KSLCore/ksl.simopt.benchmark/BenchmarkSummaryHeader

BenchmarkSummaryHeader

data class BenchmarkSummaryHeader(val experimentName: String, val macroReplications: Int, val replicationBudgetPerRun: Int, val confirmationTopK: Int?, val confirmationReplications: Int?, val verificationReplications: Int?, val numProblems: Int, val solverCaseDescriptions: Map<String, String>, val startTime: Instant, val tracesCaptured: Boolean, val solverStateCaptured: Boolean)(source)

What is known about a benchmark experiment before its first problem runs.

A sink needs this to open a record it can append to. It deliberately excludes everything that only exists once cells have run — solver configurations, traces, results — since the point of streaming is to write those as they arrive rather than at the end.

Parameters

experimentName

the experiment's name

macroReplications

the number of macro-replications this experiment runs

replicationBudgetPerRun

the per-cell replication budget

confirmationTopK

the confirmation stage's finalist count; null when confirmation is off

confirmationReplications

the confirmation stage's replications per candidate; null when confirmation is off

verificationReplications

the verification replication count; null when verification is off

numProblems

the number of problems the experiment covers, counting any a resumed run skips, so the record describes the study rather than this pass over it

solverCaseDescriptions

solver case label to its description

startTime

when the experiment's run() began

tracesCaptured

whether the experiment is capturing iteration traces

solverStateCaptured

whether the experiment is capturing solver-specific state

Constructors

BenchmarkSummaryHeader

constructor(experimentName: String, macroReplications: Int, replicationBudgetPerRun: Int, confirmationTopK: Int?, confirmationReplications: Int?, verificationReplications: Int?, numProblems: Int, solverCaseDescriptions: Map<String, String>, startTime: Instant, tracesCaptured: Boolean, solverStateCaptured: Boolean)