This is the best algorithm I could come up.

def get_primes(n):
    numbers = set(range(n, 1, -1))
    primes = []
    while numbers:
        p = numbers.pop()
        primes.append(p)
        numbers.difference_update(set(range(p*2, n+1, p)))
    return primes

>>> timeit.Timer(stmt='get_primes.get_primes(1000000)', setup='import   get_primes').timeit(1)
1.1499958793645562

Can it be made even faster?

This code has a flaw: Since numbers is an unordered set, there is no guarantee that numbers.pop() will remove the lowest number from the set. Nevertheless, it works (at least for me) for some input numbers:

>>> sum(get_primes(开发者_运维知识库2000000))
142913828922L
#That's the correct sum of all numbers below 2 million
>>> 529 in get_primes(1000)
False
>>> 529 in get_primes(530)
True

---

Warning: timeit results may vary due to differences in hardware or version of Python.

Below is a script which compares a number of implementations:

• ambi_sieve_plain,
• rwh_primes,
• rwh_primes1,
• rwh_primes2,
• sieveOfAtkin,
• sieveOfEratosthenes,
• sundaram3,
• sieve_wheel_30,
• ambi_sieve (requires numpy)
• primesfrom3to (requires numpy)
• primesfrom2to (requires numpy)

Many thanks to stephan for bringing sieve_wheel_30 to my attention. Credit goes to Robert William Hanks for primesfrom2to, primesfrom3to, rwh_primes, rwh_primes1, and rwh_primes2.

Of the plain Python methods tested, with psyco, for n=1000000, rwh_primes1 was the fastest tested.

+---------------------+-------+
| Method              | ms    |
+---------------------+-------+
| rwh_primes1         | 43.0  |
| sieveOfAtkin        | 46.4  |
| rwh_primes          | 57.4  |
| sieve_wheel_30      | 63.0  |
| rwh_primes2         | 67.8  |    
| sieveOfEratosthenes | 147.0 |
| ambi_sieve_plain    | 152.0 |
| sundaram3           | 194.0 |
+---------------------+-------+

Of the plain Python methods tested, without psyco, for n=1000000, rwh_primes2 was the fastest.

+---------------------+-------+
| Method              | ms    |
+---------------------+-------+
| rwh_primes2         | 68.1  |
| rwh_primes1         | 93.7  |
| rwh_primes          | 94.6  |
| sieve_wheel_30      | 97.4  |
| sieveOfEratosthenes | 178.0 |
| ambi_sieve_plain    | 286.0 |
| sieveOfAtkin        | 314.0 |
| sundaram3           | 416.0 |
+---------------------+-------+

Of all the methods tested, allowing numpy, for n=1000000, primesfrom2to was the fastest tested.

+---------------------+-------+
| Method              | ms    |
+---------------------+-------+
| primesfrom2to       | 15.9  |
| primesfrom3to       | 18.4  |
| ambi_sieve          | 29.3  |
+---------------------+-------+

Timings were measured using the command:

python -mtimeit -s"import primes" "primes.{method}(1000000)"

with {method} replaced by each of the method names.

primes.py:

#!/usr/bin/env python
import psyco; psyco.full()
from math import sqrt, ceil
import numpy as np

def rwh_primes(n):
    # https://stackoverflow.com/questions/2068372/fastest-way-to-list-all-primes-below-n-in-python/3035188#3035188
    """ Returns  a list of primes < n """
    sieve = [True] * n
    for i in xrange(3,int(n**0.5)+1,2):
        if sieve[i]:
            sieve[i*i::2*i]=[False]*((n-i*i-1)/(2*i)+1)
    return [2] + [i for i in xrange(3,n,2) if sieve[i]]

def rwh_primes1(n):
    # https://stackoverflow.com/questions/2068372/fastest-way-to-list-all-primes-below-n-in-python/3035188#3035188
    """ Returns  a list of primes < n """
    sieve = [True] * (n/2)
    for i in xrange(3,int(n**0.5)+1,2):
        if sieve[i/2]:
            sieve[i*i/2::i] = [False] * ((n-i*i-1)/(2*i)+1)
    return [2] + [2*i+1 for i in xrange(1,n/2) if sieve[i]]

def rwh_primes2(n):
    # https://stackoverflow.com/questions/2068372/fastest-way-to-list-all-primes-below-n-in-python/3035188#3035188
    """ Input n>=6, Returns a list of primes, 2 <= p < n """
    correction = (n%6>1)
    n = {0:n,1:n-1,2:n+4,3:n+3,4:n+2,5:n+1}[n%6]
    sieve = [True] * (n/3)
    sieve[0] = False
    for i in xrange(int(n**0.5)/3+1):
      if sieve[i]:
        k=3*i+1|1
        sieve[      ((k*k)/3)      ::2*k]=[False]*((n/6-(k*k)/6-1)/k+1)
        sieve[(k*k+4*k-2*k*(i&1))/3::2*k]=[False]*((n/6-(k*k+4*k-2*k*(i&1))/6-1)/k+1)
    return [2,3] + [3*i+1|1 for i in xrange(1,n/3-correction) if sieve[i]]

def sieve_wheel_30(N):
    # http://zerovolt.com/?p=88
    ''' Returns a list of primes <= N using wheel criterion 2*3*5 = 30

Copyright 2009 by zerovolt.com
This code is free for non-commercial purposes, in which case you can just leave this comment as a credit for my work.
If you need this code for commercial purposes, please contact me by sending an email to: info [at] zerovolt [dot] com.'''
    __smallp = ( 2, 3, 5, 7, 11, 13, 17, 19, 23, 29, 31, 37, 41, 43, 47, 53, 59,
    61, 67, 71, 73, 79, 83, 89, 97, 101, 103, 107, 109, 113, 127, 131, 137, 139,
    149, 151, 157, 163, 167, 173, 179, 181, 191, 193, 197, 199, 211, 223, 227,
    229, 233, 239, 241, 251, 257, 263, 269, 271, 277, 281, 283, 293, 307, 311,
    313, 317, 331, 337, 347, 349, 353, 359, 367, 373, 379, 383, 389, 397, 401,
    409, 419, 421, 431, 433, 439, 443, 449, 457, 461, 463, 467, 479, 487, 491,
    499, 503, 509, 521, 523, 541, 547, 557, 563, 569, 571, 577, 587, 593, 599,
    601, 607, 613, 617, 619, 631, 641, 643, 647, 653, 659, 661, 673, 677, 683,
    691, 701, 709, 719, 727, 733, 739, 743, 751, 757, 761, 769, 773, 787, 797,
    809, 811, 821, 823, 827, 829, 839, 853, 857, 859, 863, 877, 881, 883, 887,
    907, 911, 919, 929, 937, 941, 947, 953, 967, 971, 977, 983, 991, 997)

    wheel = (2, 3, 5)
    const = 30
    if N < 2:
        return []
    if N <= const:
        pos = 0
        while __smallp[pos] <= N:
            pos += 1
        return list(__smallp[:pos])
    # make the offsets list
    offsets = (7, 11, 13, 17, 19, 23, 29, 1)
    # prepare the list
    p = [2, 3, 5]
    dim = 2 + N // const
    tk1  = [True] * dim
    tk7  = [True] * dim
    tk11 = [True] * dim
    tk13 = [True] * dim
    tk17 = [True] * dim
    tk19 = [True] * dim
    tk23 = [True] * dim
    tk29 = [True] * dim
    tk1[0] = False
    # help dictionary d
    # d[a , b] = c  ==> if I want to find the smallest useful multiple of (30*pos)+a
    # on tkc, then I need the index given by the product of [(30*pos)+a][(30*pos)+b]
    # in general. If b < a, I need [(30*pos)+a][(30*(pos+1))+b]
    d = {}
    for x in offsets:
        for y in offsets:
            res = (x*y) % const
            if res in offsets:
                d[(x, res)] = y
    # another help dictionary: gives tkx calling tmptk[x]
    tmptk = {1:tk1, 7:tk7, 11:tk11, 13:tk13, 17:tk17, 19:tk19, 23:tk23, 29:tk29}
    pos, prime, lastadded, stop = 0, 0, 0, int(ceil(sqrt(N)))
    # inner functions definition
    def del_mult(tk, start, step):
        for k in xrange(start, len(tk), step):
            tk[k] = False
    # end of inner functions definition
    cpos = const * pos
    while prime < stop:
        # 30k + 7
        if tk7[pos]:
            prime = cpos + 7
            p.append(prime)
            lastadded = 7
            for off in offsets:
                tmp = d[(7, off)]
                start = (pos + prime) if off == 7 else (prime * (const * (pos + 1 if tmp < 7 else 0) + tmp) )//const
                del_mult(tmptk[off], start, prime)
        # 30k + 11
        if tk11[pos]:
            prime = cpos + 11
            p.append(prime)
            lastadded = 11
            for off in offsets:
                tmp = d[(11, off)]
                start = (pos + prime) if off == 11 else (prime * (const * (pos + 1 if tmp < 11 else 0) + tmp) )//const
                del_mult(tmptk[off], start, prime)
        # 30k + 13
        if tk13[pos]:
            prime = cpos + 13
            p.append(prime)
            lastadded = 13
            for off in offsets:
                tmp = d[(13, off)]
                start = (pos + prime) if off == 13 else (prime * (const * (pos + 1 if tmp < 13 else 0) + tmp) )//const
                del_mult(tmptk[off], start, prime)
        # 30k + 17
        if tk17[pos]:
            prime = cpos + 17
            p.append(prime)
            lastadded = 17
            for off in offsets:
                tmp = d[(17, off)]
                start = (pos + prime) if off == 17 else (prime * (const * (pos + 1 if tmp < 17 else 0) + tmp) )//const
                del_mult(tmptk[off], start, prime)
        # 30k + 19
        if tk19[pos]:
            prime = cpos + 19
            p.append(prime)
            lastadded = 19
            for off in offsets:
                tmp = d[(19, off)]
                start = (pos + prime) if off == 19 else (prime * (const * (pos + 1 if tmp < 19 else 0) + tmp) )//const
                del_mult(tmptk[off], start, prime)
        # 30k + 23
        if tk23[pos]:
            prime = cpos + 23
            p.append(prime)
            lastadded = 23
            for off in offsets:
                tmp = d[(23, off)]
                start = (pos + prime) if off == 23 else (prime * (const * (pos + 1 if tmp < 23 else 0) + tmp) )//const
                del_mult(tmptk[off], start, prime)
        # 30k + 29
        if tk29[pos]:
            prime = cpos + 29
            p.append(prime)
            lastadded = 29
            for off in offsets:
                tmp = d[(29, off)]
                start = (pos + prime) if off == 29 else (prime * (const * (pos + 1 if tmp < 29 else 0) + tmp) )//const
                del_mult(tmptk[off], start, prime)
        # now we go back to top tk1, so we need to increase pos by 1
        pos += 1
        cpos = const * pos
        # 30k + 1
        if tk1[pos]:
            prime = cpos + 1
            p.append(prime)
            lastadded = 1
            for off in offsets:
                tmp = d[(1, off)]
                start = (pos + prime) if off == 1 else (prime * (const * pos + tmp) )//const
                del_mult(tmptk[off], start, prime)
    # time to add remaining primes
    # if lastadded == 1, remove last element and start adding them from tk1
    # this way we don't need an "if" within the last while
    if lastadded == 1:
        p.pop()
    # now complete for every other possible prime
    while pos < len(tk1):
        cpos = const * pos
        if tk1[pos]: p.append(cpos + 1)
        if tk7[pos]: p.append(cpos + 7)
        if tk11[pos]: p.append(cpos + 11)
        if tk13[pos]: p.append(cpos + 13)
        if tk17[pos]: p.append(cpos + 17)
        if tk19[pos]: p.append(cpos + 19)
        if tk23[pos]: p.append(cpos + 23)
        if tk29[pos]: p.append(cpos + 29)
        pos += 1
    # remove exceeding if present
    pos = len(p) - 1
    while p[pos] > N:
        pos -= 1
    if pos < len(p) - 1:
        del p[pos+1:]
    # return p list
    return p

def sieveOfEratosthenes(n):
    """sieveOfEratosthenes(n): return the list of the primes < n."""
    # Code from: <dickinsm@gmail.com>, Nov 30 2006
    # http://groups.google.com/group/comp.lang.python/msg/f1f10ced88c68c2d
    if n <= 2:
        return []
    sieve = range(3, n, 2)
    top = len(sieve)
    for si in sieve:
        if si:
            bottom = (si*si - 3) // 2
            if bottom >= top:
                break
            sieve[bottom::si] = [0] * -((bottom - top) // si)
    return [2] + [el for el in sieve if el]

def sieveOfAtkin(end):
    """sieveOfAtkin(end): return a list of all the prime numbers <end
    using the Sieve of Atkin."""
    # Code by Steve Krenzel, <Sgk284@gmail.com>, improved
    # Code: https://web.archive.org/web/20080324064651/http://krenzel.info/?p=83
    # Info: http://en.wikipedia.org/wiki/Sieve_of_Atkin
    assert end > 0
    lng = ((end-1) // 2)
    sieve = [False] * (lng + 1)

    x_max, x2, xd = int(sqrt((end-1)/4.0)), 0, 4
    for xd in xrange(4, 8*x_max + 2, 8):
        x2 += xd
        y_max = int(sqrt(end-x2))
        n, n_diff = x2 + y_max*y_max, (y_max << 1) - 1
        if not (n & 1):
            n -= n_diff
            n_diff -= 2
        for d in xrange((n_diff - 1) << 1, -1, -8):
            m = n % 12
            if m == 1 or m == 5:
                m = n >> 1
                sieve[m] = not sieve[m]
            n -= d

    x_max, x2, xd = int(sqrt((end-1) / 3.0)), 0, 3
    for xd in xrange(3, 6 * x_max + 2, 6):
        x2 += xd
        y_max = int(sqrt(end-x2))
        n, n_diff = x2 + y_max*y_max, (y_max << 1) - 1
        if not(n & 1):
            n -= n_diff
            n_diff -= 2
        for d in xrange((n_diff - 1) << 1, -1, -8):
            if n % 12 == 7:
                m = n >> 1
                sieve[m] = not sieve[m]
            n -= d

    x_max, y_min, x2, xd = int((2 + sqrt(4-8*(1-end)))/4), -1, 0, 3
    for x in xrange(1, x_max + 1):
        x2 += xd
        xd += 6
        if x2 >= end: y_min = (((int(ceil(sqrt(x2 - end))) - 1) << 1) - 2) << 1
        n, n_diff = ((x*x + x) << 1) - 1, (((x-1) << 1) - 2) << 1
        for d in xrange(n_diff, y_min, -8):
            if n % 12 == 11:
                m = n >> 1
                sieve[m] = not sieve[m]
            n += d

    primes = [2, 3]
    if end <= 3:
        return primes[:max(0,end-2)]

    for n in xrange(5 >> 1, (int(sqrt(end))+1) >> 1):
        if sieve[n]:
            primes.append((n << 1) + 1)
            aux = (n << 1) + 1
            aux *= aux
            for k in xrange(aux, end, 2 * aux):
                sieve[k >> 1] = False

    s  = int(sqrt(end)) + 1
    if s  % 2 == 0:
        s += 1
    primes.extend([i for i in xrange(s, end, 2) if sieve[i >> 1]])

    return primes

def ambi_sieve_plain(n):
    s = range(3, n, 2)
    for m in xrange(3, int(n**0.5)+1, 2): 
        if s[(m-3)/2]: 
            for t in xrange((m*m-3)/2,(n>>1)-1,m):
                s[t]=0
    return [2]+[t for t in s if t>0]

def sundaram3(max_n):
    # https://stackoverflow.com/questions/2068372/fastest-way-to-list-all-primes-below-n-in-python/2073279#2073279
    numbers = range(3, max_n+1, 2)
    half = (max_n)//2
    initial = 4

    for step in xrange(3, max_n+1, 2):
        for i in xrange(initial, half, step):
            numbers[i-1] = 0
        initial += 2*(step+1)

        if initial > half:
            return [2] + filter(None, numbers)

################################################################################
# Using Numpy:
def ambi_sieve(n):
    # http://tommih.blogspot.com/2009/04/fast-prime-number-generator.html
    s = np.arange(3, n, 2)
    for m in xrange(3, int(n ** 0.5)+1, 2): 
        if s[(m-3)/2]: 
            s[(m*m-3)/2::m]=0
    return np.r_[2, s[s>0]]

def primesfrom3to(n):
    # https://stackoverflow.com/questions/2068372/fastest-way-to-list-all-primes-below-n-in-python/3035188#3035188
    """ Returns a array of primes, p < n """
    assert n>=2
    sieve = np.ones(n/2, dtype=np.bool)
    for i in xrange(3,int(n**0.5)+1,2):
        if sieve[i/2]:
            sieve[i*i/2::i] = False
    return np.r_[2, 2*np.nonzero(sieve)[0][1::]+1]    

def primesfrom2to(n):
    # https://stackoverflow.com/questions/2068372/fastest-way-to-list-all-primes-below-n-in-python/3035188#3035188
    """ Input n>=6, Returns a array of primes, 2 <= p < n """
    sieve = np.ones(n/3 + (n%6==2), dtype=np.bool)
    sieve[0] = False
    for i in xrange(int(n**0.5)/3+1):
        if sieve[i]:
            k=3*i+1|1
            sieve[      ((k*k)/3)      ::2*k] = False
            sieve[(k*k+4*k-2*k*(i&1))/3::2*k] = False
    return np.r_[2,3,((3*np.nonzero(sieve)[0]+1)|1)]

if __name__=='__main__':
    import itertools
    import sys

    def test(f1,f2,num):
        print('Testing {f1} and {f2} return same results'.format(
            f1=f1.func_name,
            f2=f2.func_name))
        if not all([a==b for a,b in itertools.izip_longest(f1(num),f2(num))]):
            sys.exit("Error: %s(%s) != %s(%s)"%(f1.func_name,num,f2.func_name,num))

    n=1000000
    test(sieveOfAtkin,sieveOfEratosthenes,n)
    test(sieveOfAtkin,ambi_sieve,n)
    test(sieveOfAtkin,ambi_sieve_plain,n) 
    test(sieveOfAtkin,sundaram3,n)
    test(sieveOfAtkin,sieve_wheel_30,n)
    test(sieveOfAtkin,primesfrom3to,n)
    test(sieveOfAtkin,primesfrom2to,n)
    test(sieveOfAtkin,rwh_primes,n)
    test(sieveOfAtkin,rwh_primes1,n)         
    test(sieveOfAtkin,rwh_primes2,n)

Running the script tests that all implementations give the same result.

---

Faster & more memory-wise pure Python code:

def primes(n):
    """ Returns  a list of primes < n """
    sieve = [True] * n
    for i in range(3,int(n**0.5)+1,2):
        if sieve[i]:
            sieve[i*i::2*i]=[False]*((n-i*i-1)//(2*i)+1)
    return [2] + [i for i in range(3,n,2) if sieve[i]]

or starting with half sieve

def primes1(n):
    """ Returns  a list of primes < n """
    sieve = [True] * (n//2)
    for i in range(3,int(n**0.5)+1,2):
        if sieve[i//2]:
            sieve[i*i//2::i] = [False] * ((n-i*i-1)//(2*i)+1)
    return [2] + [2*i+1 for i in range(1,n//2) if sieve[i]]

Faster & more memory-wise numpy code:

import numpy
def primesfrom3to(n):
    """ Returns a array of primes, 3 <= p < n """
    sieve = numpy.ones(n//2, dtype=bool)
    for i in range(3,int(n**0.5)+1,2):
        if sieve[i//2]:
            sieve[i*i//2::i] = False
    return 2*numpy.nonzero(sieve)[0][1::]+1

a faster variation starting with a third of a sieve:

import numpy
def primesfrom2to(n):
    """ Input n>=6, Returns a array of primes, 2 <= p < n """
    sieve = numpy.ones(n//3 + (n%6==2), dtype=bool)
    for i in range(1,int(n**0.5)//3+1):
        if sieve[i]:
            k=3*i+1|1
            sieve[       k*k//3     ::2*k] = False
            sieve[k*(k-2*(i&1)+4)//3::2*k] = False
    return numpy.r_[2,3,((3*numpy.nonzero(sieve)[0][1:]+1)|1)]

A (hard-to-code) pure-python version of the above code would be:

def primes2(n):
    """ Input n>=6, Returns a list of primes, 2 <= p < n """
    n, correction = n-n%6+6, 2-(n%6>1)
    sieve = [True] * (n//3)
    for i in range(1,int(n**0.5)//3+1):
      if sieve[i]:
        k=3*i+1|1
        sieve[      k*k//3      ::2*k] = [False] * ((n//6-k*k//6-1)//k+1)
        sieve[k*(k-2*(i&1)+4)//3::2*k] = [False] * ((n//6-k*(k-2*(i&1)+4)//6-1)//k+1)
    return [2,3] + [3*i+1|1 for i in range(1,n//3-correction) if sieve[i]]

Unfortunately pure-python don't adopt the simpler and faster numpy way of doing assignment, and calling len() inside the loop as in [False]*len(sieve[((k*k)//3)::2*k]) is too slow. So I had to improvise to correct input (& avoid more math) and do some extreme (& painful) math-magic.

Personally I think it is a shame that numpy (which is so widely used) is not part of Python standard library, and that the improvements in syntax and speed seem to be completely overlooked by Python developers.

---

There's a pretty neat sample from the Python Cookbook here -- the fastest version proposed on that URL is:

import itertools
def erat2( ):
    D = {  }
    yield 2
    for q in itertools.islice(itertools.count(3), 0, None, 2):
        p = D.pop(q, None)
        if p is None:
            D[q*q] = q
            yield q
        else:
            x = p + q
            while x in D or not (x&1):
                x += p
            D[x] = p

so that would give

def get_primes_erat(n):
  return list(itertools.takewhile(lambda p: p<n, erat2()))

Measuring at the shell prompt (as I prefer to do) with this code in pri.py, I observe:

$ python2.5 -mtimeit -s'import pri' 'pri.get_primes(1000000)'
10 loops, best of 3: 1.69 sec per loop
$ python2.5 -mtimeit -s'import pri' 'pri.get_primes_erat(1000000)'
10 loops, best of 3: 673 msec per loop

so it looks like the Cookbook solution is over twice as fast.

---

Using Sundaram's Sieve, I think I broke pure-Python's record:

def sundaram3(max_n):
    numbers = range(3, max_n+1, 2)
    half = (max_n)//2
    initial = 4

    for step in xrange(3, max_n+1, 2):
        for i in xrange(initial, half, step):
            numbers[i-1] = 0
        initial += 2*(step+1)

        if initial > half:
            return [2] + filter(None, numbers)

Comparasion:

C:\USERS>python -m timeit -n10 -s "import get_primes" "get_primes.get_primes_erat(1000000)"
10 loops, best of 3: 710 msec per loop

C:\USERS>python -m timeit -n10 -s "import get_primes" "get_primes.daniel_sieve_2(1000000)"
10 loops, best of 3: 435 msec per loop

C:\USERS>python -m timeit -n10 -s "import get_primes" "get_primes.sundaram3(1000000)"
10 loops, best of 3: 327 msec per loop

---

The algorithm is fast, but it has a serious flaw:

>>> sorted(get_primes(530))
[2, 3, 5, 7, 11, 13, 17, 19, 23, 29, 31, 37, 41, 43, 47, 53, 59, 61, 67, 71, 73,
79, 83, 89, 97, 101, 103, 107, 109, 113, 127, 131, 137, 139, 149, 151, 157, 163,
167, 173, 179, 181, 191, 193, 197, 199, 211, 223, 227, 229, 233, 239, 241, 251,
257, 263, 269, 271, 277, 281, 283, 293, 307, 311, 313, 317, 331, 337, 347, 349,
353, 359, 367, 373, 379, 383, 389, 397, 401, 409, 419, 421, 431, 433, 439, 443,
449, 457, 461, 463, 467, 479, 487, 491, 499, 503, 509, 521, 523, 527, 529]
>>> 17*31
527
>>> 23*23
529

You assume that numbers.pop() would return the smallest number in the set, but this is not guaranteed at all. Sets are unordered and pop() removes and returns an arbitrary element, so it cannot be used to select the next prime from the remaining numbers.

---

For truly fastest solution with sufficiently large N would be to download a pre-calculated list of primes, store it as a tuple and do something like:

for pos,i in enumerate(primes):
    if i > N:
        print primes[:pos]

If N > primes[-1] only then calculate more primes and save the new list in your code, so next time it is equally as fast.

Always think outside the box.

---

If you don't want to reinvent the wheel, you can install the symbolic maths library sympy (yes it's Python 3 compatible)

pip install sympy

And use the primerange function

from sympy import sieve
primes = list(sieve.primerange(1, 10**6))

---

If you accept itertools but not numpy, here is an adaptation of rwh_primes2 for Python 3 that runs about twice as fast on my machine. The only substantial change is using a bytearray instead of a list for the boolean, and using compress instead of a list comprehension to build the final list. (I'd add this as a comment like moarningsun if I were able.)

import itertools
izip = itertools.zip_longest
chain = itertools.chain.from_iterable
compress = itertools.compress
def rwh_primes2_python3(n):
    """ Input n>=6, Returns a list of primes, 2 <= p < n """
    zero = bytearray([False])
    size = n//3 + (n % 6 == 2)
    sieve = bytearray([True]) * size
    sieve[0] = False
    for i in range(int(n**0.5)//3+1):
      if sieve[i]:
        k=3*i+1|1
        start = (k*k+4*k-2*k*(i&1))//3
        sieve[(k*k)//3::2*k]=zero*((size - (k*k)//3 - 1) // (2 * k) + 1)
        sieve[  start ::2*k]=zero*((size -   start  - 1) // (2 * k) + 1)
    ans = [2,3]
    poss = chain(izip(*[range(i, n, 6) for i in (1,5)]))
    ans.extend(compress(poss, sieve))
    return ans

Comparisons:

>>> timeit.timeit('primes.rwh_primes2(10**6)', setup='import primes', number=1)
0.0652179726976101
>>> timeit.timeit('primes.rwh_primes2_python3(10**6)', setup='import primes', number=1)
0.03267321276325674

and

>>> timeit.timeit('primes.rwh_primes2(10**8)', setup='import primes', number=1)
6.394284538007014
>>> timeit.timeit('primes.rwh_primes2_python3(10**8)', setup='import primes', number=1)
3.833829450302801

---

It's instructive to write your own prime finding code, but it's also useful to have a fast reliable library at hand. I wrote a wrapper around the C++ library primesieve, named it primesieve-python

Try it pip install primesieve

import primesieve
primes = primesieve.generate_primes(10**8)

I'd be curious to see the speed compared.

---

Here is two updated (pure Python 3.6) versions of one of the fastest functions,

from itertools import compress

def rwh_primes1v1(n):
    """ Returns  a list of primes < n for n > 2 """
    sieve = bytearray([True]) * (n//2)
    for i in range(3,int(n**0.5)+1,2):
        if sieve[i//2]:
            sieve[i*i//2::i] = bytearray((n-i*i-1)//(2*i)+1)
    return [2,*compress(range(3,n,2), sieve[1:])]

def rwh_primes1v2(n):
    """ Returns a list of primes < n for n > 2 """
    sieve = bytearray([True]) * (n//2+1)
    for i in range(1,int(n**0.5)//2+1):
        if sieve[i]:
            sieve[2*i*(i+1)::2*i+1] = bytearray((n//2-2*i*(i+1))//(2*i+1)+1)
    return [2,*compress(range(3,n,2), sieve[1:])]

---

I've updated much of the code for Python 3 and threw it at perfplot (a project of mine) to see which is actually fastest. Turns out that, for large n, primesfrom{2,3}to takes the cake:

---

Code to reproduce the plot:

import perfplot
from math import sqrt, ceil
import numpy as np
import sympy


def rwh_primes(n):
    # https://stackoverflow.com/questions/2068372/fastest-way-to-list-all-primes-below-n-in-python/3035188#3035188
    """ Returns  a list of primes < n """
    sieve = [True] * n
    for i in range(3, int(n ** 0.5) + 1, 2):
        if sieve[i]:
            sieve[i * i::2 * i] = [False] * ((n - i * i - 1) // (2 * i) + 1)
    return [2] + [i for i in range(3, n, 2) if sieve[i]]


def rwh_primes1(n):
    # https://stackoverflow.com/questions/2068372/fastest-way-to-list-all-primes-below-n-in-python/3035188#3035188
    """ Returns  a list of primes < n """
    sieve = [True] * (n // 2)
    for i in range(3, int(n ** 0.5) + 1, 2):
        if sieve[i // 2]:
            sieve[i * i // 2::i] = [False] * ((n - i * i - 1) // (2 * i) + 1)
    return [2] + [2 * i + 1 for i in range(1, n // 2) if sieve[i]]


def rwh_primes2(n):
    # https://stackoverflow.com/questions/2068372/fastest-way-to-list-all-primes-below-n-in-python/3035188#3035188
    """Input n>=6, Returns a list of primes, 2 <= p < n"""
    assert n >= 6
    correction = n % 6 > 1
    n = {0: n, 1: n - 1, 2: n + 4, 3: n + 3, 4: n + 2, 5: n + 1}[n % 6]
    sieve = [True] * (n // 3)
    sieve[0] = False
    for i in range(int(n ** 0.5) // 3 + 1):
        if sieve[i]:
            k = 3 * i + 1 | 1
            sieve[((k * k) // 3)::2 * k] = [False] * (
                (n // 6 - (k * k) // 6 - 1) // k + 1
            )
            sieve[(k * k + 4 * k - 2 * k * (i & 1)) // 3::2 * k] = [False] * (
                (n // 6 - (k * k + 4 * k - 2 * k * (i & 1)) // 6 - 1) // k + 1
            )
    return [2, 3] + [3 * i + 1 | 1 for i in range(1, n // 3 - correction) if sieve[i]]


def sieve_wheel_30(N):
    # http://zerovolt.com/?p=88
    """ Returns a list of primes <= N using wheel criterion 2*3*5 = 30

Copyright 2009 by zerovolt.com
This code is free for non-commercial purposes, in which case you can just leave this comment as a credit for my work.
If you need this code for commercial purposes, please contact me by sending an email to: info [at] zerovolt [dot] com."""
    __smallp = (
        2,
        3,
        5,
        7,
        11,
        13,
        17,
        19,
        23,
        29,
        31,
        37,
        41,
        43,
        47,
        53,
        59,
        61,
        67,
        71,
        73,
        79,
        83,
        89,
        97,
        101,
        103,
        107,
        109,
        113,
        127,
        131,
        137,
        139,
        149,
        151,
        157,
        163,
        167,
        173,
        179,
        181,
        191,
        193,
        197,
        199,
        211,
        223,
        227,
        229,
        233,
        239,
        241,
        251,
        257,
        263,
        269,
        271,
        277,
        281,
        283,
        293,
        307,
        311,
        313,
        317,
        331,
        337,
        347,
        349,
        353,
        359,
        367,
        373,
        379,
        383,
        389,
        397,
        401,
        409,
        419,
        421,
        431,
        433,
        439,
        443,
        449,
        457,
        461,
        463,
        467,
        479,
        487,
        491,
        499,
        503,
        509,
        521,
        523,
        541,
        547,
        557,
        563,
        569,
        571,
        577,
        587,
        593,
        599,
        601,
        607,
        613,
        617,
        619,
        631,
        641,
        643,
        647,
        653,
        659,
        661,
        673,
        677,
        683,
        691,
        701,
        709,
        719,
        727,
        733,
        739,
        743,
        751,
        757,
        761,
        769,
        773,
        787,
        797,
        809,
        811,
        821,
        823,
        827,
        829,
        839,
        853,
        857,
        859,
        863,
        877,
        881,
        883,
        887,
        907,
        911,
        919,
        929,
        937,
        941,
        947,
        953,
        967,
        971,
        977,
        983,
        991,
        997,
    )
    # wheel = (2, 3, 5)
    const = 30
    if N < 2:
        return []
    if N <= const:
        pos = 0
        while __smallp[pos] <= N:
            pos += 1
        return list(__smallp[:pos])
    # make the offsets list
    offsets = (7, 11, 13, 17, 19, 23, 29, 1)
    # prepare the list
    p = [2, 3, 5]
    dim = 2 + N // const
    tk1 = [True] * dim
    tk7 = [True] * dim
    tk11 = [True] * dim
    tk13 = [True] * dim
    tk17 = [True] * dim
    tk19 = [True] * dim
    tk23 = [True] * dim
    tk29 = [True] * dim
    tk1[0] = False
    # help dictionary d
    # d[a , b] = c  ==> if I want to find the smallest useful multiple of (30*pos)+a
    # on tkc, then I need the index given by the product of [(30*pos)+a][(30*pos)+b]
    # in general. If b < a, I need [(30*pos)+a][(30*(pos+1))+b]
    d = {}
    for x in offsets:
        for y in offsets:
            res = (x * y) % const
            if res in offsets:
                d[(x, res)] = y
    # another help dictionary: gives tkx calling tmptk[x]
    tmptk = {1: tk1, 7: tk7, 11: tk11, 13: tk13, 17: tk17, 19: tk19, 23: tk23, 29: tk29}
    pos, prime, lastadded, stop = 0, 0, 0, int(ceil(sqrt(N)))

    # inner functions definition
    def del_mult(tk, start, step):
        for k in range(start, len(tk), step):
            tk[k] = False

    # end of inner functions definition
    cpos = const * pos
    while prime < stop:
        # 30k + 7
        if tk7[pos]:
            prime = cpos + 7
            p.append(prime)
            lastadded = 7
            for off in offsets:
                tmp = d[(7, off)]
                start = (
                    (pos + prime)
                    if off == 7
                    else (prime * (const * (pos + 1 if tmp < 7 else 0) + tmp)) // const
                )
                del_mult(tmptk[off], start, prime)
        # 30k + 11
        if tk11[pos]:
            prime = cpos + 11
            p.append(prime)
            lastadded = 11
            for off in offsets:
                tmp = d[(11, off)]
                start = (
                    (pos + prime)
                    if off == 11
                    else (prime * (const * (pos + 1 if tmp < 11 else 0) + tmp)) // const
                )
                del_mult(tmptk[off], start, prime)
        # 30k + 13
        if tk13[pos]:
            prime = cpos + 13
            p.append(prime)
            lastadded = 13
            for off in offsets:
                tmp = d[(13, off)]
                start = (
                    (pos + prime)
                    if off == 13
                    else (prime * (const * (pos + 1 if tmp < 13 else 0) + tmp)) // const
                )
                del_mult(tmptk[off], start, prime)
        # 30k + 17
        if tk17[pos]:
            prime = cpos + 17
            p.append(prime)
            lastadded = 17
            for off in offsets:
                tmp = d[(17, off)]
                start = (
                    (pos + prime)
                    if off == 17
                    else (prime * (const * (pos + 1 if tmp < 17 else 0) + tmp)) // const
                )
                del_mult(tmptk[off], start, prime)
        # 30k + 19
        if tk19[pos]:
            prime = cpos + 19
            p.append(prime)
            lastadded = 19
            for off in offsets:
                tmp = d[(19, off)]
                start = (
                    (pos + prime)
                    if off == 19
                    else (prime * (const * (pos + 1 if tmp < 19 else 0) + tmp)) // const
                )
                del_mult(tmptk[off], start, prime)
        # 30k + 23
        if tk23[pos]:
            prime = cpos + 23
            p.append(prime)
            lastadded = 23
            for off in offsets:
                tmp = d[(23, off)]
                start = (
                    (pos + prime)
                    if off == 23
                    else (prime * (const * (pos + 1 if tmp < 23 else 0) + tmp)) // const
                )
                del_mult(tmptk[off], start, prime)
        # 30k + 29
        if tk29[pos]:
            prime = cpos + 29
            p.append(prime)
            lastadded = 29
            for off in offsets:
                tmp = d[(29, off)]
                start = (
                    (pos + prime)
                    if off == 29
                    else (prime * (const * (pos + 1 if tmp < 29 else 0) + tmp)) // const
                )
                del_mult(tmptk[off], start, prime)
        # now we go back to top tk1, so we need to increase pos by 1
        pos += 1
        cpos = const * pos
        # 30k + 1
        if tk1[pos]:
            prime = cpos + 1
            p.append(prime)
            lastadded = 1
            for off in offsets:
                tmp = d[(1, off)]
                start = (
                    (pos + prime)
                    if off == 1
                    else (prime * (const * pos + tmp)) // const
                )
                del_mult(tmptk[off], start, prime)
    # time to add remaining primes
    # if lastadded == 1, remove last element and start adding them from tk1
    # this way we don't need an "if" within the last while
    if lastadded == 1:
        p.pop()
    # now complete for every other possible prime
    while pos < len(tk1):
        cpos = const * pos
        if tk1[pos]:
            p.append(cpos + 1)
        if tk7[pos]:
            p.append(cpos + 7)
        if tk11[pos]:
            p.append(cpos + 11)
        if tk13[pos]:
            p.append(cpos + 13)
        if tk17[pos]:
            p.append(cpos + 17)
        if tk19[pos]:
            p.append(cpos + 19)
        if tk23[pos]:
            p.append(cpos + 23)
        if tk29[pos]:
            p.append(cpos + 29)
        pos += 1
    # remove exceeding if present
    pos = len(p) - 1
    while p[pos] > N:
        pos -= 1
    if pos < len(p) - 1:
        del p[pos + 1 :]
    # return p list
    return p


def sieve_of_eratosthenes(n):
    """sieveOfEratosthenes(n): return the list of the primes < n."""
    # Code from: <dickinsm@gmail.com>, Nov 30 2006
    # http://groups.google.com/group/comp.lang.python/msg/f1f10ced88c68c2d
    if n <= 2:
        return []
    sieve = list(range(3, n, 2))
    top = len(sieve)
    for si in sieve:
        if si:
            bottom = (si * si - 3) // 2
            if bottom >= top:
                break
            sieve[bottom::si] = [0] * -((bottom - top) // si)
    return [2] + [el for el in sieve if el]


def sieve_of_atkin(end):
    """return a list of all the prime numbers <end using the Sieve of Atkin."""
    # Code by Steve Krenzel, <Sgk284@gmail.com>, improved
    # Code: https://web.archive.org/web/20080324064651/http://krenzel.info/?p=83
    # Info: http://en.wikipedia.org/wiki/Sieve_of_Atkin
    assert end > 0
    lng = (end - 1) // 2
    sieve = [False] * (lng + 1)

    x_max, x2, xd = int(sqrt((end - 1) / 4.0)), 0, 4
    for xd in range(4, 8 * x_max + 2, 8):
        x2 += xd
        y_max = int(sqrt(end - x2))
        n, n_diff = x2 + y_max * y_max, (y_max << 1) - 1
        if not (n & 1):
            n -= n_diff
            n_diff -= 2
        for d in range((n_diff - 1) << 1, -1, -8):
            m = n % 12
            if m == 1 or m == 5:
                m = n >> 1
                sieve[m] = not sieve[m]
            n -= d

    x_max, x2, xd = int(sqrt((end - 1) / 3.0)), 0, 3
    for xd in range(3, 6 * x_max + 2, 6):
        x2 += xd
        y_max = int(sqrt(end - x2))
        n, n_diff = x2 + y_max * y_max, (y_max << 1) - 1
        if not (n & 1):
            n -= n_diff
            n_diff -= 2
        for d in range((n_diff - 1) << 1, -1, -8):
            if n % 12 == 7:
                m = n >> 1
                sieve[m] = not sieve[m]
            n -= d

    x_max, y_min, x2, xd = int((2 + sqrt(4 - 8 * (1 - end))) / 4), -1, 0, 3
    for x in range(1, x_max + 1):
        x2 += xd
        xd += 6
        if x2 >= end:
            y_min = (((int(ceil(sqrt(x2 - end))) - 1) << 1) - 2) << 1
        n, n_diff = ((x * x + x) << 1) - 1, (((x - 1) << 1) - 2) << 1
        for d in range(n_diff, y_min, -8):
            if n % 12 == 11:
                m = n >> 1
                sieve[m] = not sieve[m]
            n += d

    primes = [2, 3]
    if end <= 3:
        return primes[: max(0, end - 2)]

    for n in range(5 >> 1, (int(sqrt(end)) + 1) >> 1):
        if sieve[n]:
            primes.append((n << 1) + 1)
            aux = (n << 1) + 1
            aux *= aux
            for k in range(aux, end, 2 * aux):
                sieve[k >> 1] = False

    s = int(sqrt(end)) + 1
    if s % 2 == 0:
        s += 1
    primes.extend([i for i in range(s, end, 2) if sieve[i >> 1]])

    return primes


def ambi_sieve_plain(n):
    s = list(range(3, n, 2))
    for m in range(3, int(n ** 0.5) + 1, 2):
        if s[(m - 3) // 2]:
            for t in range((m * m - 3) // 2, (n >> 1) - 1, m):
                s[t] = 0
    return [2] + [t for t in s if t > 0]


def sundaram3(max_n):
    # https://stackoverflow.com/questions/2068372/fastest-way-to-list-all-primes-below-n-in-python/2073279#2073279
    numbers = range(3, max_n + 1, 2)
    half = (max_n) // 2
    initial = 4

    for step in range(3, max_n + 1, 2):
        for i in range(initial, half, step):
            numbers[i - 1] = 0
        initial += 2 * (step + 1)

        if initial > half:
            return [2] + filter(None, numbers)


# Using Numpy:
def ambi_sieve(n):
    # http://tommih.blogspot.com/2009/04/fast-prime-number-generator.html
    s = np.arange(3, n, 2)
    for m in range(3, int(n ** 0.5) + 1, 2):
        if s[(m - 3) // 2]:
            s[(m * m - 3) // 2::m] = 0
    return np.r_[2, s[s > 0]]


def primesfrom3to(n):
    # https://stackoverflow.com/questions/2068372/fastest-way-to-list-all-primes-below-n-in-python/3035188#3035188
    """ Returns an array of primes, p < n """
    assert n >= 2
    sieve = np.ones(n // 2, dtype=bool)
    for i in range(3, int(n ** 0.5) + 1, 2):
        if sieve[i // 2]:
            sieve[i * i // 2::i] = False
    return np.r_[2, 2 * np.nonzero(sieve)[0][1::] + 1]


def primesfrom2to(n):
    # https://stackoverflow.com/questions/2068372/fastest-way-to-list-all-primes-below-n-in-python/3035188#3035188
    """ Input n>=6, Returns an array of primes, 2 <= p < n """
    assert n >= 6
    sieve = np.ones(n // 3 + (n % 6 == 2), dtype=bool)
    sieve[0] = False
    for i in range(int(n ** 0.5) // 3 + 1):
        if sieve[i]:
            k = 3 * i + 1 | 1
            sieve[((k * k) // 3)::2 * k] = False
            sieve[(k * k + 4 * k - 2 * k * (i & 1)) // 3::2 * k] = False
    return np.r_[2, 3, ((3 * np.nonzero(sieve)[0] + 1) | 1)]


def sympy_sieve(n):
    return list(sympy.sieve.primerange(1, n))


b = perfplot.bench(
    setup=lambda n: n,
    kernels=[
        rwh_primes,
        rwh_primes1,
        rwh_primes2,
        sieve_wheel_30,
        sieve_of_eratosthenes,
        sieve_of_atkin,
        # ambi_sieve_plain,
        # sundaram3,
        ambi_sieve,
        primesfrom3to,
        primesfrom2to,
        sympy_sieve,
    ],
    n_range=[2 ** k for k in range(3, 25)],
    xlabel="n",
)
b.save("out.png")
b.show()

---

If you have control over N, the very fastest way to list all primes is to precompute them. Seriously. Precomputing is a way overlooked optimization.

---

Here's the code I normally use to generate primes in Python:

$ python -mtimeit -s'import sieve' 'sieve.sieve(1000000)' 
10 loops, best of 3: 445 msec per loop
$ cat sieve.py
from math import sqrt

def sieve(size):
 prime=[True]*size
 rng=xrange
 limit=int(sqrt(size))

 for i in rng(3,limit+1,+2):
  if prime[i]:
   prime[i*i::+i]=[False]*len(prime[i*i::+i])

 return [2]+[i for i in rng(3,size,+2) if prime[i]]

if __name__=='__main__':
 print sieve(100)

It can't compete with the faster solutions posted here, but at least it is pure python.

Thanks for posting this question. I really learnt a lot today.

---

A slightly different implementation of a half sieve using Numpy:

http://rebrained.com/?p=458

import math
import numpy
def prime6(upto):
    primes=numpy.arange(3,upto+1,2)
    isprime=numpy.ones((upto-1)/2,dtype=bool)
    for factor in primes[:int(math.sqrt(upto))]:
        if isprime[(factor-2)/2]: isprime[(factor*3-2)/2:(upto-1)/2:factor]=0
    return numpy.insert(primes[isprime],0,2)

Can someone compare this with the other timings? On my machine it seems pretty comparable to the other Numpy half-sieve.

---

It's all written and tested. So there is no need to reinvent the wheel.

python -m timeit -r10 -s"from sympy import sieve" "primes = list(sieve.primerange(1, 10**6))"

gives us a record breaking 12.2 msec!

10 loops, best of 10: 12.2 msec per loop

If this is not fast enough, you can try PyPy:

pypy -m timeit -r10 -s"from sympy import sieve" "primes = list(sieve.primerange(1, 10**6))"

which results in:

10 loops, best of 10: 2.03 msec per loop

The answer with 247 up-votes lists 15.9 ms for the best solution. Compare this!!!

---

Fastest prime sieve in Pure Python:

from itertools import compress

def half_sieve(n):
    """
    Returns a list of prime numbers less than `n`.
    """
    if n <= 2:
        return []
    sieve = bytearray([True]) * (n // 2)
    for i in range(3, int(n ** 0.5) + 1, 2):
        if sieve[i // 2]:
            sieve[i * i // 2::i] = bytearray((n - i * i - 1) // (2 * i) + 1)
    primes = list(compress(range(1, n, 2), sieve))
    primes[0] = 2
    return primes

I optimised Sieve of Eratosthenes for speed and memory.

Benchmark

from time import clock
import platform

def benchmark(iterations, limit):
    start = clock()
    for x in range(iterations):
        half_sieve(limit)
    end = clock() - start
    print(f'{end/iterations:.4f} seconds for primes < {limit}')

if __name__ == '__main__':
    print(platform.python_version())
    print(platform.platform())
    print(platform.processor())
    it = 10
    for pw in range(4, 9):
        benchmark(it, 10**pw)

Output

>>> 3.6.7
>>> Windows-10-10.0.17763-SP0
>>> Intel64 Family 6 Model 78 Stepping 3, GenuineIntel
>>> 0.0003 seconds for primes < 10000
>>> 0.0021 seconds for primes < 100000
>>> 0.0204 seconds for primes < 1000000
>>> 0.2389 seconds for primes < 10000000
>>> 2.6702 seconds for primes < 100000000

---

A deterministic implementation of Miller-Rabin's Primality test on the assumption that N < 9,080,191

import sys

def miller_rabin_pass(a, n):
    d = n - 1
    s = 0
    while d % 2 == 0:
        d >>= 1
        s += 1

    a_to_power = pow(a, d, n)
    if a_to_power == 1:
        return True
    for i in range(s-1):
        if a_to_power == n - 1:
            return True
        a_to_power = (a_to_power * a_to_power) % n
    return a_to_power == n - 1


def miller_rabin(n):
    if n <= 2:
        return n == 2

    if n < 2_047:
        return miller_rabin_pass(2, n)

    return all(miller_rabin_pass(a, n) for a in (31, 73))


n = int(sys.argv[1])
primes = [2]
for p in range(3,n,2):
  if miller_rabin(p):
    primes.append(p)
print len(primes)

According to the article on Wikipedia (http://en.wikipedia.org/wiki/Miller–Rabin_primality_test) testing N < 9,080,191 for a = 37 and 73 is enough to decide whether N is composite or not.

And I adapted the source code from the probabilistic implementation of original Miller-Rabin's test found here: https://www.literateprograms.org/miller-rabin_primality_test__python_.html

---

For the fastest code, the numpy solution is the best. For purely academic reasons, though, I'm posting my pure python version, which is a bit less than 50% faster than the cookbook version posted above. Since I make the entire list in memory, you need enough space to hold everything, but it seems to scale fairly well.

def daniel_sieve_2(maxNumber):
    """
    Given a number, returns all numbers less than or equal to
    that number which are prime.
    """
    allNumbers = range(3, maxNumber+1, 2)
    for mIndex, number in enumerate(xrange(3, maxNumber+1, 2)):
        if allNumbers[mIndex] == 0:
            continue
        # now set all multiples to 0
        for index in xrange(mIndex+number, (maxNumber-3)/2+1, number):
            allNumbers[index] = 0
    return [2] + filter(lambda n: n!=0, allNumbers)

And the results:

>>>mine = timeit.Timer("daniel_sieve_2(1000000)",
...                    "from sieves import daniel_sieve_2")
>>>prev = timeit.Timer("get_primes_erat(1000000)",
...                    "from sieves import get_primes_erat")
>>>print "Mine: {0:0.4f} ms".format(min(mine.repeat(3, 1))*1000)
Mine: 428.9446 ms
>>>print "Previous Best {0:0.4f} ms".format(min(prev.repeat(3, 1))*1000)
Previous Best 621.3581 ms

---

I know the competition is closed for some years. …

Nonetheless this is my suggestion for a pure python prime sieve, based on omitting the multiples of 2, 3 and 5 by using appropriate steps while processing the sieve forward. Nonetheless it is actually slower for N<10^9 than @Robert William Hanks superior solutions rwh_primes2 and rwh_primes1. By using a ctypes.c_ushort sieve array above 1.5* 10^8 it is somehow adaptive to memory limits.

10^6

$ python -mtimeit -s"import primeSieveSpeedComp" "primeSieveSpeedComp.primeSieveSeq(1000000)" 10 loops, best of 3: 46.7 msec per loop

to compare:$ python -mtimeit -s"import primeSieveSpeedComp" "primeSieveSpeedComp.rwh_primes1(1000000)" 10 loops, best of 3: 43.2 msec per loop to compare: $ python -m timeit -s"import primeSieveSpeedComp" "primeSieveSpeedComp.rwh_primes2(1000000)" 10 loops, best of 3: 34.5 msec per loop

10^7

$ python -mtimeit -s"import primeSieveSpeedComp" "primeSieveSpeedComp.primeSieveSeq(10000000)" 10 loops, best of 3: 530 msec per loop

to compare:$ python -mtimeit -s"import primeSieveSpeedComp" "primeSieveSpeedComp.rwh_primes1(10000000)" 10 loops, best of 3: 494 msec per loop to compare: $ python -m timeit -s"import primeSieveSpeedComp" "primeSieveSpeedComp.rwh_primes2(10000000)" 10 loops, best of 3: 375 msec per loop

10^8

$ python -mtimeit -s"import primeSieveSpeedComp" "primeSieveSpeedComp.primeSieveSeq(100000000)" 10 loops, best of 3: 5.55 sec per loop

to compare: $ python -mtimeit -s"import primeSieveSpeedComp" "primeSieveSpeedComp.rwh_primes1(100000000)" 10 loops, best of 3: 5.33 sec per loop to compare: $ python -m timeit -s"import primeSieveSpeedComp" "primeSieveSpeedComp.rwh_primes2(100000000)" 10 loops, best of 3: 3.95 sec per loop

10^9

$ python -mtimeit -s"import primeSieveSpeedComp" "primeSieveSpeedComp.primeSieveSeq(1000000000)" 10 loops, best of 3: 61.2 sec per loop

to compare: $ python -mtimeit -n 3 -s"import primeSieveSpeedComp" "primeSieveSpeedComp.rwh_primes1(1000000000)" 3 loops, best of 3: 97.8 sec per loop

to compare: $ python -m timeit -s"import primeSieveSpeedComp" "primeSieveSpeedComp.rwh_primes2(1000000000)" 10 loops, best of 3: 41.9 sec per loop

You may copy the code below into ubuntus primeSieveSpeedComp to review this tests.

def primeSieveSeq(MAX_Int):
    if MAX_Int > 5*10**8:
        import ctypes
        int16Array = ctypes.c_ushort * (MAX_Int >> 1)
        sieve = int16Array()
        #print 'uses ctypes "unsigned short int Array"'
    else:
        sieve = (MAX_Int >> 1) * [False]
        #print 'uses python list() of long long int'
    if MAX_Int < 10**8:
        sieve[4::3] = [True]*((MAX_Int - 8)/6+1)
        sieve[12::5] = [True]*((MAX_Int - 24)/10+1)
    r = [2, 3, 5]
    n = 0
    for i in xrange(int(MAX_Int**0.5)/30+1):
        n += 3
        if not sieve[n]:
            n2 = (n << 1) + 1
            r.append(n2)
            n2q = (n2**2) >> 1
            sieve[n2q::n2] = [True]*(((MAX_Int >> 1) - n2q - 1) / n2 + 1)
        n += 2
        if not sieve[n]:
            n2 = (n << 1) + 1
            r.append(n2)
            n2q = (n2**2) >> 1
            sieve[n2q::n2] = [True]*(((MAX_Int >> 1) - n2q - 1) / n2 + 1)
        n += 1
        if not sieve[n]:
            n2 = (n << 1) + 1
            r.append(n2)
            n2q = (n2**2) >> 1
            sieve[n2q::n2] = [True]*(((MAX_Int >> 1) - n2q - 1) / n2 + 1)
        n += 2
        if not sieve[n]:
            n2 = (n << 1) + 1
            r.append(n2)
            n2q = (n2**2) >> 1
            sieve[n2q::n2] = [True]*(((MAX_Int >> 1) - n2q - 1) / n2 + 1)
        n += 1
        if not sieve[n]:
            n2 = (n << 1) + 1
            r.append(n2)
            n2q = (n2**2) >> 1
            sieve[n2q::n2] = [True]*(((MAX_Int >> 1) - n2q - 1) / n2 + 1)
        n += 2
        if not sieve[n]:
            n2 = (n << 1) + 1
            r.append(n2)
            n2q = (n2**2) >> 1
            sieve[n2q::n2] = [True]*(((MAX_Int >> 1) - n2q - 1) / n2 + 1)
        n += 3
        if not sieve[n]:
            n2 = (n << 1) + 1
            r.append(n2)
            n2q = (n2**2) >> 1
            sieve[n2q::n2] = [True]*(((MAX_Int >> 1) - n2q - 1) / n2 + 1)
        n += 1
        if not sieve[n]:
            n2 = (n << 1) + 1
            r.append(n2)
            n2q = (n2**2) >> 1
            sieve[n2q::n2] = [True]*(((MAX_Int >> 1) - n2q - 1) / n2 + 1)
    if MAX_Int < 10**8:
        return [2, 3, 5]+[(p << 1) + 1 for p in [n for n in xrange(3, MAX_Int >> 1) if not sieve[n]]]
    n = n >> 1
    try:
        for i in xrange((MAX_Int-2*n)/30 + 1):
            n += 3
            if not sieve[n]:
                r.append((n << 1) + 1)
            n += 2
            if not sieve[n]:
                r.append((n << 1) + 1)
            n += 1
            if not sieve[n]:
                r.append((n << 1) + 1)
            n += 2
            if not sieve[n]:
                r.append((n << 1) + 1)
            n += 1
            if not sieve[n]:
                r.append((n << 1) + 1)
            n += 2
            if not sieve[n]:
                r.append((n << 1) + 1)
            n += 3
            if not sieve[n]:
                r.append((n << 1) + 1)
            n += 1
            if not sieve[n]:
                r.append((n << 1) + 1)
    except:
        pass
    return r

---

I tested some unutbu's functions, i computed it with hungred millions number

The winners are the functions that use numpy library,

Note: It would also interesting make a memory utilization test :)

Sample code

Complete code on my github repository

#!/usr/bin/env python

import lib
import timeit
import sys
import math
import datetime

import prettyplotlib as ppl
import numpy as np

import matplotlib.pyplot as plt
from prettyplotlib import brewer2mpl

primenumbers_gen = [
    'sieveOfEratosthenes',
    'ambi_sieve',
    'ambi_sieve_plain',
    'sundaram3',
    'sieve_wheel_30',
    'primesfrom3to',
    'primesfrom2to',
    'rwh_primes',
    'rwh_primes1',
    'rwh_primes2',
]

def human_format(num):
    # https://stackoverflow.com/questions/579310/formatting-long-numbers-as-strings-in-python?answertab=active#tab-top
    magnitude = 0
    while abs(num) >= 1000:
        magnitude += 1
        num /= 1000.0
    # add more suffixes if you need them
    return '%.2f%s' % (num, ['', 'K', 'M', 'G', 'T', 'P'][magnitude])


if __name__=='__main__':

    # Vars
    n = 10000000 # number itereration generator
    nbcol = 5 # For decompose prime number generator
    nb_benchloop = 3 # Eliminate false positive value during the test (bench average time)
    datetimeformat = '%Y-%m-%d %H:%M:%S.%f'
    config = 'from __main__ import n; import lib'
    primenumbers_gen = {
        'sieveOfEratosthenes': {'color': 'b'},
        'ambi_sieve': {'color': 'b'},
        'ambi_sieve_plain': {'color': 'b'},
         'sundaram3': {'color': 'b'},
        'sieve_wheel_30': {'color': 'b'},
# # #        'primesfrom2to': {'color': 'b'},
        'primesfrom3to': {'color': 'b'},
        # 'rwh_primes': {'color': 'b'},
        # 'rwh_primes1': {'color': 'b'},
        'rwh_primes2': {'color': 'b'},
    }


    # Get n in command line
    if len(sys.argv)>1:
        n = int(sys.argv[1])

    step = int(math.ceil(n / float(nbcol)))
    nbs = np.array([i * step for i in range(1, int(nbcol) + 1)])
    set2 = brewer2mpl.get_map('Paired', 'qualitative', 12).mpl_colors

    print datetime.datetime.now().strftime(datetimeformat)
    print("Compute prime number to %(n)s" % locals())
    print("")

    results = dict()
    for pgen in primenumbers_gen:
        results[pgen] = dict()
        benchtimes = list()
        for n in nbs:
            t = timeit.Timer("lib.%(pgen)s(n)" % locals(), setup=config)
            execute_times = t.repeat(repeat=nb_benchloop,number=1)
            benchtime = np.mean(execute_times)
            benchtimes.append(benchtime)
        results[pgen] = {'benchtimes':np.array(benchtimes)}

fig, ax = plt.subplots(1)
plt.ylabel('Computation time (in second)')
plt.xlabel('Numbers computed')
i = 0
for pgen in primenumbers_gen:

    bench = results[pgen]['benchtimes']
    avgs = np.divide(bench,nbs)
    avg = np.average(bench, weights=nbs)

    # Compute linear regression
    A = np.vstack([nbs, np.ones(len(nbs))]).T
    a, b = np.linalg.lstsq(A, nbs*avgs)[0]

    # Plot
    i += 1
    #label="%(pgen)s" % locals()
    #ppl.plot(nbs, nbs*avgs, label=label, lw=1, linestyle='--', color=set2[i % 12])
    label="%(pgen)s avg" % locals()
    ppl.plot(nbs, a * nbs + b, label=label, lw=2, color=set2[i % 12])
print datetime.datetime.now().strftime(datetimeformat)

ppl.legend(ax, loc='upper left', ncol=4)

# Change x axis label
ax.get_xaxis().get_major_formatter().set_scientific(False)
fig.canvas.draw()
labels = [human_format(int(item.get_text())) for item in ax.get_xticklabels()]

ax.set_xticklabels(labels)
ax = plt.gca()

plt.show()

---

For Python 3

def rwh_primes2(n):
    correction = (n%6>1)
    n = {0:n,1:n-1,2:n+4,3:n+3,4:n+2,5:n+1}[n%6]
    sieve = [True] * (n//3)
    sieve[0] = False
    for i in range(int(n**0.5)//3+1):
      if sieve[i]:
        k=3*i+1|1
        sieve[      ((k*k)//3)      ::2*k]=[False]*((n//6-(k*k)//6-1)//k+1)
        sieve[(k*k+4*k-2*k*(i&1))//3::2*k]=[False]*((n//6-(k*k+4*k-2*k*(i&1))//6-1)//k+1)
    return [2,3] + [3*i+1|1 for i in range(1,n//3-correction) if sieve[i]]

---

I'm surprised nobody mentioned numba yet.

This version gets to the 1M mark in 2.47 ms ± 36.5 µs.

Years ago, pseudo-code for a version of Atkin's sieve was given on the Wikipedia page Prime number. This isn't there anymore, and a reference to the Sieve of Atkin seems to be a different algorithm. A 2007/03/01 version of the Wikipedia page, Primer number as of 2007-03-01, shows the pseudo-code I used as reference.

import numpy as np
from numba import njit

@njit
def nb_primes(n):
    # Generates prime numbers 2 <= p <= n
    # Atkin's sieve -- see https://en.wikipedia.org/w/index.php?title=Prime_number&oldid=111775466
    sqrt_n = int(np.sqrt(n)) + 1

    # initialize the sieve
    s = np.full(n + 1, -1, dtype=np.int8)
    s[2] = 1
    s[3] = 1

    # put in candidate primes:
    # integers which have an odd number of
    # representations by certain quadratic forms
    for x in range(1, sqrt_n):
        x2 = x * x
        for y in range(1, sqrt_n):
            y2 = y * y
            k = 4 * x2 + y2
            if k <= n and (k % 12 == 1 or k % 12 == 5): s[k] *= -1
            k = 3 * x2 + y2
            if k <= n and (k % 12 == 7): s[k] *= -1
            k = 3 * x2 - y2
            if k <= n and x > y and k % 12 == 11: s[k] *= -1

    # eliminate composites by sieving
    for k in range(5, sqrt_n):
        if s[k]:
            k2 = k*k
            # k is prime, omit multiples of its square; this is sufficient because
            # composites which managed to get on the list cannot be square-free
            for i in range(1, n // k2 + 1):
                j = i * k2 # j ∈ {k², 2k², 3k², ..., n}
                s[j] = -1
    return np.nonzero(s>0)[0]

# initial run for "compilation" 
nb_primes(10)

Timing

In[10]:
%timeit nb_primes(1_000_000)

Out[10]:
2.47 ms ± 36.5 µs per loop (mean ± std. dev. of 7 runs, 100 loops each)

In[11]:
%timeit nb_primes(10_000_000)

Out[11]:
33.4 ms ± 373 µs per loop (mean ± std. dev. of 7 runs, 10 loops each)

In[12]:
%timeit nb_primes(100_000_000)

Out[12]:
828 ms ± 5.64 ms per loop (mean ± std. dev. of 7 runs, 1 loop each)

---

First time using python, so some of the methods I use in this might seem a bit cumbersome. I just straight converted my c++ code to python and this is what I have (albeit a tad bit slowww in python)

#!/usr/bin/env python
import time

def GetPrimes(n):

    Sieve = [1 for x in xrange(n)]

    Done = False
    w = 3

    while not Done:

        for q in xrange (3, n, 2):
            Prod = w*q
            if Prod < n:
                Sieve[Prod] = 0
            else:
                break

        if w > (n/2):
            Done = True
        w += 2

    return Sieve



start = time.clock()

d = 10000000
Primes = GetPrimes(d)

count = 1 #This is for 2

for x in xrange (3, d, 2):
    if Primes[x]:
        count+=1

elapsed = (time.clock() - start)
print "\nFound", count, "primes in", elapsed, "seconds!\n"

pythonw Primes.py

Found 664579 primes in 12.799119 seconds!

#!/usr/bin/env python
import time

def GetPrimes2(n):

    Sieve = [1 for x in xrange(n)]

    for q in xrange (3, n, 2):
        k = q
        for y in xrange(k*3, n, k*2):
            Sieve[y] = 0

    return Sieve



start = time.clock()

d = 10000000
Primes = GetPrimes2(d)

count = 1 #This is for 2

for x in xrange (3, d, 2):
    if Primes[x]:
        count+=1

elapsed = (time.clock() - start)
print "\nFound", count, "primes in", elapsed, "seconds!\n"

pythonw Primes2.py

Found 664579 primes in 10.230172 seconds!

#!/usr/bin/env python
import time

def GetPrimes3(n):

    Sieve = [1 for x in xrange(n)]

    for q in xrange (3, n, 2):
        k = q
        for y in xrange(k*k, n, k << 1):
            Sieve[y] = 0

    return Sieve



start = time.clock()

d = 10000000
Primes = GetPrimes3(d)

count = 1 #This is for 2

for x in xrange (3, d, 2):
    if Primes[x]:
        count+=1

elapsed = (time.clock() - start)
print "\nFound", count, "primes in", elapsed, "seconds!\n"

python Primes2.py

Found 664579 primes in 7.113776 seconds!

---

The simplest way I've found of doing this is:

primes = []
for n in range(low, high + 1):
    if all(n % i for i in primes):
        primes.append(n)

---

As of (late) writing, this is the fastest working solution posted (at least it is on my machine). It uses both numpy and bitarray, and is inspired by primesfrom2to from this answer.

import numpy as np
from bitarray import bitarray


def bit_primes(n):
    bit_sieve = bitarray(n // 3 + (n % 6 == 2))
    bit_sieve.setall(1)
    bit_sieve[0] = False

    for i in range(int(n ** 0.5) // 3 + 1):
        if bit_sieve[i]:
            k = 3 * i + 1 | 1
            bit_sieve[k * k // 3::2 * k] = False
            bit_sieve[(k * k + 4 * k - 2 * k * (i & 1)) // 3::2 * k] = False

    np_sieve = np.unpackbits(np.frombuffer(bit_sieve.tobytes(), dtype=np.uint8)).view(bool)
    return np.concatenate(((2, 3), ((3 * np.flatnonzero(np_sieve) + 1) | 1)))

Here's a comparison with primesfrom2to, which was previously found to be the fastest solution in unutbu's comparison:

python3 -m timeit -s "import fast_primes" "fast_primes.bit_primes(1000000)"
200 loops, best of 5: 1.19 msec per loop

python3 -m timeit -s "import fast_primes" "fast_primes.primesfrom2to(1000000)"
200 loops, best of 5: 1.23 msec per loop

For finding primes under 1 million, bit_primes was slightly faster. For larger values of n, the difference can be more significant. In some cases, bit_primes was over twice as fast:

python3 -m timeit -s "import fast_primes" "fast_primes.bit_primes(500_000_000)"
1 loop, best of 5: 540 msec per loop

python3 -m timeit -s "import fast_primes" "fast_primes.primesfrom2to(500_000_000)"
1 loop, best of 5: 1.15 sec per loop

For reference, here's the minimally modified (to work in Python 3) version of primesfrom2to I compared with:

def primesfrom2to(n):
    # https://stackoverflow.com/questions/2068372/fastest-way-to-list-all-primes-below-n-in-python/3035188#3035188
    """ Input n>=6, Returns a array of primes, 2 <= p < n"""
    sieve = np.ones(n // 3 + (n % 6 == 2), dtype=np.bool)
    sieve[0] = False
    for i in range(int(n ** 0.5) // 3 + 1):
        if sieve[i]:
            k = 3 * i + 1 | 1
            sieve[((k * k) // 3)::2 * k] = False
            sieve[(k * k + 4 * k - 2 * k * (i & 1)) // 3::2 * k] = False
    return np.r_[2, 3, ((3 * np.nonzero(sieve)[0] + 1) | 1)]

---

My guess is that the fastest of all ways is to hard code the primes in your code.

So why not just write a slow script that generates another source file that has all numbers hardwired in it, and then import that source file when you run your actual program.

Of course, this works only if you know the upper bound of N at compile time, but thus is the case for (almost) all project Euler problems.

 

PS: I might be wrong though iff parsing the source with hard-wired primes is slower than computing them in the first place, but as far I know Python runs from compiled .pyc files so reading a binary array with all primes up to N should be bloody fast in that case.

---

Sorry to bother but erat2() has a serious flaw in the algorithm.

While searching for the next composite, we need to test odd numbers only. q,p both are odd; then q+p is even and doesn't need to be tested, but q+2*p is always odd. This eliminates the "if even" test in the while loop condition and saves about 30% of the runtime.

While we're at it: instead of the elegant 'D.pop(q,None)' get and delete method use 'if q in D: p=D[q],del D[q]' which is twice as fast! At least on my machine (P3-1Ghz). So I suggest this implementation of this clever algorithm:

def erat3( ):
    from itertools import islice, count

    # q is the running integer that's checked for primeness.
    # yield 2 and no other even number thereafter
    yield 2
    D = {}
    # no need to mark D[4] as we will test odd numbers only
    for q in islice(count(3),0,None,2):
        if q in D:                  #  is composite
            p = D[q]
            del D[q]
            # q is composite. p=D[q] is the first prime that
            # divides it. Since we've reached q, we no longer
            # need it in the map, but we'll mark the next
            # multiple of its witnesses to prepare for larger
            # numbers.
            x = q + p+p        # next odd(!) multiple
            while x in D:      # skip composites
                x += p+p
            D[x] = p
        else:                  # is prime
            # q is a new prime.
            # Yield it and mark its first multiple that isn't
            # already marked in previous iterations.
            D[q*q] = q
            yield q

---

I may be late to the party but will have to add my own code for this. It uses approximately n/2 in space because we don't need to store even numbers and I also make use of the bitarray python module, further draStically cutting down on memory consumption and enabling computing all primes up to 1,000,000,000

from bitarray import bitarray
def primes_to(n):
    size = n//2
    sieve = bitarray(size)
    sieve.setall(1)
    limit = int(n**0.5)
    for i in range(1,limit):
        if sieve[i]:
            val = 2*i+1
            sieve[(i+i*val)::val] = 0
    return [2] + [2*i+1 for i, v in enumerate(sieve) if v and i > 0]

python -m timeit -n10 -s "import euler" "euler.primes_to(1000000000)"
10 loops, best of 3: 46.5 sec per loop

This was run on a 64bit 2.4GHZ MAC OSX 10.8.3

---

Here is a numpy version of Sieve of Eratosthenes having both good complexity (lower than sorting an array of length n) and vectorization. Compared to @unutbu times this just as fast as the packages with 46 microsecons to find all primes below a million.

import numpy as np 
def generate_primes(n):
    is_prime = np.ones(n+1,dtype=bool)
    is_prime[0:2] = False
    for i in range(int(n**0.5)+1):
        if is_prime[i]:
            is_prime[i**2::i]=False
    return np.where(is_prime)[0]

Timings:

import time    
for i in range(2,10):
    timer =time.time()
    generate_primes(10**i)
    print('n = 10^',i,' time =', round(time.time()-timer,6))

>> n = 10^ 2  time = 5.6e-05
>> n = 10^ 3  time = 6.4e-05
>> n = 10^ 4  time = 0.000114
>> n = 10^ 5  time = 0.000593
>> n = 10^ 6  time = 0.00467
>> n = 10^ 7  time = 0.177758
>> n = 10^ 8  time = 1.701312
>> n = 10^ 9  time = 19.322478

---

Here is an interesting technique to generate prime numbers (yet not the most efficient) using python's list comprehensions:

noprimes = [j for i in range(2, 8) for j in range(i*2, 50, i)]
primes = [x for x in range(2, 50) if x not in noprimes]