Gain Expert Insights into Python* Parallelism Techniques

Unlock top performance for Intel® CPUs and GPUs using Python*, including how to use parallel Python to write reductions and offload them to a SYCL* device.

A reduction combines multiple values in parallel, in an unspecified order, to produce a single value. This technique uses an operator that is both associative and commutative.

The expert-level hands-on agenda includes these topics:

• Introduction to numba-dpex (ND) and examples of how to write parallel code and perform an automatic offload approach using the @numba.jit decorator and kernel decorator.
• Introduction to ND-range kernels, workgroups, and work items.
• How to write data parallel Python code using shared local memory, private memory, barriers, and atomics.
• Write a data parallel Python program to perform reductions:
  • In a single kernel
  • Using shared local memory and barriers