With all the excitement about H20 and what might or might not be added, i just thought it would be interesting to see what peoples existing favourite nodes, or bits of functionality were in Houdini?

I think my favourite node is probably the MatchSize [www.sidefx.com] node... I’d encourage anyone doing procedural work to have a good play with it. Very powerful... have used it to great time saving effect recently on a project where i needed to layout a complex group of products but then be able to swap the model of an element and everything else move accordingly

Also in recent month I’ve really started using TOPS more and more, particularly wedging to explore alternatives. Using it to automate render tasks etc..
https://www.sidefx.com/tutorials/pdg-for-design-work-pt-1-the-basics/ [www.sidefx.com]

But why is the node so slow ? And why is it not compiled in a loop ? Houdini is a program of contrasts.

HGaal

This is not true. The Match Size node is compilable and works fast

Alexey_Vanzhula

HGaal

This is not true. The Match Size node is compilable and works fast

I checked. Indeed, it is being compiled now. Less than a year ago I tried to compile it and it didn't work. It is currently being compiled. That was good news. And the bad news is that it still works slowly. I just moved the pieces to the center. A simple loop with transform and a loop with matchsize. Without compilation. Transform - 0.6 seconds, match size - 1.6 seconds. Almost 3 times. In my old project, a simple move and alignment "matchsize" took tens of seconds. So I made a wrangle and it was faster about 100 times.

HGaal

Alexey_Vanzhula

HGaal

This is not true. The Match Size node is compilable and works fast

I checked. Indeed, it is being compiled now. Less than a year ago I tried to compile it and it didn't work. It is currently being compiled. That was good news. And the bad news is that it still works slowly. I just moved the pieces to the center. A simple loop with transform and a loop with matchsize. Without compilation. Transform - 0.6 seconds, match size - 1.6 seconds. Almost 3 times. In my old project, a simple move and alignment "matchsize" took tens of seconds. So I made a wrangle and it was faster about 100 times.

The Match Size node indeed consists of multiple nodes and can be optimized since it currently calculates the result in several stages. However, no matter how you write your VEX code, it will not be able to work 100 times faster. You have measured the difference incorrectly. I will be glad to watch a scene with an example.

One of my favorite nodes is the super fast pointgenerate. I use it all the time as foundation for efficient workflows and HDAs.

Matchsize is another beloved one, although I noticed it too, it could be faster if converted to pure Vex.

Enumerate is another little gem, extremely helpful in re-enumerating spare pieces that have non-continuous indices. (I would like to understand the algorithm behind this, because I tried to do it in python and it was very slow)

And finally I should assign a mention of honor also to the Connectivity Sop, which I take it for granted now, but the very first times its usefulness was probably mind blowing for me

Attribute Wrangle SOP, OpenCL SOP