Benchmark of inplace conjugation of arrays

This is a benchmark of different ways to perform inplace conjugation of a complex numpy array.

import numpy as np
import numba as nb
import pyquickbench

numba_opt_dict = {
    'nopython':True     ,
    'cache':True        ,
    'fastmath':True     ,
    'nogil':True        ,
}

def numpy_ufunc_outofplace(x):
    x = np.conjugate(x)

def numpy_ufunc_inplace(x):
    np.conjugate(x, out=x)

def numpy_inplace_mul(x):
    x.imag *= -1

def numpy_subs(x):
    x.imag = -x.imag

@nb.jit("void(complex128[::1])", **numba_opt_dict)
def numba_loop_typedef(x):

    for i in range(x.shape[0]):
        x.imag[i] = -x.imag[i]

@nb.jit(**numba_opt_dict)
def numba_loop(x):

    for i in range(x.shape[0]):
        x.imag[i] = -x.imag[i]

@nb.jit(**numba_opt_dict, parallel=True)
def numba_loop_parallel(x):

    for i in nb.prange(x.shape[0]):
        x.imag[i] = -x.imag[i]

all_funs = [
    numpy_ufunc_outofplace ,
    numpy_ufunc_inplace ,
    numpy_inplace_mul ,
    numpy_subs ,
    numba_loop_typedef ,
    numba_loop ,
    numba_loop_parallel ,
]

all_sizes = np.array([2**n for n in range(25)])

def prepare_x(n):
    x = np.random.random(n) + 1j*np.random.random(n)
    return [('x', x)]

basename = 'Inplace_conjugation_bench'
timings_filename = os.path.join(timings_folder, basename+'.npz')

n_repeat = 10

all_values = pyquickbench.run_benchmark(
    all_sizes                       ,
    all_funs                        ,
    setup = prepare_x               ,
    n_repeat = n_repeat             ,
    filename = timings_filename     ,
    ShowProgress = True             ,
    # ForceBenchmark = True           ,
)

pyquickbench.plot_benchmark(
    all_values                      ,
    all_sizes                       ,
    all_funs                        ,
    show = True                     ,
    title = 'Inplace conjugation'   ,
)

relative_to_val = {pyquickbench.fun_ax_name:"numpy_ufunc_inplace"}

pyquickbench.plot_benchmark(
    all_values                          ,
    all_sizes                           ,
    all_funs                            ,
    relative_to_val = relative_to_val   ,
    show = True                         ,
    title = 'Inplace conjugation'       ,
    ylabel = 'Time relative to numpy_ufunc_inplace' ,
)