Towards Practical Meshlet Compression

TL;DR

The paper tackles efficient compression for meshlet-based rendering in mesh-shading pipelines by designing a codec that orders triangles into generalized triangle strips (GTS) via a mixed integer linear program (MILP). It leverages index coherence within strips and memory-aligned crack-free attribute quantization to achieve up to $16:1$ index-buffer compression relative to the traditional vertex pipeline, while supporting practical deployment on GPUs. The work demonstrates fast decompression on contemporary GPUs and discusses memory-footprint tradeoffs, including GTS-Reuse variants and the impact on meshlet buffers. Overall, the approach provides a practical, high-throughput path for meshlet-based rendering with significant compression and performance benefits, complemented by supplemental code for GPU decoding, quantization, and MILP.

Abstract

We propose a codec specifically designed for meshlet compression, optimized for rapid data-parallel GPU decompression within a mesh shader. Our compression strategy orders triangles in optimal generalized triangle strips (GTSs), which we generate by formulating the creation as a mixed integer linear program (MILP). Our method achieves index buffer compression rates of 16:1 compared to the vertex pipeline and crack-free vertex attribute quantization based on user preference. The 15.5 million triangles of our teaser image decompress and render in 0.59 ms on an AMD Radeon RX 7900 XTX.

Figures (1)

• Figure 3: We compare the performance of different versions of our compression with (a) and without (b) cone-culling against the vertex pipeline (lower is better). (c) To normalize for different mesh sizes, we compare the triangle throughput per second of our compression against the vertex pipeline (higher is better).