mcframework.benchmark.BenchmarkResult

class mcframework.benchmark.BenchmarkResult

Bases: object

Timing outcome for one (backend, size) cell.

Attributes:

backendstr

Backend label (matches BackendSpec.label).

devicestr

Device the work ran on ("cpu", "mps", "cuda").

n_simulationsint

Number of simulation draws timed.

execution_timefloat

Best (minimum) wall-clock time across repeats, in seconds.

throughputfloat

n_simulations / execution_time (simulations per second).

mean_estimatefloat

The simulation’s mean estimate (a sanity signal, not a timing metric).

peak_memory_mbfloat or None

Peak device memory in MB when available (reserved for GPU backends).

Methods

classmethod __new__(*args, **kwargs)

__init__(backend: str, device: str, n_simulations: int, execution_time: float, throughput: float, mean_estimate: float, peak_memory_mb: float | None = None) → None