bigint math

This commit is contained in:
Janis 2024-08-27 00:54:55 +02:00
parent 99bf630546
commit 42a3be8186
6 changed files with 1009 additions and 268 deletions

View file

@ -1,7 +1,5 @@
use std::num::NonZero;
use itertools::Itertools;
use crate::string_table::{self, ImmOrIndex};
pub type Node = NonZero<u32>;
@ -229,6 +227,9 @@ impl IntegralType {
bits: 32,
}
}
pub fn u64_bitmask(self) -> u64 {
(1u64 << self.bits) - 1
}
}
#[derive(Debug, Clone, Eq, Hash)]

995
src/comptime.rs Normal file
View file

@ -0,0 +1,995 @@
pub mod bigint {
use crate::lexer::Radix;
/// A base-4_294_967_295 number.
#[derive(Clone)]
pub struct BigInt(Vec<u32>);
impl BigInt {
pub fn parse_digits<C: IntoIterator<Item = char>>(text: C, radix: Radix) -> BigInt {
parse_bigint(text.into_iter(), radix)
}
pub fn one() -> BigInt {
Self(vec![1])
}
pub fn zero() -> BigInt {
Self(vec![])
}
pub fn num_digits(&self) -> usize {
self.0.iter().rposition(|&d| d != 0).map_or(0, |i| i + 1)
}
pub fn bit_width(&self) -> usize {
count_bits(&self.0)
}
pub fn is_zero(&self) -> bool {
bigint_is_zero(&self.0)
}
pub fn is_one(&self) -> bool {
bigint_is_one(&self.0)
}
pub fn is_power_of_two(&self) -> bool {
is_power_of_two(&self.0)
}
pub fn trailing_zeros(&self) -> usize {
trailing_zeros(&self.0)
}
pub fn from_bytes_le(bytes: &[u8]) -> BigInt {
let data = bytes
.chunks(4)
.map(|chunk| {
let mut int = [0u8; 4];
int[..chunk.len()].copy_from_slice(chunk);
u32::from_le_bytes(int)
})
.collect::<Vec<_>>();
BigInt(data)
}
pub fn into_bytes_le(&self) -> Vec<u8> {
let mut bytes = Vec::<u8>::new();
for d in &self.0[..] {
bytes.extend(&d.to_le_bytes());
}
let count = bytes.iter().rev().take_while(|&&b| b == 0).count();
bytes.truncate(bytes.len() - count);
bytes
}
pub fn normalise(&mut self) {
let len = self.0.iter().rposition(|&d| d != 0).map_or(0, |i| i + 1);
self.0.truncate(len);
}
pub fn normalised(mut self) -> BigInt {
self.normalise();
self
}
pub fn into_u64(&self) -> Option<u64> {
if self.0.len() <= 2 {
let mut bytes = [0u8; 8];
self.0
.get(0)
.map(|&dw| bytes[..4].copy_from_slice(&dw.to_le_bytes()));
self.0
.get(1)
.map(|&dw| bytes[4..].copy_from_slice(&dw.to_le_bytes()));
Some(u64::from_le_bytes(bytes))
} else {
None
}
}
}
impl core::cmp::PartialEq for BigInt {
fn eq(&self, other: &Self) -> bool {
cmp_bigint(&self.0, &other.0) == core::cmp::Ordering::Equal
}
}
impl core::cmp::Eq for BigInt {}
impl core::cmp::PartialOrd for BigInt {
fn partial_cmp(&self, other: &Self) -> Option<std::cmp::Ordering> {
Some(cmp_bigint(&self.0, &other.0))
}
}
impl core::cmp::Ord for BigInt {
fn cmp(&self, other: &Self) -> std::cmp::Ordering {
cmp_bigint(&self.0, &other.0)
}
}
impl core::ops::Shl<usize> for BigInt {
type Output = Self;
fn shl(mut self, rhs: usize) -> Self::Output {
shl_bitint(&mut self.0, rhs);
self
}
}
impl core::ops::Shr<usize> for BigInt {
type Output = Self;
fn shr(mut self, rhs: usize) -> Self::Output {
shr_bitint(&mut self.0, rhs);
self
}
}
impl core::ops::Add for BigInt {
type Output = Self;
fn add(mut self, mut rhs: Self) -> Self::Output {
let (mut digits, carry) = if self.0.len() > rhs.0.len() {
let c = add_bigint(&mut self.0, &rhs.0);
(self.0, c)
} else {
let c = add_bigint(&mut rhs.0, &self.0);
(rhs.0, c)
};
if carry {
digits.push(u32::from(carry));
}
BigInt(digits)
}
}
impl core::ops::Sub for BigInt {
type Output = Self;
fn sub(mut self, rhs: Self) -> Self::Output {
if self.0.len() < rhs.0.len() {
println!("extending self by {} zeroes", rhs.0.len() - self.0.len());
self.0
.extend(core::iter::repeat(0).take(rhs.0.len() - self.0.len()));
println!("self: {self:?}");
}
sub_bigint(&mut self.0, &rhs.0);
self
}
}
impl core::ops::Mul for BigInt {
type Output = Self;
fn mul(self, rhs: Self) -> Self::Output {
BigInt(mul_bigint(&self.0, &rhs.0))
}
}
impl core::ops::Div for BigInt {
type Output = Self;
fn div(self, rhs: Self) -> Self::Output {
div_rem_bigint(self, rhs).0
}
}
impl core::ops::Rem for BigInt {
type Output = Self;
fn rem(self, rhs: Self) -> Self::Output {
div_rem_bigint(self, rhs).1
}
}
impl core::ops::Not for BigInt {
type Output = Self;
fn not(mut self) -> Self::Output {
self.0.iter_mut().for_each(|c| *c = !*c);
self
}
}
impl core::fmt::Debug for BigInt {
fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
let mut list = f.debug_list();
list.entries(self.0.iter().rev()).finish()
}
}
/// counts used bits in a u32 slice, discards leading zeros in MSB.
/// `[0xff,0xff,0x00,0x00]` -> 16
/// `[0xff,0xff,0x00]` -> 16
/// `[0xff,0xff,0x0f]` -> 20
pub fn count_bits(bytes: &[u32]) -> usize {
let mut bits = bytes.len() * u32::BITS as usize;
for &d in bytes.iter().rev() {
if d == 0 {
bits -= u32::BITS as usize;
} else {
bits -= d.leading_zeros() as usize;
break;
}
}
bits
}
#[test]
fn test_count_bits() {
assert_eq!(count_bits(&[0xffffffff, 0x00, 0x00]), 32);
assert_eq!(count_bits(&[0x00, 0x00, 0x00]), 0);
assert_eq!(count_bits(&[]), 0);
assert_eq!(count_bits(&[0xffffffff, 0xff, 0x00]), 40);
assert_eq!(count_bits(&[0xffffffff, 0xff]), 40);
assert_eq!(count_bits(&[0xffffffff, 0xff, 0xffff]), 64 + 16);
}
#[test]
fn test_count_trailing_zeros() {
assert_eq!(trailing_zeros(&[0xffffffff, 0x00, 0x00]), 0);
assert_eq!(trailing_zeros(&[0x00, 0x00, 0x00]), 0);
assert_eq!(trailing_zeros(&[]), 0);
assert_eq!(trailing_zeros(&[0x00, 0xffffffff, 0xff]), 32);
assert_eq!(trailing_zeros(&[0x00, 0xffffff00, 0xff]), 40);
}
#[allow(unused)]
/// lhs <=> rhs
fn cmp_bigint(lhs: &[u32], rhs: &[u32]) -> core::cmp::Ordering {
use core::cmp::Ordering;
let lhs_bits = count_bits(lhs);
let rhs_bits = count_bits(rhs);
match lhs_bits.cmp(&rhs_bits) {
Ordering::Less => Ordering::Less,
Ordering::Greater => Ordering::Greater,
Ordering::Equal => {
for (a, b) in lhs[..(lhs_bits / u32::BITS as usize)]
.iter()
.zip(rhs[..(lhs_bits / u32::BITS as usize)].iter())
.rev()
{
let ord = a.cmp(b);
if ord != Ordering::Equal {
return ord;
}
}
return Ordering::Equal;
}
}
}
fn bigint_is_zero(lhs: &[u32]) -> bool {
if lhs.len() == 0 {
true
} else {
lhs.iter().all(|c| c == &0)
}
}
#[allow(dead_code)]
fn bigint_is_one(lhs: &[u32]) -> bool {
lhs.len() > 0 && lhs[0] == 1 && lhs[1..].iter().all(|c| c == &0)
}
#[allow(dead_code)]
fn bitnot_bigint(lhs: &mut [u32]) {
for d in lhs.iter_mut() {
*d = !*d;
}
}
#[allow(dead_code)]
fn u32_mul_bigint(lhs: &mut Vec<u32>, scalar: u32) {
match scalar {
0 => {
lhs.clear();
lhs.push(0)
}
1 => {}
_ => {
if scalar.is_power_of_two() {
lhs.push(0);
shl_bitint(lhs.as_mut_slice(), scalar.trailing_zeros() as usize);
} else {
let mut carry = 0;
for a in lhs.iter_mut() {
(*a, carry) = (*a).carrying_mul(scalar, carry);
}
if carry != 0 {
lhs.push(carry);
}
}
}
}
}
#[allow(dead_code)]
fn u64_mul_bigint(lhs: &mut Vec<u32>, scalar: u64) {
let lo = scalar as u32;
let hi = (scalar >> 32) as u32;
u32_mul_bigint(lhs, lo);
shl_bitint(lhs, 32);
u32_mul_bigint(lhs, hi);
}
#[allow(dead_code)]
fn mul_bigint(lhs: &[u32], rhs: &[u32]) -> Vec<u32> {
if bigint_is_zero(lhs) || bigint_is_zero(rhs) {
return vec![];
}
let len = lhs.len() + rhs.len() + 1;
let mut product = vec![0u32; len];
for (bth, &b) in rhs.iter().enumerate() {
let mut carry = 0u32;
for (ath, &a) in lhs.iter().enumerate() {
let prod;
(prod, carry) = a.carrying_mul(b, carry);
let (digit, c) = product[ath + bth].carrying_add(prod, false);
carry += c as u32;
product[ath + bth] = digit;
}
if carry != 0 {
product[bth + lhs.len()] += carry;
}
}
product
}
#[allow(dead_code)]
fn sum_digits(digits: &[u32]) -> u64 {
let mut sum = 0u64;
let mut carry = false;
for &d in digits {
(sum, carry) = sum.carrying_add(d as u64, carry);
}
sum + carry as u64
}
#[allow(dead_code)]
fn count_ones(lhs: &[u32]) -> usize {
lhs.iter()
.fold(0usize, |acc, c| acc + c.count_ones() as usize)
}
#[allow(dead_code)]
fn trailing_zeros(lhs: &[u32]) -> usize {
lhs.iter()
.enumerate()
.find(|(_, &c)| c != 0)
.map(|(i, &c)| i * u32::BITS as usize + c.trailing_zeros() as usize)
.unwrap_or(0)
}
#[allow(dead_code)]
fn is_power_of_two(lhs: &[u32]) -> bool {
count_ones(lhs) == 1
}
#[allow(dead_code)]
/// divident must be at least as wide as divisor
/// returns (quotient, remainder)
fn div_rem_bigint(divident: BigInt, divisor: BigInt) -> (BigInt, BigInt) {
let divident = divident.normalised();
let mut divisor = divisor.normalised();
if bigint_is_zero(&divisor.0) {
panic!("divide by zero!");
}
if bigint_is_zero(&divident.0) {
return (BigInt::zero(), BigInt::zero());
}
use core::cmp::Ordering;
match cmp_bigint(&divident.0, &divisor.0) {
Ordering::Less => return (BigInt::zero(), divident),
Ordering::Equal => {
return (BigInt::one(), BigInt::zero());
}
Ordering::Greater => {}
}
if divisor.is_power_of_two() {
let exp = divisor.trailing_zeros();
let (div, rem) = divident.0.split_at(exp.div_floor(u32::BITS as usize));
let (mut div, mut rem) = (div.to_vec(), rem.to_vec());
shr_bitint(&mut div, exp % u32::BITS as usize);
let mask = (1u32 << exp as u32 % u32::BITS) - 1;
if let Some(last) = rem.last_mut() {
*last &= mask;
}
return (BigInt(div), BigInt(rem));
}
if divisor.num_digits() == 1 {
if divisor.0[0] == 1 {
return (divident, BigInt::zero());
}
let (div, rem) = div_digit_bigint(divident, divisor.0[0]);
divisor.0.clear();
divisor.0.push(rem);
return (div, divisor);
}
let shift = divisor.0.last().unwrap().leading_zeros() as usize;
if shift == 0 {
div_rem_core(divident, &divisor.0)
} else {
let (q, r) = div_rem_core(divident << shift, &(divisor << shift).0);
(q, r >> shift)
}
}
fn scalar_div_wide(hi: u32, lo: u32, divisor: u32) -> (u32, u32) {
let (div, rem);
unsafe {
core::arch::asm! {
"div {0:e}",
in(reg) divisor,
inout("dx") hi => rem,
inout("ax") lo => div,
}
}
(div, rem)
}
#[allow(dead_code)]
fn div_digit_bigint(mut divident: BigInt, divisor: u32) -> (BigInt, u32) {
assert!(divisor != 0);
let mut rem = 0;
for d in divident.0.iter_mut().rev() {
(*d, rem) = scalar_div_wide(rem, *d, divisor);
}
(divident.normalised(), rem)
}
fn from_lo_hi(lo: u32, hi: u32) -> u64 {
lo as u64 | (hi as u64) << 32
}
fn into_lo_hi(qword: u64) -> (u32, u32) {
(qword as u32, (qword >> 32) as u32)
}
// from rust num_bigint
/// Subtract a multiple.
/// a -= b * c
/// Returns a borrow (if a < b then borrow > 0).
fn sub_mul_digit_same_len(a: &mut [u32], b: &[u32], c: u32) -> u32 {
assert!(a.len() == b.len());
// carry is between -big_digit::MAX and 0, so to avoid overflow we store
// offset_carry = carry + big_digit::MAX
let mut offset_carry = u32::MAX;
for (x, y) in a.iter_mut().zip(b) {
// We want to calculate sum = x - y * c + carry.
// sum >= -(big_digit::MAX * big_digit::MAX) - big_digit::MAX
// sum <= big_digit::MAX
// Offsetting sum by (big_digit::MAX << big_digit::BITS) puts it in DoubleBigDigit range.
let offset_sum = from_lo_hi(u32::MAX, *x) - u32::MAX as u64 + offset_carry as u64
- *y as u64 * c as u64;
let (new_x, new_offset_carry) = into_lo_hi(offset_sum);
offset_carry = new_offset_carry;
*x = new_x;
}
// Return the borrow.
u32::MAX - offset_carry
}
// from rust num_bigint
fn div_rem_core(mut a: BigInt, b: &[u32]) -> (BigInt, BigInt) {
// sanity check on fast paths
assert!(a.0.len() >= b.len() && b.len() > 1);
// a0 stores an additional extra most significant digit of the dividend, not stored in a.
let mut a0 = 0;
// [b1, b0] are the two most significant digits of the divisor. They never change.
let b0 = b[b.len() - 1];
let b1 = b[b.len() - 2];
let q_len = a.0.len() - b.len() + 1;
let mut q = BigInt(vec![0; q_len]);
for j in (0..q_len).rev() {
assert!(a.0.len() == b.len() + j);
let a1 = *a.0.last().unwrap();
let a2 = a.0[a.0.len() - 2];
// The first q0 estimate is [a1,a0] / b0. It will never be too small, it may be too large
// by at most 2.
let (mut q0, mut r) = if a0 < b0 {
let (q0, r) = scalar_div_wide(a0, a1, b0);
(q0, r as u64)
} else {
assert!(a0 == b0);
// Avoid overflowing q0, we know the quotient fits in BigDigit.
// [a1,a0] = b0 * (1<<BITS - 1) + (a0 + a1)
(u32::MAX, a0 as u64 + a1 as u64)
};
// r = [a1,a0] - q0 * b0
//
// Now we want to compute a more precise estimate [a2,a1,a0] / [b1,b0] which can only be
// less or equal to the current q0.
//
// q0 is too large if:
// [a2,a1,a0] < q0 * [b1,b0]
// (r << BITS) + a2 < q0 * b1
while r <= u32::MAX as u64 && from_lo_hi(r as u32, a2) < q0 as u64 * b1 as u64 {
q0 -= 1;
r += b0 as u64;
}
// q0 is now either the correct quotient digit, or in rare cases 1 too large.
// Subtract (q0 << j) from a. This may overflow, in which case we will have to correct.
let mut borrow = sub_mul_digit_same_len(&mut a.0[j..], b, q0);
if borrow > a0 {
// q0 is too large. We need to add back one multiple of b.
q0 -= 1;
borrow -= add_bigint(&mut a.0[j..], b) as u32;
}
// The top digit of a, stored in a0, has now been zeroed.
assert!(borrow == a0);
q.0[j] = q0;
// Pop off the next top digit of a.
a0 = a.0.pop().unwrap();
}
a.0.push(a0);
a.normalise();
assert_eq!(cmp_bigint(&a.0, b), core::cmp::Ordering::Less);
(q.normalised(), a)
}
#[allow(unused)]
fn shr_bitint(lhs: &mut [u32], shift: usize) {
if bigint_is_zero(lhs) || shift == 0 {
return;
}
let len = lhs.len();
let digit_offset = shift / 32;
let bit_shift = shift % 32;
if digit_offset != 0 {
lhs.copy_within(digit_offset..len, 0);
lhs[(len - digit_offset)..].fill(0);
}
if bit_shift != 0 {
let lo_mask = (1u32 << (u32::BITS as usize - bit_shift)) - 1;
let hi_mask = !lo_mask;
eprintln!("lhs >> {shift}");
eprintln!("\tdigit_offset: {digit_offset}");
eprintln!("\tbit_shift: {bit_shift}");
eprintln!("\tlo_mask: 0b{lo_mask:0>32b}");
eprintln!("\thi_mask: 0b{hi_mask:0>32b}");
let mut carry = 0u32;
for i in 0..lhs.len() {
let digit = ((lhs[i] as u64) << 32) >> bit_shift;
let lo = digit as u32;
let hi = (digit >> 32) as u32;
lhs[i] &= hi_mask;
lhs[i] |= hi;
if i > 0 {
lhs[i - 1] &= lo_mask;
lhs[i - 1] |= lo;
}
}
}
}
#[allow(unused)]
/// lhs must have shift / 32 + 1 digits past the last bit if shifted-past bits are desired.
fn shl_bitint(lhs: &mut [u32], shift: usize) {
if bigint_is_zero(lhs) || shift == 0 {
return;
}
let len = lhs.len();
let digit_offset = shift / 32;
let bit_shift = shift % 32;
if digit_offset != 0 {
lhs.copy_within(0..(len - digit_offset), digit_offset);
lhs[..digit_offset].fill(0);
}
if bit_shift != 0 {
let hi_mask = (1u32 << bit_shift) - 1;
let lo_mask = !hi_mask;
eprintln!("lhs << {shift}");
eprintln!("\tdigit_offset: {digit_offset}");
eprintln!("\tbit_shift: {bit_shift}");
eprintln!("\tlo_mask: 0b{lo_mask:0>32b}");
eprintln!("\thi_mask: 0b{hi_mask:0>32b}");
// example with u8 digits, shift = 3;
// hi_mask = 0b00000111
// lo_mask = 0b11111000
//
// lhs[i] as u16 = 0b00000000_01111000
// digit = 0b00000011_11000000
// hi = 0b00000011
// lo = 0b11000000
let mut carry = 0u32;
for i in (digit_offset..len).rev() {
let digit = (lhs[i] as u64) << bit_shift;
let lo = digit as u32;
let hi = (digit >> 32) as u32;
lhs[i] &= lo_mask;
lhs[i] |= lo;
if i + 1 < len {
lhs[i + 1] &= hi_mask;
lhs[i + 1] |= hi;
}
}
}
}
#[allow(unused)]
/// lhs must be bigger than rhs
fn sub_bigint(lhs: &mut [u32], rhs: &[u32]) {
if bigint_is_zero(rhs) {
return;
}
let len = lhs.len().min(rhs.len());
let (l_lo, l_hi) = lhs.split_at_mut(len);
let (r_lo, r_hi) = rhs.split_at(len);
println!("lhs: {{ lo: {l_lo:?}, hi: {l_hi:?} }}");
println!("rhs: {{ lo: {r_lo:?}, hi: {r_hi:?} }}");
let mut borrow = false;
for (lhs, rhs) in l_lo.iter_mut().zip(r_lo) {
(*lhs, borrow) = lhs.borrowing_sub(*rhs, borrow);
}
if borrow {
for lhs in l_hi {
(*lhs, borrow) = lhs.borrowing_sub(0, borrow);
}
}
if borrow || !r_hi.iter().all(|&v| v == 0) {
panic!("sub failed: borrow: {borrow}");
}
}
/// lhs must be bigger than rhs
/// returns carry
fn add_bigint(lhs: &mut [u32], rhs: &[u32]) -> bool {
if bigint_is_zero(rhs) {
return false;
}
let (l_lo, l_hi) = lhs.split_at_mut(rhs.len());
let mut carry = false;
for (lhs, rhs) in l_lo.iter_mut().zip(rhs) {
(*lhs, carry) = lhs.carrying_add(*rhs, carry);
}
if carry {
for d in l_hi.iter_mut() {
(*d, carry) = d.carrying_add(0, carry);
if !carry {
break;
}
}
}
carry
}
fn parse_bigint(text: impl Iterator<Item = char>, radix: Radix) -> BigInt {
let digits = text
.filter_map(|c| match c {
'_' => None,
c => Some(radix.map_digit(c)),
})
.collect::<Vec<_>>();
let (max, power) = {
let radix = radix.radix() as u64;
let mut power = 1;
let mut base = radix;
while let Some(b) = base.checked_mul(radix) {
if b > u32::MAX as u64 {
break;
}
base = b;
power += 1;
}
(base, power)
};
let radix = radix.radix() as u32;
let r = digits.len() % power;
let i = if r == 0 { power } else { r };
let (head, tail) = digits.split_at(i);
let first = head
.iter()
.fold(0, |acc, &digit| acc * radix + digit as u32);
let mut data = vec![first];
for chunk in tail.chunks(power) {
if data.last() != Some(&0) {
data.push(0);
}
let mut carry = 0u64;
for digit in data.iter_mut() {
carry += *digit as u64 * max as u64;
*digit = carry as u32;
carry >>= u32::BITS;
}
assert!(carry == 0);
let next = chunk
.iter()
.fold(0, |acc, &digit| acc * radix + digit as u32);
let (res, mut carry) = data[0].carrying_add(next, false);
data[0] = res;
if carry {
for digit in data[1..].iter_mut() {
(*digit, carry) = digit.carrying_add(0, carry);
if !carry {
break;
}
}
}
}
BigInt(data)
}
#[cfg(test)]
mod tests {
use super::*;
#[test]
fn parse() {
let bigint = super::parse_bigint("2_cafe_babe_dead_beef".chars(), Radix::Hex);
println!("{:#x?}", bigint);
let bigint = super::parse_bigint("f".chars(), Radix::Hex);
println!("{:#x?}", bigint);
}
#[test]
fn add() {
let a = super::parse_bigint("2_0000_0000_0000_0000".chars(), Radix::Hex);
println!("{:#x?}", a);
let b = super::parse_bigint("cafebabe".chars(), Radix::Hex);
println!("{:#x?}", b);
let sum = a + b;
println!("{:#x?}", sum);
}
#[test]
fn sub() {
let a = super::parse_bigint("deadbeef".chars(), Radix::Hex);
println!("{:#x?}", a);
let b = super::parse_bigint("56d2c".chars(), Radix::Hex);
println!("{:#x?}", b);
let sum = a - b;
println!("{:#x?}", sum);
}
#[test]
fn overflowing_sub() {
let a = super::parse_bigint("2_0000_0000_0000_0000".chars(), Radix::Hex);
println!("{:#x?}", a);
let b = super::parse_bigint("ffff_ffff".chars(), Radix::Hex);
println!("{:#x?}", b);
let sum = b - a;
println!("{:#x?}", sum);
}
#[test]
fn shr() {
let mut a = super::parse_bigint("cafe_babe_0000".chars(), Radix::Hex);
print!("{:0>8x?} >> 32 ", a);
shr_bitint(&mut a.0, 32);
println!("{:0>8x?}", a);
let mut a = super::parse_bigint("11110000".chars(), Radix::Bin);
print!("{:0>8x?} >> 32 ", a);
shr_bitint(&mut a.0, 3);
println!("{:0>8x?}", a);
}
#[test]
fn shl() {
let mut a = super::parse_bigint("ffff_ffff".chars(), Radix::Hex);
a.0.extend([0; 4]);
println!("{:0>8x?}", a);
shl_bitint(&mut a.0, 40);
println!("{:0>8x?}", a);
}
#[test]
fn div() {
let a = super::parse_bigint("cafebabe".chars(), Radix::Hex);
let b = super::parse_bigint("dead".chars(), Radix::Hex);
let (div, rem) = div_rem_bigint(a, b);
println!("div: {:0>8x?}", div);
println!("rem: {:0>8x?}", rem);
}
}
}
use std::ops::Add;
use bigint::BigInt;
use crate::ast::IntegralType;
#[derive(Debug, thiserror::Error)]
enum Error {
#[error("Incompatible Comptime Number variants.")]
IncompatibleTypes,
#[error("Integer overflow.")]
IntegerOverflow,
}
type Result<T> = core::result::Result<T, Error>;
enum ComptimeInt {
Native { bits: u64, ty: IntegralType },
BigInt { bits: BigInt, ty: IntegralType },
Comptime(BigInt),
}
impl ComptimeInt {
pub fn add(self, other: Self) -> Result<Self> {
match (self, other) {
(Self::Native { bits: a, ty: aty }, Self::Native { bits: b, ty: bty }) => {
let bits = if aty != bty {
return Err(Error::IncompatibleTypes);
} else if aty.signed {
(a as i64)
.checked_add(b as i64)
.ok_or(Error::IntegerOverflow)? as u64
} else {
(a as u64)
.checked_add(b as u64)
.ok_or(Error::IntegerOverflow)? as u64
};
if bits & !aty.u64_bitmask() != 0 {
return Err(Error::IntegerOverflow);
}
Ok(Self::Native { bits, ty: aty })
}
(Self::Comptime(a), Self::Comptime(b)) => Ok(Self::Comptime(a.clone().add(b.clone()))),
(Self::Comptime(a), Self::Native { bits, ty })
| (Self::Native { bits, ty }, Self::Comptime(a)) => {
let b = a.into_u64().ok_or(Error::IncompatibleTypes)?;
Self::Native { bits, ty }.add(Self::Native { bits: b, ty })
}
_ => Err(Error::IncompatibleTypes),
}
}
pub fn sub(self, other: Self) -> Result<Self> {
match (self, other) {
(Self::Native { bits: a, ty: aty }, Self::Native { bits: b, ty: bty }) => {
let bits = if aty != bty {
return Err(Error::IncompatibleTypes);
} else if aty.signed {
(a as i64)
.checked_sub(b as i64)
.ok_or(Error::IntegerOverflow)? as u64
} else {
(a as u64)
.checked_sub(b as u64)
.ok_or(Error::IntegerOverflow)? as u64
};
if bits & !aty.u64_bitmask() != 0 {
return Err(Error::IntegerOverflow);
}
Ok(Self::Native { bits, ty: aty })
}
(Self::Comptime(a), Self::Comptime(b)) => Ok(Self::Comptime(a - b)),
(Self::Comptime(a), b @ Self::Native { ty, .. }) => {
let a = a.into_u64().ok_or(Error::IncompatibleTypes)?;
Self::Native { bits: a, ty }.sub(b)
}
(a @ Self::Native { ty, .. }, Self::Comptime(b)) => {
let b = b.into_u64().ok_or(Error::IncompatibleTypes)?;
a.sub(Self::Native { bits: b, ty })
}
_ => Err(Error::IncompatibleTypes),
}
}
}
enum ComptimeFloat {
Binary32(f32),
Binary64(f64),
}
impl ComptimeFloat {
pub fn add(&self, other: &Self) -> Result<ComptimeFloat> {
match (self, other) {
(ComptimeFloat::Binary32(a), ComptimeFloat::Binary32(b)) => Ok(Self::Binary32(a + b)),
(ComptimeFloat::Binary64(a), ComptimeFloat::Binary64(b)) => Ok(Self::Binary64(a + b)),
_ => Err(Error::IncompatibleTypes),
}
}
pub fn sub(&self, other: &Self) -> Result<ComptimeFloat> {
match (self, other) {
(ComptimeFloat::Binary32(a), ComptimeFloat::Binary32(b)) => Ok(Self::Binary32(a - b)),
(ComptimeFloat::Binary64(a), ComptimeFloat::Binary64(b)) => Ok(Self::Binary64(a - b)),
_ => Err(Error::IncompatibleTypes),
}
}
pub fn mul(&self, other: &Self) -> Result<ComptimeFloat> {
match (self, other) {
(ComptimeFloat::Binary32(a), ComptimeFloat::Binary32(b)) => Ok(Self::Binary32(a * b)),
(ComptimeFloat::Binary64(a), ComptimeFloat::Binary64(b)) => Ok(Self::Binary64(a * b)),
_ => Err(Error::IncompatibleTypes),
}
}
pub fn div(&self, other: &Self) -> Result<ComptimeFloat> {
match (self, other) {
(ComptimeFloat::Binary32(a), ComptimeFloat::Binary32(b)) => Ok(Self::Binary32(a / b)),
(ComptimeFloat::Binary64(a), ComptimeFloat::Binary64(b)) => Ok(Self::Binary64(a / b)),
_ => Err(Error::IncompatibleTypes),
}
}
pub fn rem(&self, other: &Self) -> Result<ComptimeFloat> {
match (self, other) {
(ComptimeFloat::Binary32(a), ComptimeFloat::Binary32(b)) => Ok(Self::Binary32(a % b)),
(ComptimeFloat::Binary64(a), ComptimeFloat::Binary64(b)) => Ok(Self::Binary64(a % b)),
_ => Err(Error::IncompatibleTypes),
}
}
pub fn neg(&self) -> Result<ComptimeFloat> {
match self {
ComptimeFloat::Binary32(a) => Ok(Self::Binary32(-a)),
ComptimeFloat::Binary64(a) => Ok(Self::Binary64(-a)),
}
}
}
enum ComptimeNumber {
Integral(ComptimeInt),
Bool(bool),
Floating(ComptimeFloat),
}
impl ComptimeNumber {}

View file

@ -582,266 +582,6 @@ fn try_parse_integral_type(source: &mut Chars) -> Result<Option<()>> {
Ok(Some(()))
}
pub mod bigint {
use super::Radix;
pub struct BigInt(Vec<u32>);
impl BigInt {
pub fn parse_digits<C: IntoIterator<Item = char>>(text: C, radix: Radix) -> BigInt {
parse_bigint(text.into_iter(), radix)
}
pub fn bit_width(&self) -> u32 {
count_bits(&self.0)
}
pub fn from_bytes_le(bytes: &[u8]) -> BigInt {
let data = bytes
.chunks(4)
.map(|chunk| {
let mut int = [0u8; 4];
int[..chunk.len()].copy_from_slice(chunk);
u32::from_le_bytes(int)
})
.collect::<Vec<_>>();
BigInt(data)
}
pub fn into_bytes_le(&self) -> Vec<u8> {
let mut bytes = Vec::<u8>::new();
for d in &self.0[..] {
bytes.extend(&d.to_le_bytes());
}
let count = bytes.iter().rev().take_while(|&&b| b == 0).count();
bytes.truncate(bytes.len() - count);
bytes
}
}
impl core::ops::Add for BigInt {
type Output = Self;
fn add(mut self, mut rhs: Self) -> Self::Output {
let (mut digits, carry) = if self.0.len() > rhs.0.len() {
let c = add_bigint(&mut self.0, &rhs.0);
(self.0, c)
} else {
let c = add_bigint(&mut rhs.0, &self.0);
(rhs.0, c)
};
if carry {
digits.push(u32::from(carry));
}
BigInt(digits)
}
}
impl core::ops::Sub for BigInt {
type Output = Self;
fn sub(mut self, rhs: Self) -> Self::Output {
if self.0.len() < rhs.0.len() {
println!("extending self by {} zeroes", rhs.0.len() - self.0.len());
self.0
.extend(core::iter::repeat(0).take(rhs.0.len() - self.0.len()));
println!("self: {self:?}");
}
sub_bigint(&mut self.0, &rhs.0);
self
}
}
impl core::fmt::Debug for BigInt {
fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
let mut list = f.debug_list();
list.entries(self.0.iter()).finish()
}
}
/// counts used bits in a u32 slice, discards leading zeros in MSB.
/// `[0xff,0xff,0x00,0x00]` -> 16
/// `[0xff,0xff,0x00]` -> 16
/// `[0xff,0xff,0x0f]` -> 20
pub fn count_bits(bytes: &[u32]) -> u32 {
let mut bits = bytes.len() as u32;
for d in bytes.iter().rev() {
if *d == 0 {
bits -= u32::BITS;
} else {
bits -= d.leading_zeros();
break;
}
}
bits
}
#[test]
fn test_count_bits() {
assert_eq!(count_bits(&[0xffffffff, 0x00, 0x00]), 32);
assert_eq!(count_bits(&[0xffffffff, 0xff, 0x00]), 40);
assert_eq!(count_bits(&[0xffffffff, 0xff]), 40);
assert_eq!(count_bits(&[0xffffffff, 0xff, 0xffff]), 64 + 16);
}
#[allow(unused)]
/// lhs must be bigger than rhs
fn sub_bigint(lhs: &mut [u32], rhs: &[u32]) {
let len = lhs.len().min(rhs.len());
let (l_lo, l_hi) = lhs.split_at_mut(len);
let (r_lo, r_hi) = rhs.split_at(len);
println!("lhs: {{ lo: {l_lo:?}, hi: {l_hi:?} }}");
println!("rhs: {{ lo: {r_lo:?}, hi: {r_hi:?} }}");
let mut borrow = false;
for (lhs, rhs) in l_lo.iter_mut().zip(r_lo) {
(*lhs, borrow) = lhs.borrowing_sub(*rhs, borrow);
}
if borrow {
for lhs in l_hi {
(*lhs, borrow) = lhs.borrowing_sub(0, borrow);
}
}
if borrow || !r_hi.iter().all(|&v| v == 0) {
panic!("sub failed: borrow: {borrow}");
}
}
/// lhs must be bigger than rhs
fn add_bigint(lhs: &mut [u32], rhs: &[u32]) -> bool {
let (l_lo, l_hi) = lhs.split_at_mut(rhs.len());
let mut carry = false;
for (lhs, rhs) in l_lo.iter_mut().zip(rhs) {
(*lhs, carry) = lhs.carrying_add(*rhs, carry);
}
if carry {
for d in l_hi.iter_mut() {
(*d, carry) = d.carrying_add(0, carry);
if !carry {
break;
}
}
}
carry
}
fn parse_bigint(text: impl Iterator<Item = char>, radix: Radix) -> BigInt {
let digits = text
.filter_map(|c| match c {
'_' => None,
c => Some(radix.map_digit(c)),
})
.collect::<Vec<_>>();
let (max, power) = {
let radix = radix.radix() as u64;
let mut power = 1;
let mut base = radix;
while let Some(b) = base.checked_mul(radix) {
if b > u32::MAX as u64 {
break;
}
base = b;
power += 1;
}
(base, power)
};
let radix = radix.radix() as u32;
let r = digits.len() % power;
let i = if r == 0 { power } else { r };
let (head, tail) = digits.split_at(i);
let first = head
.iter()
.fold(0, |acc, &digit| acc * radix + digit as u32);
let mut data = vec![first];
for chunk in tail.chunks(power) {
if data.last() != Some(&0) {
data.push(0);
}
let mut carry = 0u64;
for digit in data.iter_mut() {
carry += *digit as u64 * max as u64;
*digit = carry as u32;
carry >>= u32::BITS;
}
assert!(carry == 0);
let next = chunk
.iter()
.fold(0, |acc, &digit| acc * radix + digit as u32);
let (res, mut carry) = data[0].carrying_add(next, false);
data[0] = res;
if carry {
for digit in data[1..].iter_mut() {
(*digit, carry) = digit.carrying_add(0, carry);
if !carry {
break;
}
}
}
}
BigInt(data)
}
#[cfg(test)]
mod tests {
use super::*;
#[test]
fn parse() {
let bigint = super::parse_bigint("2_cafe_babe_dead_beef".chars(), Radix::Hex);
println!("{:#x?}", bigint);
let bigint = super::parse_bigint("f".chars(), Radix::Hex);
println!("{:#x?}", bigint);
}
#[test]
fn add() {
let a = super::parse_bigint("2_0000_0000_0000_0000".chars(), Radix::Hex);
println!("{:#x?}", a);
let b = super::parse_bigint("cafebabe".chars(), Radix::Hex);
println!("{:#x?}", b);
let sum = a + b;
println!("{:#x?}", sum);
}
#[test]
fn sub() {
let a = super::parse_bigint("2_0000_0000_0000_0000".chars(), Radix::Hex);
println!("{:#x?}", a);
let b = super::parse_bigint("ffff_ffff".chars(), Radix::Hex);
println!("{:#x?}", b);
let sum = a - b;
println!("{:#x?}", sum);
}
#[test]
fn overflowing_sub() {
let a = super::parse_bigint("2_0000_0000_0000_0000".chars(), Radix::Hex);
println!("{:#x?}", a);
let b = super::parse_bigint("ffff_ffff".chars(), Radix::Hex);
println!("{:#x?}", b);
let sum = b - a;
println!("{:#x?}", sum);
}
}
}
#[derive(Debug, Clone, Copy, PartialEq, Eq)]
pub enum Radix {
Hex,
@ -873,7 +613,7 @@ impl Radix {
}
#[allow(unused)]
fn radix(self) -> u8 {
pub fn radix(self) -> u8 {
match self {
Radix::Hex => 16,
Radix::Bin => 2,

View file

@ -4,13 +4,16 @@
box_into_inner,
hash_extract_if,
bigint_helper_methods,
map_try_insert
map_try_insert,
iter_intersperse,
int_roundings
)]
#![allow(unused_macros)]
pub mod asm;
pub mod ast;
pub mod common;
pub mod comptime;
pub mod error;
pub mod lexer;
pub mod mir;

View file

@ -5,12 +5,12 @@ use itertools::Itertools;
use crate::{
ast::{FloatingType, IntegralType, LetOrVar, Node, PrimitiveType, Tag, Type},
common::NextIf,
comptime::bigint::{self, BigInt},
error::{AnalysisError, AnalysisErrorTag},
lexer::{bigint::BigInt, Radix, TokenIterator},
lexer::{Radix, TokenIterator},
string_table::{ImmOrIndex, Index, StringTable},
symbol_table::{SymbolKind, SymbolTable},
tokens::Token,
variant,
};
#[derive(Debug, thiserror::Error)]
@ -251,7 +251,7 @@ impl Tree {
.map(|(_, c)| c)
.collect::<Vec<_>>();
let value = crate::lexer::bigint::BigInt::parse_digits(digits, radix);
let value = BigInt::parse_digits(digits, radix);
let ty = match iter.clone().next() {
Some((_, 'u')) | Some((_, 'i')) => {

View file

@ -57,7 +57,7 @@ impl StringTable {
let ints = unsafe {
core::slice::from_raw_parts(bytes.as_ptr().cast::<u32>(), bytes.len() / 4)
};
crate::lexer::bigint::count_bits(ints)
bigint::count_bits(ints) as u32
}
}
}
@ -151,3 +151,5 @@ mod display {
}
pub use display::ImmOrIndexDisplay;
use crate::comptime::bigint;