in advance
as you can imagine out of the Headline, i'm trying to solve a large number of objects with the bullet solver in houdini.
(The task is to compare bullet with a solver written on my own).
Now i am doing some memory tests and found out that the memory usage which bullet has, is not growing linear with the number of objects.
Here some measurements:
Static object: grid with 6400 Spheres on it.
Tests:
Dynamic Obj. – Mem Usage
12 – 1,1 MB
98 – 110 MB
980 – 811 MB
5600 – 23.299 GB
8000 – >32 GB
Does anyone has a idea why?
I should go into detail about this in my Thesis and i have totally no clue why bullet behaves that way, in this special case…
I attached a example file.
Thanks in advance
Accuracy of Bullet Collisions Feb. 19, 2014, 8:26 a.m.
cool thanks that helped a lot
Accuracy of Bullet Collisions Feb. 18, 2014, 7:13 a.m.
Thanks that helped for the collision of the spheres itself.
The collisions with the static spheres on the other hand doesn't look as I expected. Especially they do not stay on them (as i would expect)
A perhabs complete other question, why does the simulation gets more instable if I turn up the substeps…
The collisions with the static spheres on the other hand doesn't look as I expected. Especially they do not stay on them (as i would expect)
A perhabs complete other question, why does the simulation gets more instable if I turn up the substeps…