IRON / MLIR-AIE¶
Close-to-metal Python programming for AMD Ryzen™ AI NPUs.
IRON is an open-source toolkit that lets performance engineers write Python code that compiles directly to the AI Engine array inside AMD Ryzen™ AI processors — with full control over tile placement, data movement, and vectorized compute.
Get Started Install IRON, build and run your first NPU design.
Programming Guide Tiles, ObjectFifos, data movement, vectorization, and ML examples.
API & Internals Full reference for iron, taplib, the kernel library, and the MLIR dialects.
Mini Tutorial Five short exercises — a working NPU design in minutes.
What is IRON?¶
The NPU inside Ryzen™ AI is a 2D array of AI Engine tiles — small, fast, vector-capable cores connected by programmable stream switches and DMA engines. IRON exposes that architecture directly in Python:
import aie.iron as iron
from aie.iron import In, Out, ObjectFifo, Program, Runtime, Worker
from aie.iron.controlflow import range_
import numpy as np
@iron.jit
def vector_add_one(a_in: In, b_out: Out):
# Stream data from host → compute tile → host
of_in = ObjectFifo(np.ndarray[(1024,), np.dtype[np.int32]], name="in")
of_out = ObjectFifo(np.ndarray[(1024,), np.dtype[np.int32]], name="out")
def core_fn(of_in, of_out):
ai = of_in.acquire(1)
bo = of_out.acquire(1)
for i in range_(1024):
bo[i] = ai[i] + 1
of_in.release(1)
of_out.release(1)
w = Worker(core_fn, [of_in.cons(), of_out.prod()])
rt = Runtime()
with rt.sequence(np.ndarray[(1024,), np.dtype[np.int32]],
np.ndarray[(1024,), np.dtype[np.int32]]) as (a, b):
rt.start(w)
rt.fill(of_in.prod(), a)
rt.drain(of_out.cons(), b, wait=True)
return Program(iron.get_current_device(), rt).resolve_program()
a = iron.arange(1024, dtype=np.int32, device="npu")
b = iron.zeros(1024, dtype=np.int32, device="npu")
vector_add_one(a, b) # JIT-compiles on first call, cached after
@iron.jit compiles your design to an xclbin + instruction stream using the
LLVM/MLIR-based Peano compiler and runs it
on the attached NPU. Subsequent calls hit a cache.
Architecture¶
