What's a fast way to fulfill this condition?

calccrypto · 2010-07-18 13:03:18

given:

Choose an L-bit prime modulus p such that p1 is a multiple of q.

what is the best way to find p that is both prime and (p-1) mod q = 0? right now, im doing the brute force way of:

q is some big prime number
p = randint(3,2^L)                     # i guess the bit size can be ignored for the moment
p = q * (1 + p/q) + 1                  # since the values are ints, p/q is really floor(p/q). this gets a value that is a (multiple of q) +1, like q = 7 ->p = 3*7 +1 = 22

while p is not prime:                  # the testing function is MillerRabin
   p = p - q

testing is taking forever, but since im terrible at reading other people's codes, ive been unable to find out how other people programed this to work very fast

bobbym · 2010-07-18 13:11:54

Hi calccrypto;

A question: Is randint(a,b) that you are using a python command? If so what does it return?

calccrypto · 2010-07-18 13:18:20

well the entire code block is just a pseudo code, but randint (which is also a legitimate command) is supposed to returns a random integer between a and b, inclusive. the real code is barely any different

bobbym · 2010-07-18 13:29:11

Hi;

Isn't this a Rabin Miller primality tester? It has been 15 years since I have done any computerized primality testing but this reminds me of the deterministec form of Miller Rabin?!

If I am not in error then I must say this test is much slower than the probabilitic version of it. So how big are the numbers you are trying to test for primality?

calccrypto · 2010-07-18 13:43:49

no. i guess my comment was a bit vague

i might as well put the real code:

	p = randint(3, 2**L)
	p = q * (1 + p/q) + 1
	while MillerRabin(p) == False:                # prime = True, composite = False
		p = p + q

miller rabin is not what im creating. rather, it is another function that im using to test my p values, and it is the probabilistic test.

the L value can be 1024,2048, or 3072

bobbym · 2010-07-18 16:10:01

Hi,

2^3072 are 1000 digit numbers, maybe that's where the slow down is.

calccrypto · 2010-07-19 02:47:44

yep, which is why i want to try to speed up the testing a bit

bobbym · 2010-07-19 03:15:36

Hi cal;

It may not be possible with numbers that large. Those are really large. Where is the bottleneck? Have you wrapped pieces of the code with timers to see which part is the slowest?

Also being a probabilistic primality tester each iteration reduces the probability that the number is composite by a factor of 4. So after only a couple of iterations you will have a number that is very likely to be prime. You will not have certainty.

calccrypto · 2010-07-19 06:23:27

I think the bottleneck is the testing itself. Oh well. No matter. I'll eventually figure this out.

Thanks bobbym

bobbym · 2010-07-19 06:41:05

Hi;

Yes, but which part is using most of the time. That's the part you try to speed up.

calccrypto · 2010-07-19 09:49:02

is there a better and faster method of testing?

Last edited by calccrypto (2010-07-19 11:29:47)

bobbym · 2010-07-19 10:57:59

Depends what are you trying to do? Primality testing? What?

calccrypto · 2010-07-19 11:29:20

yeah. just testing if the number is prime

bobbym · 2010-07-19 12:45:39

Hi cal;

You do know that the fastest deterministic test is the new one invented by those 3 Indian mathematicians and will determine the status of a number with 1000 digits in about 1 hour on a modern desktop.

This is the fastest known.

calccrypto · 2010-07-19 13:10:36

darn.

bobbym · 2010-07-19 19:35:45

That is the fastest for an exact answer concerning primality. If you use the probabilistic RM and run through the algorithm say 15
times you would have a probability less than .000 000 000 931 that the number is not prime.

Math Is Fun Forum

#1 2010-07-18 13:03:18

What's a fast way to fulfill this condition?

#2 2010-07-18 13:11:54

Re: What's a fast way to fulfill this condition?

#3 2010-07-18 13:18:20

Re: What's a fast way to fulfill this condition?

#4 2010-07-18 13:29:11

Re: What's a fast way to fulfill this condition?

#5 2010-07-18 13:43:49

Re: What's a fast way to fulfill this condition?

#6 2010-07-18 16:10:01

Re: What's a fast way to fulfill this condition?

#7 2010-07-19 02:47:44

Re: What's a fast way to fulfill this condition?

#8 2010-07-19 03:15:36

Re: What's a fast way to fulfill this condition?

#9 2010-07-19 06:23:27

Re: What's a fast way to fulfill this condition?

#10 2010-07-19 06:41:05

Re: What's a fast way to fulfill this condition?

#11 2010-07-19 09:49:02

Re: What's a fast way to fulfill this condition?

#12 2010-07-19 10:57:59

Re: What's a fast way to fulfill this condition?

#13 2010-07-19 11:29:20

Re: What's a fast way to fulfill this condition?

#14 2010-07-19 12:45:39

Re: What's a fast way to fulfill this condition?

#15 2010-07-19 13:10:36

Re: What's a fast way to fulfill this condition?

#16 2010-07-19 19:35:45

Re: What's a fast way to fulfill this condition?

Board footer