large numbers of points killing gh...

I am experiencing some odd behavior...I have a large point cloud, a little over 190k...and gh is managing them in the most peculiar way...there's no problem referencing them, or performing operations on them...bounding box growth, rTrees take very little time. But as soon as I try to save the definition, I get the spinny wheel of death and both gh and rhino grey out. The same thing happens when I either try to clear the point primitive references or simply delete it. Any idea what's going on?

Replies to This Discussion

Permalink Reply by David Rutten on August 14, 2013 at 4:03am

Clearing and deleting will create Undo records which is basically the same as saving. So somehow it must take a very long time to write the data. I'll see if I can replicate it here with a boatload of random points.

David Rutten

david@mcneel.com

Tirol, Austria

Permalink Reply by David Rutten on August 14, 2013 at 4:16am

Yup, I can. Now to find the bottleneck.

David Rutten

david@mcneel.com

Tirol, Austria

Permalink Reply by huaxiamengqing on August 14, 2013 at 4:44am

I have come into the same trouble.

another problem is that a geo commponent reffered large count of geometry, when the geometry is missing and left large count of null item in commponent,At this time , I use this commponent to reffer something take so much time when I right click the commponent

Permalink Reply by David Rutten on August 16, 2013 at 11:58am

I made some changes to GH_IO.dll. I can now save half a million points in ~15 seconds instead of ~20 minutes. It's still pretty slow and there are other optimizations I can add but that would result in a different GH/GHX file so I'm going to put those off until GH2.

In case you're interested, practically all the time was spend "Hey, does a point with this name and index already exist?", which obviously takes longer and longer to answer the more points are in the same collection. This is now an O(log N) operation instead of O(N).

David Rutten

david@mcneel.com

Tirol, Austria

Permalink Reply by David Stasiuk on August 16, 2013 at 12:03pm

Outstanding. I am interested, and I suspected it had something to do with checking every incremental point against all the rest...thanks so much...is this going to be available anytime soon? Onward to ridiculous point clouds!

RSS