Offload Excellence: Designing for GPU Performance

CPU-to-GPU offloading not only is increasingly common, it’s also increasingly required to ensure application performance and scalability—whether in your data center or among your customers.

The key is in efficient code design.

In this session:

• Find out how Intel® Advisor delivers the right mix of metrics and guidance to help you make informed design decisions that result in optimal performance when porting to and running code on GPUs.
• Explore GPU application performance characterization.
• Gain understanding of performance headroom against GPU limitations with automated Roofline Analysis.
• Get actionable recommendations to design efficient GPU code.

Download a stand-alone version of Intel Advisor or as part of the Intel® oneAPI Base Toolkit—a core set of tools and libraries for developing high-performance, data-centric applications across diverse architectures.