RunMat provides a GPU‑accelerated engine that compiles and runs MATLAB‑syntax code across CPU and GPU cores, delivering faster math and physics simulations without requiring language migration. It includes real‑time 2D/3D plotting that operates directly on GPU‑resident data, and built‑in version control for files and runs, enabling teams to collaborate and track changes efficiently.
Funding
Funding not disclosed
Founders
Product
Problem
MATLAB‑based engineering workflows often run on CPU only, leading to long simulation times and limited ability to explore large parameter spaces. Teams must rewrite or port existing .m scripts to other languages or frameworks to gain performance, which disrupts established codebases and collaboration practices.
Solution
RunMat delivers a GPU‑accelerated math and physics engine that compiles MATLAB‑syntax code and schedules execution across CPU and GPU cores without requiring code changes. The platform automatically manages data movement, memory residency, and dispatch fusion, allowing elementwise operations and reductions to run at maximum speed. Real‑time 2D and 3D visualizations are generated directly from GPU‑resident variables, keeping plots interactive and eliminating costly data transfers. Engineers can access the environment through a web browser, desktop application, or command‑line interface, preserving existing development habits. Built‑in file and run version history captures edits and outputs, supporting collaborative review and reproducibility.
Target Audience
RunMat targets engineering teams, scientific researchers, and analysts who rely on MATLAB for simulation, modeling, and data analysis and need higher performance and collaborative tooling without rewriting their code.
Features
- Automatic compilation of .m scripts for hybrid CPU/GPU execution with fused dispatches to minimize overhead
- Transparent handling of data movement and memory residency, freeing users from low‑level hardware management
- GPU‑accelerated 2D/3D plotting engine that renders interactive figures directly from computation results
- Integrated diagnostics and shape/type analysis that catch linear‑algebra errors before runtime
- Versioned file and run tracking enabling compare, revert, and collaborative audit of calculations
- Multi‑modal access: browser‑based, native desktop client, and lightweight CLI installer
- Agent‑in‑the‑loop support with explicit read/write policies and reversible change sets for automated workflows