MBaadsgaard Posted March 11, 2019 Share Posted March 11, 2019 I was looking into the Compile Block wondering about the exact boundaries within which to use it. I got the impression from the docs that the optimization is mostly between nodes. Not having multiple copies of the geo per node, multi threading for loops etc. But while watching the H17.5 launch event I noticed (at around 32:43) a single polyreduce node in a compile block. So... Are there just nodes we should always stuff into a compile block for speed..??? Quote Link to comment Share on other sites More sharing options...
cwhite Posted March 12, 2019 Share Posted March 12, 2019 Your thoughts about the performance characteristics are correct. I think in that launch example the TOP node expects a compiled block (very much like the existing Invoke SOP), which is why a single node would be wrapped up. The other consideration is cases where the compile block can be executed in-place on the input geometry, such as when using the option on SOP solvers to invoke a compiled block - in that the compile block is executed in-place on the DOP geometry instead of copying it back and forth between DOPs and SOPs Quote Link to comment Share on other sites More sharing options...
MBaadsgaard Posted March 12, 2019 Author Share Posted March 12, 2019 Thank you. That makes good sense. Haven't looked at the SOP Solver yet so didn't know about that option. Quote Link to comment Share on other sites More sharing options...
Recommended Posts
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.