跨设备的分布式 GPU 计算：C# 在浏览器和桌面上的应用

The SpawnDev.WebTorrent P2P network provides a natural foundation for distributed GPU compute. Connected devices exchange data over WebRTC, and the next step is to share compute workloads. SpawnDev.ILGPU introduces a new backend type, <code>AcceleratorType.P2P</code>, which transparently distributes kernels across connected devices. Developers write C# kernel code once, and it can run on the same <code>LoadAutoGroupedStreamKernel</code> API across 1 or 10 GPUs in a household. This approach simplifies the implementation of multi-device GPU computing, enhancing computational efficiency and flexibility.