Introduction
We’re thrilled to share the latest advancements by AMD in graphics research from the High-Performance Graphics (HPG) Conference 2024. Our team presented four ground-breaking papers covering work graphs, ray tracing, and compression.
In this blog, we’ll introduce each paper, and link you to additional resources to find out more.
We are incredibly honored to win the Wolfgang Straßer Award for Best Paper with our work graphs paper, so let’s start there.
Real-Time Procedural Generation with GPU Work Graphs (Best Paper winner)
Microsoft® Direct3D® 12 Work Graphs are one of the hottest new graphics topics around, generating an enormous buzz wherever they are mentioned. We are proud to be at the forefront of GPU work graph development, and our new paper at HPG (in collaboration with Coburg University) takes procedural generation to new heights. You’ll find out how easy it is to implement our system, and demonstrate with stunning procedural effects from our GDC 2024 demo.
- Paper – Real-Time Procedural Generation with GPU Work Graphs
- YouTube link to the presentation – Bastian Kuth: Real-Time Procedural Generation with GPU Work Graphs
- Demo – GDC Work Graphs blog post with video
- Related blog post – GPU Work Graphs and Mesh Nodes in Microsoft DirectX® 12
- Sample (GitHub) – Work Graphs Ivy Generation sample
We also had a follow-up presentation on work graphs and their inner workings:
- YouTube link to the presentation – Max Oberberger: Work Graphs: Hands-on with the Future of Graphics Programming
The second-placed Best Paper also went to one of our new papers, so let’s introduce H-PLOC next!
H-PLOC – Super-Fast BVH Construction (Best Paper second place)
H-PLOC revolutionizes BVH (Bounding Volume Hierarchy) construction by allowing for construction of a BVH with over millions of triangles in just a few milliseconds. This makes it ideal for efficiently ray-tracing a large amount of fully dynamic geometry per frame.
- Paper – H-PLOC: Hierarchical Parallel Locally-Ordered Clustering for Bounding Volume Hierarchy Construction
- YouTube link to the presentation – Carsten Benthin: H-PLOC Hierarchical Parallel Locally-Ordered Clustering for BVH Construction
Let’s stay with ray tracing, with our next paper on HIP RT!
HIPRT – A Ray Tracing Framework in HIP
HIPRT introduces a powerful ray tracing framework built on the HIP platform. With a focus on performance and portability, HIPRT enables developers to harness the full potential of GPUs for ray tracing applications. We’ve also released some new tutorials, including how to use HIPRT to compute direct illumination using Monte Carlo integration.
- Paper – HIPRT: A Ray Tracing Framework in HIP
- YouTube link to the presentation – Daniel Meister: HIPRT: A Ray Tracing Framework in HIP
- New tutorials – HIPRT tutorials on GitHub
- Download HIPRT now – HIPRT on GPUOpen
Finally, let’s move over to compression.
DGF – Geometry Compression for Micropolygon Meshes
DGF introduces a dense, hardware-friendly geometry format that compresses meshlets with arbitrary topologies. Just as block compression transformed texture handling, DGF enables efficient ray tracing of massive micropolygon geometry.
- Paper – DGF: A Dense, Hardware-Friendly Geometry Format for Lossily Compressing Meshlets with Arbitrary Topologies
- YouTube link to the presentation – Joshua Barczak: DGF: A Dense, Hardware-Friendly Format for Lossy Compressing Meshlets with Arbitrary Topologies
Conclusion
The HPG Conference 2024 was a fantastic platform for sharing our latest research with you all. We’re excited to see how these innovations shape the future of real-time graphics.
Make sure you follow @GPUOpen on X or @GPUOpen on Mastodon (or our RSS feed!) to keep up with the latest developments for these papers, and our other technologies.
