Minimum Browser Versions for GPU.js Support

GPU.js is a high-performance JavaScript library that accelerates complex mathematical operations by compiling user-defined functions into shader language to run directly on the GPU. Because it relies entirely on WebGL and WebGL2 to access graphics hardware, running GPU.js effectively requires a browser that fully implements these web standards alongside essential hardware-accelerated floating-point texture extensions. This guide outlines the minimum desktop and mobile browser versions required to execute GPU.js workloads reliably and at peak performance.

WebGL 2.0: The Standard for Effective GPU.js Execution

While GPU.js can downgrade to a WebGL 1.0 context or even a CPU fallback, running it "effectively" necessitates WebGL 2.0. WebGL 2.0 provides native support for 32-bit floating-point textures, multiple render targets, and broader precision in fragment shaders, eliminating the performance bottlenecks and precision limitations inherent to WebGL 1.0.

Minimum Desktop Browser Requirements

To achieve optimal performance using WebGL 2.0 with GPU.js, the following desktop browser versions are required:

Minimum Mobile Browser Requirements

Mobile devices can execute GPU.js tasks provided their underlying system drivers and mobile browsers support modern graphics pipelines:

Absolute Minimum Fallback (WebGL 1.0 Support)

If hardware or client software does not support WebGL 2.0, GPU.js can fall back to WebGL 1.0, provided that the device supports the OES_texture_float extension. The absolute minimum browser versions for this restricted fallback mode are:

Critical Configuration Requirements

Beyond matching version numbers, a browser must have hardware acceleration enabled in its system settings to allow GPU.js to interface with the graphics card. If hardware acceleration is disabled, or if the underlying GPU drivers are blacklisted by the browser, GPU.js will fail to initialize the WebGL context and will automatically switch to CPU mode, resulting in significantly slower computation times.