high performance computing on graphics processing units: hgpu.org

He Du, Qiming Ge, Jiakai Hu, Aijun Yang, Zheng Cai, Zixian Huang, Sheng Yuan, Qinxiu Cheng, Xinchen Xie, Yicheng Chen, Yining Li, Jiaxing Xie, Huanan Dong, Yaguang Wu, Xiangjun Huang, Jian Yang, Hui Wang, Bowen Zhou, Bowen Li, Qipeng Guo, Kai Chen

View

Download (PDF)

Tags: Computer science, CUDA, Heterogeneous systems, nVidia, Triton

April 13, 2026 by hgpu

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

Zhengqing Yuan, Hanchi Sun, Lichao Sun, Yanfang Ye

View

Download (PDF)

Source codes

Tags: Computer science, CUDA, LLM, nVidia, nVidia GH200, Package

April 13, 2026 by hgpu

DRTriton: Large-Scale Synthetic Data Reinforcement Learning for Triton Kernel Generation

Siqi Guo, Ming Lin, Tianbao Yang

View

Download (PDF)

Tags: Computer science, CUDA, LLM, nVidia, Triton

March 26, 2026 by hgpu

High-level Programming of Vulkan-based GPUs Through OpenMP

Ilias K. Kasmeridis, Vassilios V. Dimakopoulos

View

Download (PDF)

Tags: AMD Radeon RX 550, Computer science, CUDA, nVidia, OpenCL, OpenMP, Tesla P40, Vulkan

March 26, 2026 by hgpu

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

Evaluating CUDA Tile for AI Workloads on Hopper and Blackwell GPUs

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)

DVM: Real-Time Kernel Generation for Dynamic AI Models

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

AutoKernel: Autonomous GPU Kernel Optimization via Iterative Agent-Driven Search

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

LLM.Q: Quantized LLM training in pure CUDA/C++

LLMQ: Efficient Lower-Precision LLM Training for Consumer GPUs

SOL-ExecBench: Speed-of-Light Benchmarking for Real-World GPU Kernels Against Hardware Limits

True 4-Bit Quantized CNN Training on CPU

True 4-Bit Quantized Convolutional Neural Network Training on CPU: Achieving Full-Precision Parity

cuFuzz: A GPU-oriented coverage-guided fuzzer for userland CUDA application

Hunting CUDA Bugs at Scale with cuFuzz

See all packages

* * *

high performance computing on graphics processing units: hgpu.org

Applications

ARGUS: Agentic GPU Optimization Guided by Data-Flow Invariants

FACT: Compositional Kernel Synthesis with a Three-Stage Agentic Workflow

Revealing NVIDIA Closed-Source Driver Command Streams for CPU-GPU Runtime Behavior Insight

A Human–Machine Collaborative Tuning Framework for Triton Kernel Optimization on SIMD Platforms

Evaluating CUDA Tile for AI Workloads on Hopper and Blackwell GPUs

DVM: Real-Time Kernel Generation for Dynamic AI Models

Agentic Code Optimization via Compiler-LLM Cooperation

CuTeGen: An LLM-Based Agentic Framework for Generation and Optimization of High-Performance GPU Kernels using CuTe

Kernel-Smith: A Unified Recipe for Evolutionary Kernel Optimization

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

DRTriton: Large-Scale Synthetic Data Reinforcement Learning for Triton Kernel Generation

High-level Programming of Vulkan-based GPUs Through OpenMP

Recent source codes

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

LLM.Q: Quantized LLM training in pure CUDA/C++

SOL-ExecBench: Speed-of-Light Benchmarking for Real-World GPU Kernels Against Hardware Limits

True 4-Bit Quantized CNN Training on CPU

cuFuzz: A GPU-oriented coverage-guided fuzzer for userland CUDA application

Most viewed papers (last 30 days)