|
|
@ -0,0 +1,113 @@ |
|
|
|
/// Isolated test which gets the number of constraints for two 'naive' approaches for the
|
|
|
|
/// matrix-vector-product:
|
|
|
|
/// - handcrafted_A_by_z method
|
|
|
|
/// - mat_vec_mul_sparse_gadget
|
|
|
|
///
|
|
|
|
use ark_ff::PrimeField;
|
|
|
|
use ark_r1cs_std::fields::nonnative::NonNativeFieldVar;
|
|
|
|
use ark_r1cs_std::{alloc::AllocVar, eq::EqGadget, fields::FieldVar, R1CSVar};
|
|
|
|
use ark_relations::r1cs::{ConstraintSynthesizer, ConstraintSystemRef, SynthesisError};
|
|
|
|
use core::marker::PhantomData;
|
|
|
|
use std::ops::Mul;
|
|
|
|
|
|
|
|
use crate::utils::*;
|
|
|
|
|
|
|
|
/// - F stands for the field that we represent
|
|
|
|
/// - CF stands for the ConstraintField over which we do the operations
|
|
|
|
|
|
|
|
/// Implements the A * z matrix-vector-product by fixing the combinations of 'z'.
|
|
|
|
fn handcrafted_A_by_z<F: PrimeField, CF: PrimeField>(
|
|
|
|
cs: ConstraintSystemRef<CF>,
|
|
|
|
z: Vec<NonNativeFieldVar<F, CF>>,
|
|
|
|
) -> Result<Vec<NonNativeFieldVar<F, CF>>, SynthesisError> {
|
|
|
|
let five = NonNativeFieldVar::<F, CF>::new_constant(cs.clone(), F::from(5u32))?;
|
|
|
|
// directly hand-craft the output vector containing the operations in-place:
|
|
|
|
Ok(vec![
|
|
|
|
z[1].clone() + five.clone() * z[4].clone(),
|
|
|
|
z[1].clone() + z[3].clone(),
|
|
|
|
z[1].clone() + z[4].clone(),
|
|
|
|
five * z[0].clone() + z[4].clone() + z[5].clone(),
|
|
|
|
]
|
|
|
|
.clone())
|
|
|
|
}
|
|
|
|
|
|
|
|
/// Implements the A * z matrix-vector-product by doing the sparse matrix by vector algorithm, and
|
|
|
|
/// assuming that the elements of the matrix A are constants of the system.
|
|
|
|
pub fn mat_vec_mul_sparse_gadget<F: PrimeField, CF: PrimeField>(
|
|
|
|
m: SparseMatrixVar<F, CF>,
|
|
|
|
v: Vec<NonNativeFieldVar<F, CF>>,
|
|
|
|
) -> Vec<NonNativeFieldVar<F, CF>> {
|
|
|
|
let mut res = vec![NonNativeFieldVar::<F, CF>::zero(); m.n_rows];
|
|
|
|
for (row_i, row) in m.coeffs.iter().enumerate() {
|
|
|
|
for (value, col_i) in row.iter() {
|
|
|
|
if value.value().unwrap() == F::one() {
|
|
|
|
res[row_i] += v[*col_i].clone(); // when value==1, no need to multiply by it
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
res[row_i] += value.clone().mul(&v[*col_i].clone());
|
|
|
|
}
|
|
|
|
}
|
|
|
|
res
|
|
|
|
}
|
|
|
|
|
|
|
|
/// Circuit that takes as constants the sparse matrix A, and as inputs the vectors z and y. It
|
|
|
|
/// computes the matrix by vector product between A and z, and checks that is equal to y
|
|
|
|
/// (ie. y == A*z)
|
|
|
|
struct MatrixVectorCircuit<F: PrimeField, CF: PrimeField> {
|
|
|
|
_cf: PhantomData<CF>,
|
|
|
|
pub A: SparseMatrix<F>,
|
|
|
|
pub z: Vec<F>,
|
|
|
|
pub y: Vec<F>,
|
|
|
|
}
|
|
|
|
impl<F: PrimeField, CF: PrimeField> ConstraintSynthesizer<CF> for MatrixVectorCircuit<F, CF> {
|
|
|
|
fn generate_constraints(self, cs: ConstraintSystemRef<CF>) -> Result<(), SynthesisError> {
|
|
|
|
// set A as circuit constants
|
|
|
|
let A = SparseMatrixVar::<F, CF>::new_constant(cs.clone(), self.A)?;
|
|
|
|
// set z and y as witness (private inputs)
|
|
|
|
let z: Vec<NonNativeFieldVar<F, CF>> = Vec::new_witness(cs.clone(), || Ok(self.z.clone()))?;
|
|
|
|
let y: Vec<NonNativeFieldVar<F, CF>> = Vec::new_witness(cs.clone(), || Ok(self.y.clone()))?;
|
|
|
|
|
|
|
|
/// The next two lines are the ones that can be swapped to see the number of constraints
|
|
|
|
/// taken by the two approaches:
|
|
|
|
let Az = mat_vec_mul_sparse_gadget(A, z);
|
|
|
|
// let Az = handcrafted_A_by_z(cs, z)?;
|
|
|
|
|
|
|
|
Az.enforce_equal(&y)?;
|
|
|
|
Ok(())
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
#[cfg(test)]
|
|
|
|
mod tests {
|
|
|
|
use super::*;
|
|
|
|
use ark_pallas::{Fq, Fr};
|
|
|
|
use ark_relations::r1cs::ConstraintSystem;
|
|
|
|
|
|
|
|
#[test]
|
|
|
|
fn test_relaxed_r1cs_nonnative_matrix_vector_product() {
|
|
|
|
let A = to_F_matrix::<Fq>(vec![
|
|
|
|
vec![0, 1, 0, 0, 5, 0],
|
|
|
|
vec![0, 1, 0, 1, 0, 0],
|
|
|
|
vec![0, 1, 0, 0, 1, 0],
|
|
|
|
vec![5, 0, 0, 0, 1, 1],
|
|
|
|
]);
|
|
|
|
let z = to_F_vec(vec![1, 123, 35, 53, 80, 30]);
|
|
|
|
let y = mat_vec_mul_sparse(&A, &z); // y = A*z
|
|
|
|
println!("Matrix of size {} x {}", A.n_rows, A.n_cols);
|
|
|
|
println!("Vector of size {}", z.len());
|
|
|
|
|
|
|
|
println!(
|
|
|
|
"Build the circuit that computes the matrix-vector-product over a non-native field"
|
|
|
|
);
|
|
|
|
let cs = ConstraintSystem::<Fr>::new_ref();
|
|
|
|
let circuit = MatrixVectorCircuit::<Fq, Fr> {
|
|
|
|
_cf: PhantomData,
|
|
|
|
A,
|
|
|
|
z,
|
|
|
|
y,
|
|
|
|
};
|
|
|
|
circuit.generate_constraints(cs.clone()).unwrap();
|
|
|
|
println!("Number of constraints: {}", cs.num_constraints());
|
|
|
|
assert!(cs.is_satisfied().unwrap());
|
|
|
|
}
|
|
|
|
}
|