openethereum/src/uint.rs

683 lines
18 KiB
Rust
Raw Normal View History

2015-11-26 13:42:42 +01:00
// taken from Rust Bitcoin Library (https://github.com/apoelstra/rust-bitcoin)
// original author: Andrew Poelstra <apoelstra@wpsoftware.net>
// Rust Bitcoin Library
// Written in 2014 by
// Andrew Poelstra <apoelstra@wpsoftware.net>
//
// To the extent possible under law, the author(s) have dedicated all
// copyright and related and neighboring rights to this software to
// the public domain worldwide. This software is distributed without
// any warranty.
//
// You should have received a copy of the CC0 Public Domain Dedication
// along with this software.
// If not, see <http://creativecommons.org/publicdomain/zero/1.0/>.
//
//! Big unsigned integer types
///!
///! Implementation of a various large-but-fixed sized unsigned integer types.
///! The functions here are designed to be fast.
///!
use std::fmt;
use std::cmp::*;
2015-11-26 13:42:42 +01:00
use std::ops::*;
use std::str::FromStr;
2015-12-11 14:40:28 +01:00
use std::hash::{Hash, Hasher};
2015-11-26 13:42:42 +01:00
use rustc_serialize::hex::{FromHex, FromHexError};
pub trait FromDecStr: Sized {
type Err;
fn from_dec_str(value: &str) -> Result<Self, Self::Err>;
}
2015-11-26 13:42:42 +01:00
macro_rules! impl_map_from {
($thing:ident, $from:ty, $to:ty) => {
impl From<$from> for $thing {
fn from(value: $from) -> $thing {
From::from(value as $to)
}
}
}
2015-11-26 13:42:42 +01:00
}
macro_rules! construct_uint {
($name:ident, $n_words:expr) => (
/// Little-endian large integer type
#[derive(Copy, Clone, Eq, PartialEq)]
pub struct $name(pub [u64; $n_words]);
impl $name {
/// Conversion to u32
#[inline]
2015-12-26 15:48:41 +01:00
pub fn low_u32(&self) -> u32 {
let &$name(ref arr) = self;
arr[0] as u32
}
2015-12-26 15:48:41 +01:00
/// Conversion to u64
#[inline]
pub fn low_u64(&self) -> u64 {
let &$name(ref arr) = self;
arr[0]
}
2016-01-08 15:52:43 +01:00
/// Conversion to u32 with overflow checking
#[inline]
pub fn as_u32(&self) -> u32 {
let &$name(ref arr) = self;
if (arr[0] & (0xffffffffu64 << 32)) != 0 {
panic!("Intger overflow when casting U256")
}
self.as_u64() as u32
}
/// Conversion to u64 with overflow checking
#[inline]
pub fn as_u64(&self) -> u64 {
let &$name(ref arr) = self;
for i in 1..$n_words {
if arr[i] != 0 {
panic!("Intger overflow when casting U256")
}
}
arr[0]
}
/// Return the least number of bits needed to represent the number
#[inline]
pub fn bits(&self) -> usize {
let &$name(ref arr) = self;
for i in 1..$n_words {
if arr[$n_words - i] > 0 { return (0x40 * ($n_words - i + 1)) - arr[$n_words - i].leading_zeros() as usize; }
}
0x40 - arr[0].leading_zeros() as usize
}
#[inline]
pub fn bit(&self, index: usize) -> bool {
let &$name(ref arr) = self;
arr[index / 64] & (1 << (index % 64)) != 0
}
#[inline]
pub fn byte(&self, index: usize) -> u8 {
let &$name(ref arr) = self;
(arr[index / 8] >> ((index % 8)) * 8) as u8
}
pub fn to_bytes(&self, bytes: &mut[u8]) {
assert!($n_words * 8 == bytes.len());
let &$name(ref arr) = self;
for i in 0..bytes.len() {
let rev = bytes.len() - 1 - i;
let pos = rev / 8;
bytes[i] = (arr[pos] >> ((rev % 8) * 8)) as u8;
}
}
2015-12-09 14:50:01 +01:00
#[inline]
pub fn exp10(n: usize) -> $name {
match n {
0 => $name::from(1u64),
_ => $name::exp10(n - 1) * $name::from(10u64)
}
}
2015-12-18 10:29:06 +01:00
#[inline]
pub fn zero() -> $name {
From::from(0u64)
}
2015-12-26 15:48:41 +01:00
2015-12-18 10:29:06 +01:00
#[inline]
pub fn one() -> $name {
From::from(1u64)
}
/// Multiplication by u32
fn mul_u32(self, other: u32) -> $name {
let $name(ref arr) = self;
let mut carry = [0u64; $n_words];
let mut ret = [0u64; $n_words];
for i in 0..$n_words {
let upper = other as u64 * (arr[i] >> 32);
let lower = other as u64 * (arr[i] & 0xFFFFFFFF);
if i < 3 {
carry[i + 1] += upper >> 32;
}
ret[i] = lower + (upper << 32);
}
$name(ret) + $name(carry)
}
}
impl From<u64> for $name {
fn from(value: u64) -> $name {
let mut ret = [0; $n_words];
ret[0] = value;
$name(ret)
}
}
impl_map_from!($name, u8, u64);
impl_map_from!($name, u16, u64);
impl_map_from!($name, u32, u64);
2015-12-18 10:29:06 +01:00
impl_map_from!($name, usize, u64);
impl From<i64> for $name {
fn from(value: i64) -> $name {
match value >= 0 {
true => From::from(value as u64),
false => { panic!("Unsigned integer can't be created from negative value"); }
}
}
}
impl_map_from!($name, i8, i64);
impl_map_from!($name, i16, i64);
impl_map_from!($name, i32, i64);
impl_map_from!($name, isize, i64);
impl<'a> From<&'a [u8]> for $name {
fn from(bytes: &[u8]) -> $name {
assert!($n_words * 8 >= bytes.len());
let mut ret = [0; $n_words];
for i in 0..bytes.len() {
let rev = bytes.len() - 1 - i;
let pos = rev / 8;
ret[pos] += (bytes[i] as u64) << (rev % 8) * 8;
}
$name(ret)
}
}
impl FromStr for $name {
type Err = FromHexError;
fn from_str(value: &str) -> Result<$name, Self::Err> {
let bytes: Vec<u8> = match value.len() % 2 == 0 {
true => try!(value.from_hex()),
false => try!(("0".to_string() + value).from_hex())
};
let bytes_ref: &[u8] = &bytes;
Ok(From::from(bytes_ref))
}
}
impl Add<$name> for $name {
type Output = $name;
fn add(self, other: $name) -> $name {
let $name(ref me) = self;
let $name(ref you) = other;
let mut ret = [0u64; $n_words];
let mut carry = [0u64; $n_words];
let mut b_carry = false;
for i in 0..$n_words {
ret[i] = me[i].wrapping_add(you[i]);
if i < $n_words - 1 && ret[i] < me[i] {
carry[i + 1] = 1;
b_carry = true;
}
}
if b_carry { $name(ret) + $name(carry) } else { $name(ret) }
}
}
impl Sub<$name> for $name {
type Output = $name;
#[inline]
fn sub(self, other: $name) -> $name {
self + !other + From::from(1u64)
}
}
impl Mul<$name> for $name {
type Output = $name;
fn mul(self, other: $name) -> $name {
2015-12-09 14:50:01 +01:00
let mut res = $name::from(0u64);
// TODO: be more efficient about this
for i in 0..(2 * $n_words) {
2015-12-09 14:50:01 +01:00
res = res + (self.mul_u32((other >> (32 * i)).low_u32()) << (32 * i));
}
2015-12-09 14:50:01 +01:00
res
}
}
impl Div<$name> for $name {
type Output = $name;
fn div(self, other: $name) -> $name {
let mut sub_copy = self;
let mut shift_copy = other;
let mut ret = [0u64; $n_words];
let my_bits = self.bits();
let your_bits = other.bits();
// Check for division by 0
assert!(your_bits != 0);
// Early return in case we are dividing by a larger number than us
if my_bits < your_bits {
return $name(ret);
}
// Bitwise long division
let mut shift = my_bits - your_bits;
shift_copy = shift_copy << shift;
loop {
if sub_copy >= shift_copy {
ret[shift / 64] |= 1 << (shift % 64);
sub_copy = sub_copy - shift_copy;
}
shift_copy = shift_copy >> 1;
if shift == 0 { break; }
shift -= 1;
}
$name(ret)
}
}
impl Rem<$name> for $name {
type Output = $name;
fn rem(self, other: $name) -> $name {
let times = self / other;
self - (times * other)
}
}
impl BitAnd<$name> for $name {
type Output = $name;
#[inline]
fn bitand(self, other: $name) -> $name {
let $name(ref arr1) = self;
let $name(ref arr2) = other;
let mut ret = [0u64; $n_words];
for i in 0..$n_words {
ret[i] = arr1[i] & arr2[i];
}
$name(ret)
}
}
impl BitXor<$name> for $name {
type Output = $name;
#[inline]
fn bitxor(self, other: $name) -> $name {
let $name(ref arr1) = self;
let $name(ref arr2) = other;
let mut ret = [0u64; $n_words];
for i in 0..$n_words {
ret[i] = arr1[i] ^ arr2[i];
}
$name(ret)
}
}
impl BitOr<$name> for $name {
type Output = $name;
#[inline]
fn bitor(self, other: $name) -> $name {
let $name(ref arr1) = self;
let $name(ref arr2) = other;
let mut ret = [0u64; $n_words];
for i in 0..$n_words {
ret[i] = arr1[i] | arr2[i];
}
$name(ret)
}
}
impl Not for $name {
type Output = $name;
#[inline]
fn not(self) -> $name {
let $name(ref arr) = self;
let mut ret = [0u64; $n_words];
for i in 0..$n_words {
ret[i] = !arr[i];
}
$name(ret)
}
}
impl Shl<usize> for $name {
type Output = $name;
fn shl(self, shift: usize) -> $name {
let $name(ref original) = self;
let mut ret = [0u64; $n_words];
let word_shift = shift / 64;
let bit_shift = shift % 64;
for i in 0..$n_words {
// Shift
if bit_shift < 64 && i + word_shift < $n_words {
ret[i + word_shift] += original[i] << bit_shift;
}
// Carry
if bit_shift > 0 && i + word_shift + 1 < $n_words {
ret[i + word_shift + 1] += original[i] >> (64 - bit_shift);
}
}
$name(ret)
}
}
impl Shr<usize> for $name {
type Output = $name;
fn shr(self, shift: usize) -> $name {
let $name(ref original) = self;
let mut ret = [0u64; $n_words];
let word_shift = shift / 64;
let bit_shift = shift % 64;
for i in word_shift..$n_words {
// Shift
ret[i - word_shift] += original[i] >> bit_shift;
// Carry
if bit_shift > 0 && i < $n_words - 1 {
ret[i - word_shift] += original[i + 1] << (64 - bit_shift);
}
}
$name(ret)
}
}
impl Ord for $name {
fn cmp(&self, other: &$name) -> Ordering {
let &$name(ref me) = self;
let &$name(ref you) = other;
for i in 0..$n_words {
if me[$n_words - 1 - i] < you[$n_words - 1 - i] { return Ordering::Less; }
if me[$n_words - 1 - i] > you[$n_words - 1 - i] { return Ordering::Greater; }
}
Ordering::Equal
}
}
impl PartialOrd for $name {
fn partial_cmp(&self, other: &$name) -> Option<Ordering> {
Some(self.cmp(other))
}
}
impl fmt::Debug for $name {
fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
let &$name(ref data) = self;
try!(write!(f, "0x"));
for ch in data.iter().rev() {
try!(write!(f, "{:02x}", ch));
}
Ok(())
}
}
2015-12-11 14:40:28 +01:00
impl Hash for $name {
fn hash<H>(&self, state: &mut H) where H: Hasher {
unsafe { state.write(::std::slice::from_raw_parts(self.0.as_ptr() as *mut u8, self.0.len() * 8)); }
state.finish();
}
}
impl FromDecStr for $name {
type Err = FromHexError;
/// TODO: optimize, throw appropriate err
fn from_dec_str(value: &str) -> Result<Self, Self::Err> {
2015-12-12 13:32:42 +01:00
Ok(value.bytes()
.map(|b| b - 48)
2015-12-26 15:48:41 +01:00
.fold($name::from(0u64), | acc, c |
2015-12-12 13:32:42 +01:00
// fast multiplication by 10
// (acc << 3) + (acc << 1) => acc * 10
(acc << 3) + (acc << 1) + $name::from(c)
))
}
}
);
2015-11-26 13:42:42 +01:00
}
construct_uint!(U256, 4);
construct_uint!(U128, 2);
impl From<U128> for U256 {
fn from(value: U128) -> U256 {
let U128(ref arr) = value;
let mut ret = [0; 4];
ret[0] = arr[0];
ret[1] = arr[1];
U256(ret)
}
2015-11-26 13:42:42 +01:00
}
2015-12-26 15:48:41 +01:00
impl From<U256> for u64 {
fn from(value: U256) -> u64 {
2016-01-08 15:52:43 +01:00
value.as_u64()
2015-12-26 15:48:41 +01:00
}
}
impl From<U256> for u32 {
fn from(value: U256) -> u32 {
2016-01-08 15:52:43 +01:00
value.as_u32()
2015-12-26 15:48:41 +01:00
}
}
2015-11-27 13:17:28 +01:00
pub const ZERO_U256: U256 = U256([0x00u64; 4]);
pub const ONE_U256: U256 = U256([0x01u64, 0x00u64, 0x00u64, 0x00u64]);
pub const BAD_U256: U256 = U256([0xffffffffffffffffu64; 4]);
2015-11-26 13:42:42 +01:00
#[cfg(test)]
mod tests {
use uint::U256;
use uint::FromDecStr;
use std::str::FromStr;
#[test]
pub fn uint256_from() {
let e = U256([10, 0, 0, 0]);
// test unsigned initialization
let ua = U256::from(10u8);
let ub = U256::from(10u16);
let uc = U256::from(10u32);
let ud = U256::from(10u64);
assert_eq!(e, ua);
assert_eq!(e, ub);
assert_eq!(e, uc);
assert_eq!(e, ud);
// test initialization from bytes
let va = U256::from(&[10u8][..]);
assert_eq!(e, va);
// more tests for initialization from bytes
assert_eq!(U256([0x1010, 0, 0, 0]), U256::from(&[0x10u8, 0x10][..]));
assert_eq!(U256([0x12f0, 0, 0, 0]), U256::from(&[0x12u8, 0xf0][..]));
assert_eq!(U256([0x12f0, 0, 0, 0]), U256::from(&[0, 0x12u8, 0xf0][..]));
assert_eq!(U256([0x12f0, 0 , 0, 0]), U256::from(&[0, 0, 0, 0, 0, 0, 0, 0x12u8, 0xf0][..]));
assert_eq!(U256([0x12f0, 1 , 0, 0]), U256::from(&[1, 0, 0, 0, 0, 0, 0, 0x12u8, 0xf0][..]));
assert_eq!(U256([0x12f0, 1 , 0x0910203040506077, 0x8090a0b0c0d0e0f0]), U256::from(&[
0x80, 0x90, 0xa0, 0xb0, 0xc0, 0xd0, 0xe0, 0xf0,
0x09, 0x10, 0x20, 0x30, 0x40, 0x50, 0x60, 0x77,
0, 0, 0, 0, 0, 0, 0, 1,
0, 0, 0, 0, 0, 0, 0x12u8, 0xf0][..]));
assert_eq!(U256([0x00192437100019fa, 0x243710, 0, 0]), U256::from(&[
0x24u8, 0x37, 0x10,
0, 0x19, 0x24, 0x37, 0x10, 0, 0x19, 0xfa][..]));
// test initializtion from string
let sa = U256::from_str("0a").unwrap();
assert_eq!(e, sa);
assert_eq!(U256([0x1010, 0, 0, 0]), U256::from_str("1010").unwrap());
assert_eq!(U256([0x12f0, 0, 0, 0]), U256::from_str("12f0").unwrap());
assert_eq!(U256([0x12f0, 0, 0, 0]), U256::from_str("12f0").unwrap());
assert_eq!(U256([0x12f0, 0 , 0, 0]), U256::from_str("0000000012f0").unwrap());
assert_eq!(U256([0x12f0, 1 , 0, 0]), U256::from_str("0100000000000012f0").unwrap());
assert_eq!(U256([0x12f0, 1 , 0x0910203040506077, 0x8090a0b0c0d0e0f0]), U256::from_str("8090a0b0c0d0e0f00910203040506077000000000000000100000000000012f0").unwrap());
}
#[test]
pub fn uint256_to() {
2015-12-03 19:07:08 +01:00
let hex = "8090a0b0c0d0e0f00910203040506077583a2cf8264910e1436bda32571012f0";
let uint = U256::from_str(hex).unwrap();
let mut bytes = [0u8; 32];
uint.to_bytes(&mut bytes);
let uint2 = U256::from(&bytes[..]);
assert_eq!(uint, uint2);
}
#[test]
pub fn uint256_bits_test() {
assert_eq!(U256::from(0u64).bits(), 0);
assert_eq!(U256::from(255u64).bits(), 8);
assert_eq!(U256::from(256u64).bits(), 9);
assert_eq!(U256::from(300u64).bits(), 9);
assert_eq!(U256::from(60000u64).bits(), 16);
assert_eq!(U256::from(70000u64).bits(), 17);
//// Try to read the following lines out loud quickly
let mut shl = U256::from(70000u64);
shl = shl << 100;
assert_eq!(shl.bits(), 117);
shl = shl << 100;
assert_eq!(shl.bits(), 217);
shl = shl << 100;
assert_eq!(shl.bits(), 0);
//// Bit set check
//// 01010
assert!(!U256::from(10u8).bit(0));
assert!(U256::from(10u8).bit(1));
assert!(!U256::from(10u8).bit(2));
assert!(U256::from(10u8).bit(3));
assert!(!U256::from(10u8).bit(4));
//// byte check
assert_eq!(U256::from(10u8).byte(0), 10);
assert_eq!(U256::from(0xffu64).byte(0), 0xff);
assert_eq!(U256::from(0xffu64).byte(1), 0);
assert_eq!(U256::from(0x01ffu64).byte(0), 0xff);
assert_eq!(U256::from(0x01ffu64).byte(1), 0x1);
assert_eq!(U256([0u64, 0xfc, 0, 0]).byte(8), 0xfc);
assert_eq!(U256([0u64, 0, 0, u64::max_value()]).byte(31), 0xff);
assert_eq!(U256([0u64, 0, 0, (u64::max_value() >> 8) + 1]).byte(31), 0x01);
}
#[test]
pub fn uint256_comp_test() {
let small = U256([10u64, 0, 0, 0]);
let big = U256([0x8C8C3EE70C644118u64, 0x0209E7378231E632, 0, 0]);
let bigger = U256([0x9C8C3EE70C644118u64, 0x0209E7378231E632, 0, 0]);
let biggest = U256([0x5C8C3EE70C644118u64, 0x0209E7378231E632, 0, 1]);
assert!(small < big);
assert!(big < bigger);
assert!(bigger < biggest);
assert!(bigger <= biggest);
assert!(biggest <= biggest);
assert!(bigger >= big);
assert!(bigger >= small);
assert!(small <= small);
}
#[test]
pub fn uint256_arithmetic_test() {
let init = U256::from(0xDEADBEEFDEADBEEFu64);
let copy = init;
let add = init + copy;
assert_eq!(add, U256([0xBD5B7DDFBD5B7DDEu64, 1, 0, 0]));
// Bitshifts
let shl = add << 88;
assert_eq!(shl, U256([0u64, 0xDFBD5B7DDE000000, 0x1BD5B7D, 0]));
let shr = shl >> 40;
assert_eq!(shr, U256([0x7DDE000000000000u64, 0x0001BD5B7DDFBD5B, 0, 0]));
// Increment
let incr = shr + U256::from(1u64);
assert_eq!(incr, U256([0x7DDE000000000001u64, 0x0001BD5B7DDFBD5B, 0, 0]));
// Subtraction
let sub = incr - init;
assert_eq!(sub, U256([0x9F30411021524112u64, 0x0001BD5B7DDFBD5A, 0, 0]));
// Multiplication
let mult = sub.mul_u32(300);
assert_eq!(mult, U256([0x8C8C3EE70C644118u64, 0x0209E7378231E632, 0, 0]));
// Division
assert_eq!(U256::from(105u8) / U256::from(5u8), U256::from(21u8));
let div = mult / U256::from(300u16);
assert_eq!(div, U256([0x9F30411021524112u64, 0x0001BD5B7DDFBD5A, 0, 0]));
//// TODO: bit inversion
}
#[test]
pub fn uint256_extreme_bitshift_test() {
//// Shifting a u64 by 64 bits gives an undefined value, so make sure that
//// we're doing the Right Thing here
let init = U256::from(0xDEADBEEFDEADBEEFu64);
assert_eq!(init << 64, U256([0, 0xDEADBEEFDEADBEEF, 0, 0]));
let add = (init << 64) + init;
assert_eq!(add, U256([0xDEADBEEFDEADBEEF, 0xDEADBEEFDEADBEEF, 0, 0]));
assert_eq!(add >> 0, U256([0xDEADBEEFDEADBEEF, 0xDEADBEEFDEADBEEF, 0, 0]));
assert_eq!(add << 0, U256([0xDEADBEEFDEADBEEF, 0xDEADBEEFDEADBEEF, 0, 0]));
assert_eq!(add >> 64, U256([0xDEADBEEFDEADBEEF, 0, 0, 0]));
assert_eq!(add << 64, U256([0, 0xDEADBEEFDEADBEEF, 0xDEADBEEFDEADBEEF, 0]));
}
2015-12-09 14:50:01 +01:00
#[test]
pub fn uint256_exp10() {
assert_eq!(U256::exp10(0), U256::from(1u64));
println!("\none: {:?}", U256::from(1u64));
println!("ten: {:?}", U256::from(10u64));
assert_eq!(U256::from(2u64) * U256::from(10u64), U256::from(20u64));
assert_eq!(U256::exp10(1), U256::from(10u64));
assert_eq!(U256::exp10(2), U256::from(100u64));
assert_eq!(U256::exp10(5), U256::from(100000u64));
}
#[test]
pub fn uint256_mul32() {
assert_eq!(U256::from(0u64).mul_u32(2), U256::from(0u64));
assert_eq!(U256::from(1u64).mul_u32(2), U256::from(2u64));
assert_eq!(U256::from(10u64).mul_u32(2), U256::from(20u64));
assert_eq!(U256::from(10u64).mul_u32(5), U256::from(50u64));
assert_eq!(U256::from(1000u64).mul_u32(50), U256::from(50000u64));
}
#[test]
pub fn uint256_mul() {
assert_eq!(U256::from(1u64) * U256::from(10u64), U256::from(10u64));
}
#[test]
fn uint256_div() {
assert_eq!(U256::from(10u64) / U256::from(1u64), U256::from(10u64));
assert_eq!(U256::from(10u64) / U256::from(2u64), U256::from(5u64));
assert_eq!(U256::from(10u64) / U256::from(3u64), U256::from(3u64));
}
#[test]
fn uint256_rem() {
assert_eq!(U256::from(10u64) % U256::from(1u64), U256::from(0u64));
assert_eq!(U256::from(10u64) % U256::from(3u64), U256::from(1u64));
}
#[test]
fn uint256_from_dec_str() {
assert_eq!(U256::from_dec_str("10").unwrap(), U256::from(10u64));
assert_eq!(U256::from_dec_str("1024").unwrap(), U256::from(1024u64));
}
2015-11-26 13:42:42 +01:00
}