kompute: a cross-vendor GPU compute framework for high-performance acceleration across different graphics card brands

What it solves

Kompute is a general-purpose GPU compute framework designed to provide cross-vendor compatibility for graphics cards (including AMD, Qualcomm, NVIDIA, and others). It simplifies the process of writing GPU-accelerated applications, particularly for machine learning and data processing, without being locked into a single vendor's proprietary ecosystem.

How it works

Kompute is built on top of the Vulkan API, using a "Bring-your-own-Vulkan" design that allows it to integrate seamlessly with existing Vulkan applications. It provides a C++ SDK and a Python module for managing GPU resources. The framework uses a Manager to orchestrate device components, Tensors for structured data, and Algorithms (based on shaders) to execute logic on the GPU. It supports asynchronous and parallel processing through GPU family queues and provides explicit memory management for both host and GPU memory.

Who it’s for

It is intended for developers who need high-performance GPU acceleration across different hardware vendors, including those building machine learning models, mobile applications (via Android NDK), and game developers (e.g., using the Godot Engine).

Highlights

• Cross-Vendor Support: Works across AMD, NVIDIA, and Qualcomm GPUs.
• Mobile Enabled: Optimized for mobile environments with Android NDK support.
• Flexible Interfaces: Offers both a low-level C++ SDK for optimization and a high-level Python package for rapid experimentation.
• Asynchronous Execution: Supports parallel operations across multiple GPU queues.
• Industry Backing: Hosted project by the LF AI & Data Foundation.