High-Performance Graphics (HPG) is the leading international forum for performance-oriented graphics and imaging systems research, including innovative algorithms, efficient implementations, languages, compilers, parallelism, and hardware architectures for high-performance graphics. The conference brings together researchers, engineers, and architects to discuss the complex interactions of parallel and custom hardware, novel programming models, and efficient algorithms in the design of systems for current and future graphics and visual computing applications. HPG2025 will be co-hosted with EGSR in Copenhagen, Denmark.

Khronos Group Related Session

Slang: Bridging Graphics and Machine Learning

Description: Slang is an open-source shading language that brings automatic differentiation to real-time graphics programming. While neural techniques are increasingly vital for achieving photorealistic rendering, implementing them has traditionally required maintaining separate codebases in ML frameworks and graphics APIs. We demonstrate how Slang’s built-in automatic differentiation enables graphics developers to implement gradient-based optimization directly in shader code, while maintaining the performance characteristics needed for real-time applications. We show how Slang reduces implementation complexity by automatically generating differential code, eliminating the need to manually maintain forward and backward passes. The presentation will walk through a simplified Neural Materials implementation, demonstrating how Slang’s Python bindings enable seamless integration with ML workflows while preserving the flexibility and control required for graphics programming. By bridging the gap between graphics and machine learning ecosystems, Slang enables developers to more easily incorporate neural techniques into their rendering pipelines.
Speaker: Shannon Woods, Slang Working Group Chair, NVIDIA