high performance computing on graphics processing units: hgpu.org

Krishna Teja Chitty-Venkata, Siddhisanket Raskar, Bharat Kale, Farah Ferdaus, Aditya Tanikanti, Ken Raffenetti, Valerie Taylor, Murali Emani, Venkatram Vishwanath

View

Download (PDF)

Source codes

Tags: AI, AMD Radeon Instinct MI250, AMD Radeon Instinct MI300X, Artificial intelligence, ATI, Benchmarking, Computer science, CUDA, LLM, Machine learning, nVidia, nVidia A100, nVidia GH200, nVidia H100, OpenCL, Performance

November 10, 2024 by hgpu

NEO: Saving GPU Memory Crisis with CPU Offloading for Online LLM Inference

Xuanlin Jiang, Yang Zhou, Shiyi Cao, Ion Stoica, Minlan Yu

View

Download (PDF)

Tags: Artificial intelligence, Computer science, CUDA, LLM, Memory, nVidia, nVidia A10, nVidia H100, Tesla T4

November 10, 2024 by hgpu

Is the GPU Half-Empty or Half-Full? Practical Scheduling Techniques for LLMs

Ferdi Kossmann, Bruce Fontaine, Daya Khudia, Michael Cafarella, Samuel Madden

View

Download (PDF)

Tags: Computer science, LLM, NLP, nVidia, nVidia A100, nVidia H100, Task scheduling

November 3, 2024 by hgpu

high performance computing on graphics processing units: hgpu.org

Profiling Apple Silicon Performance for ML Training

CGP-Tuning: Structure-Aware Soft Prompt Tuning for Code Vulnerability Detection

LeetDecoding: A PyTorch Library for Exponentially Decaying Causal Linear Attention with CUDA Implementations

Finding Missed Code Size Optimizations in Compilers using LLMs

HPC-Coder-V2: Studying Code LLMs Across Low-Resource Parallel Languages

SoK: A Systems Perspective on Compound AI Threats and Countermeasures

Context Parallelism for Scalable Million-Token Inference

Improving Parallel Program Performance Through DSL-Driven Code Generation with LLM Optimizers

LLM-Inference-Bench: Inference Benchmarking of Large Language Models on AI Accelerators

NEO: Saving GPU Memory Crisis with CPU Offloading for Online LLM Inference

Is the GPU Half-Empty or Half-Full? Practical Scheduling Techniques for LLMs

Recent source codes

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

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

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

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

True 4-Bit Quantized CNN Training on CPU

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

KernelSkill: A Multi-Agent Framework for GPU Kernel Optimization

Most viewed papers (last 30 days)