Adversarr's Computer Graphics toolbox

NOTE: This project is highly expremental.

ACG is a toolbox designed for study use mainly about Physical Simulation, but not Rendering.

Why yet another Computer Graphics Tool?

Although we have seen a huge number of libraries designed for CG, (and in many areas):

1. libigl: designed for geometry processing, the header-only coding style just leads to long compile time. The types of objects that igl-viewer can render is limitted.
2. Taichi: GPU-based Parallel computing library, designed for Physical Simulation. Taichi is originally a community-driven project, and lacks Debugger support, and exception support.

On my macbook with Apple's Silicon, Taichi's crash will leads to OS's crash.

Project Structure

1. Core:

   1. General Utilities:

      1. log: based on SPDLOG
      2. god: a meta programming library for Compile Time Computing
      3. json: import nlohmann json.
      4. lots of tiny staff: such as Result<T>, unified & cross-platform time module, etc.

   2. Init: Hookers for global variables designed for shared/static libraries.

   3. Mathematics Tools: Import Eigen, and defines some types, such as Field, TriangleList.

   4. Parallel computing tools: Import Taskflow

   5. SAD(Static Auto Diff): a high performance and highly optimized compile-time AutoDiff Library.

   6. …

2. Gui:

   • (Finished) Scene visualization: Support Mesh(including instancing), Particles, and Wireframe.
   • (Finished) ImGUI & Vulkan based. high performance
   • (Finished) ImGUI based User Interface, and User Callback.
   • (Finished) Perspective/Ortho projection modes. (for 2D data-visualization, similar to Unity2d's design.)
   • Scene Editing/Storaging/...

3. Port: Data import module.

4. Plugins (not implemented)

   • libigl/cgal/... integration.

5. Examples, see below.

Examples

This repo is designed for study use. Some implemented (just demos, not optimized, debugged) algorithms are listed below:

mass_spring

mass-spring system based on Projective Dynamics. Refer to "Fast mass spring" by Tiantian Liu.

and this provides an bonus solver, which applies a single step of Jacobi method on PD, which leads to real-time simulation of traditional mass-spring at about 30fps (on my laptop).

Now, mass spring will also show the dynamic computed grid (SubdivisonAABB) for further Broad Phase CCD app.

fem-implicit

Quasi Newton solver, the algorithm is same as example "hybred" below.

hybred

Quasi newton based solver for both cloth and softbody solve. (Implicit Euler)

• Left: Mass Spring
• Right: Ogden Neo-Hookean


Other Stuffs

1. fem-implicit: first-order explicit integrator. NeoHookean or StVK FEM Soft Body Simulation.
2. mpm-explicit: APIC-Fluid simulation, toy P2G and G2P process, weakly-compressible.