runtime_estimator

Layout and arithmetic primitives for repeated-candidate runtime estimates.

Functions

candidate_slope

Recover one candidate's marginal from N-versus-2N measurements.

effective_repeat_count

Round a candidate repeat count up to a balanced PP multiple.

fixed_intercept

Recover workload overhead that does not scale with candidate count.

homogeneous_layout

Return a candidate-only benchmark layout.

median_measurement

Return a component-wise robust center for repeated overhead estimates.

scaffolded_layout

Insert one fixed scaffold before each PP stage's candidate chunk.
candidate_slope(short, long, repeat_count)

Recover one candidate’s marginal from N-versus-2N measurements.

Parameters:
Return type:

RuntimeMeasurement

effective_repeat_count(configured, pp_size)

Round a candidate repeat count up to a balanced PP multiple.

Parameters:

  • configured (int)

  • pp_size (int)

Return type:

int

fixed_intercept(short, long)

Recover workload overhead that does not scale with candidate count.

Parameters:
Return type:

RuntimeMeasurement

homogeneous_layout(candidate, repeat_count)

Return a candidate-only benchmark layout.

Parameters:
Return type:

tuple[BlockConfig, …]

median_measurement(values)

Return a component-wise robust center for repeated overhead estimates.

Parameters:

values (list[RuntimeMeasurement])

Return type:

RuntimeMeasurement

scaffolded_layout(candidate, scaffold, repeat_count, pp_size)

Insert one fixed scaffold before each PP stage’s candidate chunk.

Parameters:
Return type:

tuple[BlockConfig, …]