[wip] first pass at framing arithmetic #10
Conversation
src/finite_field.rs
Outdated
| // TODO: implement bernstein yang inversion | ||
| Self::zero_array() | ||
|
|
||
| pub const fn greater_than(&self, a: &[u64; L], b:&[u64;L])-> bool { |
There was a problem hiding this comment.
This is not truly constant time as different branches on the comparator may return at different times, we should instead always run through the whole loop and then return only afterwards.
src/finite_field.rs
Outdated
| r_squared, | ||
| n_prime, | ||
| } | ||
| correction: [0u64; L], |
There was a problem hiding this comment.
These values could be computed beforehand rather than having a mutation occur.
There was a problem hiding this comment.
agreed, this was left over from before I moved arithmetic to _internals, will remedy
src/finite_field.rs
Outdated
| // TODO (Implement Montgomery r squared) | ||
| Self::zero_array() | ||
| } | ||
| let diff = Self::subtraction_correction(modulus); |
There was a problem hiding this comment.
This could be passed in to the function rather than recomputed.
src/finite_field.rs
Outdated
| let n = modulus[0]; //need only least significant bits | ||
| let mut n_prime = 1u64; | ||
| let mut i = 0; | ||
| while i < 64 { |
There was a problem hiding this comment.
Why this number, because 64 bits?
src/finite_field.rs
Outdated
| } | ||
|
|
||
| pub const fn to_montgomery(&self, a: &[u64; L]) -> [u64; L] { | ||
| // TODO (Implement to monty form) | ||
| Self::zero_array() | ||
| self.montgomery_multiply(a, &self.r_squared) |
There was a problem hiding this comment.
Some detail on why multiplying by r_squared gives us the monty form would be helpful.
src/finite_field.rs
Outdated
| } | ||
|
|
||
| pub const fn from_montgomery(&self, a: &[u64; L]) -> [u64; L] { | ||
| // TODO (Implement from monty form) | ||
| Self::zero_array() | ||
| self.montgomery_multiply(a, &Self::one_array()) |
There was a problem hiding this comment.
Same here on why multiplying by 1 (the identity, not an array filled with ones) gives us the normal form.
src/finite_field.rs
Outdated
| @@ -198,110 +206,177 @@ impl<const L: usize, const D: usize> FinitePrimeField<L, D> { | |||
| j += 1; | |||
| } | |||
|
|
|||
| result | |||
| self.to_montgomery(&result) | |||
There was a problem hiding this comment.
Why here do we convert to_montgomery? Should this not already be in monty form at the end of this, or has it been reduced?
src/finite_field.rs
Outdated
| let mut negated = Self::zero_array(); | ||
| let mut i = 0; | ||
| while i < L { | ||
| negated[i] = self.modulus[i].wrapping_sub(a[i]); |
There was a problem hiding this comment.
I'm thinking some carry about the modulus needs to happen here.
src/finite_field.rs
Outdated
| } | ||
| } | ||
| /// The following performs the Bernstein-Yang inversion on a scalar. | ||
| pub const fn bernstein_yang_invert(&self, a: &[u64; L]) -> [u64; L] { |
There was a problem hiding this comment.
Perfect use case for dimensionals 💯
* chore: bring in template * fmt
0xAlcibiades
changed the base branch from
main
to
tristan/war-529-deep-sylow-optimization
trbritt
merged commit bc19bcb
into
tristan/war-529-deep-sylow-optimization
No description provided.