"""Auto-generated example: 3 Float, 0 Categorical (with repeats)."""
import random
import math
import bencher as bn
class HashBenchmark(bn.ParametrizedSweep):
"""Hash throughput across key size, payload size, and iterations."""
key_size = bn.FloatSweep(default=32, bounds=[8, 256], doc="Key size in bytes")
payload_size = bn.FloatSweep(default=1024, bounds=[64, 65536], doc="Payload size in bytes")
iterations = bn.FloatSweep(default=100, bounds=[10, 1000], doc="Hash iterations")
throughput = bn.ResultFloat(units="MB/s", doc="Hash throughput")
def benchmark(self):
self.throughput = 500.0 / (1.0 + 0.5 * math.log2(self.key_size / 8)) / (1.0 + 0.3 * math.log2(self.payload_size / 64)) * (self.iterations / 100)
self.throughput += random.gauss(0, 0.15 * 30)
def example_sweep_3_float_0_cat_with_repeats(run_cfg: bn.BenchRunCfg | None = None) -> bn.Bench:
"""3 Float, 0 Categorical (with repeats)."""
bench = HashBenchmark().to_bench(run_cfg)
bench.plot_sweep(input_vars=['key_size', 'payload_size', 'iterations'], result_vars=['throughput'], description='A 3 float + 0 categorical parameter sweep with multiple repeats per combination. Repeating measurements reveals the noise structure of your benchmark. If your function is deterministic, all repeats will be identical; if it has stochastic components, repeats let you estimate confidence intervals and distinguish signal from noise. The benchmark function must be pure -- if past calls affect future calls through side effects, the statistics will be invalid. A 3D float sweep produces a volumetric representation. This is useful for visualising scalar fields in 3D parameter spaces.', post_description='Swarm/violin plots show the distribution of repeated measurements. If repeat has high variance, it suggests either measurement noise or unintended side effects in the benchmark function.')
return bench
if __name__ == "__main__":
bn.run(example_sweep_3_float_0_cat_with_repeats, subsampling_divisions=4, repeats=3)