Velocity Reviews (http://www.velocityreviews.com/forums/index.php)
-   Python (http://www.velocityreviews.com/forums/f43-python.html)
-   -   Ultimate Prime Sieve -- Sieve Of Zakiya (SoZ) (http://www.velocityreviews.com/forums/t620078-ultimate-prime-sieve-sieve-of-zakiya-soz.html)

 jzakiya 06-13-2008 05:12 PM

Ultimate Prime Sieve -- Sieve Of Zakiya (SoZ)

This is to announce the release of my paper "Ultimate Prime Sieve --
Sieve of Zakiiya (SoZ)" in which I show and explain the development of
a class of Number Theory Sieves to generate prime numbers. I used
Ruby 1.9.0-1 as my development environment on a P4 2.8 Ghz laptop.

You can get the pdf of my paper and Ruby and Python source from here:

http://www.4shared.com/dir/7467736/9...1/sharing.html

Below is a sample of one of the simple prime generators. I did a
Python version of this in my paper (see Python source too). The Ruby
version below is the minimum array size version, while the Python has
array of size N (I made no attempt to optimize its implementation,
it's to show the method).

class Integer
def primesP3a
# all prime candidates > 3 are of form 6*k+1 and 6*k+5
# initialize sieve array with only these candidate values
# where sieve contains the odd integers representatives
# convert integers to array indices/vals by i = (n-3)>>1 =
(n>>1)-1
n1, n2 = -1, 1; lndx= (self-1) >>1; sieve = []
while n2 < lndx
n1 +=3; n2 += 3; sieve[n1] = n1; sieve[n2] = n2
end
#now initialize sieve array with (odd) primes < 6, resize array
sieve[0] =0; sieve[1]=1; sieve=sieve[0..lndx-1]

5.step(Math.sqrt(self).to_i, 2) do |i|
next unless sieve[(i>>1) - 1]
# p5= 5*i, k = 6*i, p7 = 7*i
# p1 = (5*i-3)>>1; p2 = (7*i-3)>>1; k = (6*i)>>1
i6 = 6*i; p1 = (i6-i-3)>>1; p2 = (i6+i-3)>>1; k = i6>>1
while p1 < lndx
sieve[p1] = nil; sieve[p2] = nil; p1 += k; p2 += k
end
end
return [2] if self < 3
[2]+([nil]+sieve).compact!.map {|i| (i<<1) +3 }
end
end

def primesP3(val):
# all prime candidates > 3 are of form 6*k+(1,5)
# initialize sieve array with only these candidate values
n1, n2 = 1, 5
sieve = [False]*(val+6)
while n2 < val:
n1 += 6; n2 += 6; sieve[n1] = n1; sieve[n2] = n2
# now load sieve with seed primes 3 < pi < 6, in this case just 5
sieve[5] = 5

for i in range( 5, int(ceil(sqrt(val))), 2) :
if not sieve[i]: continue
# p1= 5*i, k = 6*i, p2 = 7*i,
p1 = 5*i; k = p1+i; p2 = k+i
while p2 <= val:
sieve[p1] = False; sieve[p2] = False; p1 += k; p2 += k
if p1 <= val: sieve[p1] = False

primes = [2,3]
if val < 3 : return [2]
primes.extend( i for i in range(5, val+(val&1), 2) if sieve[i] )

return primes

Now to generate an array of the primes up to some N just do:

Ruby: 10000001.primesP3a
Python: primesP3a(10000001)

The paper presents benchmarks with Ruby 1.9.0-1 (YARV). I would love
to see my various prime generators benchmarked with optimized
implementations in other languages. I'm hoping Python gurus will do
better than I, though the methodology is very very simple, since all I

Have fun with the code. ;-)

Jabari Zakiya

 jzakiya 06-13-2008 05:38 PM

Re: Ultimate Prime Sieve -- Sieve Of Zakiya (SoZ)

On Jun 13, 1:12*pm, jzakiya <jzak...@gmail.com> wrote:
> This is to announce the release of my paper "Ultimate Prime Sieve --
> Sieve of Zakiiya (SoZ)" in which I show and explain the development of
> a class of Number Theory Sieves to generate prime numbers. * I used
> Ruby 1.9.0-1 as my development environment on a P4 2.8 Ghz laptop.
>
> You can get the pdf of my paper and Ruby and Python source from here:
>
> http://www.4shared.com/dir/7467736/9...1/sharing.html
>
> Below is a sample of one of the simple prime generators. I did a
> Python version of this in my paper (see Python source too). *The Ruby
> version below is the minimum array size version, while the Python has
> array of size N (I made no attempt to optimize its implementation,
> it's to show the method).
>
> class Integer
> * *def primesP3a
> * * * # all prime candidates > 3 are of form *6*k+1 and 6*k+5
> * * * # initialize sieve array with only these candidate values
> * * * # where sieve contains the odd integers representatives
> * * * # convert integers to array indices/vals by *i = (n-3)>>1 =
> (n>>1)-1
> * * * n1, n2 = -1, 1; *lndx= (self-1) >>1; *sieve = []
> * * * while n2 < lndx
> * * * * *n1 +=3; * n2 += 3; * sieve[n1] = n1; *sieve[n2] = n2
> * * * end
> * * * #now initialize sieve array with (odd) primes < 6, resize array
> * * * sieve[0] =0; *sieve[1]=1; *sieve=sieve[0..lndx-1]
>
> * * * 5.step(Math.sqrt(self).to_i, 2) do |i|
> * * * * *next unless sieve[(i>>1) - 1]
> * * * * *# p5= 5*i, *k = 6*i, *p7 = 7*i
> * * * * *# p1 = (5*i-3)>>1; *p2 = (7*i-3)>>1; *k = (6*i)>>1
> * * * * *i6 = 6*i; *p1 = (i6-i-3)>>1; *p2 = (i6+i-3)>>1; *k = i6>>1
> * * * * *while p1 < lndx
> * * * * * * *sieve[p1] = nil; *sieve[p2] = nil; *p1 += k; *p2 += k
> * * * * *end
> * * * end
> * * * return [2] if self < 3
> * * * [2]+([nil]+sieve).compact!.map {|i| (i<<1) +3 }
> * *end
> end
>
> def primesP3(val):
> * * # all prime candidates > 3 are of form *6*k+(1,5)
> * * # initialize sieve array with only these candidate values
> * * n1, n2 = 1, 5
> * * sieve = [False]*(val+6)
> * * while *n2 < val:
> * * * * n1 += 6; * n2 += 6; *sieve[n1] = n1; * sieve[n2] = n2
> * * # now load sieve with seed primes 3 < pi < 6, in this case just 5
> * * sieve[5] = 5
>
> * * for i in range( 5, int(ceil(sqrt(val))), 2) :
> * * * *if not sieve[i]: *continue
> * * * *# *p1= 5*i, *k = 6*i, *p2 = 7*i,
> * * * *p1 = 5*i; *k = p1+i; *p2 = k+i
> * * * *while p2 <= val:
> * * * * * sieve[p1] = False; *sieve[p2] = False; *p1 += k; *p2 += k
> * * * *if p1 <= val: *sieve[p1] = False
>
> * * primes = [2,3]
> * * if val < 3 : return [2]
> * * primes.extend( i for i in range(5, val+(val&1), 2) *if sieve[i] )
>
> * * return primes
>
> Now to generate an array of the primes up to some N just do:
>
> Ruby: * *10000001.primesP3a
> Python: primesP3a(10000001)
>
> The paper presents benchmarks with Ruby 1.9.0-1 (YARV). *I would love
> to see my various prime generators benchmarked with optimized
> implementations in other languages. *I'm hoping Python gurus will do
> better than I, though the methodology is very very simple, since all I
>
> Have fun with the code. *;-)
>

CORRECTION:

http://cr.yp.to/primegen.html NOT "primesgen"

Jabari Zakiya

 George Sakkis 06-18-2008 11:58 PM

Re: Ultimate Prime Sieve -- Sieve Of Zakiya (SoZ)

On Jun 13, 1:12 pm, jzakiya <jzak...@gmail.com> wrote:

> The paper presents benchmarks with Ruby 1.9.0-1 (YARV). I would love
> to see my various prime generators benchmarked with optimized
> implementations in other languages. I'm hoping Python gurus will do
> better than I, though the methodology is very very simple, since all I

After playing a little with it, I managed to get a 32-47% improvement
on average for the pure Python version, and a 230-650% improvement
with an extra "import psyco; psyco.full()" (pasted at http://codepad.org/C2nQ8syr)
The changes are:

- Replaced range() with xrange()
- Replaced x**2 with x*x
- Replaced (a,b) = (c,d) with a=c; b=d
- Replaced generator expressions with list comprehensions. This was
the big one for letting psyco do its magic.

I also tried adding type declarations and running it through Cython
but the improvement was much less impressive than Psyco. I'm not a
Pyrex/Cython expert though so I may have missed something obvious.

George

 jzakiya 07-19-2008 07:39 PM

Re: Ultimate Prime Sieve -- Sieve Of Zakiya (SoZ)

On Jun 18, 7:58*pm, George Sakkis <george.sak...@gmail.com> wrote:
> On Jun 13, 1:12 pm, jzakiya <jzak...@gmail.com> wrote:
>
> > The paper presents benchmarks with Ruby 1.9.0-1 (YARV). *I would love
> > to see my variousprimegenerators benchmarked with optimized
> > implementations in other languages. *I'm hoping Python gurus will do
> > better than I, though the methodology is very very simple, since all I

>
> After playing a little with it, I managed to get a 32-47% improvement
> on average for the pure Python version, and a 230-650% improvement
> with an extra "import psyco; psyco.full()" (pasted athttp://codepad.org/C2nQ8syr)
> The changes are:
>
> - Replaced range() with xrange()
> - Replaced x**2 with x*x
> - Replaced (a,b) = (c,d) with a=c; b=d
> - Replaced generator expressions with list comprehensions. This was
> the big one for letting psyco do its magic.
>
> I also tried adding type declarations and running it through Cython
> but the improvement was much less impressive than Psyco. I'm not a
> Pyrex/Cython expert though so I may have missed something obvious.
>
> George

George,

I took your code and included more efficient/optimized versions of
SoZ versions P3, P5, P7, and P11.

I ran the code on my PCLinuxOS, Intel P4, Python 2.4.3 system and
noted this. The SoZ code run much faster than the SoA in pure Python.
When psyco is used the SoA is significantly faster than the
pure Python version. The SoZ versions are faster too, but now
they are slower than the SoA. You can download the code from

http://www.4shared.com/dir/7467736/9...1/sharing.html

It would be interesting to see how this code runs in newer versions
of Python (Psyco).

FYI, someone else coded P7 in C on a QuadCore Intel 9650 3.67GHz
overclocked cpu, using multiple threads, and got it to be faster
than the SoA, SoE, Here's some of his results (times in seconds).

Case nPrime7x nPrime7x nPrime7x nPrime7x
Atkin Zakiya Eratosthenes Zakiya (8 core
2.5ghz)

100 billion 52.58 44.27 50.56

200 b 110.14 92.38 108.99 88.01

300 b 169.81 140.92 167.47

400 b 232.34 190.84 228.08 177.72

500 b 297.44 241.84 291.28

600 b 364.84 293.92 355.27 273.04

700 b 434.33 346.97 420.41

800 b 506.67 400.97 486.72 373.29

900 b 579.58 456.53 555.09

1 trillion 654.03 513.11 624.00 479.22

Jabari

 All times are GMT. The time now is 02:27 AM.