high performance computing on graphics processing units: hgpu.org

hgpu.org » LLM

PEAK: A Performance Engineering AI-Assistant for GPU Kernels Powered by Natural Language Transformations

Muhammad Usman Tariq, Abhinav Jangda, Angelica Moreira, Madan Musuvathi, Tyler Sorensen

View

Download (PDF)

Tags: AI, AMD, AMD Radeon Instinct MI200, ATI, Computer science, CUDA, HIP, HLSL, LLM, NLP, nVidia, nVidia RTX A6000

December 29, 2025 by hgpu

Beyond Code Pairs: Dialogue-Based Data Generation for LLM Code Translation

Le Chen1, Nuo Xu, Winson Chen, Bin Lei, Pei-Hung Lin, Dunzhi Zhou, Rajeev Thakur, Caiwen Ding, Ali Jannesari, Chunhua Liao

View

Download (PDF)

Tags: Code generation, Computer science, CUDA, LLM, nVidia, nVidia H200

December 21, 2025 by hgpu

TritonForge: Profiling-Guided Framework for Automated Triton Kernel Optimization

Haonan Li, Keyu Man, Partha Kanuparthy, Hanning Chen, Wei Sun, Sreen Tallam, Chenguang Zhu, Kevin Zhu, Zhiyun Qian

View

Download (PDF)

Source codes

Tags: Computer science, CUDA, LLM, Machine learning, nVidia, nVidia H100, Package, PyTorch, Triton

December 15, 2025 by hgpu

Chopper: A Multi-Level GPU Characterization Tool & Derived Insights Into LLM Training Inefficiency

Marco Kurzynski, Shaizeen Aga, Di Wu

View

Download (PDF)

Tags: AMD, AMD Radeon Instinct MI300X, ATI, Benchmarking, Computer science, Data parallelism, LLM, ROCm

December 15, 2025 by hgpu

QiMeng-Kernel: Macro-Thinking Micro-Coding Paradigm for LLM-Based High-Performance GPU Kernel Generation

Xinguo Zhu, Shaohui Peng, Jiaming Guo, Yunji Chen, Qi Guo, Yuanbo Wen, Hang Qin, Ruizhi Chen, Qirui Zhou, Ke Gao, Yanjun Wu, Chen Zhao, Ling Li

View

Download (PDF)

Tags: AI, Code generation, Computer science, CUDA, LLM, nVidia, nVidia A100, Triton

November 30, 2025 by hgpu

KernelBand: Boosting LLM-based Kernel Optimization with a Hierarchical and Hardware-aware Multi-armed Bandit

Dezhi Ran, Shuxiao Xie, Mingfang Ji, Ziyue Hua, Mengzhou Wu, Yuan Cao, Yuzhe Guo, Yu Hao, Linyi Li, Yitao Hu, Tao Xie

View

Download (PDF)

Tags: Code generation, Computer science, CUDA, LLM, nVidia, nVidia A100, nVidia GeForce RTX 4090, Triton

November 30, 2025 by hgpu

AIvailable: A Software-Defined Architecture for LLM-as-a-Service on Heterogeneous and Legacy GPUs

Pedro Antunes, Ana Rita Ortigoso, Gabriel Vieira, Daniel Fuentes, Luís Frazão, Nuno Costa, António Pereira

View

Download (PDF)

Tags: AMD Radeon Pro W6600, ATI, Computer science, CUDA, Heterogeneous systems, LLM, nVidia, nVidia GeForce GTX 1660, nVidia GeForce RTX 3070, ROCm

November 23, 2025 by hgpu

ProofWright: Towards Agentic Formal Verification of CUDA

Bodhisatwa Chatterjee, Drew Zagieboylo, Sana Damani, Siva Hari, Christos Kozyrakis

View

Download (PDF)

Tags: Code generation, Computer science, CUDA, LLM, nVidia

November 23, 2025 by hgpu

The Anatomy of a Triton Attention Kernel

Burkhard Ringlein, Jan van Lunteren, Radu Stoica, Thomas Parnell

View

Download (PDF)

Tags: AMD Radeon Instinct MI250, AMD Radeon Instinct MI300X, ATI, Computer science, CUDA, DSL, HIP, LLM, nVidia, nVidia H100, Performance, Programming Languages, Triton

November 23, 2025 by hgpu

PRAGMA: A Profiling-Reasoned Multi-Agent Framework for Automatic Kernel Optimization

Kelun Lei, Hailong Yang, Huaitao Zhang, Xin You, Kaige Zhang, Zhongzhi Luan, Yi Liu, Depei Qian

View

Download (PDF)

Tags: Code generation, Computer science, CUDA, LLM, nVidia, nVidia A100, Performance

November 16, 2025 by hgpu

RDMA Point-to-Point Communication for LLM Systems

Nandor Licker, Kevin Hu, Vladimir Zaytsev, Lequn Chen

View

Download (PDF)

Source codes

Tags: Computer science, CUDA, LLM, nVidia, nVidia H200, Package, Performance, RDMA

November 9, 2025 by hgpu

Serve Programs, Not Prompts

In Gim, Lin Zhong

View

Download (PDF)

Tags: Computer science, LLM, NLP

November 2, 2025 by hgpu

high performance computing on graphics processing units: hgpu.org

PEAK: A Performance Engineering AI-Assistant for GPU Kernels Powered by Natural Language Transformations

Beyond Code Pairs: Dialogue-Based Data Generation for LLM Code Translation

TritonForge: Profiling-Guided Framework for Automated Triton Kernel Optimization

Chopper: A Multi-Level GPU Characterization Tool & Derived Insights Into LLM Training Inefficiency

QiMeng-Kernel: Macro-Thinking Micro-Coding Paradigm for LLM-Based High-Performance GPU Kernel Generation

KernelBand: Boosting LLM-based Kernel Optimization with a Hierarchical and Hardware-aware Multi-armed Bandit

AIvailable: A Software-Defined Architecture for LLM-as-a-Service on Heterogeneous and Legacy GPUs

ProofWright: Towards Agentic Formal Verification of CUDA

The Anatomy of a Triton Attention Kernel

PRAGMA: A Profiling-Reasoned Multi-Agent Framework for Automatic Kernel Optimization

RDMA Point-to-Point Communication for LLM Systems

Serve Programs, Not Prompts

Recent source codes

KernelBenchX: A Comprehensive Benchmark for Evaluating LLM-Generated GPU Kernels

CUDA Kernel Fusion Benchmarks

IntelliKit: Agent-first tooling for AMD hardware

DITRON: Distributed Compiler based on Triton for Parallel Systems

CuTile Benchmark Suite: Performance and Productivity Tradeoffs for GPU Kernel Programming on Blackwell Architecture

Agentic Code Optimization via Compiler-LLM Cooperation

MegaTrain: Full Precision Training of 100B+ Parameter Large Language Models on a Single GPU

Device Virtual Machine (DVM)

AutoKernel: Autoresearch for GPU kernels. Give it any PyTorch model, go to sleep, wake up to optimized Triton kernels

Triton-Sanitizer: A Fast and Device-Agnostic Memory Sanitizer for Triton with Rich Diagnostic Context

Most viewed papers (last 30 days)