package bls12-381
Library
Module
Module type
Parameter
Class
Class type
Elliptic curve built over the field Fq
and the equation y^2 = x^3 + 4
include CURVE
exception Not_on_curve of Bytes.t
The type of the element on the curve and in the prime subgroup. The point is given in jacobian coordinates
affine_of_jacobian p
creates a new value of type affine
representing the point p
in affine coordinates
jacobian_of_affine p
creates a new value of type t
representing the point p
in jacobian coordinates
val to_affine_array : t array -> affine_array
to_affine_array pts
builds a contiguous C array and populate it with the points pts
in affine coordinates. Use it with pippenger_with_affine_array
to get better performance.
val of_affine_array : affine_array -> t array
Build a OCaml array of t
values from the contiguous C array
val size_of_affine_array : affine_array -> int
Return the number of elements in the array
module Scalar : Ff_sig.PRIME with type t = Fr.t
val check_bytes : Bytes.t -> bool
Check if a point, represented as a byte array, is on the curve and in the prime subgroup. The bytes must be of length size_in_bytes
.
Attempt to construct a point from a byte array of length size_in_bytes
.
Attempt to construct a point from a byte array of length size_in_bytes
. Raise Not_on_curve
if the point is not on the curve
Allocates a new point from a byte of length size_in_bytes / 2
array representing a point in compressed form.
Allocates a new point from a byte array of length size_in_bytes / 2
representing a point in compressed form. Raise Not_on_curve
if the point is not on the curve.
val zero : t
Zero of the elliptic curve
val one : t
A fixed generator of the elliptic curve
val is_zero : t -> bool
Return true
if the given element is zero
val random : ?state:Random.State.t -> unit -> t
Generate a random element. The function ensures the element is on the curve and in the prime subgroup.
The routines in the module Random.State
are used to generate the elements. A state can be given to the function to be used. If no state is given, Random.get_state
is used.
To create a value of type Random.State.t
, you can use Random.State.make
[|42|]
.
add_inplace a b
is the same than add
but writes the output in a
. No allocation happens.
add_bulk xs
returns the sum of the elements of xs
by performing only one allocation for the output. This method is recommended to save the allocation overhead of using n
times add
.
mul_inplace g x
is the same than mul
but writes the output in g
. No allocation happens.
fft ~domain ~points
performs a Fourier transform on points
using domain
The domain should be of the form w^{i}
where w
is a principal root of unity. If the domain is of size n
, w
must be a n
-th principal root of unity. The number of points can be smaller than the domain size, but not larger. The complexity is in O(n log(m))
where n
is the domain size and m
the number of points. A new array of size n
is allocated and is returned. The parameters are not modified.
fft_inplace ~domain ~points
performs a Fourier transform on points
using domain
The domain should be of the form w^{i}
where w
is a principal root of unity. If the domain is of size n
, w
must be a n
-th principal root of unity. The number of points must be in the same size than the domain. It does not return anything but modified the points directly. It does only perform one allocation of a scalar for the FFT. It is recommended to use this function if side-effect is acceptable.
ifft ~domain ~points
performs an inverse Fourier transform on points
using domain
. The domain should be of the form w^{-i}
(i.e the "inverse domain") where w
is a principal root of unity. If the domain is of size n
, w
must be a n
-th principal root of unity. The domain size must be exactly the same than the number of points. The complexity is O(n log(n)) where n
is the domain size. A new array of size n
is allocated and is returned. The parameters are not modified.
ifft_inplace ~domain ~points
is the same than ifft
but modifies the array points
instead of returning a new array
hash_to_curve msg dst
follows the standard Hashing to Elliptic Curves applied to BLS12-381
pippenger ?start ?len pts scalars
computes the multi scalar exponentiation/multiplication. The scalars are given in scalars
and the points in pts
. If pts
and scalars
are not of the same length, perform the computation on the first n
points where n
is the smallest size. Arguments start
and len
can be used to take advantages of multicore OCaml. Default value for start
(resp. len
) is 0
(resp. the length of the array scalars
).
val pippenger_with_affine_array :
?start:int ->
?len:int ->
affine_array ->
Scalar.t array ->
t
pippenger_with_affine_array ?start ?len pts scalars
computes the multi scalar exponentiation/multiplication. The scalars are given in scalars
and the points in pts
. If pts
and scalars
are not of the same length, perform the computation on the first n
points where n
is the smallest length. The differences with pippenger
are 1. the points are loaded in a contiguous C array to speed up the access to the elements by relying on the CPU cache 2. and the points are in affine coordinates, the form expected by the algorithm implementation, avoiding new allocations and field inversions required to convert from jacobian (representation of a points of type t
, as expected by pippenger
) to affine coordinates. Expect a speed improvement around 20% compared to pippenger
, and less allocation on the C heap. A value of affine_array
can be built using to_affine_array
. Arguments start
and len
can be used to take advantages of multicore OCaml. Default value for start
(resp. len
) is 0
(resp. the length of the array scalars
).