Does it support NVIDIA performance libraries?

Absolutely. It includes specific API references and implementation patterns for cuBLAS, cuFFT, cuSPARSE, cuRAND, cuSolver, and Thrust.

How does it assist with kernel optimization?

It provides guidance on grid/block configuration, occupancy tuning using cudaOccupancyMaxPotentialBlockSize, and techniques for ensuring coalesced global memory access.

Can this skill help debug CUDA memory issues?

Yes, it provides standard error-checking patterns, memory alignment guidance, and workflows for utilizing Unified Memory to prevent common leaks and access violations.

Which CUDA version does this skill support?

The skill is configured with references for CUDA Toolkit v13.1, providing modern best practices for NVIDIA's latest parallel computing platforms.

CUDA GPU Computing

Name: CUDA GPU Computing
Author: datathings

bydatathings

•

Data Science & ML

Accelerates high-performance computing tasks by providing expert guidance on NVIDIA CUDA kernels, memory management, and parallel programming libraries.

The CUDA skill for Claude Code enables developers to build, optimize, and debug high-performance GPU-accelerated applications. It provides domain-specific knowledge for writing custom .cu kernels, configuring thread hierarchies, and managing device memory efficiently. Beyond low-level kernel development, it offers implementation patterns for the full suite of NVIDIA libraries including cuBLAS for linear algebra, cuFFT for signal processing, and Thrust for STL-like parallel algorithms, ensuring your code maximizes throughput on NVIDIA hardware.

Key Features

01Seamless integration support for cuBLAS, cuFFT, cuSPARSE, and cuDNN

02Performance tuning for memory coalescing and GPU occupancy

03Best practices for device memory management and Unified Memory implementation

04Expert guidance on .cu kernel development and thread hierarchy optimization

057 GitHub stars

06Advanced synchronization patterns using Cooperative Groups and Streams

Use Cases

01Accelerating deep learning and machine learning inference workloads

02Optimizing real-time graphics and compute-heavy game engine logic

03Developing high-performance linear algebra and scientific simulation routines

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add datathings/marketplace cuda

For use in Claude.ai and ChatGPT

Download Skill