Pith. sign in

REVIEW 6 cited by

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

Not yet reviewed by Pith; the record is open.

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 2512.02551 v3 pith:ATFTBZLX submitted 2025-12-02 cs.LG cs.AI

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

classification cs.LG cs.AI
keywords cuda-l2kernelscublasmatmulcublaslthgemmtorchalgorithm
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved
0 comments
read the original abstract

In this paper, we propose CUDA-L2, a system that combines large language models (LLMs) and reinforcement learning (RL) to automatically optimize Half-precision General Matrix Multiply (HGEMM) CUDA kernels. Using CUDA execution speed as the RL reward, CUDA-L2 automatically optimizes HGEMM kernels across 1,000 configurations. CUDA-L2 systematically outperforms major matmul baselines to date, from the widely-used torch.matmul to state-of-the-art Nvidia's closed-source libraries, i.e., cuBLAS, cuBLASLt. In offline mode, where kernels are executed consecutively without time intervals, CUDA-L2 yields +22.0% over torch.matmul on average; +19.2% over cuBLAS using the optimal layout configuration (normal-normal NN and transposed-normal TN); +16.8% over cuBLASLt-heuristic, which queries cuBLASLt library and selects the algorithm based on the heuristic's suggestion; and +11.4% over the most competitive cuBLASLt-AutoTuning model, which selects the fastest algorithm from up to 100 candidates from cuBLASLt's suggestions. In server mode, where kernels are executed at random intervals simulating real-time inference, the speedups further increase to +28.7%, +26.0%, +22.4%, and +15.9% for torch.matmul, cuBLAS, cuBLASLt-heuristic, and cuBLASLt-AutoTuning respectively. CUDA-L2 shows that even the most performance-critical, heavily-optimized kernels like HGEMM can be improved through LLM-guided RL automation by systematically exploring configuration spaces at scales impractical for humans. Project and code can be found at github.com/deepreinforce-ai/CUDA-L2

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 6 Pith papers

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

  1. Step-TP: A Grounded, Step-Level Dataset with Chain-of-Thought Reasoning for LLM-Guided Tensor Program Optimization

    cs.LG 2026-05 unverdicted novelty 7.0

    Step-TP is a dataset providing grounded, atomic step-level IR transitions and CoT supervision to enable reliable multi-step LLM-guided tensor program optimization instead of end-to-end imitation.

  2. CODA: Rewriting Transformer Blocks as GEMM-Epilogue Programs

    cs.LG 2026-05 unverdicted novelty 7.0

    CODA re-expresses non-attention Transformer computations as GEMM epilogues using composable primitives for scaling, reductions, and transformations.

  3. CODA: Rewriting Transformer Blocks as GEMM-Epilogue Programs

    cs.LG 2026-05 unverdicted novelty 7.0

    CODA re-expresses most non-attention Transformer computations as GEMM-plus-epilogue programs using a constrained set of composable primitives to keep intermediate results on-chip and cut global memory traffic.

  4. AutoVecCoder: Teaching LLMs to Generate Explicitly Vectorized Code

    cs.CL 2026-05 unverdicted novelty 6.0

    AutoVecCoder combines VecPrompt for automated intrinsic knowledge synthesis and VecRL for efficiency-aligned RL to train an 8B LLM that achieves SOTA on SimdBench SSE/AVX subsets and sometimes exceeds -O3 compiler results.

  5. Kernel-Smith: A Unified Recipe for Evolutionary Kernel Optimization

    cs.CL 2026-03 unverdicted novelty 6.0

    Kernel-Smith combines evolutionary search with RL post-training to generate optimized GPU kernels, achieving SOTA speedups on KernelBench that beat Gemini-3.0-pro and Claude-4.6-opus on NVIDIA Triton and generalize to...

  6. AscendOptimizer: Episodic Agent for Ascend NPU Operator Optimization

    cs.LG 2026-03 unverdicted novelty 5.0

    AscendOptimizer combines kernel rewinding for reusable experience with evolutionary search on hardware feedback to optimize Ascend NPU operators, delivering 1.21x geometric-mean speedup and faster performance on 53.47...