What Intel® Processor Graphics GEN9 Unlocks in OpenCL*

If you know about integrated graphics, Intel® HD graphics, or Iris® graphics, then you know about GEN—Intel engineering shorthand for “Graphical Execution Engine Unit”.

GEN is like other GPUs: it shares the die with the regular CPU block and its primary purpose is to support low-level graphics APIs such as OpenGL*, Vulkan*, DirectX*, and Metal*. Aside from traditional graphics workloads, GEN is also a powerful general compute resource that is often underutilized … or completely overlooked. To put this in perspective, if you develop for Intel® Core™ i7 processors, over half the die area is devoted to GEN. So unless your workload is making extensive use of graphics, you’re likely leaving massive compute bandwidth on the table.

In this webinar we look at the ninth generation of graphics—GEN9—with specific focus on unlocking performance, including:

  • An overview of Intel® Processor Graphics: the architecture, driver stack, and use in general-purpose computing
  • How OpenCV* can deliver high-performance by leveraging GEN hardware through the OpenCL* API
  • How Intel® VTune™ Amplifier is an essential tool for identifying performance bottlenecks in OpenCL
  • Best approaches for OpenCL performance optimization
Adam Herr, Senior GPGPU Performance Engineer, Intel Corporation

Adam has 25 years of professional engineering experience and is currently driving the Halide* enabling project, which allows performance-portable programming across various hardware accelerators. He previously served on the Gen Compute Architecture*, OpenCL* driver, Intel® Graphics Performance Analyzers, and Larrabee Consulting* teams. Prior to joining Intel, Adam was involved in several high-tech start-ups in the areas of spherical video, photogrammetry, public safety, telematics and computer graphics. He holds several patents and has a BS in Computer Science and a BA in Philosophy from Pacific University in Forest Grove, Oregon.

For more complete information about compiler optimizations, see our Optimization Notice.