Thank you @OpenCLorg Toby Potter on your overview and demo of #OpenCL. Your high-level concepts, roadmap changes, implementations, and demonstration of how OpenCL can be used in real-world examples, truly was dazzling! Lookout for our #training sessions on OpenCL in the new year
I am absolutely loving OpenCL. I rewrote a fairly simple single threaded function that I thought didn't have enough raw compute to justify porting to a kernel:
CPU (single threaded): 217ms
GPU (plus all CPU overhead): 20ms (10ms kernel+10ms CPU overhead)
The next major release of the Basis Universal GPU texture interchange lib (v1.16) is coming along. I like to keep lots of backups.
ETC1S encoding is now 1.4x faster single threaded, 2.2x faster multithreaded, and >4x faster using OpenCL single threaded.
New OpenCL extensions provide additional interoperability, supporting an industry trend to provide open, multivendor, and multiarchitecture solutions. We are big fans. #OpenCL#IamIntel @IntelSoftware
This is a nice writeup that describes the new OpenCL extensions and the use-cases they are targeting. I'm excited to see what developers can do with this functionality!
OpenCL is announcing new extensions for two key uses cases: boosting neural network inferencing performance and providing flexible and powerful interoperability with new generation graphics APIs, including Vulkan.
https://t.co/UjQ3d0kF5w