how do you measure object resource use at i live in science land

how do you measure object resource use

there must be a way to do this, i just have no clue. two weeks ago i wrote some rough guidelines for what prims, textures, and avatars use resource wise

as you likely know, i am making field trip spots for students to use along with a workbook subQuark is working on. science education is cool to me and being able to mesh that with virtual worlds is really rewarding

i still have lots of the builds to complete and some new ones still needing to be made. plus all sorts of “kiosk” type information things to add (this will be really cool with web-on-a-prim in OpenSim)

all these things that are yet to be done will use up more RAM and we are already almost maxed out, so understanding usage is becoming very important to me

i do use textures carefully (trying to never go above 512 by 512 pixel PNG 8s) and i have actually been trying to use fewer textures (does concrete really need to look textured? how about if it was painted concrete?)

in that resource usage post, i mentioned that some prims, like a torus, take up more resources than a box does. so i imagine a sculptie can too?

is there a way to measure the “load” of an object in a relatively easy manner?

for example, the one tree i have ever made has a sculptie trunk but i don’t think it makes a big enough difference to use it if it uses more resources – i’d rather save that for more details in things like espresso machines =)

besides, i don’t think in-world beavers care =D

if you know a way to compare load, please share it here, i’m sure many people could benefit from your knowledge - thanks! =)

left = sculptie trunk | right = tapered cylinder

left = sculptie trunk | right = tapered cylinder | centre = hollow head

written by Ener Hax

August 21st, 2010 at 6:34 pm

posted in enclave harbour,virtual worlds

A good question, and the method I came up with might not be best but here it is:

Overview
——–
To focus on client experience we must place the onus of performance at their end. This is measured in Milliseconds of render time for the client.

Many things can affect this:
– Render options at the client end.
– Item build type, how the render treats occluded triangles etc.

Thus, it may not be about indidvual items. How do we test?

Render Time Graph
—————–

ctrl-shift-9

play with this, expand and collapse the left hand tree. Get it so the 6 main types are visible – Frame, System Messages, sleep, idle, render, other.
This show Milliseconds per frame to render. Obviously we have to make an assumption of moderate machine specs the client has, but optimized for one covers all. The Render Time is the biggie it sucks most time. Click the scrolling graphs to stop/start(resets the time axis) and note the render time.

Method
——

Create a single stand alone sim, preferably on a another machine.

Login. Click the graphs to stop/start. Note render time. Place some copies of your high prim lovingly built monstrosities. Click the graphs to stop/start. Note render time. Now try your two types of trees, one type at a time. Remember, start/stop to reset render time. To provide perspective, one type of tree, 2 of, costs me about 1ms of render time so its very much measurable.

Conclusion
———-

In answer to your question, you will be able to easily see a different between two types of items if any.

While milliseconds will vary for the client, you can at least say “Item A will take 2% less render time than Item B”.

Breen Whitman

21 Aug 10 at 7:50 pm

fantastic! i am going to try it right now by trying out linked groups of trees. so i’ll take like 30 trees of each type and link them as one object each. this should increase render time and reduce error (like weighing out 100 pennies versus one)

this is brilliant! and your explanation very clear, thank you!

Ener Hax

21 Aug 10 at 8:34 pm

awesome graphs! okay, so i did a set of 30 trees of each type and repeated it a few times

the prim trunk was about 26 ms, the sculpted trunk was 29 ms

while it does not sound like much, that is a 10% difference and if i plant several thousand trees, then a larger cumulative effect!

thank you Breen for the neat trick and teaching me about those graphs!

21 Aug 10 at 10:43 pm

[...] back to measuring the load of objects! Breen offered a very easy method for this last week that i tried right away and it was pure genious! i had never used fast timers and see his comment for his well explained procedure [...]