RTOri: Real Time Origami Toolkit

Making real-time origami pattern manipulation and fold simulation easy. Coming to you from the orilab.

On a high-level view, this project provides plugins, apps and bindings to manipulate and simulate origamis on a variety of platforms and with a diversity of use cases.

Goals

• open source, and open to contributions from the community
• provide software components to manipulate and simulate origamis
• modular, able to integrate new manipulations and new origami solvers & simulators seemlessly
• easy to integrate as plugins for existing apps, as well as in custom apps
• usable from multiple languages (C++, .NET, python, js, ...), crossplatform (Windows, macOS, Linux, Web, ...), and cross-architecture (x86_64, arm64, wasm)

Simulation Models

• Origami Simulator by Amanda Ghaessi
  • CPU (SIMD accelerated) 🚧
  • GPU 🚧
• ??

Apps, plugins, bindings

RTOri as Plugins

This project has been designed from the ground up to provide origami manipulation and simulation to existing apps as plugins, allowing designers, scientists, students, makers and artists around the world to access origamis in their existing workflows.

We thought about the following targets, please open an issue to express interest:

App Name	Manipulation	Simulation	Bindings
Touchdesigner	❌	🚧	C++
Blender	❌	❌	py
Rhino	❌	❌	.NET
↪ Grasshopper	❌	❌	.NET
Unity	❌	❌	.NET
Unreal	❌	❌	C++
Godot	❌	❌	rust/C++ ?

RTOri as Apps

• rtori-cli provides two operation kind:
  • convert to convert into and from the fold format
  • simulate to run simulations

RTOri as Bindings

Ecosystem	State
C++	🚧
.NET	❌
↪ Unity Burst	❌
py (cpython)	❌
Web (wasm)	❌

Software architecture

(mermaid graph)

flowchart TD
    classDef exposed fill:#f9f,stroke:#333,stroke-width:4px;


    RTORI[rtori library]:::exposed --> CORE
    RTORI --> FOLD

    subgraph "Exposed API"
        CORE[rtori-core]--> FOLD[rtori-fold]
    end

    subgraph "Origami Simulator"
    CORE[rtori-core] --> OSSIMD[rtori-os-simd]
    CORE --> OSWGPU[rtori-os-wgpu]

    OSSIMD --> OSLOADER[rtori-os-loader]
    OSWGPU --> OSLOADER
    OSLOADER --> FOLD
    end

Loading

TODO

1. Reorganize rtori-core into its own directory with supporting rust code
2. Rename rtori-os-simd (&-wgpu) to -os-simd
3. Compare performance with different x86_64 build configs (SSE2, SSE4.1, AVX, AVX2, AVX512, etc.) see https://en.wikipedia.org/wiki/X86-64#Microarchitecture_levels as inspiration (start with x86-64-v2 as a baseline)
   • See what makes sense to incorporate
4. Run the subset of targets with VTune & tune the performance (on all three)
5. Profile to see perf bottlenecks (on all three)
6. Redo step 3
7. If still there are differences, use stabby or stable_abi to create rust shared libraries that are loaded at runtime, depending on which interface is available, as well as a rtori-os-simd-loader crate that loads them on demand

Note: see https://github.com/HenrikBengtsson/x86-64-level / https://en.wikipedia.org/wiki/X86-64#Microarchitecture_levels for levels we can use as guidelines

Build Variants

Apple aarch64: M4: arm9.2-A without SVE target-cpu: "apple-m4" SHA2, FPARMv8, NEON, SME, SME2, SMEF64F64, SMEI16I64, ... https://github.com/llvm/llvm-project/blob/e5e38ddf1b8043324175868831da21e941c00aff/llvm/lib/Target/AArch64/AArch64Processors.td#L903 M3 == M2: arm8.6-A target-cpu: "apple-m2" M1: armv8.4-A target-cpu: "apple-m1"

Target Architectures

Potential support: (In LLVM/Rust triplet)

• thumbv8m.main-none-eavi: for Raspberry Pi RP2350 (rtori-core only)
• thumbv6m-none-eabi: for Raspberry Pi 2040 (rtori-core only)
• x86_64-pc-windows-msvc: for Windows on x86-64
• i686-pc-windows-msvc: for Windows on x86-32
• aarch64-pc-windows-msvc: for Windows on ARM
• x86_64-apple-darwin: for MacOs on Intel
• aarch64-apple-darwin: for MacOS on Apple Silicon
• x86_64-unknown-linux-gnu: for Linux on x86-64 (gnu)
• x86_64-unknown-linux-musl: for Linux on x86-64 (musl)
• arm-unknown-linux-gnueabihf: for Linux on armv6 (rpi 1)
• armv7-unknown-linux-gnueabihf: for Linux on rpi 2 rev 1.0
• aarch64-unknown-linux-gnu: for Linux on aarch64 (rpi 2 and later)
• aarch64-unknown-linux-musl: for Linux on aarch64 (rpi 2 and later) (musl)
• riscv32imac-unknown-none-elf: for rp2350 (riscv cores)
• wasm32-unknown-unknown: for web/wasm
• arm64ec-pc-windows-msvc: for windows on arm in "Emulation-Compatible" contexts

Contributors & Acknowledgements

Matthew Gardiner (Ars Elecronica Futurelab) Alexandre Bizri (Ars Electronica Futurelab)

Funded through the FWF Austrian Science Fund, PEEK Program (Project AR590) Origami Simulator by Amanda Ghassaei Amanda Ghassaei, Erik D. Demaine, and Neil Gershenfeld, “Fast, Interactive Origami Simulation using GPU Computation”, in Origami7: Proceedings of the 7th International Meeting on Origami in Science, Mathematics and Education (OSME 2018), volume 4, Oxford, England, September 5–7, 2018, pages 1151–1166, Tarquin.