Why Julia is so much slower than Python? Invitation to Euler Project

[gevis]

@tamasgal

The original claim that Julia is 52x slower is wrong . The best explanation is that you forgot to run the code with n=10 and thus measured the compilation time with n=100 .

No, and now I can reproduce it. Namely, running the following code

factorial_digit_sum(n::Int64) = n |> big |> factorial |> digits |> sum

n = 10
@time result = factorial_digit_sum(n)
@show result

n = 100
@time result = factorial_digit_sum(n)
@show result

for the first time in a terminal after computer reboot with the command “julia program_file.jl” produces the following output:

> 0.538902 seconds (184.37 k allocations: 9.549 MiB)
> result = 27
> 0.009945 seconds (1.42 k allocations: 30.969 KiB)
> result = 648

So, the main function was first run with n=10 and only then with n=100. However, the time it took for n=100 is exactly 9945 microseconds, what is about 52 times more than the corresponding code in Python (without the digits function, of course), which on the first run reports the following:

The sum of digits in 100! equals 648.
It took 0.0002079010009765625 seconds to get it.

The version of my Julia is 1.0.3.