feat: enable to_radix for any field element

crates/nargo_cli/tests/test_data/to_bytes_integration/Prover.toml

@@ -1 +1,2 @@
 x = "2040124"
+_y = "0x2000000000000000000000000000000000000000000000000000000000000000"

crates/nargo_cli/tests/test_data/to_bytes_integration/src/main.nr

@@ -1,6 +1,6 @@
 use dep::std;
 
-fn main(x : Field)  {
+fn main(x : Field, _y: Field)  {
     // The result of this byte array will be big-endian
     let y: Field = 2040124;
     let be_byte_array = y.to_be_bytes(31);
@@ -11,4 +11,17 @@ fn main(x : Field)  {
     assert(le_byte_array[0] == be_byte_array[30]);
     assert(le_byte_array[1] == be_byte_array[29]);
     assert(le_byte_array[2] == be_byte_array[28]);
+
+    let z = 0 - 1;
+    let p_bytes = std::field::modulus_le_bytes();
+    let z_bytes = z.to_le_bytes(32);
+    assert(p_bytes[10] == z_bytes[10]);
+    assert(p_bytes[0] == z_bytes[0] as u8 + 1 as u8);
+
+    let p_bits = std::field::modulus_le_bits();
+    let z_bits = z.to_le_bits(std::field::modulus_num_bits() as u32);
+    assert(z_bits[0] == 0);
+    assert(p_bits[100] == z_bits[100]);
+
+   _y.to_le_bits(std::field::modulus_num_bits() as u32);
 }

crates/noirc_evaluator/src/ssa/acir_gen/constraints.rs

@@ -405,25 +405,43 @@ pub(crate) fn to_radix_base(
     evaluator: &mut Evaluator,
 ) -> Vec<Witness> {
     // ensure there is no overflow
-    let mut max = BigUint::from(radix);
-    max = max.pow(limb_size) - BigUint::one();
-    assert!(max < FieldElement::modulus());
+    let rad = BigUint::from(radix);
+    let max = rad.pow(limb_size) - BigUint::one();
 
-    let (mut result, bytes) = to_radix_little(radix, limb_size, evaluator);
+    if max < FieldElement::modulus() {
+        let (mut result, bytes) = to_radix_little(radix, limb_size, evaluator);
 
-    evaluator.push_opcode(AcirOpcode::Directive(Directive::ToLeRadix {
-        a: lhs.clone(),
-        b: result.clone(),
-        radix,
-    }));
+        evaluator.push_opcode(AcirOpcode::Directive(Directive::ToLeRadix {
+            a: lhs.clone(),
+            b: result.clone(),
+            radix,
+        }));
 
-    if endianness == Endian::Big {
-        result.reverse();
-    }
+        if endianness == Endian::Big {
+            result.reverse();
+        }
 
-    evaluator.push_opcode(AcirOpcode::Arithmetic(subtract(lhs, FieldElement::one(), &bytes)));
+        evaluator.push_opcode(AcirOpcode::Arithmetic(subtract(lhs, FieldElement::one(), &bytes)));
+        result
+    } else {
+        let min = rad.pow(limb_size - 1) - BigUint::one();
+        assert!(min < FieldElement::modulus());
+
+        let max_bits = max.bits() as u32;
+        let a = evaluate_constant_modulo(lhs, radix, max_bits, evaluator)
+            .to_witness()
+            .expect("Constant expressions should already be simplified");
+        let y = subtract(lhs, FieldElement::one(), &Expression::from(a));
+        let radix_f = FieldElement::from(radix as i128);
+        let y = Expression::default().add_mul(FieldElement::one() / radix_f, &y);
+        let mut b = to_radix_base(&y, radix, limb_size - 1, endianness, evaluator);
+        match endianness {
+            Endian::Little => b.insert(0, a),
+            Endian::Big => b.push(a),
+        }
 
-    result
+        b
+    }
 }
 
 //Decomposition into b-base: \sum ai b^i, where 0<=ai<b
@@ -527,6 +545,59 @@ pub(crate) fn evaluate_truncate(
     Expression::from(b_witness)
 }
 
+//Returns b such that lhs (a number whose value requires max_bits) mod rhs is b
+pub(crate) fn evaluate_constant_modulo(
+    lhs: &Expression,
+    rhs: u32,
+    max_bits: u32,
+    evaluator: &mut Evaluator,
+) -> Expression {
+    let modulus = FieldElement::from(rhs as i128);
+    let modulus_exp = Expression::from_field(modulus);
+    assert_ne!(rhs, 0);
+    let modulus_bits = bit_size_u128((rhs - 1) as u128);
+    assert!(max_bits >= rhs, "max_bits = {max_bits}, rhs = {rhs}");
+    //0. Check for constant expression. This can happen through arithmetic simplifications
+    if let Some(a_c) = lhs.to_const() {
+        let mut a_big = BigUint::from_bytes_be(&a_c.to_be_bytes());
+        a_big %= BigUint::from_bytes_be(&modulus.to_be_bytes());
+        return Expression::from(FieldElement::from_be_bytes_reduce(&a_big.to_bytes_be()));
+    }
+
+    //1. Generate witnesses b,c
+    let b_witness = evaluator.add_witness_to_cs();
+    let c_witness = evaluator.add_witness_to_cs();
+    evaluator.push_opcode(AcirOpcode::Directive(Directive::Quotient {
+        a: lhs.clone(),
+        b: modulus_exp.clone(),
+        q: c_witness,
+        r: b_witness,
+        predicate: None,
+    }));
+    bound_constraint_with_offset(
+        &Expression::from(b_witness),
+        &modulus_exp,
+        &Expression::one(),
+        modulus_bits,
+        evaluator,
+    );
+    //if rhs is a power of 2, then we avoid this range check as it is redundant with the previous one.
+    if rhs & (rhs - 1) != 0 {
+        try_range_constraint(b_witness, modulus_bits, evaluator);
+    }
+    let c_bound = FieldElement::modulus() / BigUint::from(rhs) - BigUint::one();
+    try_range_constraint(c_witness, c_bound.bits() as u32, evaluator);
+
+    //2. Add the constraint lhs = b+q*rhs
+    let b_arith = b_witness.into();
+    let c_arith = c_witness.into();
+    let res = add(&b_arith, modulus, &c_arith);
+    let my_constraint = add(&res, -FieldElement::one(), lhs);
+    evaluator.push_opcode(AcirOpcode::Arithmetic(my_constraint));
+
+    Expression::from(b_witness)
+}
+
 pub(crate) fn evaluate_udiv(
     lhs: &Expression,
     rhs: &Expression,