Pathfinder to GPU Offload in WASM

Explore the potential of GPU offloading in WebAssembly (WASM) through this 25-minute conference talk by Atanas Atanasov and Aaron Dorney from Intel. Delve into the growing importance of parallelization in AI workloads and the challenges of integrating GPU support with WASM. Learn about the current limitations of WASM in handling multithreaded, shared memory, vectorized, and device offloaded workloads. Discover the ongoing efforts to integrate CPU-based OpenMP with WASM and investigate various GPU offloading models such as OpenMP, CUDA, and SYCL. Understand the security implications and potential benefits of incorporating GPU offloading into WASM, which could lead to more secure, portable, and cloud-compatible AI applications.