1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396
use awint_internals::*;
use const_fn::const_fn;
use crate::Bits;
/// # Division
///
/// These operations are not inplace unlike many other functions in this crate,
/// because extra mutable space is needed in order to avoid allocation.
///
/// Note that signed divisions can overflow when `duo.is_imin()` and
/// `div.is_umax()` (negative one in signed interpretation). The overflow
/// results in `quo.is_imin()` and `rem.is_zero()`.
///
/// Note about terminology: we like short three letter shorthands, but run into
/// a problem where the first three letters of "divide", "dividend", and
/// "divisor" all clash with each other. Additionally, the standard Rust
/// terminology for a function returning a quotient is things such as
/// `i64::wrapping_div`, which should have been named `i64::wrapping_quo`
/// instead. Here, we choose to type out "divide" in full whenever the operation
/// involves both quotients and remainders. We don't use "num" or "den", because
/// it may cause confusion later if an `awint` crate gains rational number
/// capabilities. We use "quo" for quotient and "rem" for remainder. We use
/// "div" for divisor. That still leaves a name clash with dividend, so we
/// choose to use the shorthand "duo". This originates from the fact that for
/// inplace division operations (which this crate does not have for performance
/// purposes and avoiding allocation), the dividend is often subtracted from in
/// the internal algorithms until it becomes the remainder, so that it serves
/// two purposes.
impl Bits {
/// Unsigned-divides `self` by `div`, sets `self` to the quotient, and
/// returns the remainder. Returns `None` if `div == 0`.
#[const_fn(cfg(feature = "const_support"))]
#[must_use]
pub const fn digit_udivide_inplace_(&mut self, div: Digit) -> Option<Digit> {
if div == 0 {
return None
}
// Note: we cannot have a carry-in because the quotient could otherwise
// overflow; `rem` needs to start as 0 and it would cause signature problems
// anyway.
let mut rem = 0;
const_for!(i in {0..self.len()}.rev() {
// Safety: we checked that `self.bw() == duo.bw()`
let y = unsafe {self.get_unchecked(i)};
let tmp = dd_division((y, rem), (div, 0));
rem = tmp.1.0;
*unsafe {self.get_unchecked_mut(i)} = tmp.0.0;
});
Some(rem)
}
// Unsigned-divides `duo` by `div`, sets `self` to the quotient, and
// returns the remainder. Returns `None` if `self.bw() != duo.bw()` or
// `div == 0`.
#[const_fn(cfg(feature = "const_support"))]
#[must_use]
pub const fn digit_udivide_(&mut self, duo: &Self, div: Digit) -> Option<Digit> {
if div == 0 || self.bw() != duo.bw() {
return None
}
let mut rem = 0;
const_for!(i in {0..self.len()}.rev() {
// Safety: we checked that `self.bw() == duo.bw()`
let y = unsafe {duo.get_unchecked(i)};
let tmp = dd_division((y, rem), (div, 0));
rem = tmp.1.0;
*unsafe {self.get_unchecked_mut(i)} = tmp.0.0;
});
Some(rem)
}
/// This function is factored out to keep code size of the division
/// functions from exploding
///
/// # Safety
///
/// Assumptions: The bitwidths all match, `div.lz() > duo.lz()`, `div.lz() -
/// duo.lz() < BITS`, and there are is at least `BITS * 2` bits worth of
/// significant bits in `duo`.
#[const_fn(cfg(feature = "const_support"))]
pub(crate) const unsafe fn two_possibility_algorithm(
quo: &mut Self,
rem: &mut Self,
duo: &Self,
div: &Self,
) {
debug_assert!(div.lz() > duo.lz());
debug_assert!((div.lz() - duo.lz()) < BITS);
debug_assert!((duo.sig()) >= (BITS * 2));
let i = duo.sig() - (BITS * 2);
let duo_sig_dd = duo.get_double_digit(i);
let div_sig_dd = div.get_double_digit(i);
// Because `lz_diff < BITS`, the quotient will fit in one `usize`
let mut small_quo = dd_division(duo_sig_dd, div_sig_dd).0 .0;
// using `rem` as a temporary
rem.copy_(div).unwrap();
let uof = rem.digit_cin_mul_(0, small_quo);
rem.rsb_(duo).unwrap();
if (uof != 0) || rem.msb() {
rem.add_(div).unwrap();
small_quo -= 1;
}
quo.digit_(small_quo);
}
/// Unsigned-divides `duo` by `div` and assigns the quotient to `quo` and
/// remainder to `rem`. Returns `None` if any bitwidths are not equal or
/// `div.is_zero()`.
#[const_fn(cfg(feature = "const_support"))]
#[must_use]
pub const fn udivide(quo: &mut Self, rem: &mut Self, duo: &Self, div: &Self) -> Option<()> {
// prevent any potential problems with the assumptions that many subroutines
// make
quo.assert_cleared_unused_bits();
rem.assert_cleared_unused_bits();
duo.assert_cleared_unused_bits();
div.assert_cleared_unused_bits();
let w = quo.bw();
if div.is_zero() || w != rem.bw() || w != duo.bw() || w != div.bw() {
return None
}
// This is a version of the "trifecta" division algorithm adapted for bigints.
// See https://github.com/AaronKutch/specialized-div-rem for better documentation.
let len = quo.len();
let mut duo_lz = duo.lz();
let div_lz = div.lz();
// quotient is 0 or 1 branch
if div_lz <= duo_lz {
if duo.uge(div).unwrap() {
quo.uone_();
rem.copy_(duo).unwrap();
rem.sub_(div).unwrap();
} else {
quo.zero_();
rem.copy_(duo).unwrap();
}
return Some(())
}
// small division branch
if (w - duo_lz) <= BITS {
let tmp_duo = duo.to_digit();
let tmp_div = div.to_digit();
quo.digit_(tmp_duo.wrapping_div(tmp_div));
rem.digit_(tmp_duo.wrapping_rem(tmp_div));
return Some(())
}
// double digit division branch. This is needed or else some branches below
// cannot rely on there being at least two digits of significant bits.
if (w - duo_lz) <= BITS * 2 {
unsafe {
let tmp = dd_division(
(duo.first(), duo.get_unchecked(1)),
(div.first(), div.get_unchecked(1)),
);
// using `digit_` to make sure other digits are zeroed
quo.digit_(tmp.0 .0);
*quo.get_unchecked_mut(1) = tmp.0 .1;
rem.digit_(tmp.1 .0);
*rem.get_unchecked_mut(1) = tmp.1 .1;
}
return Some(())
}
// TODO optimize for `trailing_zeros`. This function needs optimization in
// general such as using subslices more aggressively, need internal functions
// that handle differing lengths.
// short division branch
if w - div_lz <= BITS {
let tmp = quo.digit_udivide_(duo, div.to_digit()).unwrap();
rem.digit_(tmp);
return Some(())
}
// Two possibility division algorithm branch
let lz_diff = div_lz - duo_lz;
if lz_diff < BITS {
// Safety: we checked for bitwidth equality, the quotient is 0 or 1 branch makes
// sure `div.lz() > duo.lz()`, we just checked that `lz_diff < BITS`, and the
// double digit division branch makes sure there are at least `BITS*2`
// significant bits
unsafe {
Bits::two_possibility_algorithm(quo, rem, duo, div);
}
return Some(())
}
// We make a slight deviation from the original trifecta algorithm: instead of
// finding the quotient partial by dividing the `BITS*2` msbs of `duo` by the
// `BITS` msbs of `div`, we use `BITS*2 - 1` msbs of `duo` and `BITS` msbs of
// `div`. This is because the original partial could require `BITS + 1`
// significant bits (consider (2^(BITS*2) - 1) / (2^(BITS - 1) + 1)). This is
// possible to handle, but causes more problems than it is worth as seen from an
// earlier implementation in the `apint` crate.
let div_extra = w - div_lz - BITS;
let div_sig_d = div.get_digit(div_extra);
let div_sig_d_add1 = widen_add(div_sig_d, 1, 0);
quo.zero_();
// use `rem` as "duo" from now on
rem.copy_(duo).unwrap();
loop {
let duo_extra = w - duo_lz - (BITS * 2) + 1;
// using `<` instead of `<=` because of the change to `duo_extra`
if div_extra < duo_extra {
// Undersubtracting long division step
// `get_dd_unchecked` will not work, e.x. w = 192 and duo_lz = 0, it will
// attempt to access an imaginary zero bit beyond the bitwidth
let duo_sig_dd = unsafe {
let digits = digits_u(duo_extra);
let bits = extra_u(duo_extra);
if bits == 0 {
(rem.get_unchecked(digits), rem.get_unchecked(digits + 1))
} else {
let mid = rem.get_unchecked(digits + 1);
let last = if digits + 2 == len {
0
} else {
rem.get_unchecked(digits + 2)
};
(
(rem.get_unchecked(digits) >> bits) | (mid << (BITS - bits)),
(mid >> bits) | (last << (BITS - bits)),
)
}
};
let quo_part = dd_division(duo_sig_dd, div_sig_d_add1).0 .0;
let extra_shl = duo_extra - div_extra;
let shl_bits = extra_u(extra_shl);
let shl_digits = digits_u(extra_shl);
// Addition of `quo_part << extra_shl` to the quotient.
let (carry, next) = unsafe {
if shl_bits == 0 {
let tmp = widen_add(quo.get_unchecked(shl_digits), quo_part, 0);
*quo.get_unchecked_mut(shl_digits) = tmp.0;
(tmp.1 != 0, shl_digits + 1)
} else {
let tmp0 =
widen_add(quo.get_unchecked(shl_digits), quo_part << shl_bits, 0);
*quo.get_unchecked_mut(shl_digits) = tmp0.0;
let tmp1 = widen_add(
quo.get_unchecked(shl_digits + 1),
quo_part >> (BITS - shl_bits),
tmp0.1,
);
*quo.get_unchecked_mut(shl_digits + 1) = tmp1.0;
(tmp1.1 != 0, shl_digits + 2)
}
};
unsafe {
subdigits_mut!(quo, { next..len }, subquo, {
subquo.inc_(carry);
});
}
// Subtraction of `(div * quo_part) << extra_shl` from duo. Requires three
// different carries.
// carry for bits that wrap across digit boundaries when `<< shl_bits` is
// applied
let mut wrap_carry = 0;
// the multiplication carry
let mut mul_carry = 0;
// subtraction carry starting with the two's complement increment
let mut add_carry = 1;
unsafe {
if shl_bits == 0 {
// subdigit shift unneeded
const_for!(i in {shl_digits..len} {
let tmp1 = widen_mul_add(
div.get_unchecked(i - shl_digits),
quo_part,
mul_carry
);
mul_carry = tmp1.1;
// notice that `tmp1` is inverted for subtraction
let tmp2 = widen_add(!tmp1.0, rem.get_unchecked(i), add_carry);
add_carry = tmp2.1;
*rem.get_unchecked_mut(i) = tmp2.0;
});
} else {
const_for!(i in {shl_digits..len} {
let tmp0 = wrap_carry | (div.get_unchecked(i - shl_digits) << shl_bits);
wrap_carry = div.get_unchecked(i - shl_digits) >> (BITS - shl_bits);
let tmp1 = widen_mul_add(tmp0, quo_part, mul_carry);
mul_carry = tmp1.1;
let tmp2 = widen_add(!tmp1.0, rem.get_unchecked(i), add_carry);
add_carry = tmp2.1;
*rem.get_unchecked_mut(i) = tmp2.0;
});
}
}
} else {
// Two possibility algorithm
let i = w - duo_lz - (BITS * 2);
let duo_sig_dd = rem.get_double_digit(i);
let div_sig_dd = div.get_double_digit(i);
// Because `lz_diff < BITS`, the quotient will fit in one `usize`
let mut small_quo = dd_division(duo_sig_dd, div_sig_dd).0 .0;
// subtract `div*small_quo` from `rem` inplace
let mut mul_carry = 0;
let mut add_carry = 1;
unsafe {
const_for!(i in {0..len} {
let tmp0 = widen_mul_add(div.get_unchecked(i), small_quo, mul_carry);
mul_carry = tmp0.1;
let tmp1 = widen_add(!tmp0.0, rem.get_unchecked(i), add_carry);
add_carry = tmp1.1;
*rem.get_unchecked_mut(i) = tmp1.0;
});
}
if rem.msb() {
rem.add_(div).unwrap();
small_quo -= 1;
}
// add `quo_add` to `quo`
let tmp = widen_add(quo.first(), small_quo, 0);
*quo.first_mut() = tmp.0;
unsafe {
subdigits_mut!(quo, { 1..len }, subquo, {
subquo.inc_(tmp.1 != 0);
});
}
return Some(())
}
duo_lz = rem.lz();
if div_lz <= duo_lz {
// quotient can have 0 or 1 added to it
if div_lz == duo_lz && div.ule(rem).unwrap() {
quo.inc_(true);
rem.sub_(div).unwrap();
}
return Some(())
}
// duo fits in two digits. Only possible if `div` fits into two digits, but it
// is not worth it to unroll further
if (w - duo_lz) <= (BITS * 2) {
unsafe {
let tmp = dd_division(
(rem.first(), rem.get_unchecked(1)),
(div.first(), div.get_unchecked(1)),
);
let tmp0 = widen_add(quo.first(), tmp.0 .0, 0);
*quo.first_mut() = tmp0.0;
// tmp.0.1 is zero, just handle the carry now
subdigits_mut!(quo, { 1..len }, subquo, {
subquo.inc_(tmp0.1 != 0);
});
// using `digit_` to make sure other digits are zeroed
rem.digit_(tmp.1 .0);
*rem.get_unchecked_mut(1) = tmp.1 .1;
}
return Some(())
}
}
}
/// Signed-divides `duo` by `div` and assigns the quotient to `quo` and
/// remainder to `rem`. Returns `None` if any bitwidths are not equal or
/// `div.is_zero()`. `duo` and `div` are marked mutable but their values are
/// not changed by this function.
#[const_fn(cfg(feature = "const_support"))]
#[must_use]
pub const fn idivide(
quo: &mut Self,
rem: &mut Self,
duo: &mut Self,
div: &mut Self,
) -> Option<()> {
let w = quo.bw();
if div.is_zero() || w != rem.bw() || w != duo.bw() || w != div.bw() {
return None
}
let duo_msb = duo.msb();
let div_msb = div.msb();
duo.neg_(duo_msb);
div.neg_(div_msb);
Bits::udivide(quo, rem, duo, div).unwrap();
duo.neg_(duo_msb);
rem.neg_(duo_msb);
div.neg_(div_msb);
quo.neg_(duo_msb != div_msb);
Some(())
}
}