Fix nits in bn256.
#99
Conversation
| if power % 2 != 0 { | ||
| self.conjugate() | ||
| } |
There was a problem hiding this comment.
Can't you shift by one to the right and && by a mask? Might be harder to read but we're sure that will always be as optimized as possible.
There was a problem hiding this comment.
Not sure how to do this properly since power is usize.
We can use its methods for example:
if power.trailing_zeros() != 0 {
self.conjugate()
}but I don't think it is worth it.
| @@ -271,12 +272,14 @@ impl Fq2 { | |||
| } | |||
|
|
|||
| pub fn frobenius_map(&mut self, power: usize) { | |||
| self.c1 *= &FROBENIUS_COEFF_FQ2_C1[power % 2]; | |||
| if power % 2 != 0 { | |||
| self.conjugate() | |||
There was a problem hiding this comment.
Conjugate can't be a const fn as it is done "in place". I wonder if we can gain any speedup by not making it get an &mut and declaring the fn const.
kk since it relies in Sub which isn't const this will not work. So this is fine as is.
Just leaving the comment with my reasoning so you can double-check
| @@ -513,7 +508,7 @@ pub const FROBENIUS_COEFF_FQ6_C1: [Fq2; 6] = [ | |||
| ]; | |||
|
|
|||
| pub const FROBENIUS_COEFF_FQ6_C2: [Fq2; 6] = [ | |||
| // Fq2(u + 1)**(((2q^0) - 2) / 3) | |||
| // Fq2(u + 9)**(((2q^0) - 2) / 3) | |||
There was a problem hiding this comment.
Can you elaborate where's this 9 coming from?
There was a problem hiding this comment.
The previous value is the cubic non_residue for Bls12-381. The values for the field extensions of Bn are :
Fp2 = Fp[X] / <X^2 + 1> alpha^2 = -1
Fp6 = Fp2[X] / <X^3 - (alpha + 9> beta^3 = alpha + 9
Fp12 = Fp6[X] / <X^2 - beta> gamma^2 = beta
* Comment residues used for field extensions * Remove unusued functions and constants * Correct forb constant comment
Fixes #91 and #90.
This PR removes some unused functions and corrects some comments regarding its constants.