Engineering NotesAugust 29, 2026

Validating work on AMD XDNA1

A practical validation pattern for separating NPU execution from CPU orchestration and checking correctness before performance claims.
Accelerator work becomes difficult to reason about when correctness and performance are changed at the same time. Keep a CPU reference path that accepts the same inputs and produces a result that can be compared at meaningful boundaries. A successful program exit is not evidence that the intended kernel executed on the NPU. Keep runtime traces, device-visible counters, or other hardware evidence close to the test result. For transformer experiments, useful checkpoints include tensor shapes, selected intermediate values, and first-token decisions. Once those checks are stable, performance measurements become easier to trust.

Validation loop

  1. CPU reference
  2. NPU execution
  3. Compare outputs
  4. Record hardware evidence
A useful result includes what is not yet accelerated, what still runs on the host, the tested hardware, and the exact software/runtime assumptions. That makes later regressions and improvements measurable.