Module benchmark 

Parse, pair, aggregate, compare, and report compact benchmark markers.

The normal pipeline is parse_benchmark_events, pair_benchmark_spans, aggregate_benchmark_spans, and optionally compare_benchmark_aggregates. Native report output lives in the repository’s benchmark_report example. Marker producers can use format_marker on the host or ic_testkit::performance in canister code.

Totals and run counts remain exact integers. Means use Metrics’ exact scaled integer projection, rounded down. Callers choose the scale and handle projection errors; reports use scale one in the original counter units. Comparisons retain current and previous means instead of percentage estimates.

Structs

BenchmarkAggregateError

Overflow rejects the complete aggregation, without returning partial totals.

BenchmarkAggregateRow

Exact totals and extrema for one scope and span label.

BenchmarkAverages

Floor-rounded scaled means of exact aggregate totals and run counts.

BenchmarkComparisonRow

Current and previous measurements, preserving missing rows and measured zero.

BenchmarkCounters

Exact instruction and byte counters reported by a marker.

BenchmarkMeasurement

Run count and scaled means for one present comparison side.

BenchmarkParseReport

Parsed events, rejected markers and ignored log noise.

BenchmarkParserConfig

Marker prefixes, suite naming and noise handling selected by the caller.

BenchmarkSpan

Matched start/end markers with nonnegative counter differences.

BenchmarkSpanReport

Pairing result retaining unmatched and invalid markers.

InvalidBenchmarkSpan

Matched markers whose counters decreased.

MalformedBenchmarkMarker

Rejected marker and its parse diagnostic.

RawBenchmarkEvent

Parsed marker with its original stream and line provenance.

Enums

BenchmarkAverageError

Why a benchmark’s scaled average cannot be represented.

BenchmarkComparisonError

A comparison refuses ambiguous row identities or unrepresentable means.

BenchmarkCounter

Counter whose aggregate cannot be represented.

BenchmarkEventKind

Whether a marker opens or closes a measured span.

BenchmarkEventSource

Stream from which a marker was collected.

BenchmarkMarkerError

A marker field cannot be represented in the line-oriented marker grammar.

BenchmarkScope

Structural identity of one suite or the cross-suite aggregate.

SuiteDerivation

How a marker label determines its suite.

Constants

DEFAULT_PREFIX

Default prefix for benchmark marker lines.

Functions

aggregate_benchmark_spans

Aggregate spans by named suite and across all suites.

compare_benchmark_aggregates

Pair current and previous means at the same caller-selected scale.

format_marker

Format one marker using the supplied prefix, label and counters.

pair_benchmark_spans

Pair markers by suite and label, using the most recent unmatched start.

pair_benchmark_spans_by_origin

Pair only markers with the same prefix, stream, suite and span label.

parse_benchmark_events

Parse markers without assigning a stream identity.

parse_benchmark_events_from_captured_output

Parse separately captured stdout and stderr.

parse_benchmark_events_from_source

Parse markers while retaining the supplied stream identity.

try_format_marker

Format one complete marker after admitting its prefix and start/end label.