Intel(R) oneAPI Collective Communications Library (oneCCL) 2021.10
- Improved scaling efficiency of the Scaleup algorithms for ReduceScatter
- Optimized performance of oneCCL scaleup collectives by utilizing the new embedded Data Streaming Accelerator in Intel® 4th Generation Xeon Scalable Processors (formerly code-named Sapphire Rapids)