LuisaGroupOfficial·24 Agent Skills Included

LuisaCompute

Write GPU kernels in C++ and run them anywhere

Builds high-performance GPU compute and rendering programs from plain C++ or Python code. Compiles kernels automatically for CUDA, DirectX, Metal, Vulkan, and CPU without rewriting shaders per platform. Includes agent skills covering the DSL, runtime, compiler pipeline, testing, and build systems to speed up development.

1,043|108

↗

npx skills add LuisaGroup/LuisaCompute --all -g -y

All Skills in This Repository (24)

Pure Emerald Level Indicators

📦 In Repo

LuisaGroup

ast

Construct raw kernel and callable ASTs with FunctionBuilder in C++.

xmake

Automate cross-platform project configuration and builds with XMake.

lc_vstl

Documents vstd container APIs including HashMap, pools, queues, and smart pointers for LuisaCompute.

lc_core

Reference LuisaCompute core library APIs for types, math, logging, fibers, and STL containers.

yyjson

Parse, create, and modify JSON documents in C using the yyjson library.

lc_ast

Documents AST usage marking and CallOp propagation rules in LuisaCompute.

cmake

Configure and build LuisaCompute with CMake options, custom functions, and backend patterns.

spirv_codegen

Lowers XIR kernels to validated SPIR-V 1.5 for the Vulkan backend.

lsp

Provides clangd-powered C++ syntax checks and symbol navigation over an HTTP bridge.

lc_dsl

Write LuisaCompute DSL kernels, callables, structs, and GPU dispatch code in C++.

backend_architecture

Documents LuisaCompute backend plugin architecture, DeviceInterface API, and command encoding patterns.

glslang

Constructs SPIR-V modules programmatically using the glslang SpvBuilder API.

Official

Advanced

Frequently Asked Questions

FAQPage Schema

How to install LuisaCompute?▼

Run `npx skills add LuisaGroup/LuisaCompute --all -g -y` in your terminal to install all skills in this suite globally.

What is LuisaCompute used for?▼

It is a cross-platform GPU computing framework that lets you write kernels once in C++ or Python and run them on CUDA, DirectX, Metal, Vulkan, or CPU backends.

Do I need to learn a separate shader language?▼

No. Kernels are written directly in modern C++ using an embedded DSL, or in Python, and the framework compiles them to native GPU code automatically.

Can AI agents help me develop with LuisaCompute?▼

Yes. The included skills teach your AI agent the DSL syntax, runtime API, build systems, testing patterns, and debugging workflows for this framework.

Does LuisaCompute support ray tracing?▼

Yes. It provides hardware-accelerated ray tracing through acceleration structures, ray queries, and mesh or curve primitives on supported backends.