| 1 | // pathfinder/simd/src/scalar.rs |
| 2 | // |
| 3 | // Copyright © 2019 The Pathfinder Project Developers. |
| 4 | // |
| 5 | // Licensed under the Apache License, Version 2.0 <LICENSE-APACHE or |
| 6 | // http://www.apache.org/licenses/LICENSE-2.0> or the MIT license |
| 7 | // <LICENSE-MIT or http://opensource.org/licenses/MIT>, at your |
| 8 | // option. This file may not be copied, modified, or distributed |
| 9 | // except according to those terms. |
| 10 | |
| 11 | use std::f32; |
| 12 | use std::fmt::{self, Debug, Formatter}; |
| 13 | use std::ops::{Add, BitAnd, BitOr, Div, Index, IndexMut, Mul, Shr, Sub, Not}; |
| 14 | |
| 15 | mod swizzle_f32x4; |
| 16 | mod swizzle_i32x4; |
| 17 | |
| 18 | // Two 32-bit floats |
| 19 | |
| 20 | #[derive (Clone, Copy, Debug, Default, PartialEq)] |
| 21 | pub struct F32x2(pub [f32; 2]); |
| 22 | |
| 23 | impl F32x2 { |
| 24 | // Constructors |
| 25 | |
| 26 | #[inline ] |
| 27 | pub fn new(a: f32, b: f32) -> F32x2 { |
| 28 | F32x2([a, b]) |
| 29 | } |
| 30 | |
| 31 | #[inline ] |
| 32 | pub fn splat(x: f32) -> F32x2 { |
| 33 | F32x2([x, x]) |
| 34 | } |
| 35 | |
| 36 | // Basic operations |
| 37 | |
| 38 | #[inline ] |
| 39 | pub fn approx_recip(self) -> F32x2 { |
| 40 | F32x2([1.0 / self[0], 1.0 / self[1]]) |
| 41 | } |
| 42 | |
| 43 | #[inline ] |
| 44 | pub fn min(self, other: F32x2) -> F32x2 { |
| 45 | F32x2([f32::min(self[0], other[0]), f32::min(self[1], other[1])]) |
| 46 | } |
| 47 | |
| 48 | #[inline ] |
| 49 | pub fn max(self, other: F32x2) -> F32x2 { |
| 50 | F32x2([f32::max(self[0], other[0]), f32::max(self[1], other[1])]) |
| 51 | } |
| 52 | |
| 53 | #[inline ] |
| 54 | pub fn clamp(self, min: F32x2, max: F32x2) -> F32x2 { |
| 55 | self.max(min).min(max) |
| 56 | } |
| 57 | |
| 58 | #[inline ] |
| 59 | pub fn abs(self) -> F32x2 { |
| 60 | F32x2([self[0].abs(), self[1].abs()]) |
| 61 | } |
| 62 | |
| 63 | #[inline ] |
| 64 | pub fn floor(self) -> F32x2 { |
| 65 | F32x2([self[0].floor(), self[1].floor()]) |
| 66 | } |
| 67 | |
| 68 | #[inline ] |
| 69 | pub fn ceil(self) -> F32x2 { |
| 70 | F32x2([self[0].ceil(), self[1].ceil()]) |
| 71 | } |
| 72 | |
| 73 | #[inline ] |
| 74 | pub fn sqrt(self) -> F32x2 { |
| 75 | F32x2([self[0].sqrt(), self[1].sqrt()]) |
| 76 | } |
| 77 | |
| 78 | // Packed comparisons |
| 79 | |
| 80 | #[inline ] |
| 81 | pub fn packed_eq(self, other: F32x2) -> U32x2 { |
| 82 | U32x2([ |
| 83 | if self[0] == other[0] { !0 } else { 0 }, |
| 84 | if self[1] == other[1] { !0 } else { 0 }, |
| 85 | ]) |
| 86 | } |
| 87 | |
| 88 | #[inline ] |
| 89 | pub fn packed_gt(self, other: F32x2) -> U32x2 { |
| 90 | U32x2([ |
| 91 | if self[0] > other[0] { !0 } else { 0 }, |
| 92 | if self[1] > other[1] { !0 } else { 0 }, |
| 93 | ]) |
| 94 | } |
| 95 | |
| 96 | #[inline ] |
| 97 | pub fn packed_lt(self, other: F32x2) -> U32x2 { |
| 98 | U32x2([ |
| 99 | if self[0] < other[0] { !0 } else { 0 }, |
| 100 | if self[1] < other[1] { !0 } else { 0 }, |
| 101 | ]) |
| 102 | } |
| 103 | |
| 104 | #[inline ] |
| 105 | pub fn packed_le(self, other: F32x2) -> U32x2 { |
| 106 | U32x2([ |
| 107 | if self[0] <= other[0] { !0 } else { 0 }, |
| 108 | if self[1] <= other[1] { !0 } else { 0 }, |
| 109 | ]) |
| 110 | } |
| 111 | |
| 112 | // Conversions |
| 113 | |
| 114 | #[inline ] |
| 115 | pub fn to_f32x4(self) -> F32x4 { |
| 116 | F32x4([self[0] as f32, self[1] as f32, 0.0, 0.0]) |
| 117 | } |
| 118 | |
| 119 | /// Converts these packed floats to integers via rounding. |
| 120 | #[inline ] |
| 121 | pub fn to_i32x2(self) -> I32x2 { |
| 122 | I32x2([self[0].round() as i32, self[1].round() as i32]) |
| 123 | } |
| 124 | |
| 125 | /// Converts these packed floats to integers via rounding. |
| 126 | #[inline ] |
| 127 | pub fn to_i32x4(self) -> I32x4 { |
| 128 | I32x4([self[0].round() as i32, self[1].round() as i32, 0, 0]) |
| 129 | } |
| 130 | |
| 131 | // Swizzle |
| 132 | |
| 133 | #[inline ] |
| 134 | pub fn yx(self) -> F32x2 { |
| 135 | F32x2([self[1], self[0]]) |
| 136 | } |
| 137 | |
| 138 | // Concatenations |
| 139 | |
| 140 | #[inline ] |
| 141 | pub fn concat_xy_xy(self, other: F32x2) -> F32x4 { |
| 142 | F32x4([self[0], self[1], other[0], other[1]]) |
| 143 | } |
| 144 | } |
| 145 | |
| 146 | impl Index<usize> for F32x2 { |
| 147 | type Output = f32; |
| 148 | #[inline ] |
| 149 | fn index(&self, index: usize) -> &f32 { |
| 150 | &self.0[index] |
| 151 | } |
| 152 | } |
| 153 | |
| 154 | impl IndexMut<usize> for F32x2 { |
| 155 | #[inline ] |
| 156 | fn index_mut(&mut self, index: usize) -> &mut f32 { |
| 157 | &mut self.0[index] |
| 158 | } |
| 159 | } |
| 160 | |
| 161 | impl Add<F32x2> for F32x2 { |
| 162 | type Output = F32x2; |
| 163 | #[inline ] |
| 164 | fn add(self, other: F32x2) -> F32x2 { |
| 165 | F32x2([self[0] + other[0], self[1] + other[1]]) |
| 166 | } |
| 167 | } |
| 168 | |
| 169 | impl Div<F32x2> for F32x2 { |
| 170 | type Output = F32x2; |
| 171 | #[inline ] |
| 172 | fn div(self, other: F32x2) -> F32x2 { |
| 173 | F32x2([self[0] / other[0], self[1] / other[1]]) |
| 174 | } |
| 175 | } |
| 176 | |
| 177 | impl Mul<F32x2> for F32x2 { |
| 178 | type Output = F32x2; |
| 179 | #[inline ] |
| 180 | fn mul(self, other: F32x2) -> F32x2 { |
| 181 | F32x2([self[0] * other[0], self[1] * other[1]]) |
| 182 | } |
| 183 | } |
| 184 | |
| 185 | impl Sub<F32x2> for F32x2 { |
| 186 | type Output = F32x2; |
| 187 | #[inline ] |
| 188 | fn sub(self, other: F32x2) -> F32x2 { |
| 189 | F32x2([self[0] - other[0], self[1] - other[1]]) |
| 190 | } |
| 191 | } |
| 192 | |
| 193 | // Four 32-bit floats |
| 194 | |
| 195 | #[derive (Clone, Copy, Default, PartialEq)] |
| 196 | pub struct F32x4(pub [f32; 4]); |
| 197 | |
| 198 | impl F32x4 { |
| 199 | #[inline ] |
| 200 | pub fn new(a: f32, b: f32, c: f32, d: f32) -> F32x4 { |
| 201 | F32x4([a, b, c, d]) |
| 202 | } |
| 203 | |
| 204 | #[inline ] |
| 205 | pub fn splat(x: f32) -> F32x4 { |
| 206 | F32x4([x; 4]) |
| 207 | } |
| 208 | |
| 209 | // Basic operations |
| 210 | |
| 211 | #[inline ] |
| 212 | pub fn approx_recip(self) -> F32x4 { |
| 213 | F32x4([1.0 / self[0], 1.0 / self[1], 1.0 / self[2], 1.0 / self[3]]) |
| 214 | } |
| 215 | |
| 216 | #[inline ] |
| 217 | pub fn min(self, other: F32x4) -> F32x4 { |
| 218 | F32x4([ |
| 219 | self[0].min(other[0]), |
| 220 | self[1].min(other[1]), |
| 221 | self[2].min(other[2]), |
| 222 | self[3].min(other[3]), |
| 223 | ]) |
| 224 | } |
| 225 | |
| 226 | #[inline ] |
| 227 | pub fn max(self, other: F32x4) -> F32x4 { |
| 228 | F32x4([ |
| 229 | self[0].max(other[0]), |
| 230 | self[1].max(other[1]), |
| 231 | self[2].max(other[2]), |
| 232 | self[3].max(other[3]), |
| 233 | ]) |
| 234 | } |
| 235 | |
| 236 | #[inline ] |
| 237 | pub fn clamp(self, min: F32x4, max: F32x4) -> F32x4 { |
| 238 | self.max(min).min(max) |
| 239 | } |
| 240 | |
| 241 | #[inline ] |
| 242 | pub fn abs(self) -> F32x4 { |
| 243 | F32x4([self[0].abs(), self[1].abs(), self[2].abs(), self[3].abs()]) |
| 244 | } |
| 245 | |
| 246 | #[inline ] |
| 247 | pub fn floor(self) -> F32x4 { |
| 248 | F32x4([ |
| 249 | self[0].floor(), |
| 250 | self[1].floor(), |
| 251 | self[2].floor(), |
| 252 | self[3].floor(), |
| 253 | ]) |
| 254 | } |
| 255 | |
| 256 | #[inline ] |
| 257 | pub fn ceil(self) -> F32x4 { |
| 258 | F32x4([ |
| 259 | self[0].ceil(), |
| 260 | self[1].ceil(), |
| 261 | self[2].ceil(), |
| 262 | self[3].ceil(), |
| 263 | ]) |
| 264 | } |
| 265 | |
| 266 | #[inline ] |
| 267 | pub fn sqrt(self) -> F32x4 { |
| 268 | F32x4([ |
| 269 | self[0].sqrt(), |
| 270 | self[1].sqrt(), |
| 271 | self[2].sqrt(), |
| 272 | self[3].sqrt(), |
| 273 | ]) |
| 274 | } |
| 275 | |
| 276 | // Packed comparisons |
| 277 | |
| 278 | #[inline ] |
| 279 | pub fn packed_eq(self, other: F32x4) -> U32x4 { |
| 280 | U32x4([ |
| 281 | if self[0] == other[0] { !0 } else { 0 }, |
| 282 | if self[1] == other[1] { !0 } else { 0 }, |
| 283 | if self[2] == other[2] { !0 } else { 0 }, |
| 284 | if self[3] == other[3] { !0 } else { 0 }, |
| 285 | ]) |
| 286 | } |
| 287 | |
| 288 | #[inline ] |
| 289 | pub fn packed_gt(self, other: F32x4) -> U32x4 { |
| 290 | U32x4([ |
| 291 | if self[0] > other[0] { !0 } else { 0 }, |
| 292 | if self[1] > other[1] { !0 } else { 0 }, |
| 293 | if self[2] > other[2] { !0 } else { 0 }, |
| 294 | if self[3] > other[3] { !0 } else { 0 }, |
| 295 | ]) |
| 296 | } |
| 297 | |
| 298 | #[inline ] |
| 299 | pub fn packed_le(self, other: F32x4) -> U32x4 { |
| 300 | U32x4([ |
| 301 | if self[0] <= other[0] { !0 } else { 0 }, |
| 302 | if self[1] <= other[1] { !0 } else { 0 }, |
| 303 | if self[2] <= other[2] { !0 } else { 0 }, |
| 304 | if self[3] <= other[3] { !0 } else { 0 }, |
| 305 | ]) |
| 306 | } |
| 307 | |
| 308 | #[inline ] |
| 309 | pub fn packed_lt(self, other: F32x4) -> U32x4 { |
| 310 | U32x4([ |
| 311 | if self[0] < other[0] { !0 } else { 0 }, |
| 312 | if self[1] < other[1] { !0 } else { 0 }, |
| 313 | if self[2] < other[2] { !0 } else { 0 }, |
| 314 | if self[3] < other[3] { !0 } else { 0 }, |
| 315 | ]) |
| 316 | } |
| 317 | |
| 318 | /// Converts these packed floats to integers via rounding. |
| 319 | #[inline ] |
| 320 | pub fn to_i32x4(self) -> I32x4 { |
| 321 | I32x4([ |
| 322 | self[0].round() as i32, |
| 323 | self[1].round() as i32, |
| 324 | self[2].round() as i32, |
| 325 | self[3].round() as i32, |
| 326 | ]) |
| 327 | } |
| 328 | |
| 329 | // Swizzle conversions |
| 330 | |
| 331 | #[inline ] |
| 332 | pub fn xy(self) -> F32x2 { |
| 333 | F32x2([self[0], self[1]]) |
| 334 | } |
| 335 | |
| 336 | #[inline ] |
| 337 | pub fn xw(self) -> F32x2 { |
| 338 | F32x2([self[0], self[3]]) |
| 339 | } |
| 340 | |
| 341 | #[inline ] |
| 342 | pub fn yx(self) -> F32x2 { |
| 343 | F32x2([self[1], self[0]]) |
| 344 | } |
| 345 | |
| 346 | #[inline ] |
| 347 | pub fn zy(self) -> F32x2 { |
| 348 | F32x2([self[2], self[1]]) |
| 349 | } |
| 350 | |
| 351 | #[inline ] |
| 352 | pub fn zw(self) -> F32x2 { |
| 353 | F32x2([self[2], self[3]]) |
| 354 | } |
| 355 | |
| 356 | // Concatenations |
| 357 | |
| 358 | #[inline ] |
| 359 | pub fn concat_xy_xy(self, other: F32x4) -> F32x4 { |
| 360 | F32x4([self[0], self[1], other[0], other[1]]) |
| 361 | } |
| 362 | |
| 363 | #[inline ] |
| 364 | pub fn concat_xy_zw(self, other: F32x4) -> F32x4 { |
| 365 | F32x4([self[0], self[1], other[2], other[3]]) |
| 366 | } |
| 367 | |
| 368 | #[inline ] |
| 369 | pub fn concat_zw_zw(self, other: F32x4) -> F32x4 { |
| 370 | F32x4([self[2], self[3], other[2], other[3]]) |
| 371 | } |
| 372 | |
| 373 | #[inline ] |
| 374 | pub fn concat_wz_yx(self, other: F32x4) -> F32x4 { |
| 375 | F32x4([self[3], self[2], other[1], other[0]]) |
| 376 | } |
| 377 | } |
| 378 | |
| 379 | impl Index<usize> for F32x4 { |
| 380 | type Output = f32; |
| 381 | #[inline ] |
| 382 | fn index(&self, index: usize) -> &f32 { |
| 383 | &self.0[index] |
| 384 | } |
| 385 | } |
| 386 | |
| 387 | impl IndexMut<usize> for F32x4 { |
| 388 | #[inline ] |
| 389 | fn index_mut(&mut self, index: usize) -> &mut f32 { |
| 390 | &mut self.0[index] |
| 391 | } |
| 392 | } |
| 393 | |
| 394 | impl Debug for F32x4 { |
| 395 | #[inline ] |
| 396 | fn fmt(&self, f: &mut Formatter) -> Result<(), fmt::Error> { |
| 397 | write!(f, "< {}, {}, {}, {}>" , self[0], self[1], self[2], self[3]) |
| 398 | } |
| 399 | } |
| 400 | |
| 401 | impl Add<F32x4> for F32x4 { |
| 402 | type Output = F32x4; |
| 403 | #[inline ] |
| 404 | fn add(self, other: F32x4) -> F32x4 { |
| 405 | F32x4([ |
| 406 | self[0] + other[0], |
| 407 | self[1] + other[1], |
| 408 | self[2] + other[2], |
| 409 | self[3] + other[3], |
| 410 | ]) |
| 411 | } |
| 412 | } |
| 413 | |
| 414 | impl Div<F32x4> for F32x4 { |
| 415 | type Output = F32x4; |
| 416 | #[inline ] |
| 417 | fn div(self, other: F32x4) -> F32x4 { |
| 418 | F32x4([ |
| 419 | self[0] / other[0], |
| 420 | self[1] / other[1], |
| 421 | self[2] / other[2], |
| 422 | self[3] / other[3], |
| 423 | ]) |
| 424 | } |
| 425 | } |
| 426 | |
| 427 | impl Mul<F32x4> for F32x4 { |
| 428 | type Output = F32x4; |
| 429 | #[inline ] |
| 430 | fn mul(self, other: F32x4) -> F32x4 { |
| 431 | F32x4([ |
| 432 | self[0] * other[0], |
| 433 | self[1] * other[1], |
| 434 | self[2] * other[2], |
| 435 | self[3] * other[3], |
| 436 | ]) |
| 437 | } |
| 438 | } |
| 439 | |
| 440 | impl Sub<F32x4> for F32x4 { |
| 441 | type Output = F32x4; |
| 442 | #[inline ] |
| 443 | fn sub(self, other: F32x4) -> F32x4 { |
| 444 | F32x4([ |
| 445 | self[0] - other[0], |
| 446 | self[1] - other[1], |
| 447 | self[2] - other[2], |
| 448 | self[3] - other[3], |
| 449 | ]) |
| 450 | } |
| 451 | } |
| 452 | |
| 453 | // Two 32-bit signed integers |
| 454 | |
| 455 | #[derive (Clone, Copy, Default, Debug, PartialEq)] |
| 456 | pub struct I32x2([i32; 2]); |
| 457 | |
| 458 | impl I32x2 { |
| 459 | #[inline ] |
| 460 | pub fn new(x: i32, y: i32) -> I32x2 { |
| 461 | I32x2([x, y]) |
| 462 | } |
| 463 | |
| 464 | #[inline ] |
| 465 | pub fn splat(x: i32) -> I32x2 { |
| 466 | I32x2([x, x]) |
| 467 | } |
| 468 | |
| 469 | // Accessors |
| 470 | |
| 471 | #[inline ] |
| 472 | pub fn x(self) -> i32 { |
| 473 | self[0] |
| 474 | } |
| 475 | |
| 476 | #[inline ] |
| 477 | pub fn y(self) -> i32 { |
| 478 | self[1] |
| 479 | } |
| 480 | |
| 481 | #[inline ] |
| 482 | pub fn concat_xy_xy(self, other: I32x2) -> I32x4 { |
| 483 | I32x4([self[0], self[1], other[0], other[1]]) |
| 484 | } |
| 485 | |
| 486 | #[inline ] |
| 487 | pub fn min(self, other: I32x2) -> I32x2 { |
| 488 | I32x2([ |
| 489 | self[0].min(other[0]), |
| 490 | self[1].min(other[1]), |
| 491 | ]) |
| 492 | } |
| 493 | |
| 494 | #[inline ] |
| 495 | pub fn max(self, other: I32x2) -> I32x2 { |
| 496 | I32x2([ |
| 497 | self[0].max(other[0]), |
| 498 | self[1].max(other[1]), |
| 499 | ]) |
| 500 | } |
| 501 | |
| 502 | // Packed comparisons |
| 503 | |
| 504 | #[inline ] |
| 505 | pub fn packed_eq(self, other: I32x2) -> U32x2 { |
| 506 | U32x2([ |
| 507 | if self[0] == other[0] { !0 } else { 0 }, |
| 508 | if self[1] == other[1] { !0 } else { 0 }, |
| 509 | ]) |
| 510 | } |
| 511 | |
| 512 | #[inline ] |
| 513 | pub fn packed_gt(self, other: I32x2) -> U32x2 { |
| 514 | U32x2([ |
| 515 | if self[0] > other[0] { !0 } else { 0 }, |
| 516 | if self[1] > other[1] { !0 } else { 0 }, |
| 517 | ]) |
| 518 | } |
| 519 | |
| 520 | #[inline ] |
| 521 | pub fn packed_le(self, other: I32x2) -> U32x2 { |
| 522 | U32x2([ |
| 523 | if self[0] <= other[0] { !0 } else { 0 }, |
| 524 | if self[1] <= other[1] { !0 } else { 0 }, |
| 525 | ]) |
| 526 | } |
| 527 | |
| 528 | #[inline ] |
| 529 | pub fn packed_lt(self, other: I32x2) -> U32x2 { |
| 530 | U32x2([ |
| 531 | if self[0] < other[0] { !0 } else { 0 }, |
| 532 | if self[1] < other[1] { !0 } else { 0 }, |
| 533 | ]) |
| 534 | } |
| 535 | |
| 536 | // Conversions |
| 537 | |
| 538 | /// Converts these packed integers to floats. |
| 539 | #[inline ] |
| 540 | pub fn to_f32x2(self) -> F32x2 { |
| 541 | F32x2([self[0] as f32, self[1] as f32]) |
| 542 | } |
| 543 | } |
| 544 | |
| 545 | impl Index<usize> for I32x2 { |
| 546 | type Output = i32; |
| 547 | #[inline ] |
| 548 | fn index(&self, index: usize) -> &i32 { |
| 549 | &self.0[index] |
| 550 | } |
| 551 | } |
| 552 | |
| 553 | impl IndexMut<usize> for I32x2 { |
| 554 | #[inline ] |
| 555 | fn index_mut(&mut self, index: usize) -> &mut i32 { |
| 556 | &mut self.0[index] |
| 557 | } |
| 558 | } |
| 559 | |
| 560 | impl Add<I32x2> for I32x2 { |
| 561 | type Output = I32x2; |
| 562 | #[inline ] |
| 563 | fn add(self, other: I32x2) -> I32x2 { |
| 564 | I32x2([self[0] + other[0], self[1] + other[1]]) |
| 565 | } |
| 566 | } |
| 567 | |
| 568 | impl Sub<I32x2> for I32x2 { |
| 569 | type Output = I32x2; |
| 570 | #[inline ] |
| 571 | fn sub(self, other: I32x2) -> I32x2 { |
| 572 | I32x2([self[0] - other[0], self[1] - other[1]]) |
| 573 | } |
| 574 | } |
| 575 | |
| 576 | impl Mul<I32x2> for I32x2 { |
| 577 | type Output = I32x2; |
| 578 | #[inline ] |
| 579 | fn mul(self, other: I32x2) -> I32x2 { |
| 580 | I32x2([self[0] * other[0], self[1] * other[1]]) |
| 581 | } |
| 582 | } |
| 583 | |
| 584 | // Four 32-bit signed integers |
| 585 | |
| 586 | #[derive (Clone, Copy, Default, Debug, PartialEq)] |
| 587 | pub struct I32x4([i32; 4]); |
| 588 | |
| 589 | impl I32x4 { |
| 590 | #[inline ] |
| 591 | pub fn new(a: i32, b: i32, c: i32, d: i32) -> I32x4 { |
| 592 | I32x4([a, b, c, d]) |
| 593 | } |
| 594 | |
| 595 | #[inline ] |
| 596 | pub fn splat(x: i32) -> I32x4 { |
| 597 | I32x4([x; 4]) |
| 598 | } |
| 599 | |
| 600 | // Basic operations |
| 601 | |
| 602 | #[inline ] |
| 603 | pub fn min(self, other: I32x4) -> I32x4 { |
| 604 | I32x4([ |
| 605 | self[0].min(other[0]), |
| 606 | self[1].min(other[1]), |
| 607 | self[2].min(other[2]), |
| 608 | self[3].min(other[3]), |
| 609 | ]) |
| 610 | } |
| 611 | |
| 612 | #[inline ] |
| 613 | pub fn max(self, other: I32x4) -> I32x4 { |
| 614 | I32x4([ |
| 615 | self[0].max(other[0]), |
| 616 | self[1].max(other[1]), |
| 617 | self[2].max(other[2]), |
| 618 | self[3].max(other[3]), |
| 619 | ]) |
| 620 | } |
| 621 | |
| 622 | // Packed comparisons |
| 623 | |
| 624 | #[inline ] |
| 625 | pub fn packed_eq(self, other: I32x4) -> U32x4 { |
| 626 | U32x4([ |
| 627 | if self[0] == other[0] { !0 } else { 0 }, |
| 628 | if self[1] == other[1] { !0 } else { 0 }, |
| 629 | if self[2] == other[2] { !0 } else { 0 }, |
| 630 | if self[3] == other[3] { !0 } else { 0 }, |
| 631 | ]) |
| 632 | } |
| 633 | |
| 634 | #[inline ] |
| 635 | pub fn packed_gt(self, other: I32x4) -> U32x4 { |
| 636 | U32x4([ |
| 637 | if self[0] > other[0] { !0 } else { 0 }, |
| 638 | if self[1] > other[1] { !0 } else { 0 }, |
| 639 | if self[2] > other[2] { !0 } else { 0 }, |
| 640 | if self[3] > other[3] { !0 } else { 0 }, |
| 641 | ]) |
| 642 | } |
| 643 | |
| 644 | #[inline ] |
| 645 | pub fn packed_le(self, other: I32x4) -> U32x4 { |
| 646 | U32x4([ |
| 647 | if self[0] <= other[0] { !0 } else { 0 }, |
| 648 | if self[1] <= other[1] { !0 } else { 0 }, |
| 649 | if self[2] <= other[2] { !0 } else { 0 }, |
| 650 | if self[3] <= other[3] { !0 } else { 0 }, |
| 651 | ]) |
| 652 | } |
| 653 | |
| 654 | #[inline ] |
| 655 | pub fn packed_lt(self, other: I32x4) -> U32x4 { |
| 656 | U32x4([ |
| 657 | if self[0] < other[0] { !0 } else { 0 }, |
| 658 | if self[1] < other[1] { !0 } else { 0 }, |
| 659 | if self[2] < other[2] { !0 } else { 0 }, |
| 660 | if self[3] < other[3] { !0 } else { 0 }, |
| 661 | ]) |
| 662 | } |
| 663 | |
| 664 | // Concatenations |
| 665 | |
| 666 | #[inline ] |
| 667 | pub fn concat_xy_xy(self, other: I32x4) -> I32x4 { |
| 668 | I32x4([self[0], self[1], other[0], other[1]]) |
| 669 | } |
| 670 | |
| 671 | #[inline ] |
| 672 | pub fn concat_zw_zw(self, other: I32x4) -> I32x4 { |
| 673 | I32x4([self[2], self[3], other[2], other[3]]) |
| 674 | } |
| 675 | |
| 676 | // Swizzle conversions |
| 677 | |
| 678 | #[inline ] |
| 679 | pub fn xy(self) -> I32x2 { |
| 680 | I32x2([self[0], self[1]]) |
| 681 | } |
| 682 | |
| 683 | #[inline ] |
| 684 | pub fn xw(self) -> I32x2 { |
| 685 | I32x2([self[0], self[3]]) |
| 686 | } |
| 687 | |
| 688 | #[inline ] |
| 689 | pub fn zy(self) -> I32x2 { |
| 690 | I32x2([self[2], self[1]]) |
| 691 | } |
| 692 | |
| 693 | #[inline ] |
| 694 | pub fn zw(self) -> I32x2 { |
| 695 | I32x2([self[2], self[3]]) |
| 696 | } |
| 697 | |
| 698 | // Conversions |
| 699 | |
| 700 | /// Converts these packed integers to floats. |
| 701 | #[inline ] |
| 702 | pub fn to_f32x4(self) -> F32x4 { |
| 703 | F32x4([ |
| 704 | self[0] as f32, |
| 705 | self[1] as f32, |
| 706 | self[2] as f32, |
| 707 | self[3] as f32, |
| 708 | ]) |
| 709 | } |
| 710 | |
| 711 | /// Converts these packed signed integers to unsigned integers. |
| 712 | /// |
| 713 | /// Overflowing values will wrap around. |
| 714 | /// |
| 715 | /// FIXME(pcwalton): Should they? This will assert on overflow in debug. |
| 716 | #[inline ] |
| 717 | pub fn to_u32x4(self) -> U32x4 { |
| 718 | U32x4([self[0] as u32, self[1] as u32, self[2] as u32, self[3] as u32]) |
| 719 | } |
| 720 | } |
| 721 | |
| 722 | impl Index<usize> for I32x4 { |
| 723 | type Output = i32; |
| 724 | #[inline ] |
| 725 | fn index(&self, index: usize) -> &i32 { |
| 726 | &self.0[index] |
| 727 | } |
| 728 | } |
| 729 | |
| 730 | impl IndexMut<usize> for I32x4 { |
| 731 | #[inline ] |
| 732 | fn index_mut(&mut self, index: usize) -> &mut i32 { |
| 733 | &mut self.0[index] |
| 734 | } |
| 735 | } |
| 736 | |
| 737 | impl Add<I32x4> for I32x4 { |
| 738 | type Output = I32x4; |
| 739 | #[inline ] |
| 740 | fn add(self, other: I32x4) -> I32x4 { |
| 741 | I32x4([ |
| 742 | self[0] + other[0], |
| 743 | self[1] + other[1], |
| 744 | self[2] + other[2], |
| 745 | self[3] + other[3], |
| 746 | ]) |
| 747 | } |
| 748 | } |
| 749 | |
| 750 | impl Sub<I32x4> for I32x4 { |
| 751 | type Output = I32x4; |
| 752 | #[inline ] |
| 753 | fn sub(self, other: I32x4) -> I32x4 { |
| 754 | I32x4([ |
| 755 | self[0] - other[0], |
| 756 | self[1] - other[1], |
| 757 | self[2] - other[2], |
| 758 | self[3] - other[3], |
| 759 | ]) |
| 760 | } |
| 761 | } |
| 762 | |
| 763 | impl Mul<I32x4> for I32x4 { |
| 764 | type Output = I32x4; |
| 765 | #[inline ] |
| 766 | fn mul(self, other: I32x4) -> I32x4 { |
| 767 | I32x4([ |
| 768 | self[0] * other[0], |
| 769 | self[1] * other[1], |
| 770 | self[2] * other[2], |
| 771 | self[3] * other[3], |
| 772 | ]) |
| 773 | } |
| 774 | } |
| 775 | |
| 776 | impl BitAnd<I32x4> for I32x4 { |
| 777 | type Output = I32x4; |
| 778 | #[inline ] |
| 779 | fn bitand(self, other: I32x4) -> I32x4 { |
| 780 | I32x4([self[0] & other[0], self[1] & other[1], self[2] & other[2], self[3] & other[3]]) |
| 781 | } |
| 782 | } |
| 783 | |
| 784 | impl BitOr<I32x4> for I32x4 { |
| 785 | type Output = I32x4; |
| 786 | #[inline ] |
| 787 | fn bitor(self, other: I32x4) -> I32x4 { |
| 788 | I32x4([self[0] | other[0], self[1] | other[1], self[2] | other[2], self[3] | other[3]]) |
| 789 | } |
| 790 | } |
| 791 | |
| 792 | impl Shr<I32x4> for I32x4 { |
| 793 | type Output = I32x4; |
| 794 | #[inline ] |
| 795 | fn shr(self, other: I32x4) -> I32x4 { |
| 796 | I32x4([ |
| 797 | self[0] >> other[0], |
| 798 | self[1] >> other[1], |
| 799 | self[2] >> other[2], |
| 800 | self[3] >> other[3], |
| 801 | ]) |
| 802 | } |
| 803 | } |
| 804 | |
| 805 | // Two 32-bit unsigned integers |
| 806 | |
| 807 | #[derive (Clone, Copy)] |
| 808 | pub struct U32x2(pub [u32; 2]); |
| 809 | |
| 810 | impl U32x2 { |
| 811 | #[inline ] |
| 812 | pub fn new(x: u32, y: u32) -> U32x2 { |
| 813 | U32x2([x, y]) |
| 814 | } |
| 815 | |
| 816 | #[inline ] |
| 817 | pub fn splat(x: u32) -> U32x2 { |
| 818 | U32x2::new(x, x) |
| 819 | } |
| 820 | |
| 821 | /// Returns true if both booleans in this vector are true. |
| 822 | /// |
| 823 | /// The result is *undefined* if both values in this vector are not booleans. A boolean is a |
| 824 | /// value with all bits set or all bits clear (i.e. !0 or 0). |
| 825 | #[inline ] |
| 826 | pub fn all_true(&self) -> bool { |
| 827 | self[0] == !0 && self[1] == !0 |
| 828 | } |
| 829 | |
| 830 | /// Returns true if both booleans in this vector are false. |
| 831 | /// |
| 832 | /// The result is *undefined* if both values in this vector are not booleans. A boolean is a |
| 833 | /// value with all bits set or all bits clear (i.e. !0 or 0). |
| 834 | #[inline ] |
| 835 | pub fn all_false(&self) -> bool { |
| 836 | self[0] == 0 && self[1] == 0 |
| 837 | } |
| 838 | |
| 839 | #[inline ] |
| 840 | pub fn to_i32x2(self) -> I32x2 { |
| 841 | I32x2::new(self[0] as i32, self[1] as i32) |
| 842 | } |
| 843 | |
| 844 | } |
| 845 | |
| 846 | impl BitAnd<U32x2> for U32x2 { |
| 847 | type Output = U32x2; |
| 848 | #[inline ] |
| 849 | fn bitand(self, other: U32x2) -> U32x2 { |
| 850 | U32x2([self[0] & other[0], self[1] & other[1]]) |
| 851 | } |
| 852 | } |
| 853 | |
| 854 | impl BitOr<U32x2> for U32x2 { |
| 855 | type Output = U32x2; |
| 856 | #[inline ] |
| 857 | fn bitor(self, other: U32x2) -> U32x2 { |
| 858 | U32x2([self[0] | other[0], self[1] | other[1]]) |
| 859 | } |
| 860 | } |
| 861 | |
| 862 | impl Not for U32x2 { |
| 863 | type Output = U32x2; |
| 864 | #[inline ] |
| 865 | fn not(self) -> U32x2 { |
| 866 | U32x2([!self[0], !self[1]]) |
| 867 | } |
| 868 | } |
| 869 | |
| 870 | impl Index<usize> for U32x2 { |
| 871 | type Output = u32; |
| 872 | #[inline ] |
| 873 | fn index(&self, index: usize) -> &u32 { |
| 874 | &self.0[index] |
| 875 | } |
| 876 | } |
| 877 | |
| 878 | // Four 32-bit unsigned integers |
| 879 | |
| 880 | #[derive (Clone, Copy, PartialEq, Eq, Debug)] |
| 881 | pub struct U32x4(pub [u32; 4]); |
| 882 | |
| 883 | impl U32x4 { |
| 884 | pub fn new(a: u32, b: u32, c: u32, d: u32) -> U32x4 { |
| 885 | U32x4([a, b, c, d]) |
| 886 | } |
| 887 | |
| 888 | // Conversions |
| 889 | |
| 890 | /// Converts these packed unsigned integers to signed integers. |
| 891 | /// |
| 892 | /// Overflowing values will wrap around. |
| 893 | /// |
| 894 | /// FIXME(pcwalton): Should they? This will assert on overflow in debug. |
| 895 | #[inline ] |
| 896 | pub fn to_i32x4(self) -> I32x4 { |
| 897 | I32x4([self[0] as i32, self[1] as i32, self[2] as i32, self[3] as i32]) |
| 898 | } |
| 899 | |
| 900 | // Basic operations |
| 901 | |
| 902 | /// Returns true if all four booleans in this vector are true. |
| 903 | /// |
| 904 | /// The result is *undefined* if all four values in this vector are not booleans. A boolean is |
| 905 | /// a value with all bits set or all bits clear (i.e. !0 or 0). |
| 906 | #[inline ] |
| 907 | pub fn all_true(&self) -> bool { |
| 908 | self[0] == !0 && self[1] == !0 && self[2] == !0 && self[3] == !0 |
| 909 | } |
| 910 | |
| 911 | /// Returns true if all four booleans in this vector are false. |
| 912 | /// |
| 913 | /// The result is *undefined* if all four values in this vector are not booleans. A boolean is |
| 914 | /// a value with all bits set or all bits clear (i.e. !0 or 0). |
| 915 | #[inline ] |
| 916 | pub fn all_false(&self) -> bool { |
| 917 | self[0] == 0 && self[1] == 0 && self[2] == 0 && self[3] == 0 |
| 918 | } |
| 919 | } |
| 920 | |
| 921 | impl Index<usize> for U32x4 { |
| 922 | type Output = u32; |
| 923 | #[inline ] |
| 924 | fn index(&self, index: usize) -> &u32 { |
| 925 | &self.0[index] |
| 926 | } |
| 927 | } |
| 928 | |
| 929 | impl Shr<u32> for U32x4 { |
| 930 | type Output = U32x4; |
| 931 | #[inline ] |
| 932 | fn shr(self, amount: u32) -> U32x4 { |
| 933 | U32x4([self[0] >> amount, self[1] >> amount, self[2] >> amount, self[3] >> amount]) |
| 934 | } |
| 935 | } |
| 936 | |