Open Source RoboticsRobots, software & learning

Stiff-GIPC

Software

CUDA contact-simulation research implementation for coupled stiff, soft and cloth scenes, with interactive and bounded batch execution. [1]

Installation & first result →Official websiteGetting started

For
Researchers investigating stiff/soft contact simulation [1]
Source rights
Some licenses are unresolved. Review licenses
Unknowns
compute requirements not fully established. Review gaps

Source researched · Not tested by us
Sources checked · Research details

Licenses & related hardware
License
not fully assessed for software, documentation · Details by part

Requirements · Research details and dates

On this page

Overview

Stiff-GIPC is a CUDA implementation of incremental potential contact methods for stiff, soft and coupled simulation scenes. Its value for robotics research is access to the numerical contact implementation and examples, not a ready-made robot-control stack. Upstream performance claims were not reproduced for this profile. [1]

Is this for you?

Known limitations
  • Batch mode still creates a GLUT window before hiding it.
  • Finite final vertices are not a convergence or physical-accuracy guarantee.
  • README commercial-negotiation wording remains unresolved against the MPL-2.0 license file.

Editorial assessment from cited documentation; no runtime testing.

[1] [3] [7] · Editorial assessment

What you need

Operating system
  • Windows
  • Linux

[1]

Dependencies
  • NVIDIA GPU and CUDA toolkit
  • C++17/CUDA17 toolchain and CMake
  • Eigen, GLEW, GLUT, OpenGL and nlohmann-json

[1] [2]

Installation & first result

Open the upstream getting-started guide [1]

  • Prepare the documented NVIDIA/CUDA and graphics build environment.
  • Configure with cmake -S . -B build and build with cmake --build build.
  • Select a case and bounded step count using GIPC_CASE and GIPC_STEPS as described in the overview.

Editorial assessment from cited documentation; no runtime testing.

[1] [2] [3] · Editorial assessment

Installation

Configure and build the pinned source with CMake after installing the CUDA/graphics dependencies; target name is gipc.

Editorial assessment from cited documentation; no runtime testing.

[1] [2] · Editorial assessment

Hardware, models & data

Inputs and outputs

Built-in scene selection advances simulated mesh vertices; batch mode can optionally dump final vertex data.

[3]

Project notes

Build for the target GPU and inspect a bounded run

The reviewed snapshot is bb2849a7b292099581907937860d96ecfdf42588, with no GitHub release listed. Upstream lists Windows/Linux, NVIDIA GPUs and CUDA 11 or newer. CMake additionally requires Eigen, GLEW, GLUT, OpenGL and nlohmann-json, sets C++/CUDA 17, and builds the gipc executable with native CUDA architecture selection. Check the actual toolkit, CMake and GPU combination rather than treating the README's CUDA minimum as a tested compatibility matrix. [1] [2] [6]

After provisioning those dependencies, a source-derived build route is cmake -S . -B build followed by cmake --build build. Assets and output locations are compiled as paths into the checkout, so preserve its directory layout. No build was run during this review. [2]

The program defaults to case 6, a large box-pipe-and-cloth scene. Set GIPC_CASE to an integer from 1 through 6 to choose another built-in case. On a Linux build, GIPC_CASE=1 GIPC_STEPS=10 ./build/gipc requests a short ten-step run of case 1. This is a proposed smoke check derived from the entry point, not a reproduced benchmark. [3]

Batch execution still initializes GLUT and creates a window before hiding it, so the step option does not remove the display/graphics-context dependency. At the end it checks vertices for non-finite values and can write their raw data if GIPC_DUMP_STATE is set. Passing that check would not establish solver convergence, collision accuracy or agreement with a physical experiment. [3]

Clarify commercial terms before relying on them

The license file contains MPL-2.0, while the README separately requests negotiation for commercial use. This profile leaves the intended relationship between those statements unresolved rather than asserting either blanket commercial permission or a settled prohibition. METIS and GKlib components also retain their own authors' rights. Source inspection does not establish deployment or redistribution clearance for a complete application. [1] [7]

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

Upstream last updated
Aug 31, 2026 [6]
Latest release
Not established
Profile updated
Oct 17, 2026
Sources last checked
Oct 17, 2026
Next review
Nov 17, 2026
Reviewed revision
commit snapshot; no GitHub releases listed / bb2849a7b292099581907937860d96ecfdf42588 · software repository

Source files and licenses

Available GPU simulation source; bounded-run behavior and unresolved commercial-use wording explicitly scoped.

LayerAvailability / license
softwareavailable · License not fully assessed
License file MPL-2.0; README requests commercial-use negotiation. Intended scope unresolved; external libraries retain their own terms. [1] [7]
documentationavailable · License not fully assessed
Root MPL-2.0 and README commercial-use negotiation wording need clarification. [1] [7]

The simulator does not require trained model weights.

The described path uses built-in scenes; individual asset rights remain separate.

References

Each source links to the exact revision we read. “Used for” shows where this page relies on it.

Correction guidance · Discussion is not enabled · Profile format 0.7.0-draft.5

Change history