我知道比赛已经结束好几年了。.．.

尽管如此，这是我对纯python质数筛子的建议，基于在向前处理筛子时使用适当的步骤省略2、3和5的倍数。尽管如此，在N<10^9时，它实际上比@Robert William Hanks的优解rwh_primes2和rwh_primes1要慢。通过使用大于1.5* 10^8的ctypes.c_ushort筛分数组，可以在某种程度上适应内存限制。

10^6

$ python -mtimeit -s"import primeSieveSpeedComp" "primeSieveSpeedComp. primesieveseq (1000000)" 10个循环，最好的3:46.7毫秒每循环

import primeSieveSpeedComp (primeSieveSpeedComp) “primeSieveSpeedComp.rwh_primes1(1000000)”10个循环，最好的3:43.2 每回路Msec $ python -m timeit -s"import primeSieveSpeedComp" “primeSieveSpeedComp.rwh_primes2(1000000)”10圈，最好成绩是3:34.5 每回路Msec

10^7

$ python -mtimeit -s"import primeSieveSpeedComp" "primeSieveSpeedComp. primesieveseq (10000000)" 10个循环，最好是3:530毫秒每循环

import primeSieveSpeedComp (primeSieveSpeedComp) “primeSieveSpeedComp.rwh_primes1(10000000)”10圈，3:494的最佳成绩 每回路Msec $ python -m timeit -s"import primeSieveSpeedComp" “primeSieveSpeedComp.rwh_primes2(10000000)”10圈，最好的3:375 每回路Msec

10^8

$ python -mtimeit -s"import primeSieveSpeedComp" "primeSieveSpeedComp. primesieveseq (100000000)" 10圈，最好的3:5.55秒每圈

import primeSieveSpeedComp (primeSieveSpeedComp) “primeSieveSpeedComp.rwh_primes1(100000000)”10圈，最好成绩是3:5.33 秒/循环 $ python -m timeit -s"import primeSieveSpeedComp" “primeSieveSpeedComp.rwh_primes2(100000000)”10圈，最好的3:3.95 秒/循环

10^9

$ python -mtimeit -s"import primeSieveSpeedComp" "primeSieveSpeedComp. primesieveseq (1000000000)" 10圈，最好的3圈:每圈61.2秒

$ python -mtimeit -n 3 -s"import primeSieveSpeedComp" “primeSieveSpeedComp.rwh_primes1(1000000000)”3圈，最好的3:97.8 秒/循环 $ python -m timeit -s"import primeSieveSpeedComp" “primeSieveSpeedComp.rwh_primes2(1000000000)”10个循环，3个最好: 每循环41.9秒

您可以将下面的代码复制到ubuntu primeSieveSpeedComp中以查看此测试。

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

对于最快的代码，numpy解决方案是最好的。不过，出于纯粹的学术原因，我发布了我的纯python版本，它比上面发布的食谱版本快不到50%。由于我将整个列表放在内存中，所以需要足够的空间来容纳所有内容，但它的可伸缩性似乎相当好。

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)

结果是:

>>>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

使用Sundaram的Sieve，我想我打破了pure-Python的记录:

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

对于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]]

很抱歉打扰，但erat2()在算法中有一个严重的缺陷。

在搜索下一个合成时，我们只需要测试奇数。 Q p都是奇数;那么q+p是偶数，不需要检验，但q+2*p总是奇数。这消除了while循环条件中的“if even”测试，并节省了大约30%的运行时。

当我们在它:而不是优雅的'D.pop(q,None)'获取和删除方法，使用'if q in D: p=D[q]，del D[q]'，这是两倍的速度!至少在我的机器上(P3-1Ghz)。 所以我建议这个聪明算法的实现:

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

如果你接受itertools，但不接受numpy，这里有一个针对Python 3的rwh_primes2的改编版本，它在我的机器上运行速度大约是原来的两倍。唯一的实质性变化是使用bytearray而不是列表来表示布尔值，并使用压缩而不是列表推导来构建最终列表。(如果可以的话，我会把这句话作为moarningsun之类的评论。)

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

比较:

>>> 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