Designing Custom GPU Algorithms to Accelerate Graphical and Machine Vision Processing

Please note: This session is part of AI Day. Learn more and register for the full event here.

Algorithms for processing graphical objects such as raster images and 3D point clouds can quickly become very compute-intensive. Many of these algorithms are excellent choices for parallel processing optimizations. Developing custom general-purpose GPU code in C++ is easier than you might think, and well-designed GPU algorithms can yield game-changing performance improvements. This is especially critical for real-time applications such as autonomous navigation, machine vision, and medical imaging.

This talk will cover:

• Using nVidia and Intel APIs with C++ to create high-performance native GPU kernels.

• Techniques to maximize GPU throughput and avoid performance bottlenecks.

• Combining GPU and CPU-based optimizations (SIMD, multithreading) for additional performance.