| 1 |
#![allow(non_camel_case_types)] |
| 2 |
|
| 3 |
use crate::soft::{x2, x4}; |
| 4 |
use crate::types::*; |
| 5 |
use core::ops::*; |
| 6 |
use zerocopy::{FromBytes, IntoBytes}; |
| 7 |
|
| 8 |
zerocopy::cryptocorrosion_derive_traits! { |
| 9 |
#[repr(C)] |
| 10 |
#[derive(Clone, Copy)] |
| 11 |
pub union vec128_storage { |
| 12 |
d: [u32; 4], |
| 13 |
q: [u64; 2], |
| 14 |
} |
| 15 |
} |
| 16 |
|
| 17 |
impl From<[u32; 4]> for vec128_storage { |
| 18 |
#[inline(always)] |
| 19 |
fn from(d: [u32; 4]) -> Self { |
| 20 |
Self { d } |
| 21 |
} |
| 22 |
} |
| 23 |
impl From<vec128_storage> for [u32; 4] { |
| 24 |
#[inline(always)] |
| 25 |
fn from(d: vec128_storage) -> Self { |
| 26 |
unsafe { d.d } |
| 27 |
} |
| 28 |
} |
| 29 |
impl From<[u64; 2]> for vec128_storage { |
| 30 |
#[inline(always)] |
| 31 |
fn from(q: [u64; 2]) -> Self { |
| 32 |
Self { q } |
| 33 |
} |
| 34 |
} |
| 35 |
impl From<vec128_storage> for [u64; 2] { |
| 36 |
#[inline(always)] |
| 37 |
fn from(q: vec128_storage) -> Self { |
| 38 |
unsafe { q.q } |
| 39 |
} |
| 40 |
} |
| 41 |
impl Default for vec128_storage { |
| 42 |
#[inline(always)] |
| 43 |
fn default() -> Self { |
| 44 |
Self { q: [0, 0] } |
| 45 |
} |
| 46 |
} |
| 47 |
impl Eq for vec128_storage {} |
| 48 |
impl PartialEq<vec128_storage> for vec128_storage { |
| 49 |
#[inline(always)] |
| 50 |
fn eq(&self, rhs: &Self) -> bool { |
| 51 |
unsafe { self.q == rhs.q } |
| 52 |
} |
| 53 |
} |
| 54 |
#[derive(Clone, Copy, PartialEq, Eq, Default)] |
| 55 |
pub struct vec256_storage { |
| 56 |
v128: [vec128_storage; 2], |
| 57 |
} |
| 58 |
impl vec256_storage { |
| 59 |
#[inline(always)] |
| 60 |
pub fn new128(v128: [vec128_storage; 2]) -> Self { |
| 61 |
Self { v128 } |
| 62 |
} |
| 63 |
#[inline(always)] |
| 64 |
pub fn split128(self) -> [vec128_storage; 2] { |
| 65 |
self.v128 |
| 66 |
} |
| 67 |
} |
| 68 |
impl From<vec256_storage> for [u64; 4] { |
| 69 |
#[inline(always)] |
| 70 |
fn from(q: vec256_storage) -> Self { |
| 71 |
let [a, b]: [u64; 2] = q.v128[0].into(); |
| 72 |
let [c, d]: [u64; 2] = q.v128[1].into(); |
| 73 |
[a, b, c, d] |
| 74 |
} |
| 75 |
} |
| 76 |
impl From<[u64; 4]> for vec256_storage { |
| 77 |
#[inline(always)] |
| 78 |
fn from([a, b, c, d]: [u64; 4]) -> Self { |
| 79 |
Self { |
| 80 |
v128: [[a, b].into(), [c, d].into()], |
| 81 |
} |
| 82 |
} |
| 83 |
} |
| 84 |
#[derive(Clone, Copy, PartialEq, Eq, Default)] |
| 85 |
pub struct vec512_storage { |
| 86 |
v128: [vec128_storage; 4], |
| 87 |
} |
| 88 |
impl vec512_storage { |
| 89 |
#[inline(always)] |
| 90 |
pub fn new128(v128: [vec128_storage; 4]) -> Self { |
| 91 |
Self { v128 } |
| 92 |
} |
| 93 |
#[inline(always)] |
| 94 |
pub fn split128(self) -> [vec128_storage; 4] { |
| 95 |
self.v128 |
| 96 |
} |
| 97 |
} |
| 98 |
|
| 99 |
#[inline(always)] |
| 100 |
fn dmap<T, F>(t: T, f: F) -> T |
| 101 |
where |
| 102 |
T: Store<vec128_storage> + Into<vec128_storage>, |
| 103 |
F: Fn(u32) -> u32, |
| 104 |
{ |
| 105 |
let t: vec128_storage = t.into(); |
| 106 |
let d = unsafe { t.d }; |
| 107 |
let d = vec128_storage { |
| 108 |
d: [f(d[0]), f(d[1]), f(d[2]), f(d[3])], |
| 109 |
}; |
| 110 |
unsafe { T::unpack(d) } |
| 111 |
} |
| 112 |
|
| 113 |
fn dmap2<T, F>(a: T, b: T, f: F) -> T |
| 114 |
where |
| 115 |
T: Store<vec128_storage> + Into<vec128_storage>, |
| 116 |
F: Fn(u32, u32) -> u32, |
| 117 |
{ |
| 118 |
let a: vec128_storage = a.into(); |
| 119 |
let b: vec128_storage = b.into(); |
| 120 |
let ao = unsafe { a.d }; |
| 121 |
let bo = unsafe { b.d }; |
| 122 |
let d = vec128_storage { |
| 123 |
d: [ |
| 124 |
f(ao[0], bo[0]), |
| 125 |
f(ao[1], bo[1]), |
| 126 |
f(ao[2], bo[2]), |
| 127 |
f(ao[3], bo[3]), |
| 128 |
], |
| 129 |
}; |
| 130 |
unsafe { T::unpack(d) } |
| 131 |
} |
| 132 |
|
| 133 |
#[inline(always)] |
| 134 |
fn qmap<T, F>(t: T, f: F) -> T |
| 135 |
where |
| 136 |
T: Store<vec128_storage> + Into<vec128_storage>, |
| 137 |
F: Fn(u64) -> u64, |
| 138 |
{ |
| 139 |
let t: vec128_storage = t.into(); |
| 140 |
let q = unsafe { t.q }; |
| 141 |
let q = vec128_storage { |
| 142 |
q: [f(q[0]), f(q[1])], |
| 143 |
}; |
| 144 |
unsafe { T::unpack(q) } |
| 145 |
} |
| 146 |
|
| 147 |
#[inline(always)] |
| 148 |
fn qmap2<T, F>(a: T, b: T, f: F) -> T |
| 149 |
where |
| 150 |
T: Store<vec128_storage> + Into<vec128_storage>, |
| 151 |
F: Fn(u64, u64) -> u64, |
| 152 |
{ |
| 153 |
let a: vec128_storage = a.into(); |
| 154 |
let b: vec128_storage = b.into(); |
| 155 |
let ao = unsafe { a.q }; |
| 156 |
let bo = unsafe { b.q }; |
| 157 |
let q = vec128_storage { |
| 158 |
q: [f(ao[0], bo[0]), f(ao[1], bo[1])], |
| 159 |
}; |
| 160 |
unsafe { T::unpack(q) } |
| 161 |
} |
| 162 |
|
| 163 |
#[inline(always)] |
| 164 |
fn o_of_q(q: [u64; 2]) -> u128 { |
| 165 |
u128::from(q[0]) | (u128::from(q[1]) << 64) |
| 166 |
} |
| 167 |
|
| 168 |
#[inline(always)] |
| 169 |
fn q_of_o(o: u128) -> [u64; 2] { |
| 170 |
[o as u64, (o >> 64) as u64] |
| 171 |
} |
| 172 |
|
| 173 |
#[inline(always)] |
| 174 |
fn omap<T, F>(a: T, f: F) -> T |
| 175 |
where |
| 176 |
T: Store<vec128_storage> + Into<vec128_storage>, |
| 177 |
F: Fn(u128) -> u128, |
| 178 |
{ |
| 179 |
let a: vec128_storage = a.into(); |
| 180 |
let ao = o_of_q(unsafe { a.q }); |
| 181 |
let o = vec128_storage { q: q_of_o(f(ao)) }; |
| 182 |
unsafe { T::unpack(o) } |
| 183 |
} |
| 184 |
|
| 185 |
#[inline(always)] |
| 186 |
fn omap2<T, F>(a: T, b: T, f: F) -> T |
| 187 |
where |
| 188 |
T: Store<vec128_storage> + Into<vec128_storage>, |
| 189 |
F: Fn(u128, u128) -> u128, |
| 190 |
{ |
| 191 |
let a: vec128_storage = a.into(); |
| 192 |
let b: vec128_storage = b.into(); |
| 193 |
let ao = o_of_q(unsafe { a.q }); |
| 194 |
let bo = o_of_q(unsafe { b.q }); |
| 195 |
let o = vec128_storage { |
| 196 |
q: q_of_o(f(ao, bo)), |
| 197 |
}; |
| 198 |
unsafe { T::unpack(o) } |
| 199 |
} |
| 200 |
|
| 201 |
impl RotateEachWord128 for u128x1_generic {} |
| 202 |
impl BitOps128 for u128x1_generic {} |
| 203 |
impl BitOps64 for u128x1_generic {} |
| 204 |
impl BitOps64 for u64x2_generic {} |
| 205 |
impl BitOps32 for u128x1_generic {} |
| 206 |
impl BitOps32 for u64x2_generic {} |
| 207 |
impl BitOps32 for u32x4_generic {} |
| 208 |
impl BitOps0 for u128x1_generic {} |
| 209 |
impl BitOps0 for u64x2_generic {} |
| 210 |
impl BitOps0 for u32x4_generic {} |
| 211 |
|
| 212 |
macro_rules! impl_bitops { |
| 213 |
($vec:ident) => { |
| 214 |
impl Not for $vec { |
| 215 |
type Output = Self; |
| 216 |
#[inline(always)] |
| 217 |
fn not(self) -> Self::Output { |
| 218 |
omap(self, |x| !x) |
| 219 |
} |
| 220 |
} |
| 221 |
impl BitAnd for $vec { |
| 222 |
type Output = Self; |
| 223 |
#[inline(always)] |
| 224 |
fn bitand(self, rhs: Self) -> Self::Output { |
| 225 |
omap2(self, rhs, |x, y| x & y) |
| 226 |
} |
| 227 |
} |
| 228 |
impl BitOr for $vec { |
| 229 |
type Output = Self; |
| 230 |
#[inline(always)] |
| 231 |
fn bitor(self, rhs: Self) -> Self::Output { |
| 232 |
omap2(self, rhs, |x, y| x | y) |
| 233 |
} |
| 234 |
} |
| 235 |
impl BitXor for $vec { |
| 236 |
type Output = Self; |
| 237 |
#[inline(always)] |
| 238 |
fn bitxor(self, rhs: Self) -> Self::Output { |
| 239 |
omap2(self, rhs, |x, y| x ^ y) |
| 240 |
} |
| 241 |
} |
| 242 |
impl AndNot for $vec { |
| 243 |
type Output = Self; |
| 244 |
#[inline(always)] |
| 245 |
fn andnot(self, rhs: Self) -> Self::Output { |
| 246 |
omap2(self, rhs, |x, y| !x & y) |
| 247 |
} |
| 248 |
} |
| 249 |
impl BitAndAssign for $vec { |
| 250 |
#[inline(always)] |
| 251 |
fn bitand_assign(&mut self, rhs: Self) { |
| 252 |
*self = *self & rhs |
| 253 |
} |
| 254 |
} |
| 255 |
impl BitOrAssign for $vec { |
| 256 |
#[inline(always)] |
| 257 |
fn bitor_assign(&mut self, rhs: Self) { |
| 258 |
*self = *self | rhs |
| 259 |
} |
| 260 |
} |
| 261 |
impl BitXorAssign for $vec { |
| 262 |
#[inline(always)] |
| 263 |
fn bitxor_assign(&mut self, rhs: Self) { |
| 264 |
*self = *self ^ rhs |
| 265 |
} |
| 266 |
} |
| 267 |
|
| 268 |
impl Swap64 for $vec { |
| 269 |
#[inline(always)] |
| 270 |
fn swap1(self) -> Self { |
| 271 |
qmap(self, |x| { |
| 272 |
((x & 0x5555555555555555) << 1) | ((x & 0xaaaaaaaaaaaaaaaa) >> 1) |
| 273 |
}) |
| 274 |
} |
| 275 |
#[inline(always)] |
| 276 |
fn swap2(self) -> Self { |
| 277 |
qmap(self, |x| { |
| 278 |
((x & 0x3333333333333333) << 2) | ((x & 0xcccccccccccccccc) >> 2) |
| 279 |
}) |
| 280 |
} |
| 281 |
#[inline(always)] |
| 282 |
fn swap4(self) -> Self { |
| 283 |
qmap(self, |x| { |
| 284 |
((x & 0x0f0f0f0f0f0f0f0f) << 4) | ((x & 0xf0f0f0f0f0f0f0f0) >> 4) |
| 285 |
}) |
| 286 |
} |
| 287 |
#[inline(always)] |
| 288 |
fn swap8(self) -> Self { |
| 289 |
qmap(self, |x| { |
| 290 |
((x & 0x00ff00ff00ff00ff) << 8) | ((x & 0xff00ff00ff00ff00) >> 8) |
| 291 |
}) |
| 292 |
} |
| 293 |
#[inline(always)] |
| 294 |
fn swap16(self) -> Self { |
| 295 |
dmap(self, |x| x.rotate_left(16)) |
| 296 |
} |
| 297 |
#[inline(always)] |
| 298 |
fn swap32(self) -> Self { |
| 299 |
qmap(self, |x| x.rotate_left(32)) |
| 300 |
} |
| 301 |
#[inline(always)] |
| 302 |
fn swap64(self) -> Self { |
| 303 |
omap(self, |x| (x << 64) | (x >> 64)) |
| 304 |
} |
| 305 |
} |
| 306 |
}; |
| 307 |
} |
| 308 |
impl_bitops!(u32x4_generic); |
| 309 |
impl_bitops!(u64x2_generic); |
| 310 |
impl_bitops!(u128x1_generic); |
| 311 |
|
| 312 |
impl RotateEachWord32 for u32x4_generic { |
| 313 |
#[inline(always)] |
| 314 |
fn rotate_each_word_right7(self) -> Self { |
| 315 |
dmap(self, |x| x.rotate_right(7)) |
| 316 |
} |
| 317 |
#[inline(always)] |
| 318 |
fn rotate_each_word_right8(self) -> Self { |
| 319 |
dmap(self, |x| x.rotate_right(8)) |
| 320 |
} |
| 321 |
#[inline(always)] |
| 322 |
fn rotate_each_word_right11(self) -> Self { |
| 323 |
dmap(self, |x| x.rotate_right(11)) |
| 324 |
} |
| 325 |
#[inline(always)] |
| 326 |
fn rotate_each_word_right12(self) -> Self { |
| 327 |
dmap(self, |x| x.rotate_right(12)) |
| 328 |
} |
| 329 |
#[inline(always)] |
| 330 |
fn rotate_each_word_right16(self) -> Self { |
| 331 |
dmap(self, |x| x.rotate_right(16)) |
| 332 |
} |
| 333 |
#[inline(always)] |
| 334 |
fn rotate_each_word_right20(self) -> Self { |
| 335 |
dmap(self, |x| x.rotate_right(20)) |
| 336 |
} |
| 337 |
#[inline(always)] |
| 338 |
fn rotate_each_word_right24(self) -> Self { |
| 339 |
dmap(self, |x| x.rotate_right(24)) |
| 340 |
} |
| 341 |
#[inline(always)] |
| 342 |
fn rotate_each_word_right25(self) -> Self { |
| 343 |
dmap(self, |x| x.rotate_right(25)) |
| 344 |
} |
| 345 |
} |
| 346 |
|
| 347 |
impl RotateEachWord32 for u64x2_generic { |
| 348 |
#[inline(always)] |
| 349 |
fn rotate_each_word_right7(self) -> Self { |
| 350 |
qmap(self, |x| x.rotate_right(7)) |
| 351 |
} |
| 352 |
#[inline(always)] |
| 353 |
fn rotate_each_word_right8(self) -> Self { |
| 354 |
qmap(self, |x| x.rotate_right(8)) |
| 355 |
} |
| 356 |
#[inline(always)] |
| 357 |
fn rotate_each_word_right11(self) -> Self { |
| 358 |
qmap(self, |x| x.rotate_right(11)) |
| 359 |
} |
| 360 |
#[inline(always)] |
| 361 |
fn rotate_each_word_right12(self) -> Self { |
| 362 |
qmap(self, |x| x.rotate_right(12)) |
| 363 |
} |
| 364 |
#[inline(always)] |
| 365 |
fn rotate_each_word_right16(self) -> Self { |
| 366 |
qmap(self, |x| x.rotate_right(16)) |
| 367 |
} |
| 368 |
#[inline(always)] |
| 369 |
fn rotate_each_word_right20(self) -> Self { |
| 370 |
qmap(self, |x| x.rotate_right(20)) |
| 371 |
} |
| 372 |
#[inline(always)] |
| 373 |
fn rotate_each_word_right24(self) -> Self { |
| 374 |
qmap(self, |x| x.rotate_right(24)) |
| 375 |
} |
| 376 |
#[inline(always)] |
| 377 |
fn rotate_each_word_right25(self) -> Self { |
| 378 |
qmap(self, |x| x.rotate_right(25)) |
| 379 |
} |
| 380 |
} |
| 381 |
impl RotateEachWord64 for u64x2_generic { |
| 382 |
#[inline(always)] |
| 383 |
fn rotate_each_word_right32(self) -> Self { |
| 384 |
qmap(self, |x| x.rotate_right(32)) |
| 385 |
} |
| 386 |
} |
| 387 |
|
| 388 |
// workaround for koute/cargo-web#52 (u128::rotate_* broken with cargo web) |
| 389 |
#[inline(always)] |
| 390 |
fn rotate_u128_right(x: u128, i: u32) -> u128 { |
| 391 |
(x >> i) | (x << (128 - i)) |
| 392 |
} |
| 393 |
#[test] |
| 394 |
fn test_rotate_u128() { |
| 395 |
const X: u128 = 0x0001_0203_0405_0607_0809_0a0b_0c0d_0e0f; |
| 396 |
assert_eq!(rotate_u128_right(X, 17), X.rotate_right(17)); |
| 397 |
} |
| 398 |
|
| 399 |
impl RotateEachWord32 for u128x1_generic { |
| 400 |
#[inline(always)] |
| 401 |
fn rotate_each_word_right7(self) -> Self { |
| 402 |
Self([rotate_u128_right(self.0[0], 7)]) |
| 403 |
} |
| 404 |
#[inline(always)] |
| 405 |
fn rotate_each_word_right8(self) -> Self { |
| 406 |
Self([rotate_u128_right(self.0[0], 8)]) |
| 407 |
} |
| 408 |
#[inline(always)] |
| 409 |
fn rotate_each_word_right11(self) -> Self { |
| 410 |
Self([rotate_u128_right(self.0[0], 11)]) |
| 411 |
} |
| 412 |
#[inline(always)] |
| 413 |
fn rotate_each_word_right12(self) -> Self { |
| 414 |
Self([rotate_u128_right(self.0[0], 12)]) |
| 415 |
} |
| 416 |
#[inline(always)] |
| 417 |
fn rotate_each_word_right16(self) -> Self { |
| 418 |
Self([rotate_u128_right(self.0[0], 16)]) |
| 419 |
} |
| 420 |
#[inline(always)] |
| 421 |
fn rotate_each_word_right20(self) -> Self { |
| 422 |
Self([rotate_u128_right(self.0[0], 20)]) |
| 423 |
} |
| 424 |
#[inline(always)] |
| 425 |
fn rotate_each_word_right24(self) -> Self { |
| 426 |
Self([rotate_u128_right(self.0[0], 24)]) |
| 427 |
} |
| 428 |
#[inline(always)] |
| 429 |
fn rotate_each_word_right25(self) -> Self { |
| 430 |
Self([rotate_u128_right(self.0[0], 25)]) |
| 431 |
} |
| 432 |
} |
| 433 |
impl RotateEachWord64 for u128x1_generic { |
| 434 |
#[inline(always)] |
| 435 |
fn rotate_each_word_right32(self) -> Self { |
| 436 |
Self([rotate_u128_right(self.0[0], 32)]) |
| 437 |
} |
| 438 |
} |
| 439 |
|
| 440 |
#[derive(Copy, Clone)] |
| 441 |
pub struct GenericMachine; |
| 442 |
impl Machine for GenericMachine { |
| 443 |
type u32x4 = u32x4_generic; |
| 444 |
type u64x2 = u64x2_generic; |
| 445 |
type u128x1 = u128x1_generic; |
| 446 |
type u32x4x2 = u32x4x2_generic; |
| 447 |
type u64x2x2 = u64x2x2_generic; |
| 448 |
type u64x4 = u64x4_generic; |
| 449 |
type u128x2 = u128x2_generic; |
| 450 |
type u32x4x4 = u32x4x4_generic; |
| 451 |
type u64x2x4 = u64x2x4_generic; |
| 452 |
type u128x4 = u128x4_generic; |
| 453 |
#[inline(always)] |
| 454 |
unsafe fn instance() -> Self { |
| 455 |
Self |
| 456 |
} |
| 457 |
} |
| 458 |
|
| 459 |
zerocopy::cryptocorrosion_derive_traits! { |
| 460 |
#[repr(transparent)] |
| 461 |
#[derive(Copy, Clone, Debug, PartialEq)] |
| 462 |
pub struct u32x4_generic([u32; 4]); |
| 463 |
} |
| 464 |
|
| 465 |
zerocopy::cryptocorrosion_derive_traits! { |
| 466 |
#[repr(transparent)] |
| 467 |
#[derive(Copy, Clone, Debug, PartialEq)] |
| 468 |
pub struct u64x2_generic([u64; 2]); |
| 469 |
} |
| 470 |
|
| 471 |
zerocopy::cryptocorrosion_derive_traits! { |
| 472 |
#[repr(transparent)] |
| 473 |
#[derive(Copy, Clone, Debug, PartialEq)] |
| 474 |
pub struct u128x1_generic([u128; 1]); |
| 475 |
} |
| 476 |
|
| 477 |
impl From<u32x4_generic> for vec128_storage { |
| 478 |
#[inline(always)] |
| 479 |
fn from(d: u32x4_generic) -> Self { |
| 480 |
Self { d: d.0 } |
| 481 |
} |
| 482 |
} |
| 483 |
impl From<u64x2_generic> for vec128_storage { |
| 484 |
#[inline(always)] |
| 485 |
fn from(q: u64x2_generic) -> Self { |
| 486 |
Self { q: q.0 } |
| 487 |
} |
| 488 |
} |
| 489 |
impl From<u128x1_generic> for vec128_storage { |
| 490 |
#[inline(always)] |
| 491 |
fn from(o: u128x1_generic) -> Self { |
| 492 |
Self { q: q_of_o(o.0[0]) } |
| 493 |
} |
| 494 |
} |
| 495 |
|
| 496 |
impl Store<vec128_storage> for u32x4_generic { |
| 497 |
#[inline(always)] |
| 498 |
unsafe fn unpack(s: vec128_storage) -> Self { |
| 499 |
Self(s.d) |
| 500 |
} |
| 501 |
} |
| 502 |
impl Store<vec128_storage> for u64x2_generic { |
| 503 |
#[inline(always)] |
| 504 |
unsafe fn unpack(s: vec128_storage) -> Self { |
| 505 |
Self(s.q) |
| 506 |
} |
| 507 |
} |
| 508 |
impl Store<vec128_storage> for u128x1_generic { |
| 509 |
#[inline(always)] |
| 510 |
unsafe fn unpack(s: vec128_storage) -> Self { |
| 511 |
Self([o_of_q(s.q); 1]) |
| 512 |
} |
| 513 |
} |
| 514 |
|
| 515 |
impl ArithOps for u32x4_generic {} |
| 516 |
impl ArithOps for u64x2_generic {} |
| 517 |
impl ArithOps for u128x1_generic {} |
| 518 |
|
| 519 |
impl Add for u32x4_generic { |
| 520 |
type Output = Self; |
| 521 |
#[inline(always)] |
| 522 |
fn add(self, rhs: Self) -> Self::Output { |
| 523 |
dmap2(self, rhs, |x, y| x.wrapping_add(y)) |
| 524 |
} |
| 525 |
} |
| 526 |
impl Add for u64x2_generic { |
| 527 |
type Output = Self; |
| 528 |
#[inline(always)] |
| 529 |
fn add(self, rhs: Self) -> Self::Output { |
| 530 |
qmap2(self, rhs, |x, y| x.wrapping_add(y)) |
| 531 |
} |
| 532 |
} |
| 533 |
impl Add for u128x1_generic { |
| 534 |
type Output = Self; |
| 535 |
#[inline(always)] |
| 536 |
fn add(self, rhs: Self) -> Self::Output { |
| 537 |
omap2(self, rhs, |x, y| x.wrapping_add(y)) |
| 538 |
} |
| 539 |
} |
| 540 |
impl AddAssign for u32x4_generic { |
| 541 |
#[inline(always)] |
| 542 |
fn add_assign(&mut self, rhs: Self) { |
| 543 |
*self = *self + rhs |
| 544 |
} |
| 545 |
} |
| 546 |
impl AddAssign for u64x2_generic { |
| 547 |
#[inline(always)] |
| 548 |
fn add_assign(&mut self, rhs: Self) { |
| 549 |
*self = *self + rhs |
| 550 |
} |
| 551 |
} |
| 552 |
impl AddAssign for u128x1_generic { |
| 553 |
#[inline(always)] |
| 554 |
fn add_assign(&mut self, rhs: Self) { |
| 555 |
*self = *self + rhs |
| 556 |
} |
| 557 |
} |
| 558 |
impl BSwap for u32x4_generic { |
| 559 |
#[inline(always)] |
| 560 |
fn bswap(self) -> Self { |
| 561 |
dmap(self, |x| x.swap_bytes()) |
| 562 |
} |
| 563 |
} |
| 564 |
impl BSwap for u64x2_generic { |
| 565 |
#[inline(always)] |
| 566 |
fn bswap(self) -> Self { |
| 567 |
qmap(self, |x| x.swap_bytes()) |
| 568 |
} |
| 569 |
} |
| 570 |
impl BSwap for u128x1_generic { |
| 571 |
#[inline(always)] |
| 572 |
fn bswap(self) -> Self { |
| 573 |
omap(self, |x| x.swap_bytes()) |
| 574 |
} |
| 575 |
} |
| 576 |
impl StoreBytes for u32x4_generic { |
| 577 |
#[inline(always)] |
| 578 |
unsafe fn unsafe_read_le(input: &[u8]) -> Self { |
| 579 |
let x = u32x4_generic::read_from_bytes(input).unwrap(); |
| 580 |
dmap(x, |x| x.to_le()) |
| 581 |
} |
| 582 |
#[inline(always)] |
| 583 |
unsafe fn unsafe_read_be(input: &[u8]) -> Self { |
| 584 |
let x = u32x4_generic::read_from_bytes(input).unwrap(); |
| 585 |
dmap(x, |x| x.to_be()) |
| 586 |
} |
| 587 |
#[inline(always)] |
| 588 |
fn write_le(self, out: &mut [u8]) { |
| 589 |
let x = dmap(self, |x| x.to_le()); |
| 590 |
x.write_to(out).unwrap(); |
| 591 |
} |
| 592 |
#[inline(always)] |
| 593 |
fn write_be(self, out: &mut [u8]) { |
| 594 |
let x = dmap(self, |x| x.to_be()); |
| 595 |
x.write_to(out).unwrap(); |
| 596 |
} |
| 597 |
} |
| 598 |
impl StoreBytes for u64x2_generic { |
| 599 |
#[inline(always)] |
| 600 |
unsafe fn unsafe_read_le(input: &[u8]) -> Self { |
| 601 |
let x = u64x2_generic::read_from_bytes(input).unwrap(); |
| 602 |
qmap(x, |x| x.to_le()) |
| 603 |
} |
| 604 |
#[inline(always)] |
| 605 |
unsafe fn unsafe_read_be(input: &[u8]) -> Self { |
| 606 |
let x = u64x2_generic::read_from_bytes(input).unwrap(); |
| 607 |
qmap(x, |x| x.to_be()) |
| 608 |
} |
| 609 |
#[inline(always)] |
| 610 |
fn write_le(self, out: &mut [u8]) { |
| 611 |
let x = qmap(self, |x| x.to_le()); |
| 612 |
x.write_to(out).unwrap(); |
| 613 |
} |
| 614 |
#[inline(always)] |
| 615 |
fn write_be(self, out: &mut [u8]) { |
| 616 |
let x = qmap(self, |x| x.to_be()); |
| 617 |
x.write_to(out).unwrap(); |
| 618 |
} |
| 619 |
} |
| 620 |
|
| 621 |
#[derive(Copy, Clone)] |
| 622 |
pub struct G0; |
| 623 |
#[derive(Copy, Clone)] |
| 624 |
pub struct G1; |
| 625 |
pub type u32x4x2_generic = x2<u32x4_generic, G0>; |
| 626 |
pub type u64x2x2_generic = x2<u64x2_generic, G0>; |
| 627 |
pub type u64x4_generic = x2<u64x2_generic, G1>; |
| 628 |
pub type u128x2_generic = x2<u128x1_generic, G0>; |
| 629 |
pub type u32x4x4_generic = x4<u32x4_generic>; |
| 630 |
pub type u64x2x4_generic = x4<u64x2_generic>; |
| 631 |
pub type u128x4_generic = x4<u128x1_generic>; |
| 632 |
|
| 633 |
impl Vector<[u32; 16]> for u32x4x4_generic { |
| 634 |
fn to_scalars(self) -> [u32; 16] { |
| 635 |
let [a, b, c, d] = self.0; |
| 636 |
let a = a.0; |
| 637 |
let b = b.0; |
| 638 |
let c = c.0; |
| 639 |
let d = d.0; |
| 640 |
[ |
| 641 |
a[0], a[1], a[2], a[3], // |
| 642 |
b[0], b[1], b[2], b[3], // |
| 643 |
c[0], c[1], c[2], c[3], // |
| 644 |
d[0], d[1], d[2], d[3], // |
| 645 |
] |
| 646 |
} |
| 647 |
} |
| 648 |
|
| 649 |
impl MultiLane<[u32; 4]> for u32x4_generic { |
| 650 |
#[inline(always)] |
| 651 |
fn to_lanes(self) -> [u32; 4] { |
| 652 |
self.0 |
| 653 |
} |
| 654 |
#[inline(always)] |
| 655 |
fn from_lanes(xs: [u32; 4]) -> Self { |
| 656 |
Self(xs) |
| 657 |
} |
| 658 |
} |
| 659 |
impl MultiLane<[u64; 2]> for u64x2_generic { |
| 660 |
#[inline(always)] |
| 661 |
fn to_lanes(self) -> [u64; 2] { |
| 662 |
self.0 |
| 663 |
} |
| 664 |
#[inline(always)] |
| 665 |
fn from_lanes(xs: [u64; 2]) -> Self { |
| 666 |
Self(xs) |
| 667 |
} |
| 668 |
} |
| 669 |
impl MultiLane<[u64; 4]> for u64x4_generic { |
| 670 |
#[inline(always)] |
| 671 |
fn to_lanes(self) -> [u64; 4] { |
| 672 |
let (a, b) = (self.0[0].to_lanes(), self.0[1].to_lanes()); |
| 673 |
[a[0], a[1], b[0], b[1]] |
| 674 |
} |
| 675 |
#[inline(always)] |
| 676 |
fn from_lanes(xs: [u64; 4]) -> Self { |
| 677 |
let (a, b) = ( |
| 678 |
u64x2_generic::from_lanes([xs[0], xs[1]]), |
| 679 |
u64x2_generic::from_lanes([xs[2], xs[3]]), |
| 680 |
); |
| 681 |
x2::new([a, b]) |
| 682 |
} |
| 683 |
} |
| 684 |
impl MultiLane<[u128; 1]> for u128x1_generic { |
| 685 |
#[inline(always)] |
| 686 |
fn to_lanes(self) -> [u128; 1] { |
| 687 |
self.0 |
| 688 |
} |
| 689 |
#[inline(always)] |
| 690 |
fn from_lanes(xs: [u128; 1]) -> Self { |
| 691 |
Self(xs) |
| 692 |
} |
| 693 |
} |
| 694 |
impl Vec4<u32> for u32x4_generic { |
| 695 |
#[inline(always)] |
| 696 |
fn extract(self, i: u32) -> u32 { |
| 697 |
self.0[i as usize] |
| 698 |
} |
| 699 |
#[inline(always)] |
| 700 |
fn insert(mut self, v: u32, i: u32) -> Self { |
| 701 |
self.0[i as usize] = v; |
| 702 |
self |
| 703 |
} |
| 704 |
} |
| 705 |
impl Vec4<u64> for u64x4_generic { |
| 706 |
#[inline(always)] |
| 707 |
fn extract(self, i: u32) -> u64 { |
| 708 |
let d: [u64; 4] = self.to_lanes(); |
| 709 |
d[i as usize] |
| 710 |
} |
| 711 |
#[inline(always)] |
| 712 |
fn insert(self, v: u64, i: u32) -> Self { |
| 713 |
self.0[(i / 2) as usize].insert(v, i % 2); |
| 714 |
self |
| 715 |
} |
| 716 |
} |
| 717 |
impl Vec2<u64> for u64x2_generic { |
| 718 |
#[inline(always)] |
| 719 |
fn extract(self, i: u32) -> u64 { |
| 720 |
self.0[i as usize] |
| 721 |
} |
| 722 |
#[inline(always)] |
| 723 |
fn insert(mut self, v: u64, i: u32) -> Self { |
| 724 |
self.0[i as usize] = v; |
| 725 |
self |
| 726 |
} |
| 727 |
} |
| 728 |
|
| 729 |
impl Words4 for u32x4_generic { |
| 730 |
#[inline(always)] |
| 731 |
fn shuffle2301(self) -> Self { |
| 732 |
self.swap64() |
| 733 |
} |
| 734 |
#[inline(always)] |
| 735 |
fn shuffle1230(self) -> Self { |
| 736 |
let x = self.0; |
| 737 |
Self([x[3], x[0], x[1], x[2]]) |
| 738 |
} |
| 739 |
#[inline(always)] |
| 740 |
fn shuffle3012(self) -> Self { |
| 741 |
let x = self.0; |
| 742 |
Self([x[1], x[2], x[3], x[0]]) |
| 743 |
} |
| 744 |
} |
| 745 |
impl LaneWords4 for u32x4_generic { |
| 746 |
#[inline(always)] |
| 747 |
fn shuffle_lane_words2301(self) -> Self { |
| 748 |
self.shuffle2301() |
| 749 |
} |
| 750 |
#[inline(always)] |
| 751 |
fn shuffle_lane_words1230(self) -> Self { |
| 752 |
self.shuffle1230() |
| 753 |
} |
| 754 |
#[inline(always)] |
| 755 |
fn shuffle_lane_words3012(self) -> Self { |
| 756 |
self.shuffle3012() |
| 757 |
} |
| 758 |
} |
| 759 |
|
| 760 |
impl Words4 for u64x4_generic { |
| 761 |
#[inline(always)] |
| 762 |
fn shuffle2301(self) -> Self { |
| 763 |
x2::new([self.0[1], self.0[0]]) |
| 764 |
} |
| 765 |
#[inline(always)] |
| 766 |
fn shuffle1230(self) -> Self { |
| 767 |
unimplemented!() |
| 768 |
} |
| 769 |
#[inline(always)] |
| 770 |
fn shuffle3012(self) -> Self { |
| 771 |
unimplemented!() |
| 772 |
} |
| 773 |
} |
| 774 |
|
| 775 |
impl u32x4<GenericMachine> for u32x4_generic {} |
| 776 |
impl u64x2<GenericMachine> for u64x2_generic {} |
| 777 |
impl u128x1<GenericMachine> for u128x1_generic {} |
| 778 |
impl u32x4x2<GenericMachine> for u32x4x2_generic {} |
| 779 |
impl u64x2x2<GenericMachine> for u64x2x2_generic {} |
| 780 |
impl u64x4<GenericMachine> for u64x4_generic {} |
| 781 |
impl u128x2<GenericMachine> for u128x2_generic {} |
| 782 |
impl u32x4x4<GenericMachine> for u32x4x4_generic {} |
| 783 |
impl u64x2x4<GenericMachine> for u64x2x4_generic {} |
| 784 |
impl u128x4<GenericMachine> for u128x4_generic {} |
| 785 |
|
| 786 |
#[macro_export] |
| 787 |
macro_rules! dispatch { |
| 788 |
($mach:ident, $MTy:ident, { $([$pub:tt$(($krate:tt))*])* fn $name:ident($($arg:ident: $argty:ty),* $(,)*) -> $ret:ty $body:block }) => { |
| 789 |
#[inline(always)] |
| 790 |
$($pub$(($krate))*)* fn $name($($arg: $argty),*) -> $ret { |
| 791 |
let $mach = unsafe { $crate::generic::GenericMachine::instance() }; |
| 792 |
#[inline(always)] |
| 793 |
fn fn_impl<$MTy: $crate::Machine>($mach: $MTy, $($arg: $argty),*) -> $ret $body |
| 794 |
fn_impl($mach, $($arg),*) |
| 795 |
} |
| 796 |
}; |
| 797 |
($mach:ident, $MTy:ident, { $([$pub:tt $(($krate:tt))*])* fn $name:ident($($arg:ident: $argty:ty),* $(,)*) $body:block }) => { |
| 798 |
dispatch!($mach, $MTy, { |
| 799 |
$([$pub $(($krate))*])* fn $name($($arg: $argty),*) -> () $body |
| 800 |
}); |
| 801 |
} |
| 802 |
} |
| 803 |
#[macro_export] |
| 804 |
macro_rules! dispatch_light128 { |
| 805 |
($mach:ident, $MTy:ident, { $([$pub:tt$(($krate:tt))*])* fn $name:ident($($arg:ident: $argty:ty),* $(,)*) -> $ret:ty $body:block }) => { |
| 806 |
#[inline(always)] |
| 807 |
$($pub$(($krate))*)* fn $name($($arg: $argty),*) -> $ret { |
| 808 |
let $mach = unsafe { $crate::generic::GenericMachine::instance() }; |
| 809 |
#[inline(always)] |
| 810 |
fn fn_impl<$MTy: $crate::Machine>($mach: $MTy, $($arg: $argty),*) -> $ret $body |
| 811 |
fn_impl($mach, $($arg),*) |
| 812 |
} |
| 813 |
}; |
| 814 |
($mach:ident, $MTy:ident, { $([$pub:tt $(($krate:tt))*])* fn $name:ident($($arg:ident: $argty:ty),* $(,)*) $body:block }) => { |
| 815 |
dispatch!($mach, $MTy, { |
| 816 |
$([$pub $(($krate))*])* fn $name($($arg: $argty),*) -> () $body |
| 817 |
}); |
| 818 |
} |
| 819 |
} |
| 820 |
#[macro_export] |
| 821 |
macro_rules! dispatch_light256 { |
| 822 |
($mach:ident, $MTy:ident, { $([$pub:tt$(($krate:tt))*])* fn $name:ident($($arg:ident: $argty:ty),* $(,)*) -> $ret:ty $body:block }) => { |
| 823 |
#[inline(always)] |
| 824 |
$($pub$(($krate))*)* fn $name($($arg: $argty),*) -> $ret { |
| 825 |
let $mach = unsafe { $crate::generic::GenericMachine::instance() }; |
| 826 |
#[inline(always)] |
| 827 |
fn fn_impl<$MTy: $crate::Machine>($mach: $MTy, $($arg: $argty),*) -> $ret $body |
| 828 |
fn_impl($mach, $($arg),*) |
| 829 |
} |
| 830 |
}; |
| 831 |
($mach:ident, $MTy:ident, { $([$pub:tt $(($krate:tt))*])* fn $name:ident($($arg:ident: $argty:ty),* $(,)*) $body:block }) => { |
| 832 |
dispatch!($mach, $MTy, { |
| 833 |
$([$pub $(($krate))*])* fn $name($($arg: $argty),*) -> () $body |
| 834 |
}); |
| 835 |
} |
| 836 |
} |
| 837 |
#[macro_export] |
| 838 |
macro_rules! dispatch_light512 { |
| 839 |
($mach:ident, $MTy:ident, { $([$pub:tt$(($krate:tt))*])* fn $name:ident($($arg:ident: $argty:ty),* $(,)*) -> $ret:ty $body:block }) => { |
| 840 |
#[inline(always)] |
| 841 |
$($pub$(($krate))*)* fn $name($($arg: $argty),*) -> $ret { |
| 842 |
let $mach = unsafe { $crate::generic::GenericMachine::instance() }; |
| 843 |
#[inline(always)] |
| 844 |
fn fn_impl<$MTy: $crate::Machine>($mach: $MTy, $($arg: $argty),*) -> $ret $body |
| 845 |
fn_impl($mach, $($arg),*) |
| 846 |
} |
| 847 |
}; |
| 848 |
($mach:ident, $MTy:ident, { $([$pub:tt $(($krate:tt))*])* fn $name:ident($($arg:ident: $argty:ty),* $(,)*) $body:block }) => { |
| 849 |
dispatch!($mach, $MTy, { |
| 850 |
$([$pub $(($krate))*])* fn $name($($arg: $argty),*) -> () $body |
| 851 |
}); |
| 852 |
} |
| 853 |
} |
| 854 |
|
| 855 |
#[cfg(test)] |
| 856 |
mod test { |
| 857 |
use super::*; |
| 858 |
|
| 859 |
#[test] |
| 860 |
fn test_bswap32() { |
| 861 |
let xs = [0x0f0e_0d0c, 0x0b0a_0908, 0x0706_0504, 0x0302_0100]; |
| 862 |
let ys = [0x0c0d_0e0f, 0x0809_0a0b, 0x0405_0607, 0x0001_0203]; |
| 863 |
|
| 864 |
let m = unsafe { GenericMachine::instance() }; |
| 865 |
|
| 866 |
let x: <GenericMachine as Machine>::u32x4 = m.vec(xs); |
| 867 |
let x = x.bswap(); |
| 868 |
|
| 869 |
let y = m.vec(ys); |
| 870 |
assert_eq!(x, y); |
| 871 |
} |
| 872 |
} |