voiage.main_backends.benchmark_mps_vs_cpu

benchmark_mps_vs_cpu

benchmark_mps_vs_cpu([positional or keyword] net_benefit_array: ArrayLike = None, [positional or keyword] repeats: int = 1000, [positional or keyword] warmup_runs: int = 1, [positional or keyword] benchmark: Callable[[Backend, ArrayLike, int, int], dict[str, object]] = benchmark_memory_throughput) -> dict[str, object]

Compare a CPU benchmark payload against an optional Apple Metal payload.

The returned payload includes deterministic workload and runtime metadata for review packets, including enough information to produce a Phase-3 handoff artifact without any additional environment probing.

Parameters:

• net_benefit_array ArrayLike
• repeats int (default: 1000)
• warmup_runs int (default: 1)
• benchmark Callable[[Backend, ArrayLike, int, int], dict[str, object]] (default: benchmark_memory_throughput)

Returns: dict[str, object]