Where can i find huge lists of primes up to as close to 2^128 as possible of SPRP's

LarsNet · 2021-04-27, 19:28

For example i can find 2-SPRP-2-to-64.zip at https://miller-rabin.appspot.com/ Are there any other repos or places i can find large numbers like this. Wikipedia often mention tests at higher than 2^64 but fails to link to them, i'm wondering if anyone here knows of any places to find numbers like these to test against.

Also, if such lists don't exist publicly, if i wanted to make a 2-SPRP-2 list to say 2^79, what would be the minimum investment required to do something like this and what kind of equipment what be needed.

paulunderwood · 2021-04-27, 20:49

I would imagine a list up to 2^65 would take twice as much disk space as the list up to 2^64. Thus such a file would be 2^64 times as big as the the list up to 2^64. That would be about 1,000,000,000,000,000,000,000 Gigabytes.

LarsNet · 2021-04-27, 22:15

Quote:

Originally Posted by paulunderwood

I would imagine a list up to 2^65 would take twice as much disk space as the list up to 2^64. Thus such a file would be 2^64 times as big as the the list up to 2^64. That would be about 1,000,000,000,000,000,000,000 Gigabytes.

The lists i'm looking to build would only be a smaller subset of the numbers, like for this list i have from the miller rabin site is composite 2-SPRP only, so the size isn't too bad:

Code:

$ ls -al 2-SPRP-2-to-64.txt
-rw-rw-r-- 1 user user 663879564 Apr 19  2011 2-SPRP-2-to-64.txt

VBCurtis · 2021-04-27, 22:57

Quote:

Originally Posted by LarsNet

so the size isn't too bad:

Code:

$ ls -al 2-SPRP-2-to-64.txt
-rw-rw-r-- 1 user user 663879564 Apr 19  2011 2-SPRP-2-to-64.txt

663MB for 2^64 list. 2^79 is 32000x bigger than 2^64, so such a list would be.... big.

I think you'd be better off generating such a list live and testing them as they're generated. I'd rather spend 100x running time than 100x disk space! Of course, I've no idea how fast they are to generate vs read from disk, nor how many times you'll want to use the list- maybe you should find out before you buy 20TB of storage.

tuckerkao · 2021-04-27, 22:57

This website allows the input of up to 128 digits of the numbers to test for its primality, the format can be composed of a short formula as well -
https://www.numberempire.com/primenumbers.php

Uncwilly · 2021-04-27, 23:00

Quote:

Originally Posted by tuckerkao

This website allows the input of up to 128 digits of numbers to test for its primality,

And how do you generate a "huge list" from that? TheOP wants a huge list.

charybdis · 2021-04-27, 23:00

Feitsma's data suggests a rough 4.5-times increase in the number of 2-PSPs for each additional 5 powers of 2. So the file for 2^79 would be ~90 times the size of the file for 2^64: very large for a download but not hard to store. 2^128 would of course be totally impractical. The proportion of 2-PSPs below 2^64 that are strong pseudoprimes is about 27%, and this appears to increase as the numbers get bigger, so it's not a whole load easier to only store the SPSPs.

The larger problem would be finding all these PSPs. The search up to 2^64 took tens of CPU-years. Granted, that was in 2009, and GPUs would likely be useful as it appears that the most difficult part of the search involved doing a lot of trial-factoring of Mersenne numbers. Going up to 2^79 ought to be ~2^15 times harder than 2^64 (as we just have to trial-divide 15 bits higher, right?). Even if using modern GPUs gives a speedup of factor 2^15 relative to CPUs from 2009 - others can enlighten me on what a realistic number would be - this would still suggest tens of GPU-years for a search to 2^79.

And that's before we get on to this part, where the algorithm that Feitsma used would require over a terabyte of memory, so a slower method would likely be required.

tuckerkao · 2021-04-27, 23:04

I have a formula that can generate some 128 digits primes with each smaller prime, but I still haven't figured out a way to generate large amount of primes with very short period of time.

I only have around 200 of them.

The examples use 103374113, I can always substitute that number with another prime of the similar size, then try to adjust the 2^m and 3^n.

paulunderwood · 2021-04-27, 23:11

Quote:

Originally Posted by tuckerkao

I have a formula that can generate some primes
8<--------------

The OP is interested in (strong) pseudoprimes not primes

LarsNet · 2021-04-28, 04:07

Quote:

Originally Posted by charybdis

Feitsma's data suggests a rough 4.5-times increase in the number of 2-PSPs for each additional 5 powers of 2. So the file for 2^79 would be ~90 times the size of the file for 2^64: very large for a download but not hard to store. 2^128 would of course be totally impractical. The proportion of 2-PSPs below 2^64 that are strong pseudoprimes is about 27%, and this appears to increase as the numbers get bigger, so it's not a whole load easier to only store the SPSPs.

The larger problem would be finding all these PSPs. The search up to 2^64 took tens of CPU-years. Granted, that was in 2009, and GPUs would likely be useful as it appears that the most difficult part of the search involved doing a lot of trial-factoring of Mersenne numbers. Going up to 2^79 ought to be ~2^15 times harder than 2^64 (as we just have to trial-divide 15 bits higher, right?). Even if using modern GPUs gives a speedup of factor 2^15 relative to CPUs from 2009 - others can enlighten me on what a realistic number would be - this would still suggest tens of GPU-years for a search to 2^79.

And that's before we get on to this part, where the algorithm that Feitsma used would require over a terabyte of memory, so a slower method would likely be required.

Charybdis, this is awesome to me, i'm looking for these SPRP's, is there really no place to download them at? Obcoisouly i'm hoping for sizes larger than 64 but it really seems like if i want to do this that i'm going to have to buy the hardware and make them myself

LaurV · 2021-04-28, 04:20

What's the purpose of such a list? You can check the primality of any such small N extremely fast, by doing few divisions (very low TF) and then 1, or 2, or few PRP tests. That is because somebody else already did all the work for you, and such tests at this size take microseconds on a modern computer. That would be much faster than reading from a many-GB-sized file.

2021-04-27, 19:28	#1
LarsNet Mar 2021 2²·11 Posts	Where can i find huge lists of primes up to as close to 2^128 as possible of SPRP's For example i can find 2-SPRP-2-to-64.zip at https://miller-rabin.appspot.com/ Are there any other repos or places i can find large numbers like this. Wikipedia often mention tests at higher than 2^64 but fails to link to them, i'm wondering if anyone here knows of any places to find numbers like these to test against. Also, if such lists don't exist publicly, if i wanted to make a 2-SPRP-2 list to say 2^79, what would be the minimum investment required to do something like this and what kind of equipment what be needed. Last fiddled with by LarsNet on 2021-04-27 at 19:45

2021-04-27, 20:49	#2
paulunderwood Sep 2002 Database er0rr 2·3·691 Posts	I would imagine a list up to 2^65 would take twice as much disk space as the list up to 2^64. Thus such a file would be 2^64 times as big as the the list up to 2^64. That would be about 1,000,000,000,000,000,000,000 Gigabytes. Last fiddled with by paulunderwood on 2021-04-27 at 20:49

2021-04-27, 22:57	#5
tuckerkao "Tucker Kao" Jan 2020 Head Base M168202123 2·367 Posts	This website allows the input of up to 128 digits of the numbers to test for its primality, the format can be composed of a short formula as well - https://www.numberempire.com/primenumbers.php Last fiddled with by tuckerkao on 2021-04-27 at 23:00

2021-04-27, 23:04	#8
tuckerkao "Tucker Kao" Jan 2020 Head Base M168202123 2·367 Posts	I have a formula that can generate some 128 digits primes with each smaller prime, but I still haven't figured out a way to generate large amount of primes with very short period of time. I only have around 200 of them. The examples use 103374113, I can always substitute that number with another prime of the similar size, then try to adjust the 2^m and 3^n. Attached Thumbnails Last fiddled with by tuckerkao on 2021-04-27 at 23:12

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Where to find Prime Gap lists?	jaydfox	Prime Gap Searches	43	2021-07-19 17:56
Using CUDA to find better SPRP classifiers	SPWorley	Computer Science & Computational Number Theory	11	2012-11-21 20:13
Why arent there many softwares for finding Huge Primes	blistervol	Math	2	2012-08-20 17:26
An aliquot sequence with huge, huge, huge tracts of sand !!!	garambois	Aliquot Sequences	50	2012-01-19 18:25
Why Search for these Huge Primes?	Unregistered	Math	8	2005-04-27 00:55

2021-04-27, 23:00	#7
charybdis Apr 2020 2C4₁₆ Posts	Feitsma's data suggests a rough 4.5-times increase in the number of 2-PSPs for each additional 5 powers of 2. So the file for 2^79 would be ~90 times the size of the file for 2^64: very large for a download but not hard to store. 2^128 would of course be totally impractical. The proportion of 2-PSPs below 2^64 that are strong pseudoprimes is about 27%, and this appears to increase as the numbers get bigger, so it's not a whole load easier to only store the SPSPs. The larger problem would be finding all these PSPs. The search up to 2^64 took tens of CPU-years. Granted, that was in 2009, and GPUs would likely be useful as it appears that the most difficult part of the search involved doing a lot of trial-factoring of Mersenne numbers. Going up to 2^79 ought to be ~2^15 times harder than 2^64 (as we just have to trial-divide 15 bits higher, right?). Even if using modern GPUs gives a speedup of factor 2^15 relative to CPUs from 2009 - others can enlighten me on what a realistic number would be - this would still suggest tens of GPU-years for a search to 2^79. And that's before we get on to this part, where the algorithm that Feitsma used would require over a terabyte of memory, so a slower method would likely be required.

2021-04-28, 04:20	#11
LaurV Romulan Interpreter "name field" Jun 2011 Thailand 9963₁₀ Posts	What's the purpose of such a list? You can check the primality of any such small N extremely fast, by doing few divisions (very low TF) and then 1, or 2, or few PRP tests. That is because somebody else already did all the work for you, and such tests at this size take microseconds on a modern computer. That would be much faster than reading from a many-GB-sized file.