NVIDIA/CUDALibrarySamples

What it solves

This repository provides practical, GPU-accelerated examples for a wide range of high-performance computing tasks. It solves the problem of how to implement complex mathematical and data processing operations efficiently using NVIDIA's specialized CUDA libraries.

How it works

The project consists of a collection of sample applications that demonstrate the use of various NVIDIA CUDA libraries. These libraries handle the heavy lifting of GPU acceleration for specific domains such as linear algebra, signal processing, and image manipulation.

Who it’s for

It is designed for developers and researchers who need to perform high-performance computations on NVIDIA GPUs, specifically those working in fields like math, image processing, signal processing, and linear algebra.

Highlights

• Comprehensive Library Coverage: Includes examples for cuBLAS, cuTENSOR, cuSPARSE, cuSOLVER, cuFFT, cuRAND, and others.
• GPU-Accelerated Math: Provides implementations for basic linear algebra, tensor operations, and linear solvers.
• Image and Video Processing: Features tools for JPEG, JPEG2000, and TIFF encoding/decoding via nvJPEG and NPP.
• Specialized Computations: Includes support for post-quantum cryptography (cuPQC) and quantum chemistry (cuEST).
• Data Compression: Demonstrates efficient data compression and decompression using nvCOMP.