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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521
// Copyright © 2018–2024 Trevor Spiteri
// This library is free software: you can redistribute it and/or
// modify it under the terms of either
//
// * the Apache License, Version 2.0 or
// * the MIT License
//
// at your option.
//
// You should have recieved copies of the Apache License and the MIT
// License along with the library. If not, see
// <https://www.apache.org/licenses/LICENSE-2.0> and
// <https://opensource.org/licenses/MIT>.
use core::num::{NonZeroU128, NonZeroU16, NonZeroU32, NonZeroU64, NonZeroU8};
// The mathematics below is based on the comments from FreeBSD's
// /usr/src/lib/msun/src/e_sqrt.c.
// q_i = sqrt(y) truncated to i bits after point.
// q_0 = 1
// y_i = 2^i (y - q_i^2)
// y_0 = y - 1
//
// If (q_i + 1>>(i+1))^2 <= y:
// q_(i+1) = q_i + 1>>(i+1)
// Else:
// q_(i+1) = q_i
//
// Equivalently:
//
// If q_i + 1>>(i+2) <= y_i:
// q_(i+1) = q_i + 1>>(i+1)
// y_(i+1) = 2 (y_i - q_i - 1>>(i+2))
// Else:
// q_i+1 = q_i
// y_i+1 = 2 y_i
//
// * Iterations do not include q_0, y_0 as they are initialization.
// * i goes from 1 to iter.
// * Both q and y are stored with 2 integer bits. q is in range [1, 2); y is
// in range [1, 4).
// * 1>>(i+2) needs special code when i + 2 > nbits - 2. Since maximum iter is
// nbits - 1, i + 2 can be nbits + 1 which is > nbits - 2 by 2.
//
// Some examples for u8.
//
// frac_nbits == 0:
// sip = 4 - leading / 2
// 4 significant int pairs: 0100 0000. -> 0000 1000. (y << 0, 3 iter, q >> 3)
// 3 significant int pairs: 0001 0000. -> 0000 0100. (y << 2, 2 iter, q >> 4)
// 2 significant int pairs: 0000 0100. -> 0000 0010. (y << 4, 1 iter, q >> 5)
// 1 significant int pairs: 0000 0001. -> 0000 0001. (y << 6, 0 iter, q >> 6)
// General: y << 8 - 2sip, -1 + sip iter, q >> 7 - sip
//
// frac_nbits == 1:
// sip = 4 - (leading + 1) / 2
// 4 significant int pairs: 100 0000.0 -> 000 1000.0 (y >> 1, 4 iter, q >> 2)
// 3 significant int pairs: 001 0000.0 -> 000 0100.0 (y << 1, 3 iter, q >> 3)
// 2 significant int pairs: 000 0100.0 -> 000 0010.0 (y << 3, 2 iter, q >> 4)
// 1 significant int pairs: 000 0001.0 -> 000 0001.0 (y << 5, 1 iter, q >> 5)
// 0 significant int pairs: 000 0000.1 -> 000 0000.1 (y << 7, 0 iter, q >> 6)
// General: y << 7 - 2sip, sip iter, q >> 6 - sip
//
// frac_nbits == 2:
// sip = 3 - leading / 2
// 3 significant int pairs: 01 0000.00 -> 00 0100.00 (y << 0, 4 iter, q >> 2)
// 2 significant int pairs: 00 0100.00 -> 00 0010.00 (y << 2, 3 iter, q >> 3)
// 1 significant int pairs: 00 0001.00 -> 00 0001.00 (y << 4, 2 iter, q >> 4)
// 0 significant int pairs: 00 0000.01 -> 00 0000.10 (y << 6, 1 iter, q >> 5)
// General: y << 6 - 2sip, 1 + sip iter, q >> 5 - sip
//
// frac_nbits = 3:
// sip = 3 - (leading + 1) / 2
// 3 significant int pairs: 1 0000.000 -> 0 0100.000 (y >> 1, 5 iter, q >> 1)
// 2 significant int pairs: 0 0100.000 -> 0 0010.000 (y << 1, 4 iter, q >> 2)
// 1 significant int pairs: 0 0001.000 -> 0 0001.000 (y << 3, 3 iter, q >> 3)
// 0 significant int pairs: 0 0000.010 -> 0 0000.100 (y << 5, 2 iter, q >> 4)
// -1 significant int pairs: 0 0000.001 -> 0 0000.010 (y << 7, 1 iter, q >> 5)
// General: y << 5 - 2sip, 2 + sip iter, q >> 4 - sip
//
// frac_nbits == 4:
// sip = 2 - leading / 2
// 2 significant int pairs: 0100.0000 -> 0010.0000 (y << 0, 5 iter, q >> 1)
// 1 significant int pairs: 0001.0000 -> 0001.0000 (y << 2, 4 iter, q >> 2)
// 0 significant int pairs: 0000.0100 -> 0000.1000 (y << 4, 3 iter, q >> 3)
// -1 significant int pairs: 0000.0001 -> 0000.0100 (y << 6, 2 iter, q >> 4)
// General: y << 4 - 2sip, 3 + sip iter, q >> 3 - sip
//
// frac_nbits = 5:
// sip = 2 - (leading + 1) / 2
// 2 significant int pairs: 100.0000 0 -> 010.0000 0 (y >> 1, 6 iter, q >> 0)
// 1 significant int pairs: 001.0000 0 -> 001.0000 0 (y << 1, 5 iter, q >> 1)
// 0 significant int pairs: 000.0100 0 -> 000.1000 0 (y << 3, 4 iter, q >> 2)
// -1 significant int pairs: 000.0001 0 -> 000.0100 0 (y << 5, 3 iter, q >> 3)
// -2 significant int pairs: 000.0000 1 -> 000.0010 1 (y << 7, 2 iter, q >> 4)
// General: y << 3 - 2sip, 4 + sip iter, q >> 2 - sip
//
// frac_nbits == 6:
// sip = 1 - leading / 2
// 1 significant int pairs: 01.0000 00 -> 01.0000 00 (y << 0, 6 iter, q >> 0)
// 0 significant int pairs: 00.0100 00 -> 00.1000 00 (y << 2, 5 iter, q >> 1)
// -1 significant int pairs: 00.0001 00 -> 00.0100 00 (y << 4, 4 iter, q >> 2)
// -2 significant int pairs: 00.0000 01 -> 00.0010 00 (y << 6, 3 iter, q >> 3)
// General: y << 2 - 2sip, 5 + sip iter, q >> 1 - sip
//
// frac_nbits == 7:
// sip = 1 - (leading + 1) / 2
// 1 significant int pairs: 1.0000 000 -> 1.0000 000 (y >> 1, 7 iter, q << 1)
// 0 significant int pairs: 0.0100 000 -> 0.1000 000 (y << 1, 6 iter, q >> 0)
// -1 significant int pairs: 0.0001 000 -> 0.0100 000 (y << 3, 5 iter, q >> 1)
// -2 significant int pairs: 0.0000 010 -> 0.0010 000 (y << 5, 4 iter, q >> 2)
// -3 significant int pairs: 0.0000 001 -> 0.0001 011 (y << 7, 3 iter, q >> 3)
// General: y << 1 - 2sip, 6 + sip iter, q >> -sip
//
// frac_nbits == 8:
// sip = 0 - leading / 2
// 0 significant int pairs: .0100 0000 -> .1000 0000 (y << 0, 7 iter, q << 1)
// -1 significant int pairs: .0001 0000 -> .0100 0000 (y << 2, 6 iter, q >> 0)
// -2 significant int pairs: .0000 0100 -> .0010 0000 (y << 4, 5 iter, q >> 1)
// -3 significant int pairs: .0000 0001 -> .0001 0000 (y << 6, 4 iter, q >> 2)
// General: y << -2sip, 7 + sip iter, q >> -1 - sip
//
// General:
// Even frac_nbits:
// sip = int_nbits / 2 - leading / 2
// Odd frac_nbits:
// sip = (int_nbits + 1) / 2 - (leading + 1) / 2
// Then:
// y << int_nbits - 2sip, frac_nbits - 1 + sip iter, q >> int_nbits - 1 - sip
macro_rules! impl_sqrt {
($u:ident, $NZ:ident) => {
pub const fn $u(val: $NZ, frac_nbits: u32) -> $u {
let int_nbits = $u::BITS - frac_nbits;
let odd_frac_nbits = frac_nbits % 2 != 0;
let leading = val.leading_zeros();
let sig_int_pairs = if odd_frac_nbits {
((int_nbits + 1) / 2) as i32 - ((leading + 1) / 2) as i32
} else {
(int_nbits / 2) as i32 - (leading / 2) as i32
};
let mut i = 1;
let mut q_i = 1 << ($u::BITS - 2);
let mut next_bit = q_i >> 1;
let mut y_i = val.get();
let input_shl = int_nbits as i32 - sig_int_pairs * 2;
if input_shl < 0 {
// This can only happen when we have odd frac_nbits and the most
// significant bit is set. We would need to shift right by 1.
debug_assert!(input_shl == -1);
// Do one iteration here as this is a special case.
// In this special case, y is in the range [1, 2) instead of [1, 4),
// and q is in the range [1, √2) instead of [1, 2).
// Therefore, q_1 is always 0b1.0, and never 0b1.1.
// Since q_0 = q_1 = 1, y_1 = 2 × (y - q_1^2) = 2 × y - 2 × q_i.
// Since input_shl is -1, its effect is cancelled out by 2 × y,
// and we only need to subtract 2 × q_i from y_i.
y_i -= 2 * q_i;
next_bit >>= 1;
i += 1;
} else {
y_i <<= input_shl;
y_i -= q_i;
};
let iters = (frac_nbits as i32 - 1 + sig_int_pairs) as u32;
while i <= iters {
let d = next_bit >> 1;
if d == 0 {
if i == iters {
// Here result_shr must be 0, otherwise we wouldn't have
// room to potentially insert one extra bit.
debug_assert!(int_nbits as i32 - 1 - sig_int_pairs == 0);
// d == 0.5, so we really need q_i + 0.5 <= y_i,
// which can be obtained with q_i < y_i
if q_i < y_i {
q_i += 1;
}
return q_i;
}
debug_assert!(i == iters - 1);
// Here result_shr must be -1, otherwise we wouldn't have
// room to potentially insert two extra bits.
debug_assert!(int_nbits as i32 - 1 - sig_int_pairs == -1);
// d == 0.5, so we really need q_i + 0.5 <= y_i,
// which can be obtained with q_i < y_i
if q_i < y_i {
// We cannot subtract d == 0.5 from y_i immediately, so
// we subtract 1 from y_i before the multiplication by 2
// and then add 1 back. (There may be a potential overflow
// if we multiply y_i by 2 and then subtract 1.)
y_i -= q_i + 1;
y_i *= 2;
y_i += 1;
q_i += 1;
} else {
y_i *= 2;
}
// d == 0.25, so we really need q_i + 0.25 <= y_i,
// which can be obtained with q_i < y_i
if q_i < y_i {
// We cannot add next_bit == 0.5 to q_i immediately, so
// we add 1 to q_i after the left shift.
q_i = (q_i << 1) + 1;
} else {
q_i <<= 1;
}
return q_i;
}
if q_i + d <= y_i {
y_i -= q_i + d;
q_i += next_bit;
}
y_i *= 2;
next_bit = d;
i += 1;
}
let result_shr = int_nbits as i32 - 1 - sig_int_pairs;
q_i >> result_shr
}
};
}
impl_sqrt! { u8, NonZeroU8 }
impl_sqrt! { u16, NonZeroU16 }
impl_sqrt! { u32, NonZeroU32 }
impl_sqrt! { u64, NonZeroU64 }
impl_sqrt! { u128, NonZeroU128 }
#[cfg(test)]
mod tests {
use crate::types::extra::{
U0, U1, U125, U126, U127, U128, U13, U14, U15, U16, U17, U29, U3, U30, U31, U32, U33, U4,
U5, U6, U61, U62, U63, U64, U65, U7, U8, U9,
};
use crate::{
FixedI128, FixedI16, FixedI32, FixedI64, FixedI8, FixedU128, FixedU16, FixedU32, FixedU64,
FixedU8,
};
macro_rules! check_sqrt {
($val:expr) => {{
let sqrt = $val.sqrt();
assert!(sqrt * sqrt <= $val);
let delta = $val.wrapping_neg().wrapping_sub(!$val);
if let Some(sqrt_delta) = sqrt.checked_add(delta) {
if let Some(prod) = sqrt_delta.checked_mul(sqrt_delta) {
assert!(prod >= $val);
}
}
}};
}
#[test]
fn check_max_8() {
check_sqrt!(FixedU8::<U0>::MAX);
check_sqrt!(FixedU8::<U1>::MAX);
check_sqrt!(FixedU8::<U3>::MAX);
check_sqrt!(FixedU8::<U4>::MAX);
check_sqrt!(FixedU8::<U5>::MAX);
check_sqrt!(FixedU8::<U7>::MAX);
check_sqrt!(FixedU8::<U8>::MAX);
assert_eq!(FixedU8::<U8>::MAX.sqrt(), FixedU8::<U8>::MAX);
check_sqrt!(FixedI8::<U0>::MAX);
check_sqrt!(FixedI8::<U1>::MAX);
check_sqrt!(FixedI8::<U3>::MAX);
check_sqrt!(FixedI8::<U4>::MAX);
check_sqrt!(FixedI8::<U5>::MAX);
check_sqrt!(FixedI8::<U7>::MAX);
assert!(FixedI8::<U8>::MAX.checked_sqrt().is_none());
}
#[test]
fn check_max_16() {
check_sqrt!(FixedU16::<U0>::MAX);
check_sqrt!(FixedU16::<U1>::MAX);
check_sqrt!(FixedU16::<U7>::MAX);
check_sqrt!(FixedU16::<U8>::MAX);
check_sqrt!(FixedU16::<U9>::MAX);
check_sqrt!(FixedU16::<U15>::MAX);
check_sqrt!(FixedU16::<U16>::MAX);
assert_eq!(FixedU16::<U16>::MAX.sqrt(), FixedU16::<U16>::MAX);
check_sqrt!(FixedI16::<U0>::MAX);
check_sqrt!(FixedI16::<U1>::MAX);
check_sqrt!(FixedI16::<U7>::MAX);
check_sqrt!(FixedI16::<U8>::MAX);
check_sqrt!(FixedI16::<U9>::MAX);
check_sqrt!(FixedI16::<U15>::MAX);
assert!(FixedI16::<U16>::MAX.checked_sqrt().is_none());
}
#[test]
fn check_max_32() {
check_sqrt!(FixedU32::<U0>::MAX);
check_sqrt!(FixedU32::<U1>::MAX);
check_sqrt!(FixedU32::<U15>::MAX);
check_sqrt!(FixedU32::<U16>::MAX);
check_sqrt!(FixedU32::<U17>::MAX);
check_sqrt!(FixedU32::<U31>::MAX);
check_sqrt!(FixedU32::<U32>::MAX);
assert_eq!(FixedU32::<U32>::MAX.sqrt(), FixedU32::<U32>::MAX);
check_sqrt!(FixedI32::<U0>::MAX);
check_sqrt!(FixedI32::<U1>::MAX);
check_sqrt!(FixedI32::<U15>::MAX);
check_sqrt!(FixedI32::<U16>::MAX);
check_sqrt!(FixedI32::<U17>::MAX);
check_sqrt!(FixedI32::<U31>::MAX);
assert!(FixedI32::<U32>::MAX.checked_sqrt().is_none());
}
#[test]
fn check_max_64() {
check_sqrt!(FixedU64::<U0>::MAX);
check_sqrt!(FixedU64::<U1>::MAX);
check_sqrt!(FixedU64::<U31>::MAX);
check_sqrt!(FixedU64::<U32>::MAX);
check_sqrt!(FixedU64::<U33>::MAX);
check_sqrt!(FixedU64::<U63>::MAX);
check_sqrt!(FixedU64::<U64>::MAX);
assert_eq!(FixedU64::<U64>::MAX.sqrt(), FixedU64::<U64>::MAX);
check_sqrt!(FixedI64::<U0>::MAX);
check_sqrt!(FixedI64::<U1>::MAX);
check_sqrt!(FixedI64::<U31>::MAX);
check_sqrt!(FixedI64::<U32>::MAX);
check_sqrt!(FixedI64::<U33>::MAX);
check_sqrt!(FixedI64::<U63>::MAX);
assert!(FixedI64::<U64>::MAX.checked_sqrt().is_none());
}
#[test]
fn check_max_128() {
check_sqrt!(FixedU128::<U0>::MAX);
check_sqrt!(FixedU128::<U1>::MAX);
check_sqrt!(FixedU128::<U63>::MAX);
check_sqrt!(FixedU128::<U64>::MAX);
check_sqrt!(FixedU128::<U65>::MAX);
check_sqrt!(FixedU128::<U127>::MAX);
check_sqrt!(FixedU128::<U128>::MAX);
assert_eq!(FixedU128::<U128>::MAX.sqrt(), FixedU128::<U128>::MAX);
check_sqrt!(FixedI128::<U0>::MAX);
check_sqrt!(FixedI128::<U1>::MAX);
check_sqrt!(FixedI128::<U63>::MAX);
check_sqrt!(FixedI128::<U64>::MAX);
check_sqrt!(FixedI128::<U65>::MAX);
check_sqrt!(FixedI128::<U127>::MAX);
assert!(FixedI128::<U128>::MAX.checked_sqrt().is_none());
}
#[test]
fn check_two_8() {
assert_eq!(FixedU8::<U0>::from_num(2).sqrt(), FixedU8::<U0>::SQRT_2);
assert_eq!(FixedU8::<U1>::from_num(2).sqrt(), FixedU8::<U1>::SQRT_2);
assert_eq!(FixedU8::<U3>::from_num(2).sqrt(), FixedU8::<U3>::SQRT_2);
assert_eq!(FixedU8::<U4>::from_num(2).sqrt(), FixedU8::<U4>::SQRT_2);
assert_eq!(FixedU8::<U5>::from_num(2).sqrt(), FixedU8::<U5>::SQRT_2);
assert_eq!(FixedU8::<U6>::from_num(2).sqrt(), FixedU8::<U6>::SQRT_2);
assert!(
FixedU8::<U7>::MAX.sqrt() == FixedU8::<U7>::SQRT_2 - FixedU8::<U7>::DELTA
|| FixedU8::<U7>::MAX.sqrt() == FixedU8::<U7>::SQRT_2
);
assert_eq!(FixedI8::<U0>::from_num(2).sqrt(), FixedI8::<U0>::SQRT_2);
assert_eq!(FixedI8::<U1>::from_num(2).sqrt(), FixedI8::<U1>::SQRT_2);
assert_eq!(FixedI8::<U3>::from_num(2).sqrt(), FixedI8::<U3>::SQRT_2);
assert_eq!(FixedI8::<U4>::from_num(2).sqrt(), FixedI8::<U4>::SQRT_2);
assert_eq!(FixedI8::<U5>::from_num(2).sqrt(), FixedI8::<U5>::SQRT_2);
assert!(
FixedI8::<U6>::MAX.sqrt() == FixedI8::<U6>::SQRT_2 - FixedI8::<U6>::DELTA
|| FixedI8::<U6>::MAX.sqrt() == FixedI8::<U6>::SQRT_2
);
}
#[test]
fn check_two_16() {
assert_eq!(FixedU16::<U0>::from_num(2).sqrt(), FixedU16::<U0>::SQRT_2);
assert_eq!(FixedU16::<U1>::from_num(2).sqrt(), FixedU16::<U1>::SQRT_2);
assert_eq!(FixedU16::<U7>::from_num(2).sqrt(), FixedU16::<U7>::SQRT_2);
assert_eq!(FixedU16::<U8>::from_num(2).sqrt(), FixedU16::<U8>::SQRT_2);
assert_eq!(FixedU16::<U9>::from_num(2).sqrt(), FixedU16::<U9>::SQRT_2);
assert_eq!(FixedU16::<U13>::from_num(2).sqrt(), FixedU16::<U13>::SQRT_2);
assert_eq!(FixedU16::<U14>::from_num(2).sqrt(), FixedU16::<U14>::SQRT_2);
assert!(
FixedU16::<U15>::MAX.sqrt() == FixedU16::<U15>::SQRT_2 - FixedU16::<U15>::DELTA
|| FixedU16::<U15>::MAX.sqrt() == FixedU16::<U15>::SQRT_2
);
assert_eq!(FixedI16::<U0>::from_num(2).sqrt(), FixedI16::<U0>::SQRT_2);
assert_eq!(FixedI16::<U1>::from_num(2).sqrt(), FixedI16::<U1>::SQRT_2);
assert_eq!(FixedI16::<U7>::from_num(2).sqrt(), FixedI16::<U7>::SQRT_2);
assert_eq!(FixedI16::<U8>::from_num(2).sqrt(), FixedI16::<U8>::SQRT_2);
assert_eq!(FixedI16::<U9>::from_num(2).sqrt(), FixedI16::<U9>::SQRT_2);
assert_eq!(FixedI16::<U13>::from_num(2).sqrt(), FixedI16::<U13>::SQRT_2);
assert!(
FixedI16::<U14>::MAX.sqrt() == FixedI16::<U14>::SQRT_2 - FixedI16::<U14>::DELTA
|| FixedI16::<U14>::MAX.sqrt() == FixedI16::<U14>::SQRT_2
);
}
#[test]
fn check_two_32() {
assert_eq!(FixedU32::<U0>::from_num(2).sqrt(), FixedU32::<U0>::SQRT_2);
assert_eq!(FixedU32::<U1>::from_num(2).sqrt(), FixedU32::<U1>::SQRT_2);
assert_eq!(FixedU32::<U15>::from_num(2).sqrt(), FixedU32::<U15>::SQRT_2);
assert_eq!(FixedU32::<U16>::from_num(2).sqrt(), FixedU32::<U16>::SQRT_2);
assert_eq!(FixedU32::<U17>::from_num(2).sqrt(), FixedU32::<U17>::SQRT_2);
assert_eq!(FixedU32::<U29>::from_num(2).sqrt(), FixedU32::<U29>::SQRT_2);
assert_eq!(FixedU32::<U30>::from_num(2).sqrt(), FixedU32::<U30>::SQRT_2);
assert!(
FixedU32::<U31>::MAX.sqrt() == FixedU32::<U31>::SQRT_2 - FixedU32::<U31>::DELTA
|| FixedU32::<U31>::MAX.sqrt() == FixedU32::<U31>::SQRT_2
);
assert_eq!(FixedI32::<U0>::from_num(2).sqrt(), FixedI32::<U0>::SQRT_2);
assert_eq!(FixedI32::<U1>::from_num(2).sqrt(), FixedI32::<U1>::SQRT_2);
assert_eq!(FixedI32::<U15>::from_num(2).sqrt(), FixedI32::<U15>::SQRT_2);
assert_eq!(FixedI32::<U16>::from_num(2).sqrt(), FixedI32::<U16>::SQRT_2);
assert_eq!(FixedI32::<U17>::from_num(2).sqrt(), FixedI32::<U17>::SQRT_2);
assert_eq!(FixedI32::<U29>::from_num(2).sqrt(), FixedI32::<U29>::SQRT_2);
assert!(
FixedI32::<U30>::MAX.sqrt() == FixedI32::<U30>::SQRT_2 - FixedI32::<U30>::DELTA
|| FixedI32::<U30>::MAX.sqrt() == FixedI32::<U30>::SQRT_2
);
}
#[test]
fn check_two_64() {
assert_eq!(FixedU64::<U0>::from_num(2).sqrt(), FixedU64::<U0>::SQRT_2);
assert_eq!(FixedU64::<U1>::from_num(2).sqrt(), FixedU64::<U1>::SQRT_2);
assert_eq!(FixedU64::<U31>::from_num(2).sqrt(), FixedU64::<U31>::SQRT_2);
assert_eq!(FixedU64::<U32>::from_num(2).sqrt(), FixedU64::<U32>::SQRT_2);
assert_eq!(FixedU64::<U33>::from_num(2).sqrt(), FixedU64::<U33>::SQRT_2);
assert_eq!(FixedU64::<U61>::from_num(2).sqrt(), FixedU64::<U61>::SQRT_2);
assert_eq!(FixedU64::<U62>::from_num(2).sqrt(), FixedU64::<U62>::SQRT_2);
assert!(
FixedU64::<U63>::MAX.sqrt() == FixedU64::<U63>::SQRT_2 - FixedU64::<U63>::DELTA
|| FixedU64::<U63>::MAX.sqrt() == FixedU64::<U63>::SQRT_2
);
assert_eq!(FixedI64::<U0>::from_num(2).sqrt(), FixedI64::<U0>::SQRT_2);
assert_eq!(FixedI64::<U1>::from_num(2).sqrt(), FixedI64::<U1>::SQRT_2);
assert_eq!(FixedI64::<U31>::from_num(2).sqrt(), FixedI64::<U31>::SQRT_2);
assert_eq!(FixedI64::<U32>::from_num(2).sqrt(), FixedI64::<U32>::SQRT_2);
assert_eq!(FixedI64::<U33>::from_num(2).sqrt(), FixedI64::<U33>::SQRT_2);
assert_eq!(FixedI64::<U61>::from_num(2).sqrt(), FixedI64::<U61>::SQRT_2);
assert!(
FixedI64::<U62>::MAX.sqrt() == FixedI64::<U62>::SQRT_2 - FixedI64::<U62>::DELTA
|| FixedI64::<U62>::MAX.sqrt() == FixedI64::<U62>::SQRT_2
);
}
#[test]
fn check_two_128() {
assert_eq!(FixedU128::<U0>::from_num(2).sqrt(), FixedU128::<U0>::SQRT_2);
assert_eq!(FixedU128::<U1>::from_num(2).sqrt(), FixedU128::<U1>::SQRT_2);
assert_eq!(
FixedU128::<U63>::from_num(2).sqrt(),
FixedU128::<U63>::SQRT_2
);
assert_eq!(
FixedU128::<U64>::from_num(2).sqrt(),
FixedU128::<U64>::SQRT_2
);
assert_eq!(
FixedU128::<U65>::from_num(2).sqrt(),
FixedU128::<U65>::SQRT_2
);
assert_eq!(
FixedU128::<U125>::from_num(2).sqrt(),
FixedU128::<U125>::SQRT_2
);
assert_eq!(
FixedU128::<U126>::from_num(2).sqrt(),
FixedU128::<U126>::SQRT_2
);
assert!(
FixedU128::<U127>::MAX.sqrt() == FixedU128::<U127>::SQRT_2 - FixedU128::<U127>::DELTA
|| FixedU128::<U127>::MAX.sqrt() == FixedU128::<U127>::SQRT_2
);
assert_eq!(FixedI128::<U0>::from_num(2).sqrt(), FixedI128::<U0>::SQRT_2);
assert_eq!(FixedI128::<U1>::from_num(2).sqrt(), FixedI128::<U1>::SQRT_2);
assert_eq!(
FixedI128::<U63>::from_num(2).sqrt(),
FixedI128::<U63>::SQRT_2
);
assert_eq!(
FixedI128::<U64>::from_num(2).sqrt(),
FixedI128::<U64>::SQRT_2
);
assert_eq!(
FixedI128::<U65>::from_num(2).sqrt(),
FixedI128::<U65>::SQRT_2
);
assert_eq!(
FixedI128::<U125>::from_num(2).sqrt(),
FixedI128::<U125>::SQRT_2
);
assert!(
FixedI128::<U126>::MAX.sqrt() == FixedI128::<U126>::SQRT_2 - FixedI128::<U126>::DELTA
|| FixedI128::<U126>::MAX.sqrt() == FixedI128::<U126>::SQRT_2
);
}
}