[Paper Review] Bringing Together Dynamic Geometry Software and the Graphics Processing Unit

Aaron Montag, Jürgen Richter-Gebert|arXiv (Cornell University)|Aug 14, 2018

Computer Graphics and Visualization Techniques20 references4 citations

TL;DR

This paper presents a framework that seamlessly integrates GPU computing into dynamic geometry software (DGS) via automatic transcompilation of a high-level scripting language (CindyScript) into GPU-compatible GLSL shaders. By enabling direct GPU execution within a DGS environment using WebGL, it allows non-experts to perform real-time, parallel computations for mathematical visualizations, simulations, and exploratory research without learning low-level shader programming—demonstrated through applications like raytraced algebraic surfaces, fluid simulations, and interactive complex function visualizations.

ABSTRACT

We equip dynamic geometry software (DGS) with a user-friendly method that enables massively parallel calculations on the graphics processing unit (GPU). This interplay of DGS and GPU opens up various applications in education and mathematical research. The GPU-aided discovery of mathematical properties, interactive visualizations of algebraic surfaces (raycasting), the mathematical deformation of images and footage in real-time, and computationally demanding numerical simulations of PDEs are examples from the long and versatile list of new domains that our approach makes accessible within a DGS. We ease the development of complex (mathematical) visualizations and provide a rapid-prototyping scheme for general-purpose computations (GPGPU). The possibility to program both CPU and GPU with the use of only one high-level (scripting) programming language is a crucial aspect of our concept. We embed shader programming seamlessly within a high-level (scripting) programming environment. The aforementioned requires the symbolic process of the transcompilation of a high-level programming language into shader programming language for GPU and, in this article, we address the challenge of the automatic translation of a high-level programming language to a shader language of the GPU. To maintain platform independence and the possibility to use our technology on modern devices, we focus on a realization through WebGL.

Motivation & Objective

• To bridge the gap between CPU-based dynamic geometry software and the high-performance parallelism of GPUs for mathematical exploration and education.
• To enable non-expert users to perform complex, GPU-accelerated computations within a familiar scripting environment without learning low-level shader programming.
• To develop a platform-independent, browser-based solution using WebGL that supports real-time rendering and simulation on modern devices like tablets and mobile phones.
• To provide a rapid-prototyping environment for general-purpose GPU computing (GPGPU) within a DGS framework.
• To demonstrate the feasibility and educational value of GPU-accelerated mathematical visualizations through practical implementations in complex analysis, physics simulations, and numerical methods.

Proposed method

• The core method is automatic symbolic transcompilation of an untyped high-level scripting language (CindyScript) into OpenGL Shading Language (GLSL) for GPU execution.
• A recursive syntax tree analysis detects code segments suitable for GPU parallelization, while a fixed-point algorithm infers types to ensure correctness during transpilation.
• The framework leverages WebGL to ensure cross-platform compatibility and browser-based deployment without plugins, enabling use on modern devices including tablets and mobile phones.
• GPU computations are triggered via colorplot commands that read from textures, allowing data-parallel operations such as particle simulations and PDE solvers to be executed in parallel.
• The system integrates GPU-accelerated rendering and computation within a DGS environment, allowing interactive manipulation of geometric and algebraic objects.
• The approach supports advanced visualizations such as raycasting of algebraic surfaces, phase portraits of complex functions, and stereographic projections on the Riemann sphere.

Experimental results

Research questions

• RQ1 Can a high-level scripting language in a DGS be automatically transpiled into GPU-optimized GLSL code to enable general-purpose GPU computing?
• RQ2 To what extent can GPU-accelerated numerical simulations (e.g., n-body problems, Navier-Stokes equations) be integrated into a DGS environment without sacrificing usability?
• RQ3 How effectively can non-expert users create complex mathematical visualizations using a unified scripting interface that abstracts away low-level GPU programming?
• RQ4 What performance and usability improvements does GPU acceleration bring to dynamic geometry software on modern web-based platforms?
• RQ5 Can the integration of GPU computing in DGS enhance mathematical discovery and educational engagement in STEM fields?

Key findings

• The transcompiler successfully converts CindyScript into GLSL, enabling GPU execution within a DGS environment without requiring users to write shader code directly.
• Real-time simulations of 500 positively and 500 negatively charged particles with 10^6 interactions were achieved, demonstrating significant performance gains over CPU-only implementations.
• The simulation of the Navier-Stokes equations on the GPU was implemented with approximately one-fourth the lines of code required for a native WebGL implementation.
• Raycasting of algebraic surfaces and interactive visualizations of complex functions (e.g., phase portraits, Taylor expansions, Riemann sphere projections) were rendered in real time using GPU acceleration.
• Two bachelor’s theses demonstrated that students with no prior shader programming experience could create advanced GPU-accelerated visualizations and teaching materials using CindyGL.
• The framework enables deployment of interactive, GPU-accelerated mathematical content via a single HTML file, supporting broad accessibility on modern devices including tablets and mobile phones.