All Tools
X
Dev ToolsFreeOpen Source
XNNPACK
High-efficiency NN operators for mobile, server, and Web inference
BSD-3-Clause
ABOUT
Mobile and Web inference needs fast conv, pooling, and elementwise kernels across ARM, x86, and WASM. XNNPACK supplies those operators so TFLite, ONNX Runtime, PyTorch, and ExecuTorch share one optimized backend instead of each framework rewriting SIMD kernels.
INTEGRATION GUIDE
1. Accelerate TensorFlow Lite models on ARM phones and x86 servers
2. Speed up ONNX Runtime and PyTorch mobile inference with XNNPACK kernels
3. Run WebAssembly neural nets in the browser via WASM SIMD
4. Back ExecuTorch and MediaPipe on-device pipelines with shared operators
TAGS
cinferencemobilewasmarmonnxtfliteexecutorchperformance