LOD

[DaJuice]

Hey folks,

I was reading up on Level-of-Detail today, and started playing with PolyReduce. I'd like to know how long a frame of this scene takes to cook on your machines, and what kind of processors they have. It takes about 4 seconds on my computer (800mhz P3). If you got any ideas for speeding it up I'd love to hear them. I'm still experimenting myself.

http://www.geocities.com/stehrani3d/3dfiles/LOD_terrain.zip (50kb, Right Click > Save)

[aracid]

hey DaJuice

Im running a dual xeon 2GHz. 2 gig ram, and FX5800.

and it takes about 2 seconds to cook the first frame.

my machine is a bit wierd,

by ticking use original points, theres a slight increase.

anyways hope this helps.

all the best

aracid

stat.txt

statOriginalPoints.txt

[DaJuice]

Thanks aracid!

The more the merrier, keep 'em coming.

[keltuzar]

question.. how did u get that text o/p for the on the fly calculations what aracid showed?

Thanks

[michael]

it's the Performance Monitor...look under Windows...

[keltuzar]

here u go dajuice

I have a dual 2 ghz xeon and 1 gig of ram

thanks arctor

stats.txt

[DaJuice]

Thank you. Also saving everything before the PolyReduce SOPs and reading off disk speeds things up a bit.

[edward]

About 4 seconds as well on a single P4 1.6 GHz on 384 Mb of RAM (which clearly isn't the issue as houdini.exe only uses 113 Mb right now in the task manager.)

I was going to suggest that if you want the first cook in the hip file to be faster, you can try setting the Threading parameter on the VEX SOPs to use 1 per processor if you have more than one processor. In my mind, we don't really care about that though as any other subsequent cooks will be mostly the polyreduce sops anyhow.

[crunch]

Here are two sets of stats:

stats-opteron.txt - Dual 2.1GHz Opteron running Debian Linux

stats-centrino.txt - Centrino 1.6GHz running Fedora Core 1

stats_centrino.txt

stats_opteron.txt

[DaJuice]

Thanks guys.

you can try setting the Threading parameter on the VEX SOPs to use 1 per processor if you have more than one processor.

Ahh, it is actually a dual processor machine. Hmm, I just noticed that when you use either '1 Thread per Processor' or '2 Threads' it automatically gives your model normals.

[sibarrick]

You should be able to get round this by adding the line

computenormal(0);

to your vex code.

[Mr.Ed]

Hello,

does anyone still have a copy of this file id like to have a look at it

Cheers

Ed