SonicSampler: Unified Tile-Aware Kernels for LLM Sampling and Speculative Verification

[Submitted on 24 May 2026]

View PDF HTML (experimental)

Abstract:Sampling in LLM inference comprises a combinatorial set of logit processing, token selection, and verification operations for speculative decoding. However, existing implementations either accelerate only subsets of this pipeline, rely on multiple kernel launches, or assume homogeneous sampling behavior across a batch, limiting support for dynamic serving workloads and preventing efficient CUDA Graph execution. We present $\textbf{SonicSampler}$, a unified suite of tile-aware Triton kernels that vertically fuses the complete sampling pipeline into a fixed, workload-aware execution model. Our kernels support dynamic per-request sampling behaviors, including grammar-constrained decoding, repetition, frequency and presence penalties, logit bias, temperature scaling, top-$k$ / top-$p$ / min-$p$ filtering, and speculative verification – within a single batched kernel while remaining fully CUDA Graph-compatible. Central to our approach is a novel hierarchical two-stage top-$k$ algorithm that achieves up to $\textbf{10x speedup}$ over competitive baselines and exploits the low-entropy structure of LLM outputs to enable efficient selection over large vocabularies. Across heterogeneous speculative decoding workloads, SonicSampler achieves up to $\textbf{16x speedup}$ over state-of-the-art baselines while preserving flexible batched execution.

Submission history

From: Pragaash Ponnusamy [view email]
[v1] Sun, 24 May 2026 23:01:50 UTC (3,581 KB)