Quadrants
SoftwarePython kernel compiler for physics and robotics workloads, with platform-specific CPU and GPU backends and no promised Taichi backward compatibility. [1]
Installation & first result →Official websiteGetting started
- For
- Simulation developers writing numerical kernels [1]
- Source rights
- Named licenses are documented by part. Review licenses
- Unknowns
- compute requirements not fully established. Review gaps
Source researched · Not tested by us
Sources checked · Research details
Licenses & related hardware
- License
- Apache-2.0 · Details by part
Overview
Quadrants compiles numerical Python kernels for physics simulation and robotics workloads. It is useful when you are implementing the computation inside a simulator, rather than looking for robot models or a ready-made simulation scene. The project originated as a Taichi fork but explicitly does not promise backward compatibility. [1]
Is this for you?
What you need
Installation & first result
Open the upstream getting-started guide [3]
- Install the reviewed package version in a supported Python environment.
- Initialize a supported backend and run the documented first kernel.
- Synchronize before and after timing and separate compilation from repeated execution.
Editorial assessment from cited documentation; no runtime testing.
[3] · Editorial assessment
Hardware, models & data
- Inputs and outputs
Typed Python kernels operate on arrays or fields and produce numerical results on the chosen backend.
Project notes
Choose a backend for the actual platform
This review covers v1.3.3. Package metadata requires Python 3.10 through 3.13; install quadrants==1.3.3 to match the reviewed release. The upstream platform table is more specific than a general CPU/GPU support claim: CUDA is available on Linux x64 and Windows x86, AMD GPU support uses Linux x64 ROCm, and Metal is for Apple hardware. The ARM64 rows do not provide CUDA support even where vendor CUDA toolchains exist. [2] [4] [6]
The README requires ROCm 5.2 or newer for AMD GPU support. CUDA requires a host driver and runtime; the documented NVIDIA architecture range is SM 6.0 through SM 12.0. Check the pinned backend matrix against your machine before installing a larger simulator around it. These are upstream support declarations, not platforms tested by us. [1] [4]
Use the first kernel to check execution and timing
The getting-started guide initializes a GPU backend and implements a small numerical kernel over an array. Use that to check compilation and execution before a robotics workload. Kernel launches are asynchronous: the guide calls qd.sync() around timed work so the reported duration includes completion. Its later copy example performs a warm-up before repeated timings. Keep first compilation and repeated execution measurements separate. [3]
Existing Taichi applications need an API audit. Quadrants lists GUI/GGUI, the C API, ahead-of-time compilation, DirectX, mobile targets and OpenGL among removed components. A similar repository layout does not imply an old visualization or deployment path will work unchanged. [1]
Repository code and documentation use Apache-2.0 terms; dependencies retain their own terms. The basic kernel workflow needs no trained model or dataset. We inspected source at commit 00423caa93b620f6f3ebbb5d3f26e74cf56ecf5b without installing the compiler or reproducing upstream speed comparisons. [7] [6]
What we checked
Source researched only; no installation, training, benchmark or physical hardware test performed. No affiliate links or reused media.
Not yet established
- Requirements: host computer
- Version-specific hardware compatibility (ecosystem tags alone do not establish it)
Updates
Source files and licenses
Apache-2.0 compiler dependency with explicit platform and API boundaries.
| Layer | Availability / license |
|---|---|
| software | available · Apache-2.0 Repository materials only; external components retain their own terms. [7] |
| documentation | available · Apache-2.0 Repository materials only; external components retain their own terms. [7] |
Compiler kernel workflow does not require model weights.
First-kernel inputs are created by the example.
References
Each source links to the exact revision we read. “Used for” shows where this page relies on it.
- S001 Compiler scope and removed APIs
Revision, date & where cited
Genesis-Embodied-AI/quadrants contributors · github.com · revision
00423ca· Accessed Oct 17, 2026
Used for: Overview, Is this for you?, Installation & first result, Choose a backend for the actual platform and 1 more - S002 Python requirements
Revision, date & where cited
Genesis-Embodied-AI/quadrants contributors · github.com · revision
00423ca· Accessed Oct 17, 2026
Used for: What you need, Choose a backend for the actual platform - S003 First kernel and timing
Revision, date & where cited
Genesis-Embodied-AI/quadrants contributors · github.com · revision
00423ca· Accessed Oct 17, 2026
Used for: Installation & first result, Hardware, models & data, Use the first kernel to check execution and timing - S004 Platform and backend matrix
Revision, date & where cited
Genesis-Embodied-AI/quadrants contributors · github.com · revision
00423ca· Accessed Oct 17, 2026
Used for: Is this for you?, What you need, Choose a backend for the actual platform - S007 Apache license
Revision, date & where cited
Genesis-Embodied-AI/quadrants contributors · github.com · revision
00423ca· Accessed Oct 17, 2026
Used for: Use the first kernel to check execution and timing, Source files and licenses - S006 Reviewed release and revision
Revision, date & where cited
Genesis-Embodied-AI/quadrants / GitHub · github.com · revision
00423ca· v1.3.3 / · Accessed Oct 17, 2026
Used for: Installation & first result, Choose a backend for the actual platform, Use the first kernel to check execution and timing, What we checked
Correction guidance · Discussion is not enabled · Profile format 0.7.0-draft.5