Documentation

Benchmark name

Whether to include header information.

Command-line parser for Header

Levels of geometric difference in compared performance that triggers reporting.

>>> defaultCompareLevels
CompareLevels {errorLevel = 0.2, warningLevel = 5.0e-2, improvedLevel = 5.0e-2}

Command-line parser for CompareLevels

Options for production of a performance report.

Default options

>>> defaultReportOptions
ReportOptions {reportN = 1000, reportStatDType = StatAverage, reportMeasureType = MeasureTime, reportGolden = Golden {golden = "other/bench.perf", check = True, record = False}, reportHeader = Header, reportCompare = CompareLevels {errorLevel = 0.2, warningLevel = 5.0e-2, improvedLevel = 5.0e-2}}

Command-line parser for ReportOptions

Default command-line parser.

Report results

If a goldenFile is checked, and performance has degraded, the function will exit with ExitFailure so that 'cabal bench' and other types of processes can signal performance issues.

Run and report a benchmark to the console. For example,

reportMain "foo" (fap "sum" sum [1..1000]) would:

• run a benchmark for summing the numbers 1 to a thousand.
• look for saved performance data in other/foo-1000-MeasureTime-StatAverage.perf
• report on performance in isolation or versus the canned data file if it exists.
• exit with failure if the performace had degraded.

Run and report a benchmark to the console with the supplied options.

Write results to file

Read results from a file.

Comparison data between two results.

Compare two results and produce some notes given level triggers.

Format a result as a table.

Golden file options.

Default filepath is "other/bench.perf"

Parse command-line golden file options.

Replace the Golden file path with the suggested stem, but only if the user did not specify a specific file path at the command line.