class true_north.Group(name: str | None = None)

Collection of benchmarks.

If name is not specified, file name and line number will be used instead.

add(func: Func | None = None, *, name: str | None = None, loops: int | None = None, repeats: int = 5, min_time: float = 0.2, timer: Timer = <built-in function perf_counter>) Callable[[Func], Check]

Register a new benchmark function in the group.

The first registered benchmark will be used as the baseline for all others.

  • name – if not specified, the function name will be used.

  • loops – how many times to run the benchmark in each repeat. If not specified, will be automatically detected to make each repeat last at least min_time seconds.

  • repeats – how many times repeat the benchmark (all loops). The results will show only the best repeat to reduce how external factors affect the results.

  • min_time – the minimum run time to target if loops is not specified.

  • timer – function used to get the current time.

print(config: ~true_north._config.Config = Config(stream=<_io.TextIOWrapper name='<stdout>' mode='w' encoding='utf-8'>, opcodes=False, allocations=False, histogram_lines=None)) None

Run all benchmarks in the group and print their results.

  • stream – the stream where to write all output. Default is stdout.

  • opcodes – count opcodes. Slow but reproducible.

  • allocations – track memory allocations. Slow but interesting.

class true_north.types.Check(name: str, func: Func, loops: int | None, repeats: int, min_time: float, timer: Timer)

A single benchmark.

Don’t instancinate directly, use Group.add decorator instead.

check_mallocs(lines: int, loops: int = 1) MallocResult

Run the benchmark and trace memory allocations.

check_opcodes(loops: int = 1, best: float = 0) OpcodesResult

Run the benchmark and count executed opcodes.

check_timing() TimingResult

Run benchmarks for the check.


class true_north.types.BaseResult
format_text() str

Represent the result as a human-friendly text.

class true_north.types.TimingResult(total_timings: list[float], each_timings: list[float])

The result of benchmarking a code execution time.

property best: float

The best of all total timings (repeats).

format_histogram(limit: int = 64, lines: int = 2) str

Histogram of timings (repeats).

format_text() str

Represent the timing result as a human-friendly text.

property loop_timings: list[float]

Execution time of each loop in a single repeat (bechmarking function call).

property stdev: float

Standard deviation of loops in a single repeat.

If there is only one loop in each repeat, use all repeats instead.

property total_timings: list[float]

Average time per loop for each repeat (bechmarking function call).

class true_north.types.OpcodesResult(opcodes: int, lines: int, timings: list[float], best: float)

The result of benchmarking opcodes executed by a code.

property durations: list[float]

How long it took to execute each opcode.

format_text() str

Generate a human-friendly representation of opcodes.

property lines: int

Number of lines of code executed.

See lnotab_notes.txt in CPython to learn more what is considered line.

property opcodes_count: int

Number of opcodes executed.

property timings: list[float]

The time when each opcode was executed.

class true_north.types.MallocResult(totals: list[int], allocs: list[Counter[str]])

The result of benchmarking memory allocations of a code.

property allocs: list[Counter[str]]

Memory allocations in each file for each sample.

Each item of the list is a Counter for a single sample. The Counter holds the number of allocations in each file.

format_text() str

Generate a human-friendly representation of memory allocations.

property total_allocs: int

Total memory allocations during the code execution.

property totals: list[int]

Total memory used by the code on each sample.