how to optimize some algorithm in cuda.

NCCL Tests

RAFT contains fundamental widely-used algorithms and primitives for machine learning and information retrieval. The algorithms are CUDA-accelerated and form building blocks for more easily writing …

CUDA Kernel Benchmarking Library

CUDA Data Parallel Primitives Library

A simple high performance CUDA GEMM implementation.

Optimizing SGEMM kernel functions on NVIDIA GPUs to a close-to-cuBLAS performance.

Matrix Multiply-Accumulate with CUDA and WMMA( Tensor Core)

Instructions, Docker images, and examples for Nsight Compute and Nsight Systems

PyTorch-Based Fast and Efficient Processing for Various Machine Learning Applications with Diverse Sparsity

A tool for examining GPU scheduling behavior.

A repository where GPU applications are aggregated using a common build flow that supports multiple CUDA versions.

Sparse matrix computation library for GPU

Graphiler is a compiler stack built on top of DGL and TorchScript which compiles GNNs defined using user-defined functions (UDFs) into efficient execution plans.

High-Performance Streaming Graph Analytics on GPUs

HeteroSync is a benchmark suite for performing fine-grained synchronization on tightly coupled GPUs

Custom SpMM operations integrated into PyTorch

A class project for CS508 from UIUC to implement the classic BLAST algorithm using different GPU computing techniques.

Detect Strongly Connected Components in a directed graph with low average degree using parallel programming.

Matrix multiplication on CUDA