Python Profiling

Basic Timing

import time

start = time.perf_counter()
result = my_function(data)
elapsed = time.perf_counter() - start
print(f"Time: {elapsed:.4f}s")

Benchmarking

import timeit

time_taken = timeit.timeit(
    lambda: my_function(data),
    number=100
)
print(f"Average: {time_taken / 100:.6f}s")

Function Profiling

import cProfile, pstats

profiler = cProfile.Profile()
profiler.enable()
result = my_function(data)
profiler.disable()

stats = pstats.Stats(profiler)
stats.sort_stats('cumulative').print_stats(20)

CLI: python -m cProfile -s cumulative script.py

Numba-Specific

from numba import njit, prange

@njit
def my_func(arr):
    return arr.sum()

my_func(np.array([1, 2, 3]))  # Compile
print(my_func.signatures)      # View types

# Parallel diagnostics
@njit(parallel=True)
def parallel_func(arr):
    result = 0.0
    for i in prange(len(arr)):
        result += arr[i]
    return result

parallel_func.parallel_diagnostics(level=4)

Benchmark JIT Functions

def benchmark_numba(func, data, n_warmup=3, n_runs=10):
    for _ in range(n_warmup):
        func(data)  # Warm-up
    times = []
    for _ in range(n_runs):
        start = time.perf_counter()
        func(data)
        times.append(time.perf_counter() - start)
    return np.mean(times), np.std(times)

Quick Decorator

import functools

def profile_time(func):
    @functools.wraps(func)
    def wrapper(*args, **kwargs):
        start = time.perf_counter()
        result = func(*args, **kwargs)
        print(f"{func.__name__}: {time.perf_counter() - start:.4f}s")
        return result
    return wrapper