benchmark Module (stx.benchmark)

scitex-benchmark — performance benchmarking + monitoring + profiling (standalone).

This module provides tools for benchmarking and monitoring the performance of SciTeX functions.

scitex.benchmark.benchmark_function(func, args=(), kwargs=None, iterations=10, warmup=2, input_size=None, measure_memory=False)

Benchmark a single function.

Parameters:

• func (Callable) – Function to benchmark

• args (tuple) – Arguments to pass to function

• kwargs (dict) – Keyword arguments to pass to function

• iterations (int) – Number of benchmark iterations

• warmup (int) – Number of warmup iterations

• input_size (str, optional) – Description of input size

• measure_memory (bool) – Whether to measure memory usage

Returns:

Benchmark results

Return type:

BenchmarkResult

scitex.benchmark.benchmark_module(module_name, pattern='test_*')

Create a benchmark suite for all matching functions in a module.

Parameters:

• module_name (str) – Name of module to benchmark

• pattern (str) – Pattern to match function names

Returns:

Suite containing all matching benchmarks

Return type:

BenchmarkSuite

class scitex.benchmark.BenchmarkResult(function_name, module, mean_time, std_time, min_time, max_time, iterations, input_size=None, memory_usage=None, notes=None)

Bases: object

Results from a benchmark run.

function_name: str

module: str

mean_time: float

std_time: float

min_time: float

max_time: float

iterations: int

input_size: str | None = None

memory_usage: float | None = None

notes: str | None = None

to_dict()

Convert to dictionary for easy serialization.

class scitex.benchmark.BenchmarkSuite(name)

Bases: object

Collection of benchmarks for a module or set of functions.

add_benchmark(func, test_data_generator, name=None, sizes=None)

Add a benchmark to the suite.

run(iterations=10, verbose=True)

Run all benchmarks in the suite.

Return type:

DataFrame

save_results(path)

Save benchmark results to CSV.

compare_with_baseline(baseline_path)

Compare current results with baseline.

Return type:

DataFrame

scitex.benchmark.run_all_benchmarks(output_dir='./benchmark_results')

Run all pre-defined benchmark suites.

Parameters:

output_dir (str) – Directory to save results

Returns:

Dictionary mapping suite names to results

Return type:

dict

scitex.benchmark.compare_implementations(implementations, test_data_generator, iterations=10, sizes=None)

Compare multiple implementations of the same functionality.

Parameters:

• implementations (dict) – Dictionary mapping implementation names to functions

• test_data_generator (callable) – Function that returns (args, kwargs) for testing

• iterations (int) – Number of iterations per implementation

• sizes (list, optional) – List of input sizes to test

Returns:

Comparison results

Return type:

pd.DataFrame

scitex.benchmark.profile_function(func)

Decorator to profile a function using the global profiler.

Example

>>> @profile_function
... def my_function(x):
...     return sum(range(x))

Return type:

Callable

scitex.benchmark.profile_module(module_name, pattern='*')

Profile all matching functions in a module.

Parameters:

• module_name (str) – Name of module to profile

• pattern (str) – Pattern to match function names

Returns:

Profiling results

Return type:

dict

scitex.benchmark.get_profile_report()

Get profiling report from global profiler.

Return type:

Dict[str, Any]

class scitex.benchmark.PerformanceMonitor(max_history=1000)

Bases: object

Monitor performance metrics for SciTeX functions.

Example

>>> monitor = PerformanceMonitor()
>>> monitor.start()
>>> # Your code here
>>> stats = monitor.get_stats()

start()

Start monitoring.

stop()

Stop monitoring.

record_metric(metric)

Record a performance metric.

add_alert_callback(callback)

Add a callback for performance alerts.

get_stats(function=None)

Get performance statistics.

Parameters:

function (str, optional) – Specific function to get stats for

Returns:

Performance statistics

Return type:

dict

get_recent_metrics(n=100)

Get n most recent metrics.

Return type:

List[PerformanceMetric]

save_metrics(path)

Save metrics to file.

load_metrics(path)

Load metrics from file.

clear()

Clear all metrics.

scitex.benchmark.track_performance(func)

Decorator to track function performance.

Example

>>> @track_performance
... def my_function(x):
...     return x ** 2

Return type:

Callable

scitex.benchmark.get_performance_stats(function=None)

Get performance statistics from global monitor.

Return type:

Dict[str, Any]