grassrootseconomics/openethereum
5
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

Merge pull request #629 from ethcore/uint_opt

Browse Source 
Optimizing uint operations (architecture independent)
This commit is contained in:
Gav Wood 2016-03-15 11:24:00 +01:00
commit e98cfd84a9
4 changed files with 255 additions and 162 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

10
test.sh
View File

@ -1,4 +1,12 @@
#!/bin/sh
# Running Parity Full Test Sute
cargo test --features ethcore/json-tests $1 -p ethash -p ethcore-util -p ethcore -p ethsync -p ethcore-rpc -p parity -p ethminer
cargo test --features ethcore/json-tests $1 \
-p ethash \
-p ethcore-util \
-p ethcore \
-p ethsync \
-p ethcore-rpc \
-p parity \
-p ethminer \
-p bigint

15
util/benches/bigint.rs
View File

@ -50,9 +50,16 @@ fn u256_sub(b: &mut Bencher) {
fn u512_sub(b: &mut Bencher) {
b.iter(|| {
let n = black_box(10000);
(0..n).fold(U512([rand::random::<u64>(), rand::random::<u64>(), rand::random::<u64>(), rand::random::<u64>(),
rand::random::<u64>(), rand::random::<u64>(), rand::random::<u64>(), rand::random::<u64>()]),
|old, new| { old.overflowing_sub(U512([0, 0, 0, 0, 0, 0, 0, new])).0 })
(0..n).fold(
U512([
rand::random::<u64>(), rand::random::<u64>(), rand::random::<u64>(), rand::random::<u64>(),
rand::random::<u64>(), rand::random::<u64>(), rand::random::<u64>(), rand::random::<u64>()
]),
|old, new| {
let p = new % 2;
old.overflowing_sub(U512([p, p, p, p, p, p, p, new])).0
}
)
});
}
@ -79,7 +86,7 @@ fn u256_full_mul(b: &mut Bencher) {
b.iter(|| {
let n = black_box(10000);
(0..n).fold(U256([rand::random::<u64>(), rand::random::<u64>(), rand::random::<u64>(), rand::random::<u64>()]),
|old, new| {
|old, _new| {
let U512(ref u512words) = old.full_mul(U256([rand::random::<u64>(), rand::random::<u64>(), rand::random::<u64>(), rand::random::<u64>()]));
U256([u512words[0], u512words[2], u512words[2], u512words[3]])
})

334
util/bigint/src/uint.rs
View File

@ -72,29 +72,19 @@ macro_rules! uint_overflowing_add_reg {
($name:ident, $n_words:expr, $self_expr: expr, $other: expr) => ({
let $name(ref me) = $self_expr;
let $name(ref you) = $other;
let mut ret = [0u64; $n_words];
let mut carry = [0u64; $n_words];
let mut b_carry = false;
let mut overflow = false;
let mut carry = 0u64;
for i in 0..$n_words {
ret[i] = me[i].wrapping_add(you[i]);
let (res1, overflow1) = me[i].overflowing_add(you[i]);
let (res2, overflow2) = res1.overflowing_add(carry);
if ret[i] < me[i] {
if i < $n_words - 1 {
carry[i + 1] = 1;
b_carry = true;
} else {
overflow = true;
}
}
}
if b_carry {
let ret = overflowing!($name(ret).overflowing_add($name(carry)), overflow);
(ret, overflow)
} else {
($name(ret), overflow)
ret[i] = res2;
carry = overflow1 as u64 + overflow2 as u64;
}
($name(ret), carry > 0)
})
}
@ -177,9 +167,28 @@ macro_rules! uint_overflowing_add {
#[cfg(not(all(asm_available, target_arch="x86_64")))]
macro_rules! uint_overflowing_sub {
($name:ident, $n_words: expr, $self_expr: expr, $other: expr) => ({
let res = overflowing!((!$other).overflowing_add(From::from(1u64)));
let res = overflowing!($self_expr.overflowing_add(res));
(res, $self_expr < $other)
uint_overflowing_sub_reg!($name, $n_words, $self_expr, $other)
})
}
macro_rules! uint_overflowing_sub_reg {
($name:ident, $n_words: expr, $self_expr: expr, $other: expr) => ({
let $name(ref me) = $self_expr;
let $name(ref you) = $other;
let mut ret = [0u64; $n_words];
let mut carry = 0u64;
for i in 0..$n_words {
let (res1, overflow1) = me[i].overflowing_sub(you[i]);
let (res2, overflow2) = res1.overflowing_sub(carry);
ret[i] = res2;
carry = overflow1 as u64 + overflow2 as u64;
}
($name(ret), carry > 0)
})
}
@ -253,9 +262,7 @@ macro_rules! uint_overflowing_sub {
(U512(result), overflow != 0)
});
($name:ident, $n_words: expr, $self_expr: expr, $other: expr) => ({
let res = overflowing!((!$other).overflowing_add(From::from(1u64)));
let res = overflowing!($self_expr.overflowing_add(res));
(res, $self_expr < $other)
uint_overflowing_sub_reg!($name, $n_words, $self_expr, $other)
})
}
@ -388,14 +395,50 @@ macro_rules! uint_overflowing_mul {
macro_rules! uint_overflowing_mul_reg {
($name:ident, $n_words: expr, $self_expr: expr, $other: expr) => ({
let mut res = $name::from(0u64);
let mut overflow = false;
for i in 0..(2 * $n_words) {
let v = overflowing!($self_expr.overflowing_mul_u32(($other >> (32 * i)).low_u32()), overflow);
let res2 = overflowing!(v.overflowing_shl(32 * i as u32), overflow);
res = overflowing!(res.overflowing_add(res2), overflow);
let $name(ref me) = $self_expr;
let $name(ref you) = $other;
let mut ret = [0u64; 2*$n_words];
for i in 0..$n_words {
if you[i] == 0 {
continue;
}
(res, overflow)
let mut carry2 = 0u64;
let (b_u, b_l) = split(you[i]);
for j in 0..$n_words {
if me[j] == 0 && carry2 == 0 {
continue;
}
let a = split(me[j]);
// multiply parts
let (c_l, overflow_l) = mul_u32(a, b_l, ret[i + j]);
let (c_u, overflow_u) = mul_u32(a, b_u, c_l >> 32);
ret[i + j] = (c_l & 0xFFFFFFFF) + (c_u << 32);
// Only single overflow possible here
let carry = (c_u >> 32) + (overflow_u << 32) + overflow_l + carry2;
let (carry, o) = carry.overflowing_add(ret[i + j + 1]);
ret[i + j + 1] = carry;
carry2 = o as u64;
}
}
let mut res = [0u64; $n_words];
let mut overflow = false;
for i in 0..$n_words {
res[i] = ret[i];
}
for i in $n_words..2*$n_words {
overflow |= ret[i] != 0;
}
($name(res), overflow)
})
}
@ -423,6 +466,23 @@ macro_rules! panic_on_overflow {
}
}
#[inline(always)]
fn mul_u32(a: (u64, u64), b: u64, carry: u64) -> (u64, u64) {
let upper = b * a.0;
let lower = b * a.1;
let (res1, overflow1) = lower.overflowing_add(upper << 32);
let (res2, overflow2) = res1.overflowing_add(carry);
let carry = (upper >> 32) + overflow1 as u64 + overflow2 as u64;
(res2, carry)
}
#[inline(always)]
fn split(a: u64) -> (u64, u64) {
(a >> 32, a & 0xFFFFFFFF)
}
/// Large, fixed-length unsigned integer type.
pub trait Uint: Sized + Default + FromStr + From<u64> + fmt::Debug + fmt::Display + PartialOrd + Ord + PartialEq + Eq + Hash {
@ -481,9 +541,6 @@ pub trait Uint: Sized + Default + FromStr + From<u64> + fmt::Debug + fmt::Displa
/// Returns negation of this `Uint` and overflow (always true)
fn overflowing_neg(self) -> (Self, bool);
/// Shifts this `Uint` and returns overflow
fn overflowing_shl(self, shift: u32) -> (Self, bool);
}
macro_rules! construct_uint {
@ -672,92 +729,32 @@ macro_rules! construct_uint {
fn overflowing_neg(self) -> ($name, bool) {
(!self, true)
}
fn overflowing_shl(self, shift32: u32) -> ($name, bool) {
let $name(ref original) = self;
let mut ret = [0u64; $n_words];
let shift = shift32 as usize;
let word_shift = shift / 64;
let bit_shift = shift % 64;
for i in 0..$n_words {
// Shift
if 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);
}
}
// Detecting overflow
let last = $n_words - word_shift - if bit_shift > 0 { 1 } else { 0 };
let overflow = if bit_shift > 0 {
(original[last] >> (64 - bit_shift)) > 0
} else if word_shift > 0 {
original[last] > 0
} else {
false
};
for i in last+1..$n_words-1 {
if original[i] > 0 {
return ($name(ret), true);
}
}
($name(ret), overflow)
}
}
impl $name {
#[allow(dead_code)] // not used when multiplied with inline assembly
/// Multiplication by u32
fn mul_u32(self, other: u32) -> Self {
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);
ret[i] = lower.wrapping_add(upper << 32);
if i < $n_words - 1 {
carry[i + 1] = upper >> 32;
if ret[i] < lower {
carry[i + 1] += 1;
}
}
}
$name(ret) + $name(carry)
let (ret, overflow) = self.overflowing_mul_u32(other);
panic_on_overflow!(overflow);
ret
}
#[allow(dead_code)] // not used when multiplied with inline assembly
/// Overflowing multiplication by u32
fn overflowing_mul_u32(self, other: u32) -> (Self, bool) {
let $name(ref arr) = self;
let mut carry = [0u64; $n_words];
let mut ret = [0u64; $n_words];
let mut overflow = false;
let mut carry = 0;
let o = other as u64;
for i in 0..$n_words {
let upper = other as u64 * (arr[i] >> 32);
let lower = other as u64 * (arr[i] & 0xFFFFFFFF);
let (res, carry2) = mul_u32(split(arr[i]), o, carry);
ret[i] = res;
carry = carry2;
}
ret[i] = lower.wrapping_add(upper << 32);
if i < $n_words - 1 {
carry[i + 1] = upper >> 32;
if ret[i] < lower {
carry[i + 1] += 1;
}
} else if (upper >> 32) > 0 || ret[i] < lower {
overflow = true
}
}
let result = overflowing!(
$name(ret).overflowing_add($name(carry)),
overflow
);
(result, overflow)
($name(ret), carry > 0)
}
}
@ -1006,14 +1003,15 @@ macro_rules! construct_uint {
let mut ret = [0u64; $n_words];
let word_shift = shift / 64;
let bit_shift = shift % 64;
for i in 0..$n_words {
// Shift
if i + word_shift < $n_words {
ret[i + word_shift] += original[i] << bit_shift;
// shift
for i in word_shift..$n_words {
ret[i] += original[i - word_shift] << bit_shift;
}
// Carry
if bit_shift > 0 && i + word_shift + 1 < $n_words {
ret[i + word_shift + 1] += original[i] >> (64 - bit_shift);
// carry
if bit_shift > 0 {
for i in word_shift+1..$n_words {
ret[i] += original[i - 1 - word_shift] >> (64 - bit_shift);
}
}
$name(ret)
@ -1028,6 +1026,7 @@ macro_rules! construct_uint {
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;
@ -1044,9 +1043,11 @@ macro_rules! construct_uint {
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; }
let mut i = $n_words;
while i > 0 {
i -= 1;
if me[i] < you[i] { return Ordering::Less; }
if me[i] > you[i] { return Ordering::Greater; }
}
Ordering::Equal
}
@ -1259,10 +1260,40 @@ impl U256 {
/// No overflow possible
#[cfg(not(all(asm_available, target_arch="x86_64")))]
pub fn full_mul(self, other: U256) -> U512 {
let self_512 = U512::from(self);
let other_512 = U512::from(other);
let (result, _) = self_512.overflowing_mul(other_512);
result
let U256(ref me) = self;
let U256(ref you) = other;
let mut ret = [0u64; 8];
for i in 0..4 {
if you[i] == 0 {
continue;
}
let mut carry2 = 0u64;
let (b_u, b_l) = split(you[i]);
for j in 0..4 {
if me[j] == 0 && carry2 == 0 {
continue;
}
let a = split(me[j]);
// multiply parts
let (c_l, overflow_l) = mul_u32(a, b_l, ret[i + j]);
let (c_u, overflow_u) = mul_u32(a, b_u, c_l >> 32);
ret[i + j] = (c_l & 0xFFFFFFFF) + (c_u << 32);
// Only single overflow possible here
let carry = (c_u >> 32) + (overflow_u << 32) + overflow_l + carry2;
let (carry, o) = carry.overflowing_add(ret[i + j + 1]);
ret[i + j + 1] = carry;
carry2 = o as u64;
}
}
U512(ret)
}
}
@ -1528,6 +1559,18 @@ mod tests {
//// TODO: bit inversion
}
#[test]
pub fn uint256_simple_mul() {
let a = U256::from_str("10000000000000000").unwrap();
let b = U256::from_str("10000000000000000").unwrap();
let c = U256::from_str("100000000000000000000000000000000").unwrap();
println!("Multiplying");
let result = a.overflowing_mul(b);
println!("Got result");
assert_eq!(result, (c, false))
}
#[test]
pub fn uint256_extreme_bitshift_test() {
//// Shifting a u64 by 64 bits gives an undefined value, so make sure that
@ -1595,6 +1638,14 @@ mod tests {
assert_eq!(U256::from(1u64) * U256::from(10u64), U256::from(10u64));
}
#[test]
pub fn uint256_mul2() {
let a = U512::from_str("10000000000000000fffffffffffffffe").unwrap();
let b = U512::from_str("ffffffffffffffffffffffffffffffff").unwrap();
assert_eq!(a * b, U512::from_str("10000000000000000fffffffffffffffcffffffffffffffff0000000000000002").unwrap());
}
#[test]
pub fn uint256_overflowing_mul() {
assert_eq!(
@ -1690,54 +1741,25 @@ mod tests {
}
#[test]
pub fn uint256_shl_overflow() {
pub fn uint256_shl() {
assert_eq!(
U256::from_str("7fffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffff").unwrap()
.overflowing_shl(4),
(U256::from_str("fffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffff0").unwrap(), true)
<< 4,
U256::from_str("fffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffff0").unwrap()
);
}
#[test]
pub fn uint256_shl_overflow_words() {
pub fn uint256_shl_words() {
assert_eq!(
U256::from_str("0000000000000001ffffffffffffffffffffffffffffffffffffffffffffffff").unwrap()
.overflowing_shl(64),
(U256::from_str("ffffffffffffffffffffffffffffffffffffffffffffffff0000000000000000").unwrap(), true)
<< 64,
U256::from_str("ffffffffffffffffffffffffffffffffffffffffffffffff0000000000000000").unwrap()
);
assert_eq!(
U256::from_str("0000000000000000ffffffffffffffffffffffffffffffffffffffffffffffff").unwrap()
.overflowing_shl(64),
(U256::from_str("ffffffffffffffffffffffffffffffffffffffffffffffff0000000000000000").unwrap(), false)
);
}
#[test]
pub fn uint256_shl_overflow_words2() {
assert_eq!(
U256::from_str("00000000000000000000000000000001ffffffffffffffffffffffffffffffff").unwrap()
.overflowing_shl(128),
(U256::from_str("ffffffffffffffffffffffffffffffff00000000000000000000000000000000").unwrap(), true)
);
assert_eq!(
U256::from_str("00000000000000000000000000000000ffffffffffffffffffffffffffffffff").unwrap()
.overflowing_shl(128),
(U256::from_str("ffffffffffffffffffffffffffffffff00000000000000000000000000000000").unwrap(), false)
);
assert_eq!(
U256::from_str("00000000000000000000000000000000ffffffffffffffffffffffffffffffff").unwrap()
.overflowing_shl(129),
(U256::from_str("fffffffffffffffffffffffffffffffe00000000000000000000000000000000").unwrap(), true)
);
}
#[test]
pub fn uint256_shl_overflow2() {
assert_eq!(
U256::from_str("0fffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffff").unwrap()
.overflowing_shl(4),
(U256::from_str("fffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffff0").unwrap(), false)
<< 64,
U256::from_str("ffffffffffffffffffffffffffffffffffffffffffffffff0000000000000000").unwrap()
);
}

56
util/src/lib.rs
View File

@ -169,3 +169,59 @@ pub use io::*;
pub use log::*;
pub use kvdb::*;
#[cfg(test)]
mod tests {
use super::numbers::*;
use std::str::FromStr;
#[test]
fn u256_multi_muls() {
let (result, _) = U256([0, 0, 0, 0]).overflowing_mul(U256([0, 0, 0, 0]));
assert_eq!(U256([0, 0, 0, 0]), result);
let (result, _) = U256([1, 0, 0, 0]).overflowing_mul(U256([1, 0, 0, 0]));
assert_eq!(U256([1, 0, 0, 0]), result);
let (result, _) = U256([5, 0, 0, 0]).overflowing_mul(U256([5, 0, 0, 0]));
assert_eq!(U256([25, 0, 0, 0]), result);
let (result, _) = U256([0, 5, 0, 0]).overflowing_mul(U256([0, 5, 0, 0]));
assert_eq!(U256([0, 0, 25, 0]), result);
let (result, _) = U256([0, 0, 0, 1]).overflowing_mul(U256([1, 0, 0, 0]));
assert_eq!(U256([0, 0, 0, 1]), result);
let (result, _) = U256([0, 0, 0, 5]).overflowing_mul(U256([2, 0, 0, 0]));
assert_eq!(U256([0, 0, 0, 10]), result);
let (result, _) = U256([0, 0, 1, 0]).overflowing_mul(U256([0, 5, 0, 0]));
assert_eq!(U256([0, 0, 0, 5]), result);
let (result, _) = U256([0, 0, 8, 0]).overflowing_mul(U256([0, 0, 7, 0]));
assert_eq!(U256([0, 0, 0, 0]), result);
let (result, _) = U256([2, 0, 0, 0]).overflowing_mul(U256([0, 5, 0, 0]));
assert_eq!(U256([0, 10, 0, 0]), result);
let (result, _) = U256([1, 0, 0, 0]).overflowing_mul(U256([0, 0, 0, ::std::u64::MAX]));
assert_eq!(U256([0, 0, 0, ::std::u64::MAX]), result);
let x1 = U256::from_str("0000000000000000000000000000000000000000000000000000012365124623").unwrap();
let x2sqr_right = U256::from_str("000000000000000000000000000000000000000000014baeef72e0378e2328c9").unwrap();
let x1sqr = x1 * x1;
assert_eq!(H256::from(x2sqr_right), H256::from(x1sqr));
let x1cube = x1sqr * x1;
let x1cube_right = U256::from_str("0000000000000000000000000000000001798acde139361466f712813717897b").unwrap();
assert_eq!(H256::from(x1cube_right), H256::from(x1cube));
let x1quad = x1cube * x1;
let x1quad_right = U256::from_str("000000000000000000000001adbdd6bd6ff027485484b97f8a6a4c7129756dd1").unwrap();
assert_eq!(H256::from(x1quad_right), H256::from(x1quad));
let x1penta = x1quad * x1;
let x1penta_right = U256::from_str("00000000000001e92875ac24be246e1c57e0507e8c46cc8d233b77f6f4c72993").unwrap();
assert_eq!(H256::from(x1penta_right), H256::from(x1penta));
let x1septima = x1penta * x1;
let x1septima_right = U256::from_str("00022cca1da3f6e5722b7d3cc5bbfb486465ebc5a708dd293042f932d7eee119").unwrap();
assert_eq!(H256::from(x1septima_right), H256::from(x1septima));
}
}