A specialized migration tool for converting NVIDIA CUDA source code into Moore Threads MUSA architecture using automated text-based API mapping.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install musify
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install musify using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is Musify: CUDA to MUSA Migration Tool?
Musify is a powerful utility designed to bridge the gap between NVIDIA's CUDA platform and the Moore Threads MUSA architecture. As part of the ecosystem of Openclaw Skills, it automates the tedious process of renaming APIs, prefixes, and constants, allowing developers to port high-performance GPU kernels with minimal manual intervention.
By leveraging sophisticated text-based mapping, Musify identifies CUDA-specific patterns and replaces them with their MUSA equivalents. This is essential for developers looking to expand their software's hardware compatibility without rewriting entire libraries from scratch. The tool ensures that the core logic of kernels remains intact while the underlying hardware abstractions are correctly translated for the MUSA toolkit.
Musify: CUDA to MUSA Migration Tool Use Cases
- Porting existing NVIDIA CUDA kernels to Moore Threads GPUs for high-performance computing.
- Batch converting legacy C++ and CUDA projects to ensure MUSA compatibility across entire directories.
- Automating the migration of AI training or inference workloads to alternative hardware architectures.
- Maintaining dual-platform codebases with bi-directional conversion capabilities between CUDA and MUSA platforms.
How Musify: CUDA to MUSA Migration Tool Works
- Source Identification: The tool scans specified directories for source files with .cu, .cuh, .cpp, and .h extensions.
- API Mapping Application: It executes a text-based search-and-replace using a dictionary of CUDA to MUSA mappings, such as transforming cudaMalloc to musaMalloc.
- Exclusion Logic: The engine respects inline comments and markers to skip specific code segments that require manual preservation.
- File Transformation: Depending on the flags provided, Musify either modifies the files in-place or generates new, converted versions.
- Post-Processing: Developers review the automated changes and finalize the migration using the MUSA compiler (mcc) for verification.
Musify: CUDA to MUSA Migration Tool Setup
To begin using this tool within the Openclaw Skills framework, ensure you have the MUSA toolkit installed and then install the required Python dependency:
pip install ahocorapy
The musify-text utility will then be available for command-line execution. You can verify the installation by running musify-text --help.
Musify: CUDA to MUSA Migration Tool Data Schema & Taxonomy
Musify processes standard source code files and optional configuration files to manage the migration logic:
| Data Element | Type | Purpose |
|---|---|---|
| Source Code | Files (.cu, .cpp) | The input CUDA code to be converted. |
| Mapping File | JSON | Custom API translation rules to extend the default dictionary. |
| Exclusion Markers | Code Comments | Inline directives like MUSIFY_EXCL_LINE to prevent conversion. |
| Logs | Text Output | DEBUG, INFO, or WARNING level logs regarding the conversion process. |
Musify: CUDA to MUSA Migration Tool Advanced Features
- Support for custom JSON mapping files to handle proprietary or third-party library translations.
- Bi-directional conversion support allowing for MUSA to CUDA (m2c) migration paths.
- High-speed batch processing via integration with external tools like ripgrep and find.
- Granular block-level exclusion markers (START/STOP) for complex code sections that must remain unchanged.
- In-place file modification for seamless integration into automated build and migration pipelines.
SKILL.md
Loading
Related Openclaw Skills
Waiting Tips
A bilingual learning utility that displays curated AI and productivity tips to users while the agent processes their requests.
Social Copy Generator
A multi-platform content engine that transforms a single product description into 14+ optimized social media posts with a one-click copy dashboard.
Mac AI Optimizer
A comprehensive toolset to transform macOS into a high-performance AI server node by reclaiming system resources and optimizing background processes.
ClawMe Browser Interaction Skill
A powerful browser automation skill that lets AI agents perform actions in your live Chrome session with manual user confirmation.
Corpus
A powerful integration for AI agents to interact with the Corpus AI library for content retrieval, web bookmarking, and reminder management.
Joan Workflow
A workspace-based knowledge and task management system designed specifically for AI-assisted development workflows.
Featured*
