Why Python is so slow for a simple for loop?

Because you mention scientific code, have a look at numpy. What you're doing has probably already been done (or rather, it uses LAPACK for things like SVD). When you hear about python being used for scientific code, people probably aren't referring to using it in the way you do in your example.

As a quick example:

(If you're using python3, your example would use float division. My example assumes you're using python2.x, and therefore integer division. If not, specify i = np.arange(9999, dtype=np.float), etc)

import numpy as np
i = np.arange(9999)
j = np.arange(1, 9999)
print np.divide.outer(i,j).sum()

To give some idea of timing... (I'll use floating point division here, instead of integer division as in your example):

import numpy as np

def f1(num):
    total = 0.0
    for i in range(num): 
        for j in range(1, num):
            total += (float(i) / j)
    return total

def f2(num):
    i = np.arange(num, dtype=np.float)
    j = np.arange(1, num, dtype=np.float)
    return np.divide.outer(i, j).sum()

def f3(num):
    """Less memory-hungry (and faster) version of f2."""
    total = 0.0
    j = np.arange(1, num, dtype=np.float)
    for i in xrange(num):
        total += (i / j).sum()
    return total

If we compare timings:

In [30]: %timeit f1(9999)
1 loops, best of 3: 27.2 s per loop

In [31]: %timeit f2(9999)
1 loops, best of 3: 1.46 s per loop

In [32]: %timeit f3(9999)
1 loops, best of 3: 915 ms per loop

Why is Java faster than Python on this example loops?

Novice Explanation: Think of a program like a freight train that lays its own train-track as it moves forward. Track must be laid before the train can move. The Java Freight train can send thousands of track-layers ahead of the train, all working in parallel laying track many miles in advance, wheras python can only send one laboror at a time, and can only lay track 10 feet in front of where the train is.

Java has strong types and that affords the compiler to use JIT features: (https://en.wikipedia.org/wiki/Just-in-time_compilation) which enable the CPU to fetch memory and execute instructions in the future in parallel, before the instruction is needed. Java can 'sort of' run the instructions in your for loop in parallel with itself. Python has no concrete types and so the nature of the work to be done has to be decided at every instruction. This causes your entire computer to stop and wait for all the memory in all of your variables to be re-scanned. Meaning loops in python are polynomial O(n^2) time, wheras Java loops can be, and often are linear time O(n), due to strong types.

I think I am making mistakes because I know Python is used by lots of scientific projects.

They're heavily using SciPy (NumPy being the most prominent component, but I've heard the ecosystem that developed around NumPy's API is even more important) which vastly speeds up all kinds operations these projects need. There's what you are doing wrong: You aren't writing your critical code in C. Python is great for developing in general, but well-placed extension modules are a vital optimization in its own right (at least when you're crunching numbers). Python is a really crappy language to implement tight inner loops in.

The default (and for the time being most popular and widely-supported) implementation is a simple bytecode interpreter. Even the simplest operations, like an integer division, can take hundreds of CPU cycles, multiple memory accesses (type checks being a popular example), several C function calls, etc. instead of a few (or even single, in the case of integer division) instruction. Moreover, the language is designed with many abstractions which add overhead. Your loop allocates 9999 objects on the heap if you use xrange - far more if you use range (99999999 integer minus around 256256 for small integers which are cached). Also, the xrange version calls a method on each iteration to advance - the range version would too if iteration over sequences hadn't been optimized specifically. It still takes a whole bytecode dispatch though, which is itself vastly complex (compared to an integer division, of course).

It would be interesting to see what a JIT (I'd recommend PyPy over Psyco, the latter isn't actively developed anymore and very limited in scope anyway - it might work well for this simple example though). After a tiny fraction of iterations, it should produce a nigh-optimal machine code loop augmented with a few guards - simple integer comparisions, jumping if they fail - to maintain correctness in case you got a string in that list. Java can do the same thing, only sooner (it doesn't have to trace first) and with fewer guards (at least if you use ints). That's why it's so much faster.


I think NumPy can be faster than CPython for loops (I didn't test in PyPy).

I want to start from Joe Kington's code because this answer used NumPy.

%timeit f3(9999)
704 ms ± 2.33 ms per loop (mean ± std. dev. of 7 runs, 1 loop each)

by myself:

def f4(num):
    x=np.ones(num-1)
    y=np.arange(1,num)
    return np.sum(np.true_divide(x,y))*np.sum(y)

155 µs ± 284 ns per loop (mean ± std. dev. of 7 runs, 10000 loops each)

In addition, High School Mathematics can simplify the problem to computer.

Problem= (1+2+...+(num-1)) * (1/1+1/2+...+1/(num-1))
1+2+...+(num-1)=np.sum(np.arange(1,num))=num*(num-1)/2
1/1+1/2+...+1/(num-1)=np.true_divide (1,y)=np.reciprocal(y.astype(np.float64))

Therefore,

def f5(num):
    return np.sum(np.reciprocal(np.arange(1, num).astype(np.float64))) * num*(num-1)/2
%timeit f5(9999)
106 µs ± 615 ns per loop (mean ± std. dev. of 7 runs, 10000 loops each)

In addition, University Mathematics can simplify the problem to computer more.

1/1+1/2+...+1/(num-1)=np.log(num-1)+1/(2*num-2)+np.euler_gamma
(n>2)

np.euler_gamma: Euler-Mascheroni constant (0.57721566...)

Because of inaccuracy of Euler-Mascheroni constant in NumPy, You lose accuracy like 489223499.9991845 -> 489223500.0408554. If You can ignore 0.0000000085% inaccuracy, You can save more time.

def f6(num):
    return (np.log(num-1)+1/(2*num-2)+np.euler_gamma)* num*(num-1)/2
%timeit f6(9999)
4.82 µs ± 29.1 ns per loop (mean ± std. dev. of 7 runs, 100000 loops each)

Benefit of NumPy becomes larger with larger input.

%timeit f3(99999)
56.7 s ± 590 ms per loop (mean ± std. dev. of 7 runs, 1 loop each)
%timeit f5(99999)
534 µs ± 86.5 µs per loop (mean ± std. dev. of 7 runs, 1000 loops each)
%timeit f5(99999999)
1.42 s ± 15.4 ms per loop (mean ± std. dev. of 7 runs, 1 loop each)
9.498947911958**416**e+16
%timeit f6(99999999)
4.88 µs ± 26.7 ns per loop (mean ± std. dev. of 7 runs, 100000 loops each)
9.498947911958**506**e+16
%timeit f6(9999999999999999999)
17.9 µs ± 921 ns per loop (mean ± std. dev. of 7 runs, 10000 loops each)

In special case, You can use numba (Unfortunately not always).

from numba import jit
@jit
def f7(num):
    return (np.log(num-1)+1/(2*num-2)+np.euler_gamma)* num*(num-1)/2
# same code with f6(num)

%timeit f6(999999999999999)
5.63 µs ± 29.6 ns per loop (mean ± std. dev. of 7 runs, 100000 loops each)
f7(123) # compile f7(num)
%timeit f7(999999999999999)
331 ns ± 1.9 ns per loop (mean ± std. dev. of 7 runs, 1000000 loops each)
%timeit f7(9999)
286 ns ± 3.09 ns per loop (mean ± std. dev. of 7 runs, 1000000 loops each)

So, I recommend to use NumPy, mathematics and numba together.