thinkboxsoftware.com/genome-faq
We have started a FAQ page in the Genome docs, please post questions that you think should be answered in the FAQ in this thread!
Just opened the forum (back from a week of rest), first question that pops on my mind: is a Genome “modifier” multithreaded?
Will be in office tomorrow, can’t wait to install this! Wonder if I could create a modifier that reads vertex selection from the stack and assigns a vertex colour to it?
Yes, it is multi-threaded, but that does not mean that you will get 8 times the performance on 8 cores. It also depends on the complexity of the flow - the higher the complexity, the better it scales on multi-core machines. Also, some operations benefit from multi-threading more than others. For example, evaluating a texture map benefits more.
Benchmark: Taking a Teapot with 64 segments and pushing vertices along the normals based on a Cellular Texture Map using Genome vs. VolumeSelect+Push modifier shows the following performance differences - the texture had animated U/V offset and the viewport was played back without skipping frames over 100 frames:
3ds Max 2012 64bit, Nitrous
Genome, MT: 26.628 seconds
Genome, ST: 34.904 seconds
Vol.Select+Push: 31.149 seconds
3ds Max 2012 64bit, D3D
Genome, MT: 19.690 seconds
Genome, ST: 28.368 seconds
Vol.Select+Push: 24.480 seconds
From this you can see that a Multi-threaded Genome is about 5 seconds faster than Volume Select, but about 4 seconds slower in Single-threaded mode. So in general Genome is supposed to be faster than comparable Max modifiers thanks to multi-threading, but would have been slower if it were not multi-threaded due to the overhead of converting Max data structures to our own data channels and back. We are looking at the situations where this is not the case and would like to optimize Genome to be the faster and more flexible alternative to some Max modifiers. And then there is also this…
I’m super excited about this. Seems a dream that comes true.