This workshop covers SYCL* programming and optimizations for multi-GPU applications.

Hosted on Intel® Tiber™ AI Cloud with access to 4th generation Intel® Xeon® Scalable processors and Intel® Data Center GPU Max Series 1100, this workshop highlights multi-GPU programming and optimization techniques using SYCL. Geared to intermediate and expert programmers, the session focuses on using all available GPUs on the system efficiently.

This session covers:

• Discovering all GPUs available for compute offload

• Writing SYCL code to offload compute kernels to multiple GPUs

• Writing SYCL code to split computation between multiple GPUs

• Writing portable SYCL code that offloads the workload to all available GPUs

• Optimizing SYCL code so that all kernels run concurrently on multiple GPUs