Low hanging performance fixes and some comment cleanup #4
Conversation
Clean up some old code comments that were distracting Use a new factorization algorithm so that the tests actually finish. This will make working much less tedious.
…dulus the prime and return maybe prime although we can just as easily check if it actually that prime right here. This was causing small primes to go through the more expensive primality tests although we didn't need to. Add a test for factoring two prime numbers
| // Settings | ||
| // NIST recomends 5 rounds for miller rabin. This implementation does 8. Apple uses 16. Three iterations has a probability of 2^80 of failing | ||
| const MILLER_RABIN_ROUNDS: usize = 8usize; | ||
| const MILLER_RABIN_ROUNDS: usize = 64usize; |
There was a problem hiding this comment.
I bumped this to 64 to match openssl but we can set it back to 8. Just noticed the comment so I will update it later this evening.
| //candidate.set_bit(0, true); | ||
| //candidate.set_bit((n-1) as u32, true); | ||
| let mut candidate: BigUint = rng.gen_biguint(n); | ||
| candidate.set_bit((n-1) as u64, true); |
There was a problem hiding this comment.
Composite numbers should be able to be even(?) so I didn't set the 0th bit. Let me know if we only want to generate odd composites.
| } | ||
|
|
||
| if &d == &org { | ||
| return Some(brute_force_prime_factor(d.try_into().unwrap())); |
There was a problem hiding this comment.
If we fail to find a factor, fall back to brute force search up to the sqrt.
| for mut i in 3..n_sqrt { | ||
| while n.divides(&BigUint::from(i)) { | ||
| n = n / BigUint::from(i); | ||
| for mut i in 3..n_sqrt + 1 { |
There was a problem hiding this comment.
We want to check the n_sqrt as well in this case since it is possible that n = p^2
| MaybePrime | ||
| } | ||
|
|
||
| // if true, then is not prime |
There was a problem hiding this comment.
I'll remove this comments tonight.
|
|
||
| let (d,s) = rewrite(&candidate); | ||
| let step = s.sub(&one).to_usize().unwrap(); | ||
| let one_usize = one.to_usize().unwrap(); |
There was a problem hiding this comment.
Just moved these outside of the loop since they are constants.
| else { | ||
| return true | ||
| } | ||
| return true |
There was a problem hiding this comment.
This was unreachable code.
| assert_eq!(prime_factor, q); | ||
| } | ||
|
|
||
| } |
There was a problem hiding this comment.
I'll add a newline tonight :D
1 participant
Bump dependencies and move allocations outside of the hotpath.
Clean up some old code comments that were distracting
Use a new factorization algorithm so that the tests actually finish. This will make working much less tedious. It may be buggy still but its a very simple algorithm and buggy is better than an infinite run time.
Set bits properly now that set_bit is available. This way the integers will be guaranteed to be large enough to use in cryptography.