high performance computing on graphics processing units: hgpu.org

hgpu.org » paper

KernelBench: Can LLMs Write Efficient GPU Kernels?

Anne Ouyang, Simon Guo, Simran Arora, Alex L. Zhang, William Hu, Christopher Ré, Azalia Mirhoseini

View

Download (PDF)

Source codes

Tags: AI, Benchmarking, Computer science, CUDA, LLM, Machine learning, nVidia, nVidia L40s, Package, PyTorch

February 24, 2025 by hgpu

Seamless acceleration of Fortran intrinsics via AMD AI engines

Nick Brown, Gabriel Rodríguez Canal

View

Download (PDF)

Source codes

Tags: AI, AMD, Computer science, Fortran, Linear Algebra, Package, Performance

February 24, 2025 by hgpu

Forecasting time series with constraints

Nathan Doumèche, Francis Bach, Éloi Bedek, Gérard Biau, Claire Boyer, Yannig Goude

View

Download (PDF)

Source codes

Tags: AI, Benchmarking, Computer science, Linear Algebra, Machine learning, nVidia, nVidia L4, Package

February 24, 2025 by hgpu

Evaluating the Performance of the DeepSeek Model in Confidential Computing Environment

Ben Dong, Qian Wang

View

Download (PDF)

Tags: Benchmarking, Cloud, Computer science, LLM, nVidia, nVidia A100, Performance, Security

February 24, 2025 by hgpu

The AI CUDA Engineer: Agentic CUDA Kernel Discovery, Optimization and Composition

Robert Tjarko Lange, Aaditya Prasad, Qi Sun, Maxence Faldor, Yujin Tang, David Ha

View

Download (PDF)

Source codes

Tags: AI, Computer science, CUDA, LLM, nVidia, nVidia H100, Package, Performance

February 24, 2025 by hgpu

cuSZp2: A GPU Lossy Compressor with Extreme Throughput and Optimized Compression Ratio

Yafan Huang, Sheng Di, Guanpeng Li, Franck Cappello

View

Download (PDF)

Source codes

Tags: Compression, Computer science, CUDA, nVidia, nVidia A100, nVidia GeForce RTX 3080, nVidia GeForce RTX 3090, Package, PTX

February 16, 2025 by hgpu

Leveraging LLVM OpenMP GPU Offload Optimizations for Kokkos Applications

Rahulkumar Gayatri, Shilei Tian, Stephen Olivier, Johannes Doerfert, Eric Wright

View

Download (PDF)

Source codes

Tags: AMD Radeon Instinct MI250X, ATI, Computer science, CUDA, HIP, MPI, nVidia, nVidia A100, OpenMP, Package, performance portability

February 16, 2025 by hgpu

InfiniteHiP: Extending Language Model Context Up to 3 Million Tokens on a Single GPU

Heejun Lee, Geon Park, Jaduk Suh, Sung Ju Hwang

View

Download (PDF)

Source codes

Tags: Computer science, LLM, NLP, nVidia, nVidia GeForce RTX 4090, Package

February 16, 2025 by hgpu

Teaching An Old Dog New Tricks: Porting Legacy Code to Heterogeneous Compute Architectures With Automated Code Translation

Nicolas Nytko, Andrew Reisner, J. David Moulton, Luke N. Olson, Matthew West

View

Download (PDF)

Tags: Code generation, Computer science, CUDA, Fortran, Heterogeneous systems, MPI, nVidia, nVidia A100, OpenCL

February 16, 2025 by hgpu

Vortex: Overcoming Memory Capacity Limitations in GPU-Accelerated Large-Scale Data Analytics

Yichao Yuan, Advait Iyer, Lin Ma, Nishil Talati

View

Download (PDF)

Source codes

Tags: AMD Radeon Instinct MI100, ATI, Computer science, CUDA, Databases, HIP, nVidia, nVidia A40, Package

February 16, 2025 by hgpu

Optimizing the optimizer increasing performance efficiency of modern compilers

Hafsah Shahzad

View

Download (PDF)

Tags: Compilers, Computer science, FPGA, HLS, Intel, nVidia, nVidia GeForce RTX 4070, Thesis

February 10, 2025 by hgpu

Demystifying Cost-Efficiency in LLM Serving over Heterogeneous GPUs

Youhe Jiang, Fangcheng Fu, Xiaozhe Yao, Guoliang He, Xupeng Miao, Ana Klimovic, Bin Cui, Binhang Yuan, Eiko Yoneki

View

Download (PDF)

Source codes

Tags: Benchmarking, Cloud, Computer science, Heterogeneous systems, LLM, nVidia, nVidia GeForce RTX 4090, nVidia H100, nVidia RTX A6000, Package

February 10, 2025 by hgpu

Data-efficient LLM Fine-tuning for Code Generation

DeepSpeed is a deep learning optimization library that makes distributed training and inference easy, efficient, and effective.

DeepCompile: A Compiler-Driven Approach to Optimizing Distributed Deep Learning Training

Large Language Model Powered C-to-CUDA Code Translation: A Novel Auto-Parallelization Framework

GigaAPI: a user-space API that simplifies multi-GPU programming, bridging the gap between the capabilities of parallel GPU systems and the ability of developers to harness their full potential

GigaAPI for GPU Parallelization

Coccinelle: a C code transformation engine using SmPL for matches, refactorings, and bug fixing

Advances in Semantic Patching for HPC-oriented Refactorings with Coccinelle

DuoReduce: MLIR's benchmark

Hardware-Assisted Software Testing and Debugging for Heterogeneous Computing

See all packages

* * *

high performance computing on graphics processing units: hgpu.org

Applications

KernelBench: Can LLMs Write Efficient GPU Kernels?

Seamless acceleration of Fortran intrinsics via AMD AI engines

Forecasting time series with constraints

The AI CUDA Engineer: Agentic CUDA Kernel Discovery, Optimization and Composition

cuSZp2: A GPU Lossy Compressor with Extreme Throughput and Optimized Compression Ratio

Leveraging LLVM OpenMP GPU Offload Optimizations for Kokkos Applications

InfiniteHiP: Extending Language Model Context Up to 3 Million Tokens on a Single GPU

Teaching An Old Dog New Tricks: Porting Legacy Code to Heterogeneous Compute Architectures With Automated Code Translation

Vortex: Overcoming Memory Capacity Limitations in GPU-Accelerated Large-Scale Data Analytics

Optimizing the optimizer increasing performance efficiency of modern compilers

Recent source codes

PELSI: Power-Efficient Layer-Switched Inference

Ouroboros: Virtualized Queues for dynamic memory management

MSCCL++: A GPU-driven communication stack for scalable AI applications

Benchmark compute shader of Unity against InteropUnityCUDA

Data-efficient LLM Fine-tuning for Code Generation

DeepSpeed is a deep learning optimization library that makes distributed training and inference easy, efficient, and effective.

Large Language Model Powered C-to-CUDA Code Translation: A Novel Auto-Parallelization Framework

GigaAPI: a user-space API that simplifies multi-GPU programming, bridging the gap between the capabilities of parallel GPU systems and the ability of developers to harness their full potential

Coccinelle: a C code transformation engine using SmPL for matches, refactorings, and bug fixing

DuoReduce: MLIR's benchmark

Most viewed papers (last 30 days)