arXiv subjects

Sean Siddens

Publications and source records attributed to Sean Siddens.

4 recordsLinked to original sources

arXiv2026

Fleet: Hierarchical Task-based Abstraction for Megakernels on Multi-Die GPUs

Modern GPUs adopt chiplet-based designs with multiple private cache hierarchies, but current programming models (CUDA/HIP) expose a flat execution hierarchy that cannot express chiplet-level locality or synchronization. This mismatch leads to redundant memory traffic and poor cache utilization in memory-bound workloads such as LLM inference. We present Fleet, a multi-level task model that maps computation to memory scopes. Fleet introduces Chiplet-tasks, a new abstraction that binds work and data to a chiplet and enables coordination through its shared L2 cache. Wavefront-level, CU-level, and device-level tasks align with existing abstractions, while Chiplet-tasks expose a previously unaddressed level of the hierarchy. Fleet is implemented as a persistent kernel runtime with per-chiplet scheduling, allowing workers within a chiplet to cooperatively execute tasks with coordinated cache reuse. On AMD Instinct MI350 with Qwen3-8B, Fleet achieves 1.3-1.5x lower decode latency than vLLM at batch sizes 1-8 through persistent kernel execution and per-chiplet scheduling. At larger batch sizes, cooperative weight tiling increases L2 hit rate (from 12% to 54% at batch size 32 and from 39% to 61% at batch size 64), reducing HBM traffic by up to 37% and delivering 1.27-1.30x speedup over a chiplet-unaware megakernel baseline.

cs.AR↗

arXiv2026

Memory DisOrder: Memory Re-orderings as a Timerless Side-channel

To improve efficiency, nearly all parallel processing units (CPUs and GPUs) implement relaxed memory models in which memory operations may be re-ordered, i.e., executed out-of-order. Prior testing work in this area found that memory re-orderings are observed more frequently when other cores are active, e.g., stressing the memory system, which likely triggers aggressive hardware optimizations. In this work, we present Memory DisOrder: a timerless side-channel that uses memory re-orderings to infer activity on other processes. We first perform a fuzzing campaign and show that many mainstream processors (X86/Arm/Apple CPUs, NVIDIA/AMD/Apple GPUs) are susceptible to cross-process signals. We then show how the vulnerability can be used to implement classic attacks, including a covert channel, achieving up to 16 bits/second with 95% accuracy on an Apple M3 GPU, and application fingerprinting, achieving reliable closed-world DNN architecture fingerprinting on several CPUs and an Apple M3 GPU. Finally, we explore how low-level system details can be exploited to increase re-orderings, showing the potential for a covert channel to achieve nearly 30K bits/second on X86 CPUs. More precise attacks can likely be developed as the vulnerability becomes better understood.

cs.CR↗

arXiv2025

tritonBLAS: Triton-based Analytical Approach for GEMM Kernel Parameter Selection

We present tritonBLAS, a fast and deterministic analytical model that uses architectural parameters like the cache hierarchy, and relative code and data placement to generate performant GPU GEMM kernels. tritonBLAS explicitly models the relationship between architectural topology, matrix shapes, and algorithmic blocking behavior to predict near-optimal configurations without runtime autotuning. Based on this model, we developed and implemented a lightweight GEMM framework entirely within Triton. We evaluate the performance of tritonBLAS across a diverse set of GEMM problem sizes on modern GPUs. tritonBLAS achieves over 95% of the performance of autotuning solutions, while reducing autotuning time to zero. This makes tritonBLAS a practical drop-in replacement for empirical tuning in production HPC and ML workloads.

cs.DC↗

arXiv2025

HipKittens: Fast and Furious AMD Kernels

AMD GPUs offer state-of-the-art compute and memory bandwidth; however, peak performance AMD kernels are written in raw assembly. To address the difficulty of mapping AI algorithms to hardware, recent work proposes C++ embedded and PyTorch-inspired domain-specific languages like ThunderKittens (TK) to simplify high performance AI kernel development on NVIDIA hardware. We explore the extent to which such primitives -- for explicit tile-based programming with optimized memory accesses and fine-grained asynchronous execution across workers -- are NVIDIA-specific or general. We provide the first detailed study of the programming primitives that lead to performant AMD AI kernels, and we encapsulate these insights in the HipKittens (HK) programming framework. We find that tile-based abstractions used in prior DSLs generalize to AMD GPUs, however we need to rethink the algorithms that instantiate these abstractions for AMD. We validate the HK primitives across CDNA3 and CDNA4 AMD platforms. In evaluations, HK kernels compete with AMD's hand-optimized assembly kernels for GEMMs and attention, and consistently outperform compiler baselines. Moreover, assembly is difficult to scale to the breadth of AI workloads; reflecting this, in some settings HK outperforms all available kernel baselines by $1.2-2.4\times$ (e.g., $d=64$ attention, GQA backwards, memory-bound kernels). These findings help pave the way for a single, tile-based software layer for high-performance AI kernels that translates across GPU vendors. HipKittens is released at: https://github.com/HazyResearch/HipKittens.

cs.LG↗