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
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
//! Types and traits associated with masking lanes of vectors.
//! Types representing
#![allow(non_camel_case_types)]

#[cfg_attr(
    not(all(target_arch = "x86_64", target_feature = "avx512f")),
    path = "masks/full_masks.rs"
)]
#[cfg_attr(
    all(target_arch = "x86_64", target_feature = "avx512f"),
    path = "masks/bitmask.rs"
)]
mod mask_impl;

use crate::simd::intrinsics;
use crate::simd::{LaneCount, Simd, SimdElement, SupportedLaneCount};
use core::cmp::Ordering;
use core::{fmt, mem};

mod sealed {
    use super::*;

    /// Not only does this seal the `MaskElement` trait, but these functions prevent other traits
    /// from bleeding into the parent bounds.
    ///
    /// For example, `eq` could be provided by requiring `MaskElement: PartialEq`, but that would
    /// prevent us from ever removing that bound, or from implementing `MaskElement` on
    /// non-`PartialEq` types in the future.
    pub trait Sealed {
        fn valid<const LANES: usize>(values: Simd<Self, LANES>) -> bool
        where
            LaneCount<LANES>: SupportedLaneCount,
            Self: SimdElement;

        fn eq(self, other: Self) -> bool;

        const TRUE: Self;

        const FALSE: Self;
    }
}
use sealed::Sealed;

/// Marker trait for types that may be used as SIMD mask elements.
pub unsafe trait MaskElement: SimdElement + Sealed {}

macro_rules! impl_element {
    { $ty:ty } => {
        impl Sealed for $ty {
            fn valid<const LANES: usize>(value: Simd<Self, LANES>) -> bool
            where
                LaneCount<LANES>: SupportedLaneCount,
            {
                (value.lanes_eq(Simd::splat(0)) | value.lanes_eq(Simd::splat(-1))).all()
            }

            fn eq(self, other: Self) -> bool { self == other }

            const TRUE: Self = -1;
            const FALSE: Self = 0;
        }

        unsafe impl MaskElement for $ty {}
    }
}

impl_element! { i8 }
impl_element! { i16 }
impl_element! { i32 }
impl_element! { i64 }
impl_element! { isize }

/// A SIMD vector mask for `LANES` elements of width specified by `Element`.
///
/// The layout of this type is unspecified.
#[repr(transparent)]
pub struct Mask<T, const LANES: usize>(mask_impl::Mask<T, LANES>)
where
    T: MaskElement,
    LaneCount<LANES>: SupportedLaneCount;

impl<T, const LANES: usize> Copy for Mask<T, LANES>
where
    T: MaskElement,
    LaneCount<LANES>: SupportedLaneCount,
{
}

impl<T, const LANES: usize> Clone for Mask<T, LANES>
where
    T: MaskElement,
    LaneCount<LANES>: SupportedLaneCount,
{
    fn clone(&self) -> Self {
        *self
    }
}

impl<T, const LANES: usize> Mask<T, LANES>
where
    T: MaskElement,
    LaneCount<LANES>: SupportedLaneCount,
{
    /// Construct a mask by setting all lanes to the given value.
    pub fn splat(value: bool) -> Self {
        Self(mask_impl::Mask::splat(value))
    }

    /// Converts an array of bools to a SIMD mask.
    pub fn from_array(array: [bool; LANES]) -> Self {
        // SAFETY: Rust's bool has a layout of 1 byte (u8) with a value of
        //     true:    0b_0000_0001
        //     false:   0b_0000_0000
        // Thus, an array of bools is also a valid array of bytes: [u8; N]
        // This would be hypothetically valid as an "in-place" transmute,
        // but these are "dependently-sized" types, so copy elision it is!
        unsafe {
            let bytes: [u8; LANES] = mem::transmute_copy(&array);
            let bools: Simd<i8, LANES> =
                intrinsics::simd_ne(Simd::from_array(bytes), Simd::splat(0u8));
            Mask::from_int_unchecked(intrinsics::simd_cast(bools))
        }
    }

    /// Converts a SIMD mask to an array of bools.
    pub fn to_array(self) -> [bool; LANES] {
        // This follows mostly the same logic as from_array.
        // SAFETY: Rust's bool has a layout of 1 byte (u8) with a value of
        //     true:    0b_0000_0001
        //     false:   0b_0000_0000
        // Thus, an array of bools is also a valid array of bytes: [u8; N]
        // Since our masks are equal to integers where all bits are set,
        // we can simply convert them to i8s, and then bitand them by the
        // bitpattern for Rust's "true" bool.
        // This would be hypothetically valid as an "in-place" transmute,
        // but these are "dependently-sized" types, so copy elision it is!
        unsafe {
            let mut bytes: Simd<i8, LANES> = intrinsics::simd_cast(self.to_int());
            bytes &= Simd::splat(1i8);
            mem::transmute_copy(&bytes)
        }
    }

    /// Converts a vector of integers to a mask, where 0 represents `false` and -1
    /// represents `true`.
    ///
    /// # Safety
    /// All lanes must be either 0 or -1.
    #[inline]
    #[must_use = "method returns a new mask and does not mutate the original value"]
    pub unsafe fn from_int_unchecked(value: Simd<T, LANES>) -> Self {
        unsafe { Self(mask_impl::Mask::from_int_unchecked(value)) }
    }

    /// Converts a vector of integers to a mask, where 0 represents `false` and -1
    /// represents `true`.
    ///
    /// # Panics
    /// Panics if any lane is not 0 or -1.
    #[inline]
    #[must_use = "method returns a new mask and does not mutate the original value"]
    pub fn from_int(value: Simd<T, LANES>) -> Self {
        assert!(T::valid(value), "all values must be either 0 or -1",);
        unsafe { Self::from_int_unchecked(value) }
    }

    /// Converts the mask to a vector of integers, where 0 represents `false` and -1
    /// represents `true`.
    #[inline]
    #[must_use = "method returns a new vector and does not mutate the original value"]
    pub fn to_int(self) -> Simd<T, LANES> {
        self.0.to_int()
    }

    /// Tests the value of the specified lane.
    ///
    /// # Safety
    /// `lane` must be less than `LANES`.
    #[inline]
    #[must_use = "method returns a new bool and does not mutate the original value"]
    pub unsafe fn test_unchecked(&self, lane: usize) -> bool {
        unsafe { self.0.test_unchecked(lane) }
    }

    /// Tests the value of the specified lane.
    ///
    /// # Panics
    /// Panics if `lane` is greater than or equal to the number of lanes in the vector.
    #[inline]
    #[must_use = "method returns a new bool and does not mutate the original value"]
    pub fn test(&self, lane: usize) -> bool {
        assert!(lane < LANES, "lane index out of range");
        unsafe { self.test_unchecked(lane) }
    }

    /// Sets the value of the specified lane.
    ///
    /// # Safety
    /// `lane` must be less than `LANES`.
    #[inline]
    pub unsafe fn set_unchecked(&mut self, lane: usize, value: bool) {
        unsafe {
            self.0.set_unchecked(lane, value);
        }
    }

    /// Sets the value of the specified lane.
    ///
    /// # Panics
    /// Panics if `lane` is greater than or equal to the number of lanes in the vector.
    #[inline]
    pub fn set(&mut self, lane: usize, value: bool) {
        assert!(lane < LANES, "lane index out of range");
        unsafe {
            self.set_unchecked(lane, value);
        }
    }

    /// Convert this mask to a bitmask, with one bit set per lane.
    #[cfg(feature = "generic_const_exprs")]
    #[inline]
    #[must_use = "method returns a new array and does not mutate the original value"]
    pub fn to_bitmask(self) -> [u8; LaneCount::<LANES>::BITMASK_LEN] {
        self.0.to_bitmask()
    }

    /// Convert a bitmask to a mask.
    #[cfg(feature = "generic_const_exprs")]
    #[inline]
    #[must_use = "method returns a new mask and does not mutate the original value"]
    pub fn from_bitmask(bitmask: [u8; LaneCount::<LANES>::BITMASK_LEN]) -> Self {
        Self(mask_impl::Mask::from_bitmask(bitmask))
    }

    /// Returns true if any lane is set, or false otherwise.
    #[inline]
    #[must_use = "method returns a new bool and does not mutate the original value"]
    pub fn any(self) -> bool {
        self.0.any()
    }

    /// Returns true if all lanes are set, or false otherwise.
    #[inline]
    #[must_use = "method returns a new bool and does not mutate the original value"]
    pub fn all(self) -> bool {
        self.0.all()
    }
}

// vector/array conversion
impl<T, const LANES: usize> From<[bool; LANES]> for Mask<T, LANES>
where
    T: MaskElement,
    LaneCount<LANES>: SupportedLaneCount,
{
    fn from(array: [bool; LANES]) -> Self {
        Self::from_array(array)
    }
}

impl<T, const LANES: usize> From<Mask<T, LANES>> for [bool; LANES]
where
    T: MaskElement,
    LaneCount<LANES>: SupportedLaneCount,
{
    fn from(vector: Mask<T, LANES>) -> Self {
        vector.to_array()
    }
}

impl<T, const LANES: usize> Default for Mask<T, LANES>
where
    T: MaskElement,
    LaneCount<LANES>: SupportedLaneCount,
{
    #[inline]
    #[must_use = "method returns a defaulted mask with all lanes set to false (0)"]
    fn default() -> Self {
        Self::splat(false)
    }
}

impl<T, const LANES: usize> PartialEq for Mask<T, LANES>
where
    T: MaskElement + PartialEq,
    LaneCount<LANES>: SupportedLaneCount,
{
    #[inline]
    #[must_use = "method returns a new bool and does not mutate the original value"]
    fn eq(&self, other: &Self) -> bool {
        self.0 == other.0
    }
}

impl<T, const LANES: usize> PartialOrd for Mask<T, LANES>
where
    T: MaskElement + PartialOrd,
    LaneCount<LANES>: SupportedLaneCount,
{
    #[inline]
    #[must_use = "method returns a new Ordering and does not mutate the original value"]
    fn partial_cmp(&self, other: &Self) -> Option<Ordering> {
        self.0.partial_cmp(&other.0)
    }
}

impl<T, const LANES: usize> fmt::Debug for Mask<T, LANES>
where
    T: MaskElement + fmt::Debug,
    LaneCount<LANES>: SupportedLaneCount,
{
    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
        f.debug_list()
            .entries((0..LANES).map(|lane| self.test(lane)))
            .finish()
    }
}

impl<T, const LANES: usize> core::ops::BitAnd for Mask<T, LANES>
where
    T: MaskElement,
    LaneCount<LANES>: SupportedLaneCount,
{
    type Output = Self;
    #[inline]
    #[must_use = "method returns a new mask and does not mutate the original value"]
    fn bitand(self, rhs: Self) -> Self {
        Self(self.0 & rhs.0)
    }
}

impl<T, const LANES: usize> core::ops::BitAnd<bool> for Mask<T, LANES>
where
    T: MaskElement,
    LaneCount<LANES>: SupportedLaneCount,
{
    type Output = Self;
    #[inline]
    #[must_use = "method returns a new mask and does not mutate the original value"]
    fn bitand(self, rhs: bool) -> Self {
        self & Self::splat(rhs)
    }
}

impl<T, const LANES: usize> core::ops::BitAnd<Mask<T, LANES>> for bool
where
    T: MaskElement,
    LaneCount<LANES>: SupportedLaneCount,
{
    type Output = Mask<T, LANES>;
    #[inline]
    #[must_use = "method returns a new mask and does not mutate the original value"]
    fn bitand(self, rhs: Mask<T, LANES>) -> Mask<T, LANES> {
        Mask::splat(self) & rhs
    }
}

impl<T, const LANES: usize> core::ops::BitOr for Mask<T, LANES>
where
    T: MaskElement,
    LaneCount<LANES>: SupportedLaneCount,
{
    type Output = Self;
    #[inline]
    #[must_use = "method returns a new mask and does not mutate the original value"]
    fn bitor(self, rhs: Self) -> Self {
        Self(self.0 | rhs.0)
    }
}

impl<T, const LANES: usize> core::ops::BitOr<bool> for Mask<T, LANES>
where
    T: MaskElement,
    LaneCount<LANES>: SupportedLaneCount,
{
    type Output = Self;
    #[inline]
    #[must_use = "method returns a new mask and does not mutate the original value"]
    fn bitor(self, rhs: bool) -> Self {
        self | Self::splat(rhs)
    }
}

impl<T, const LANES: usize> core::ops::BitOr<Mask<T, LANES>> for bool
where
    T: MaskElement,
    LaneCount<LANES>: SupportedLaneCount,
{
    type Output = Mask<T, LANES>;
    #[inline]
    #[must_use = "method returns a new mask and does not mutate the original value"]
    fn bitor(self, rhs: Mask<T, LANES>) -> Mask<T, LANES> {
        Mask::splat(self) | rhs
    }
}

impl<T, const LANES: usize> core::ops::BitXor for Mask<T, LANES>
where
    T: MaskElement,
    LaneCount<LANES>: SupportedLaneCount,
{
    type Output = Self;
    #[inline]
    #[must_use = "method returns a new mask and does not mutate the original value"]
    fn bitxor(self, rhs: Self) -> Self::Output {
        Self(self.0 ^ rhs.0)
    }
}

impl<T, const LANES: usize> core::ops::BitXor<bool> for Mask<T, LANES>
where
    T: MaskElement,
    LaneCount<LANES>: SupportedLaneCount,
{
    type Output = Self;
    #[inline]
    #[must_use = "method returns a new mask and does not mutate the original value"]
    fn bitxor(self, rhs: bool) -> Self::Output {
        self ^ Self::splat(rhs)
    }
}

impl<T, const LANES: usize> core::ops::BitXor<Mask<T, LANES>> for bool
where
    T: MaskElement,
    LaneCount<LANES>: SupportedLaneCount,
{
    type Output = Mask<T, LANES>;
    #[inline]
    #[must_use = "method returns a new mask and does not mutate the original value"]
    fn bitxor(self, rhs: Mask<T, LANES>) -> Self::Output {
        Mask::splat(self) ^ rhs
    }
}

impl<T, const LANES: usize> core::ops::Not for Mask<T, LANES>
where
    T: MaskElement,
    LaneCount<LANES>: SupportedLaneCount,
{
    type Output = Mask<T, LANES>;
    #[inline]
    #[must_use = "method returns a new mask and does not mutate the original value"]
    fn not(self) -> Self::Output {
        Self(!self.0)
    }
}

impl<T, const LANES: usize> core::ops::BitAndAssign for Mask<T, LANES>
where
    T: MaskElement,
    LaneCount<LANES>: SupportedLaneCount,
{
    #[inline]
    fn bitand_assign(&mut self, rhs: Self) {
        self.0 = self.0 & rhs.0;
    }
}

impl<T, const LANES: usize> core::ops::BitAndAssign<bool> for Mask<T, LANES>
where
    T: MaskElement,
    LaneCount<LANES>: SupportedLaneCount,
{
    #[inline]
    fn bitand_assign(&mut self, rhs: bool) {
        *self &= Self::splat(rhs);
    }
}

impl<T, const LANES: usize> core::ops::BitOrAssign for Mask<T, LANES>
where
    T: MaskElement,
    LaneCount<LANES>: SupportedLaneCount,
{
    #[inline]
    fn bitor_assign(&mut self, rhs: Self) {
        self.0 = self.0 | rhs.0;
    }
}

impl<T, const LANES: usize> core::ops::BitOrAssign<bool> for Mask<T, LANES>
where
    T: MaskElement,
    LaneCount<LANES>: SupportedLaneCount,
{
    #[inline]
    fn bitor_assign(&mut self, rhs: bool) {
        *self |= Self::splat(rhs);
    }
}

impl<T, const LANES: usize> core::ops::BitXorAssign for Mask<T, LANES>
where
    T: MaskElement,
    LaneCount<LANES>: SupportedLaneCount,
{
    #[inline]
    fn bitxor_assign(&mut self, rhs: Self) {
        self.0 = self.0 ^ rhs.0;
    }
}

impl<T, const LANES: usize> core::ops::BitXorAssign<bool> for Mask<T, LANES>
where
    T: MaskElement,
    LaneCount<LANES>: SupportedLaneCount,
{
    #[inline]
    fn bitxor_assign(&mut self, rhs: bool) {
        *self ^= Self::splat(rhs);
    }
}

/// Vector of eight 8-bit masks
pub type mask8x8 = Mask<i8, 8>;

/// Vector of 16 8-bit masks
pub type mask8x16 = Mask<i8, 16>;

/// Vector of 32 8-bit masks
pub type mask8x32 = Mask<i8, 32>;

/// Vector of 16 8-bit masks
pub type mask8x64 = Mask<i8, 64>;

/// Vector of four 16-bit masks
pub type mask16x4 = Mask<i16, 4>;

/// Vector of eight 16-bit masks
pub type mask16x8 = Mask<i16, 8>;

/// Vector of 16 16-bit masks
pub type mask16x16 = Mask<i16, 16>;

/// Vector of 32 16-bit masks
pub type mask16x32 = Mask<i16, 32>;

/// Vector of two 32-bit masks
pub type mask32x2 = Mask<i32, 2>;

/// Vector of four 32-bit masks
pub type mask32x4 = Mask<i32, 4>;

/// Vector of eight 32-bit masks
pub type mask32x8 = Mask<i32, 8>;

/// Vector of 16 32-bit masks
pub type mask32x16 = Mask<i32, 16>;

/// Vector of two 64-bit masks
pub type mask64x2 = Mask<i64, 2>;

/// Vector of four 64-bit masks
pub type mask64x4 = Mask<i64, 4>;

/// Vector of eight 64-bit masks
pub type mask64x8 = Mask<i64, 8>;

/// Vector of two pointer-width masks
pub type masksizex2 = Mask<isize, 2>;

/// Vector of four pointer-width masks
pub type masksizex4 = Mask<isize, 4>;

/// Vector of eight pointer-width masks
pub type masksizex8 = Mask<isize, 8>;

macro_rules! impl_from {
    { $from:ty  => $($to:ty),* } => {
        $(
        impl<const LANES: usize> From<Mask<$from, LANES>> for Mask<$to, LANES>
        where
            LaneCount<LANES>: SupportedLaneCount,
        {
            fn from(value: Mask<$from, LANES>) -> Self {
                Self(value.0.convert())
            }
        }
        )*
    }
}
impl_from! { i8 => i16, i32, i64, isize }
impl_from! { i16 => i32, i64, isize, i8 }
impl_from! { i32 => i64, isize, i8, i16 }
impl_from! { i64 => isize, i8, i16, i32 }
impl_from! { isize => i8, i16, i32, i64 }