cc

jit.py

"""
$ pip install numba
$ python jit.py

resultados:
$ python test_cc.py
Tempos com JIT
cc_0_jit - Min: 0.15884569298941642, Max: 0.1812134649953805, Mean: 0.16276181136781814
cc_1_jit - Min: 0.15603229100815952, Max: 0.18515240401029587, Mean: 0.16477074182126672
cc_2_jit - Min: 0.1618276800145395, Max: 0.25311970300390385, Mean: 0.17394073442847002
cc_3_jit - Min: 0.1659998889954295, Max: 0.2548448049928993, Mean: 0.17590742760221473
cc_3_loop_jit - Min: 0.1471281429985538, Max: 0.26632445299765095, Mean: 0.16636071566026658

Tempos sem JIT
cc_0 - Min: 0.06313921601395123, Max: 0.09617432300001383, Mean: 0.06712803481175797
cc_1 - Min: 0.07579782500397414, Max: 0.0936408819979988, Mean: 0.07817688702954911
cc_2 - Min: 0.08451158300158568, Max: 0.09043807699345052, Mean: 0.08629662264720536
cc_3 - Min: 0.0900561090093106, Max: 0.14318020700011402, Mean: 0.09299033465969841
cc_3_loop - Min: 58.21612359400024, Max: 64.1921016240085, Mean: 60.712612782407085
"""

from statistics import mean
from timeit import repeat

from numba import njit


def means(stmt, r=100):
    result = repeat(stmt, globals=globals(), repeat=r)
    return f'- Min: {min(result)}, Max: {max(result)}, Mean: {mean(result)}'


def cc_0(x):
    return x


def cc_1(x):
    if x:
        return x


def cc_2(x):
    if x:
        if x:
            return x


def cc_3(x):
    if x:
        if x:
            if x:
                return x


def cc_3_loop(x):
    for x in range(x):
        if x ** 2:
            if x ** 3:
                x + 3


cc_0_jit = njit(cc_0)
cc_1_jit = njit(cc_1)
cc_2_jit = njit(cc_2)
cc_3_jit = njit(cc_3)
cc_3_loop_jit = njit(cc_3_loop)


cc_0_jit(1000000000)  # primeira execução para compilação
cc_1_jit(1000000000)  # primeira execução para compilação
cc_2_jit(1000000000)  # primeira execução para compilação
cc_3_loop_jit(1000000000)  # primeira execução para compilação

print('Tempos com JIT')


print(f"cc_0_jit {means('cc_0_jit(1000000)')}")
print(f"cc_1_jit {means('cc_1_jit(1000000)')}")
print(f"cc_2_jit {means('cc_2_jit(1000000)')}")
print(f"cc_3_jit {means('cc_3_jit(1000000)')}")
print(f"cc_3_loop_jit {means('cc_3_loop_jit(1000000)')}")


print('Tempos sem JIT')
print(f"cc_0 {means('cc_0(1000000)')}")
print(f"cc_1 {means('cc_1(1000000)')}")
print(f"cc_2 {means('cc_2(1000000)')}")
print(f"cc_3 {means('cc_3(1000000)')}")
print(f"cc_3_loop {means('cc_3_loop(100)', r=5)}")

results.txt

$ python -mtimeit -s'import test' 'test.cc_0(10000000)'
5000000 loops, best of 5: 78.2 nsec per loop

$ python -mtimeit -s'import test' 'test.cc_1(10000000)'
5000000 loops, best of 5: 86.4 nsec per loop

$ python -mtimeit -s'import test' 'test.cc_2(10000000)'
5000000 loops, best of 5: 90.9 nsec per loop

$ python -mtimeit -s'import test' 'test.cc_3(10000000)'
5000000 loops, best of 5: 95.9 nsec per loop

test.py

def cc_0(x):
    return x


def cc_1(x):
    if x:
        return x


def cc_2(x):
    if x:
        if x:
            return x

def cc_3(x):
    if x:
        if x:
            if x:
                return x