Page 1 of 2

Re: Challenge: Make the fastest island selection compound

Posted: 13 Feb 2012, 19:27
by Chris_TC
Mr.Core wrote:I have also attached my scene,2012 sp1\ x64
I opened your scene, and all 4 islands are extruded. If I remove the Array From Set, I'm back to the 3 islands from my screenshot.

Someone else is going to have to take a look, maybe something is broken on my end.

Re: Challenge: Make the fastest island selection compound

Posted: 13 Feb 2012, 19:51
by milanvasek
i tried Mr.Core's scene and got the same result: http://clip2net.com/s/1AC56

2012 SAP, 64bit

Re: Challenge: Make the fastest island selection compound

Posted: 22 Feb 2012, 18:14
by Chris_TC
I also tested the compound at work, but I get the same erratic behavior. I'm not sure why :-s
Regardless, everybody is free to add their compounds to the mix...

Re: Challenge: Make the fastest island selection compound

Posted: 01 Apr 2012, 23:42
by gustavoeb
Im currently building something related, but not quite the same... Im currently using a setup that is a slight variation of the method on Gui's Lafforge blog: http://frenchdog.wordpress.com/2010/03/ ... ng-arrays/

Im using a slightly different while loop, and a different way to organize the indexes... Anyhow it should be able to select from a PolyInput, and I dont find it that slow... Is anybody doing somehting completaly different?

I found a lot of broken links in this thread, so my question is:
What is the fastest working compound until now? So I can compare mine...

Re: Challenge: Make the fastest island selection compound

Posted: 03 Apr 2012, 19:53
by grahamef
gustavoeb wrote:Im currently building something related, but not quite the same... Im currently using a setup that is a slight variation of the method on Gui's Lafforge blog: http://frenchdog.wordpress.com/2010/03/ ... ng-arrays/

Im using a slightly different while loop, and a different way to organize the indexes... Anyhow it should be able to select from a PolyInput, and I dont find it that slow... Is anybody doing somehting completaly different?

I found a lot of broken links in this thread, so my question is:
What is the fastest working compound until now? So I can compare mine...
I'm interested in knowing that too. I was thinking of trying that technique myself but I haven't had the time lately.

Re: Challenge: Make the fastest island selection compound

Posted: 18 Apr 2012, 15:05
by gustavoeb
my similar technique doesnt work on islands with very messed up geo ~x(
I had to make it slower, but even then it wont support islands with veeeeeeeeery messed up geo :))

Ill just have to copy and paste Guis work and start from there again, but Im being lazy lately

Re: Challenge: Make the fastest island selection compound

Posted: 18 Apr 2012, 16:16
by iamVFX
I really like Oleg's compound. Result of it in a per point context and its fast

Image

Re: Challenge: Make the fastest island selection compound

Posted: 18 Apr 2012, 18:42
by Chris_TC
iamVFX wrote:I really like Oleg's compound. Result of it in a per point context and its fast

Image
Where can this compound be found? The speed looks great, I'd love to try it out.

Re: Challenge: Make the fastest island selection compound

Posted: 18 Apr 2012, 19:16
by iamVFX
Chris_TC wrote:Where can this compound be found? The speed looks great, I'd love to try it out.
Here, the fifth post, but you have already tested it, didn't you?

Re: Challenge: Make the fastest island selection compound

Posted: 19 Apr 2012, 17:00
by Chris_TC
Oh sorry, yes of course I have. I thought this was a different one.

Re: Challenge: Make the fastest island selection compound

Posted: 10 Nov 2012, 01:52
by rray
some compounds here I made recently, work reasonably fast, based of Guillaume's method, but simplified
#1: "Initialize Island Lookups"
#2: "Get Island Locations": per point, fetches array of locations of all island points for subsequent attribute lookups, requires Compound 1 executed

Don't know how much this is already covered in this thread, just needed some quick search&paste place to post this in :)

Re: Challenge: Make the fastest island selection compound

Posted: 11 Nov 2012, 13:19
by julian johnson
I never seemed to have any problems with Oleg's compound so I'm not sure this variant will work on other machines but in the attached scene I've used the same basic technique as Oleg (i.e. growing outwards from the initial vertices attached to the target poly) but tried to filter the search vertices so that you're not iterating over the same points. I set up a variant of the test scene with highly subdivided topology and my tree executes in about 1.75s and Oleg's in about 7.5s (on my machine).

http://www.exch.demon.co.uk/xsi/files/i ... nge_JJ.rar

Let me know if this a) works and b) if my timings are grossly wrong! On the surface it seems to do the job. The scene is 2013 SP1.

Re: Challenge: Make the fastest island selection compound

Posted: 12 Nov 2012, 14:32
by julian johnson
For completeness, just updated the scene to include a comparison of the two basic techniques a) iterating outwards as Oleg does and b) identifying all the polygon islands first and then deriving the island-to-be-actioned from the polygon index (similar to Reinhard's compounds above). The former is always about 30% quicker but the latter does give you the polygon island indices as a bonus.

http://www.exch.demon.co.uk/xsi/files/i ... island.rar