shawn_kearney Posted October 29, 2017 Share Posted October 29, 2017 (edited) I wanted to play around with PyOpenCL in SOPs so I had to change out the symbolic link from HFS OpenCL to OpenCL 2.1 (from the Intel SDK) to get it to work. Upon testing to make sure I didn't completely break Houdini, I noticed that OpenCL performance in Heightfields and FLIP seemed noticeably faster. Is this possible, or is it just my imagination? (cross posted on Facebook Houdini Artists) Edited October 30, 2017 by shawn_kearney 1 Quote Link to comment Share on other sites More sharing options...
symek Posted October 30, 2017 Share Posted October 30, 2017 You can always confront your imagination with reality measuring Houdini's performance in both cases. I wouldn't be surprised if lso OCL 1.2 code could run faster on OCL 2.0 driver. Aside of language's changes, new versions usually evolves compilation for better vectorization (I also imagine that dynamic parallelism might be used for work stealing). 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.