Pith. sign in

REVIEW 5 cited by

SwizzlePerf: Hardware-Aware LLMs for GPU Kernel Performance Optimization

Not yet reviewed by Pith; the record is open.

Re-run · record.json Download PDF Read on arXiv ↗

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2508.20258 v1 pith:DARY36CQ submitted 2025-08-27 cs.DC cs.AI

Arya Tschand , Muhammad Awad , Ryan Swann , Kesavan Ramakrishnan , Jeffrey Ma , Keith Lowery , Ganesh Dasika , Vijay Janapa Reddi This is my paper
classification cs.DCcs.AI
keywords performanceswizzleperfkernelkernelsllmsengineeringengineersgenerate
verification ladder T0 review T1 audit T2 compute T3 formal
0 comments
read the original abstract

Large language models (LLMs) have shown progress in GPU kernel performance engineering using inefficient search-based methods that optimize around runtime. Any existing approach lacks a key characteristic that human performance engineers rely on for near-optimal utilization -- hardware-awareness. By leveraging the workload's specific memory access patterns, architecture specifications, filtered profiling logs, and reflections on historical performance, we can make software-level optimizations that are tailored to the underlying hardware. SwizzlePerf automatically generates spatial optimizations for GPU kernels on disaggregated architectures by giving LLMs explicit hardware-awareness. For a GEMM kernel, SwizzlePerf takes less than 5 minutes to generate the same hardware-specific optimal swizzling pattern that took expert performance engineers 2 weeks to find. On a suite of 10 diverse ML and Science kernels, SwizzlePerf can generate swizzling patterns for 9 of the kernels that achieve up to a 2.06x speedup and 70% improvement in L2 hit rate. This work is the first of many steps toward systematically creating hardware-aware LLM performance engineering agents.

Discussion (0). Sign in to comment.

Forward citations

Cited by 5 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. JAXBench: Benchmarking Autonomous TPU Kernel Optimization

    cs.AI 2026-05 conditional novelty 6.0 of 10

    On a new 50-workload TPU benchmark (JAXBench), documentation-conditioned LLM agents reach 37.3% per-sample correctness and 1.28-1.36x geomean speedups over XLA, with curated context beating model scale.

  2. Kernel Foundry: A Diagnosis-driven Evolutionary Kernel Optimizer with Multi-Experts

    cs.NE 2026-05 unverdicted novelty 6.0 of 10

    Kernel Foundry improves GPU kernel correctness and performance on KernelBench via expert-guided evolutionary search with diagnostic feedback and an experience library, reaching 100% correctness on Level 2.

  3. LEO: Tracing GPU Stall Root Causes via Cross-Vendor Backward Slicing

    cs.DC 2026-04 unverdicted novelty 6.0 of 10

    LEO performs cross-vendor backward slicing from stalled GPU instructions to attribute root causes to source code, enabling optimizations that produce geometric-mean speedups of 1.73-1.82x on 21 workloads.

  4. CUDA-L2: Surpassing cuBLAS Performance for Matrix Multiplication through Reinforcement Learning

    cs.LG 2025-12 conditional novelty 5.0 of 10

    CUDA-L2's RL-generated HGEMM kernels are claimed to beat cuBLASLt autotuning by 11.4-15.9% on 1000 A100 shapes, though no independent artifacts are supplied.

  5. Towards Automated Kernel Generation in the Era of LLMs

    cs.LG 2026-01 conditional novelty 4.0 of 10

    A structured survey of LLM-based and agentic approaches for GPU kernel generation, plus a catalog of datasets and benchmarks for the field.

Pith tools