# LuisaGroup/LuisaCompute

High-Performance Rendering Framework on Stream Architectures

Repository: https://github.com/LuisaGroup/LuisaCompute
Canonical: https://ross.abutalabs.com/products/luisacompute
Language: C++
License: Apache-2.0
License Family: permissive
Topics: cpu, gpu, high-performance, cross-platform, cuda, directx, graphics, ispc, llvm, metal, raytracing, rendering, rtx, dxr, optix, dsl, siggraph-asia-2022, huawei-ascend
Last push: 2026-08-31T06:57:17+00:00

## Health v2 (maintenance only)
Score: 77/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 100, release rhythm 35, longevity 100
- inputs: {"age_days": 2112, "days_push": 2, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1043, forks 108 (observed 2026-09-01T02:13:55.522766+00:00)

## What it is
LuisaCompute is a high-performance cross-platform computing framework for graphics and beyond, featuring a C++-embedded DSL for GPU kernel programming with JIT code generation. It provides a unified runtime over multiple backends including CUDA, DirectX, Metal, LLVM, ISPC, and Huawei Ascend.

## Use cases
- write gpu kernels in c++ without writing cuda or metal directly
- build a cross-platform gpu renderer
- run the same compute code on nvidia, amd, apple, and cpu backends
- prototype high-performance graphics algorithms with a dsl
- use automatic differentiation on gpu for simulation or inverse rendering
- power robotics or physics simulation frameworks with gpu compute

## When to choose
- you need portable high-performance gpu compute across multiple vendor backends
- you are building a renderer, simulator, or physics engine in c++ or python
- you want jit-compiled kernels from an embedded dsl instead of managing per-backend code
- you need differentiable gpu computation

## When to avoid
- you only target a single gpu backend and prefer native cuda or metal directly
- you need a general-purpose ml training framework rather than a compute/graphics framework
- your project is not performance-sensitive and simpler libraries suffice
- you need stable long-term api guarantees for production enterprise software

## Facets
- artifact type: framework
- maturity: active
- function: graphics, simulation, compiler, gpu-computing, programming-language
- domain: graphics, gpu-computing, simulation, cross-platform
- platform: cpp, python, cross-platform, windows
- tags: dsl, raytracing, jit-compilation, rendering, high-performance-computing, llvm, metal, directx, vulkan, automatic-differentiation, game-development, cuda, linux, macos, gpu

## Member repositories
- LuisaGroup/LuisaCompute (main) score 77

## Provenance
- Observed fields: from GitHub, fetched 2026-09-01T02:13:55.522766+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T07:02:33.918163+00:00, confidence not recorded.
  - readme: https://github.com/LuisaGroup/LuisaCompute (fetched 2026-09-01T02:13:55.522766+00:00, sha f33e927ee940)
- Data as of 2026-08-30T08:39:29.467469+00:00.