Reference LuisaCompute core library APIs for types, math, logging, fibers, and STL containers.
npx skills add https://github.com/LuisaGroup/LuisaCompute --skill lc-core
Please help me install this Agent Skill. Skill: lc_core Source: https://github.com/LuisaGroup/LuisaCompute/tree/main/.agents/skills/lc_core Command: npx skills add https://github.com/LuisaGroup/LuisaCompute --skill lc-core
Developers working with the LuisaCompute C++ framework need quick, accurate access to its core library APIs—basic types, math functions, logging, fiber-based parallelism, and custom STL containers—without digging through headers or unit tests.
Ask how to use a specific LuisaCompute core API, such as creating a float4x4 transformation matrix or running a fiber-based parallel loop, and receive the correct header and code pattern.
High-intent search queries and answers about installing and using this skill.
Use aliases like float2, float3, float4 for vectors and float2x2, float3x3, float4x4 for matrices from <luisa/core/basic_types.h>. Construct them with make_float3(1,2,3) or make_float4x4 with row-major values; matrices default to identity.
Create a luisa::fiber::scheduler, then call luisa::fiber::parallel(count, lambda) for blocking execution or luisa::fiber::async_parallel for asynchronous execution returning a counter. The header is <luisa/core/fiber.h>, built on marl.
LuisaCompute provides LUISA_VERBOSE, LUISA_INFO, and LUISA_WARNING macros with fmt-style {} formatting, plus _WITH_LOCATION variants that include source location. Set verbosity with luisa::log_level_info() and related functions from <luisa/core/logging.h>.
Yes, the luisa namespace wraps std or EASTL containers under include/luisa/core/stl/, including vector, unordered_map, fixed_vector, lru_cache, optional, and variant. The implementation depends on the LUISA_USE_SYSTEM_STL build option.
Use luisa::DynamicModule::load(name) from <luisa/core/dynamic_module.h>, then retrieve symbols with address() or function<T>(). You can add search paths with DynamicModule::add_search_path before loading.
The mathematics header provides transpose, inverse, and determinant for 2x2/3x3/4x4 matrices, plus float4x4 builders like translation, scaling, and rotation. Vector operations include dot, cross, normalize, length, and distance.
yd5768365-hueAutomate C++ testing workflows with GoogleTest and CTest in CMake.
dongg622Compile and run Triton GPU kernels on曦云GPU systems.
cncorpEnforce test-first development using the red-green-refactor cycle.
copyleftdevRefactor C++ code to replace raw loops with STL algorithms.
