Page 2 of 2

Re: Animesh vs opensim NPCs

Posted: Sun Mar 22, 2020 3:35 pm
by Graham Mills
Hope these are what you want, Ilan. Much of the activity on the network chart correlates with shift-copy rezzing of avatars although the final peak on the 256 chart corresponds to spinning my avatar in place to watch the animesh in motion around it.

256 avatars:
Image

310 avatars:
Image

Ilan Tochner wrote:
Sat Mar 21, 2020 9:15 pm
Hi Graham,

Can you please post the CPU and network graphs from the Active World panel after 10 minutes of running this test?

Re: Animesh vs opensim NPCs

Posted: Sun Mar 22, 2020 4:45 pm
by Graham Mills
In case it's of any interest, viewer data from yesterday's 355 animesh bot run. It is actually quite a prim-heavy sim with a lot of plants beyond the main conservatory. For anyone interested, it represents the Liverpool Botanic Garden which opened in 1802/3.

Image

Re: Animesh vs opensim NPCs

Posted: Sun Mar 22, 2020 5:07 pm
by Ilan Tochner
Thank you Graham,

Those were indeed the graphs I wanted to see. It looks like the CPU load isn't high with this test but network load on the viewer may be an issue for some people. It looks like it required 500 Kbps just to send the movement updates for your 355 animesh bot test. Having them move in less linear ways with more direction / animation changes will likely result in even higher network usage.

Re: Animesh vs opensim NPCs

Posted: Tue Mar 24, 2020 6:24 pm
by Graham Mills
Quick update...

The issue of being unable to rez animesh avatars was resolved by simply stopping any and all existing animations even though these were not visibly playing in the rezzed animesh. The necessary code is here:

http://wiki.secondlife.com/wiki/LlGetOb ... ationNames

It was then straightforward to select a random walk animation at rez time from those added to the animesh.

Branching was effected by providing a comma-separated list of alternative destination markers to the current marker's description field with a random choice being made from those on the list.

I should add that the CPU usage under these conditions was much higher with 250 avatars, approx 60%