Benchmark of Error-Free Transforms for summation

This benchmark compares accuracy and efficiency of several summation algorithms in floating point arithmetics

def naive_sum(x):
    return choreo.segm.cython.eft_lib.SumK(x,0)

def np_sum(x):
    return np.sum(x)

def m_fsum(x):
    return m.fsum(x)

def SumK_1(x):
    return choreo.segm.cython.eft_lib.SumK(x,1)

def SumK_2(x):
    return choreo.segm.cython.eft_lib.SumK(x,2)

def SumK_3(x):
    return choreo.segm.cython.eft_lib.SumK(x,3)

def SumK_4(x):
    return choreo.segm.cython.eft_lib.SumK(x,4)

def SumK_5(x):
    return choreo.segm.cython.eft_lib.SumK(x,5)

def prepare_x(n):
    x = np.random.random(n)
    return {'x': x}


pyquickbench.plot_benchmark(
    all_times   ,
    all_sizes   ,
    all_funs    ,
    show = True ,
    title = "Measured CPU time"   ,
)

pyquickbench.plot_benchmark(
    all_times   ,
    all_sizes   ,
    all_funs    ,
    show = True ,
    relative_to_val = {pyquickbench.fun_ax_name: "naive_sum"},
    title = "Relative measured CPU time compared to naive_sum"   ,
)