Merge pull request #1554 from nbraud/factor/faster/montgomery32

factor: Refactor and improve performance (plus a few bug fixes)

Cargo.lock

@@ -649,6 +649,23 @@ dependencies = [
  "pkg-config 0.3.17 (registry+https://github.com/rust-lang/crates.io-index)",
 ]
 
+[[package]]
+name = "paste"
+version = "0.1.18"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+dependencies = [
+ "paste-impl 0.1.18 (registry+https://github.com/rust-lang/crates.io-index)",
+ "proc-macro-hack 0.5.16 (registry+https://github.com/rust-lang/crates.io-index)",
+]
+
+[[package]]
+name = "paste-impl"
+version = "0.1.18"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+dependencies = [
+ "proc-macro-hack 0.5.16 (registry+https://github.com/rust-lang/crates.io-index)",
+]
+
 [[package]]
 name = "pkg-config"
 version = "0.3.17"
@@ -668,6 +685,11 @@ name = "ppv-lite86"
 version = "0.2.8"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 
+[[package]]
+name = "proc-macro-hack"
+version = "0.5.16"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+
 [[package]]
 name = "proc-macro2"
 version = "1.0.18"
@@ -1335,6 +1357,8 @@ dependencies = [
 name = "uu_factor"
 version = "0.0.1"
 dependencies = [
+ "num-traits 0.2.12 (registry+https://github.com/rust-lang/crates.io-index)",
+ "paste 0.1.18 (registry+https://github.com/rust-lang/crates.io-index)",
  "quickcheck 0.9.2 (registry+https://github.com/rust-lang/crates.io-index)",
  "rand 0.5.6 (registry+https://github.com/rust-lang/crates.io-index)",
  "uucore 0.0.4 (git+https://github.com/uutils/uucore.git?branch=canary)",
@@ -2253,9 +2277,12 @@ dependencies = [
 "checksum numtoa 0.1.0 (registry+https://github.com/rust-lang/crates.io-index)" = "b8f8bdf33df195859076e54ab11ee78a1b208382d3a26ec40d142ffc1ecc49ef"
 "checksum onig 4.3.3 (registry+https://github.com/rust-lang/crates.io-index)" = "8518fcb2b1b8c2f45f0ad499df4fda6087fc3475ca69a185c173b8315d2fb383"
 "checksum onig_sys 69.1.0 (registry+https://github.com/rust-lang/crates.io-index)" = "388410bf5fa341f10e58e6db3975f4bea1ac30247dd79d37a9e5ced3cb4cc3b0"
+"checksum paste 0.1.18 (registry+https://github.com/rust-lang/crates.io-index)" = "45ca20c77d80be666aef2b45486da86238fabe33e38306bd3118fe4af33fa880"
+"checksum paste-impl 0.1.18 (registry+https://github.com/rust-lang/crates.io-index)" = "d95a7db200b97ef370c8e6de0088252f7e0dfff7d047a28528e47456c0fc98b6"
 "checksum pkg-config 0.3.17 (registry+https://github.com/rust-lang/crates.io-index)" = "05da548ad6865900e60eaba7f589cc0783590a92e940c26953ff81ddbab2d677"
 "checksum platform-info 0.0.1 (registry+https://github.com/rust-lang/crates.io-index)" = "f2fd076acdc7a98374de6e300bf3af675997225bef21aecac2219553f04dd7e8"
 "checksum ppv-lite86 0.2.8 (registry+https://github.com/rust-lang/crates.io-index)" = "237a5ed80e274dbc66f86bd59c1e25edc039660be53194b5fe0a482e0f2612ea"
+"checksum proc-macro-hack 0.5.16 (registry+https://github.com/rust-lang/crates.io-index)" = "7e0456befd48169b9f13ef0f0ad46d492cf9d2dbb918bcf38e01eed4ce3ec5e4"
 "checksum proc-macro2 1.0.18 (registry+https://github.com/rust-lang/crates.io-index)" = "beae6331a816b1f65d04c45b078fd8e6c93e8071771f41b8163255bbd8d7c8fa"
 "checksum quick-error 1.2.3 (registry+https://github.com/rust-lang/crates.io-index)" = "a1d01941d82fa2ab50be1e79e6714289dd7cde78eba4c074bc5a4374f650dfe0"
 "checksum quickcheck 0.9.2 (registry+https://github.com/rust-lang/crates.io-index)" = "a44883e74aa97ad63db83c4bf8ca490f02b2fc02f92575e720c8551e843c945f"

src/uu/factor/Cargo.toml

@@ -11,12 +11,18 @@ keywords = ["coreutils", "uutils", "cross-platform", "cli", "utility"]
 categories = ["command-line-utilities"]
 edition = "2018"
 
+[build-dependencies]
+num-traits = "0.2" # used in src/numerics.rs, which is included by build.rs
+
+
 [dependencies]
+num-traits = "0.2"
 rand = "0.5"
 uucore = { version="0.0.4", package="uucore", git="https://github.com/uutils/uucore.git", branch="canary" }
 uucore_procs = { version="0.0.4", package="uucore_procs", git="https://github.com/uutils/uucore.git", branch="canary" }
 
 [dev-dependencies]
+paste = "0.1.18"
 quickcheck = "0.9.2"
 
 [[bin]]

src/uu/factor/build.rs

@@ -29,7 +29,7 @@ use miller_rabin::is_prime;
 
 #[path = "src/numeric.rs"]
 mod numeric;
-use numeric::inv_mod_u64;
+use numeric::modular_inverse;
 
 mod sieve;
 
@@ -57,7 +57,7 @@ fn main() {
     let mut x = primes.next().unwrap();
     for next in primes {
         // format the table
-        let outstr = format!("({}, {}, {}),", x, inv_mod_u64(x), std::u64::MAX / x);
+        let outstr = format!("({}, {}, {}),", x, modular_inverse(x), std::u64::MAX / x);
         if cols + outstr.len() > MAX_WIDTH {
             write!(file, "\n    {}", outstr).unwrap();
             cols = 4 + outstr.len();

src/uu/factor/src/factor.rs

@@ -54,12 +54,16 @@ impl fmt::Display for Factors {
     }
 }
 
-fn _factor<A: Arithmetic>(num: u64, f: Factors) -> Factors {
+fn _factor<A: Arithmetic + miller_rabin::Basis>(num: u64, f: Factors) -> Factors {
     use miller_rabin::Result::*;
+
     // Shadow the name, so the recursion automatically goes from “Big” arithmetic to small.
     let _factor = |n, f| {
-        // TODO: Optimise with 32 and 64b versions
+        if n < (1 << 32) {
-        _factor::<A>(n, f)
+            _factor::<Montgomery<u32>>(n, f)
+        } else {
+            _factor::<A>(n, f)
+        }
     };
 
     if num == 1 {
@@ -101,8 +105,11 @@ pub fn factor(mut n: u64) -> Factors {
 
     let (factors, n) = table::factor(n, factors);
 
-    // TODO: Optimise with 32 and 64b versions
+    if n < (1 << 32) {
-    _factor::<Montgomery>(n, factors)
+        _factor::<Montgomery<u32>>(n, factors)
+    } else {
+        _factor::<Montgomery<u64>>(n, factors)
+    }
 }
 
 #[cfg(test)]

src/uu/factor/src/miller_rabin.rs

@@ -2,10 +2,27 @@
 
 use crate::numeric::*;
 
-// Small set of bases for the Miller-Rabin prime test, valid for all 64b integers;
+pub(crate) trait Basis {
-//  discovered by Jim Sinclair on 2011-04-20, see miller-rabin.appspot.com
+    const BASIS: &'static [u64];
-#[allow(clippy::unreadable_literal)]
+}
-const BASIS: [u64; 7] = [2, 325, 9375, 28178, 450775, 9780504, 1795265022];
+
+impl Basis for Montgomery<u64> {
+    // Small set of bases for the Miller-Rabin prime test, valid for all 64b integers;
+    //  discovered by Jim Sinclair on 2011-04-20, see miller-rabin.appspot.com
+    #[allow(clippy::unreadable_literal)]
+    const BASIS: &'static [u64] = &[2, 325, 9375, 28178, 450775, 9780504, 1795265022];
+}
+
+impl Basis for Montgomery<u32> {
+    // Small set of bases for the Miller-Rabin prime test, valid for all 32b integers;
+    //  discovered by Steve Worley on 2013-05-27, see miller-rabin.appspot.com
+    #[allow(clippy::unreadable_literal)]
+    const BASIS: &'static [u64] = &[
+        4230279247111683200,
+        14694767155120705706,
+        16641139526367750375,
+    ];
+}
 
 #[derive(Eq, PartialEq)]
 pub(crate) enum Result {
@@ -23,16 +40,12 @@ impl Result {
 // Deterministic Miller-Rabin primality-checking algorithm, adapted to extract
 // (some) dividers; it will fail to factor strong pseudoprimes.
 #[allow(clippy::many_single_char_names)]
-pub(crate) fn test<A: Arithmetic>(m: A) -> Result {
+pub(crate) fn test<A: Arithmetic + Basis>(m: A) -> Result {
     use self::Result::*;
 
     let n = m.modulus();
-    if n < 2 {
+    debug_assert!(n > 1);
-        return Pseudoprime;
+    debug_assert!(n % 2 != 0);
-    }
-    if n % 2 == 0 {
-        return if n == 2 { Prime } else { Composite(2) };
-    }
 
     // n-1 = r 2ⁱ
     let i = (n - 1).trailing_zeros();
@@ -41,10 +54,10 @@ pub(crate) fn test<A: Arithmetic>(m: A) -> Result {
     let one = m.one();
     let minus_one = m.minus_one();
 
-    for _a in BASIS.iter() {
+    for _a in A::BASIS.iter() {
         let _a = _a % n;
         if _a == 0 {
-            break;
+            continue;
         }
 
         let a = m.from_u64(_a);
@@ -87,48 +100,113 @@ pub(crate) fn test<A: Arithmetic>(m: A) -> Result {
 // Used by build.rs' tests and debug assertions
 #[allow(dead_code)]
 pub(crate) fn is_prime(n: u64) -> bool {
-    if n % 2 == 0 {
+    if n < 2 {
+        false
+    } else if n % 2 == 0 {
         n == 2
     } else {
-        test::<Montgomery>(Montgomery::new(n)).is_prime()
+        test::<Montgomery<u64>>(Montgomery::new(n)).is_prime()
     }
 }
 
 #[cfg(test)]
 mod tests {
-    use super::is_prime;
+    use super::*;
+    use crate::numeric::{Arithmetic, Montgomery};
+    use quickcheck::quickcheck;
+    use std::iter;
     const LARGEST_U64_PRIME: u64 = 0xFFFFFFFFFFFFFFC5;
 
+    fn primes() -> impl Iterator<Item = u64> {
+        iter::once(2).chain(odd_primes())
+    }
+
+    fn odd_primes() -> impl Iterator<Item = u64> {
+        use crate::table::{NEXT_PRIME, P_INVS_U64};
+        P_INVS_U64
+            .iter()
+            .map(|(p, _, _)| *p)
+            .chain(iter::once(NEXT_PRIME))
+    }
+
     #[test]
     fn largest_prime() {
         assert!(is_prime(LARGEST_U64_PRIME));
     }
 
     #[test]
-    fn first_primes() {
+    fn largest_composites() {
-        use crate::table::{NEXT_PRIME, P_INVS_U64};
+        for i in LARGEST_U64_PRIME + 1..=u64::MAX {
-        for (p, _, _) in P_INVS_U64.iter() {
+            assert!(!is_prime(i), "2⁶⁴ - {} reported prime", u64::MAX - i + 1);
-            assert!(is_prime(*p), "{} reported composite", p);
         }
-        assert!(is_prime(NEXT_PRIME));
     }
 
+    #[test]
+    fn two() {
+        assert!(is_prime(2));
+    }
+
+    // TODO: Deduplicate with macro in numeric.rs
+    macro_rules! parametrized_check {
+        ( $f:ident ) => {
+            paste::item! {
+                #[test]
+                fn [< $f _ u32 >]() {
+                    $f::<Montgomery<u32>>()
+                }
+                #[test]
+                fn [< $f _ u64 >]() {
+                    $f::<Montgomery<u64>>()
+                }
+            }
+        };
+    }
+
+    fn first_primes<A: Arithmetic + Basis>() {
+        for p in odd_primes() {
+            assert!(test(A::new(p)).is_prime(), "{} reported composite", p);
+        }
+    }
+    parametrized_check!(first_primes);
+
+    #[test]
+    fn one() {
+        assert!(!is_prime(1));
+    }
+    #[test]
+    fn zero() {
+        assert!(!is_prime(0));
+    }
+
+    fn first_composites<A: Arithmetic + Basis>() {
+        for (p, q) in primes().zip(odd_primes()) {
+            for i in p + 1..q {
+                assert!(!is_prime(i), "{} reported prime", i);
+            }
+        }
+    }
+    parametrized_check!(first_composites);
+
     #[test]
     fn issue_1556() {
         // 10 425 511 = 2441 × 4271
         assert!(!is_prime(10_425_511));
     }
 
-    #[test]
+    fn small_semiprimes<A: Arithmetic + Basis>() {
-    fn small_composites() {
+        for p in odd_primes() {
-        use crate::table::P_INVS_U64;
+            for q in odd_primes().take_while(|q| *q <= p) {
-
-        for i in 0..P_INVS_U64.len() {
-            let (p, _, _) = P_INVS_U64[i];
-            for (q, _, _) in &P_INVS_U64[0..i] {
                 let n = p * q;
-                assert!(!is_prime(n), "{} = {} × {} reported prime", n, p, q);
+                let m = A::new(n);
+                assert!(!test(m).is_prime(), "{} = {} × {} reported prime", n, p, q);
             }
         }
     }
+    parametrized_check!(small_semiprimes);
+
+    quickcheck! {
+        fn composites(i: u64, j: u64) -> bool {
+            i < 2 || j < 2 || !is_prime(i*j)
+        }
+    }
 }

src/uu/factor/src/numeric.rs

@@ -6,6 +6,12 @@
 // * For the full copyright and license information, please view the LICENSE file
 // * that was distributed with this source code.
 
+use num_traits::{
+    identities::{One, Zero},
+    int::PrimInt,
+    ops::wrapping::{WrappingMul, WrappingNeg, WrappingSub},
+};
+use std::fmt::{Debug, Display};
 use std::mem::swap;
 
 // This is incorrectly reported as dead code,
@@ -20,16 +26,17 @@ pub(crate) fn gcd(mut a: u64, mut b: u64) -> u64 {
 }
 
 pub(crate) trait Arithmetic: Copy + Sized {
-    type I: Copy + Sized + Eq;
+    // The type of integers mod m, in some opaque representation
+    type ModInt: Copy + Sized + Eq;
 
     fn new(m: u64) -> Self;
     fn modulus(&self) -> u64;
-    fn from_u64(&self, n: u64) -> Self::I;
+    fn from_u64(&self, n: u64) -> Self::ModInt;
-    fn to_u64(&self, n: Self::I) -> u64;
+    fn to_u64(&self, n: Self::ModInt) -> u64;
-    fn add(&self, a: Self::I, b: Self::I) -> Self::I;
+    fn add(&self, a: Self::ModInt, b: Self::ModInt) -> Self::ModInt;
-    fn mul(&self, a: Self::I, b: Self::I) -> Self::I;
+    fn mul(&self, a: Self::ModInt, b: Self::ModInt) -> Self::ModInt;
 
-    fn pow(&self, mut a: Self::I, mut b: u64) -> Self::I {
+    fn pow(&self, mut a: Self::ModInt, mut b: u64) -> Self::ModInt {
         let (_a, _b) = (a, b);
         let mut result = self.one();
         while b > 0 {
@@ -54,75 +61,90 @@ pub(crate) trait Arithmetic: Copy + Sized {
         result
     }
 
-    fn one(&self) -> Self::I {
+    fn one(&self) -> Self::ModInt {
         self.from_u64(1)
     }
-    fn minus_one(&self) -> Self::I {
+    fn minus_one(&self) -> Self::ModInt {
         self.from_u64(self.modulus() - 1)
     }
-    fn zero(&self) -> Self::I {
+    fn zero(&self) -> Self::ModInt {
         self.from_u64(0)
     }
 }
 
 #[derive(Clone, Copy, Debug)]
-pub(crate) struct Montgomery {
+pub(crate) struct Montgomery<T: DoubleInt> {
-    a: u64,
+    a: T,
-    n: u64,
+    n: T,
 }
 
-impl Montgomery {
+impl<T: DoubleInt> Montgomery<T> {
     /// computes x/R mod n efficiently
-    fn reduce(&self, x: u128) -> u64 {
+    fn reduce(&self, x: T::DoubleWidth) -> T {
-        debug_assert!(x < (self.n as u128) << 64);
+        let t_bits = T::zero().count_zeros() as usize;
+
+        debug_assert!(x < (self.n.as_double_width()) << t_bits);
         // TODO: optimiiiiiiise
         let Montgomery { a, n } = self;
-        let m = (x as u64).wrapping_mul(*a);
+        let m = T::from_double_width(x).wrapping_mul(a);
-        let nm = (*n as u128) * (m as u128);
+        let nm = (n.as_double_width()) * (m.as_double_width());
-        let (xnm, overflow) = (x as u128).overflowing_add(nm); // x + n*m
+        let (xnm, overflow) = x.overflowing_add_(nm); // x + n*m
-        debug_assert_eq!(xnm % (1 << 64), 0);
+        debug_assert_eq!(
+            xnm % (T::DoubleWidth::one() << T::zero().count_zeros() as usize),
+            T::DoubleWidth::zero()
+        );
 
         // (x + n*m) / R
         // in case of overflow, this is (2¹²⁸ + xnm)/2⁶⁴ - n = xnm/2⁶⁴ + (2⁶⁴ - n)
-        let y = (xnm >> 64) as u64 + if !overflow { 0 } else { n.wrapping_neg() };
+        let y = T::from_double_width(xnm >> t_bits)
+            + if !overflow {
+                T::zero()
+            } else {
+                n.wrapping_neg()
+            };
 
         if y >= *n {
-            y - n
+            y - *n
         } else {
             y
         }
     }
 }
 
-impl Arithmetic for Montgomery {
+impl<T: DoubleInt> Arithmetic for Montgomery<T> {
     // Montgomery transform, R=2⁶⁴
     // Provides fast arithmetic mod n (n odd, u64)
-    type I = u64;
+    type ModInt = T;
 
     fn new(n: u64) -> Self {
-        let a = inv_mod_u64(n).wrapping_neg();
+        debug_assert!(T::zero().count_zeros() >= 64 || n < (1 << T::zero().count_zeros() as usize));
-        debug_assert_eq!(n.wrapping_mul(a), 1_u64.wrapping_neg());
+        let n = T::from_u64(n);
+        let a = modular_inverse(n).wrapping_neg();
+        debug_assert_eq!(n.wrapping_mul(&a), T::one().wrapping_neg());
         Montgomery { a, n }
     }
 
     fn modulus(&self) -> u64 {
-        self.n
+        self.n.as_u64()
     }
 
-    fn from_u64(&self, x: u64) -> Self::I {
+    fn from_u64(&self, x: u64) -> Self::ModInt {
         // TODO: optimise!
-        assert!(x < self.n);
+        debug_assert!(x < self.n.as_u64());
-        let r = (((x as u128) << 64) % self.n as u128) as u64;
+        let r = T::from_double_width(
+            ((T::DoubleWidth::from_u64(x)) << T::zero().count_zeros() as usize)
+                % self.n.as_double_width(),
+        );
         debug_assert_eq!(x, self.to_u64(r));
         r
     }
 
-    fn to_u64(&self, n: Self::I) -> u64 {
+    fn to_u64(&self, n: Self::ModInt) -> u64 {
-        self.reduce(n as u128)
+        self.reduce(n.as_double_width()).as_u64()
     }
 
-    fn add(&self, a: Self::I, b: Self::I) -> Self::I {
+    fn add(&self, a: Self::ModInt, b: Self::ModInt) -> Self::ModInt {
-        let (r, overflow) = a.overflowing_add(b);
+        let (r, overflow) = a.overflowing_add_(b);
 
         // In case of overflow, a+b = 2⁶⁴ + r = (2⁶⁴ - n) + r (working mod n)
         let r = if !overflow {
@@ -138,10 +160,10 @@ impl Arithmetic for Montgomery {
         // a+b % n
         #[cfg(debug_assertions)]
         {
-            let a_r = self.to_u64(a);
+            let a_r = self.to_u64(a) as u128;
-            let b_r = self.to_u64(b);
+            let b_r = self.to_u64(b) as u128;
             let r_r = self.to_u64(r);
-            let r_2 = (((a_r as u128) + (b_r as u128)) % (self.n as u128)) as u64;
+            let r_2 = ((a_r + b_r) % self.n.as_u128()) as u64;
             debug_assert_eq!(
                 r_r, r_2,
                 "[{}] = {} ≠ {} = {} + {} = [{}] + [{}] mod {}; a = {}",
@@ -151,17 +173,17 @@ impl Arithmetic for Montgomery {
         r
     }
 
-    fn mul(&self, a: Self::I, b: Self::I) -> Self::I {
+    fn mul(&self, a: Self::ModInt, b: Self::ModInt) -> Self::ModInt {
-        let r = self.reduce((a as u128) * (b as u128));
+        let r = self.reduce(a.as_double_width() * b.as_double_width());
 
         // Check that r (reduced back to the usual representation) equals
         // a*b % n
         #[cfg(debug_assertions)]
         {
-            let a_r = self.to_u64(a);
+            let a_r = self.to_u64(a) as u128;
-            let b_r = self.to_u64(b);
+            let b_r = self.to_u64(b) as u128;
             let r_r = self.to_u64(r);
-            let r_2 = (((a_r as u128) * (b_r as u128)) % (self.n as u128)) as u64;
+            let r_2: u64 = ((a_r * b_r) % self.n.as_u128()) as u64;
             debug_assert_eq!(
                 r_r, r_2,
                 "[{}] = {} ≠ {} = {} * {} = [{}] * [{}] mod {}; a = {}",
@@ -172,35 +194,112 @@ impl Arithmetic for Montgomery {
     }
 }
 
+// NOTE: Trait can be removed once num-traits adds a similar one;
+//       see https://github.com/rust-num/num-traits/issues/168
+pub(crate) trait OverflowingAdd: Sized {
+    fn overflowing_add_(self, n: Self) -> (Self, bool);
+}
+macro_rules! overflowing {
+    ($x:ty) => {
+        impl OverflowingAdd for $x {
+            fn overflowing_add_(self, n: Self) -> (Self, bool) {
+                self.overflowing_add(n)
+            }
+        }
+    };
+}
+overflowing!(u32);
+overflowing!(u64);
+overflowing!(u128);
+
+pub(crate) trait Int:
+    Display + Debug + PrimInt + OverflowingAdd + WrappingNeg + WrappingSub + WrappingMul
+{
+    fn as_u64(&self) -> u64;
+    fn from_u64(n: u64) -> Self;
+
+    #[cfg(debug_assertions)]
+    fn as_u128(&self) -> u128;
+}
+
+pub(crate) trait DoubleInt: Int {
+    /// An integer type with twice the width of `Self`.
+    /// In particular, multiplications (of `Int` values) can be performed in
+    ///  `Self::DoubleWidth` without possibility of overflow.
+    type DoubleWidth: Int;
+
+    fn as_double_width(self) -> Self::DoubleWidth;
+    fn from_double_width(n: Self::DoubleWidth) -> Self;
+}
+
+macro_rules! int {
+    ( $x:ty ) => {
+        impl Int for $x {
+            fn as_u64(&self) -> u64 {
+                *self as u64
+            }
+            fn from_u64(n: u64) -> Self {
+                n as _
+            }
+            #[cfg(debug_assertions)]
+            fn as_u128(&self) -> u128 {
+                *self as u128
+            }
+        }
+    };
+}
+macro_rules! double_int {
+    ( $x:ty, $y:ty ) => {
+        int!($x);
+        impl DoubleInt for $x {
+            type DoubleWidth = $y;
+
+            fn as_double_width(self) -> $y {
+                self as _
+            }
+            fn from_double_width(n: $y) -> $x {
+                n as _
+            }
+        }
+    };
+}
+
+double_int!(u32, u64);
+double_int!(u64, u128);
+int!(u128);
+
 // extended Euclid algorithm
 // precondition: a is odd
-pub(crate) fn inv_mod_u64(a: u64) -> u64 {
+pub(crate) fn modular_inverse<T: Int>(a: T) -> T {
-    assert!(a % 2 == 1, "{} is not odd", a);
+    let zero = T::zero();
-    let mut t = 0u64;
+    let one = T::one();
-    let mut newt = 1u64;
+    debug_assert!(a % (one + one) == one, "{:?} is not odd", a);
-    let mut r = 0u64;
+
+    let mut t = zero;
+    let mut newt = one;
+    let mut r = zero;
     let mut newr = a;
 
-    while newr != 0 {
+    while newr != zero {
-        let quot = if r == 0 {
+        let quot = if r == zero {
             // special case when we're just starting out
             // This works because we know that
             // a does not divide 2^64, so floor(2^64 / a) == floor((2^64-1) / a);
-            std::u64::MAX
+            T::max_value()
         } else {
             r
         } / newr;
 
-        let newtp = t.wrapping_sub(quot.wrapping_mul(newt));
+        let newtp = t.wrapping_sub(&quot.wrapping_mul(&newt));
         t = newt;
         newt = newtp;
 
-        let newrp = r.wrapping_sub(quot.wrapping_mul(newr));
+        let newrp = r.wrapping_sub(&quot.wrapping_mul(&newr));
         r = newr;
         newr = newrp;
     }
 
-    assert_eq!(r, 1);
+    debug_assert_eq!(r, one);
     t
 }
 
@@ -208,19 +307,37 @@ pub(crate) fn inv_mod_u64(a: u64) -> u64 {
 mod tests {
     use super::*;
 
-    #[test]
+    macro_rules! parametrized_check {
-    fn test_inverter() {
+        ( $f:ident ) => {
-        // All odd integers from 1 to 20 000
+            paste::item! {
-        let mut test_values = (0..10_000u64).map(|i| 2 * i + 1);
+                #[test]
-
+                fn [< $f _ u32 >]() {
-        assert!(test_values.all(|x| x.wrapping_mul(inv_mod_u64(x)) == 1));
+                    $f::<u32>()
+                }
+                #[test]
+                fn [< $f _ u64 >]() {
+                    $f::<u64>()
+                }
+            }
+        };
     }
 
-    #[test]
+    fn test_inverter<T: Int>() {
-    fn test_montgomery_add() {
+        // All odd integers from 1 to 20 000
+        let one = T::from(1).unwrap();
+        let two = T::from(2).unwrap();
+        let mut test_values = (0..10_000)
+            .map(|i| T::from(i).unwrap())
+            .map(|i| two * i + one);
+
+        assert!(test_values.all(|x| x.wrapping_mul(&modular_inverse(x)) == one));
+    }
+    parametrized_check!(test_inverter);
+
+    fn test_add<A: DoubleInt>() {
         for n in 0..100 {
             let n = 2 * n + 1;
-            let m = Montgomery::new(n);
+            let m = Montgomery::<A>::new(n);
             for x in 0..n {
                 let m_x = m.from_u64(x);
                 for y in 0..=x {
@@ -231,12 +348,12 @@ mod tests {
             }
         }
     }
+    parametrized_check!(test_add);
 
-    #[test]
+    fn test_mult<A: DoubleInt>() {
-    fn test_montgomery_mult() {
         for n in 0..100 {
             let n = 2 * n + 1;
-            let m = Montgomery::new(n);
+            let m = Montgomery::<A>::new(n);
             for x in 0..n {
                 let m_x = m.from_u64(x);
                 for y in 0..=x {
@@ -246,16 +363,17 @@ mod tests {
             }
         }
     }
+    parametrized_check!(test_mult);
 
-    #[test]
+    fn test_roundtrip<A: DoubleInt>() {
-    fn test_montgomery_roundtrip() {
         for n in 0..100 {
             let n = 2 * n + 1;
-            let m = Montgomery::new(n);
+            let m = Montgomery::<A>::new(n);
             for x in 0..n {
                 let x_ = m.from_u64(x);
                 assert_eq!(x, m.to_u64(x_));
             }
         }
     }
+    parametrized_check!(test_roundtrip);
 }