Npunlock - Custom C kernels for Intel NPUs

Show HN: Npunlock – Run custom C kernels for Intel NPUs

▲ 41hsfzxjy10 HN discussion

Npunlock unlocks the programmable SHAVE cores inside Intel NPUs, letting you write and run custom C kernels directly on the hardware. While Intel's public stack only exposes graph-level programming, this tool reconstructs the missing path from C code to a runnable NPU kernel. Verified on Windows x64 with Meteor Lake / NPU3720, it supports FP32 and FP16 custom operations like GELU, and even mixed-precision branches in a single graph. With Python, CLI, and native C APIs, npunlock opens up new possibilities for NPU-accelerated computing. Contributions for Linux and newer NPUs are welcome.

Intel ships programmable SHAVE cores inside its NPUs, but the public stack exposes only graph-level programming. npunlock reconstructs the missing path from custom C code to a runnable NPU kernel.
Read original →HN discussionRSS

HN discussion

10
  1. Bayard_ne

    Finally, bare-metal NPU programming! Immediately thought of those custom, weird inference tasks Intel won't ever officially support.

  2. hsfzxjy

    Author here. Feel free to ask any questions :)

  3. Gigachad

    What is this useful for?

More from this day

2026-09-23