core/stdarch/crates/core_arch/src/x86_64/
amx.rs

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
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
/// Load tile configuration from a 64-byte memory location specified by mem_addr.
/// The tile configuration format is specified below, and includes the tile type pallette,
/// the number of bytes per row, and the number of rows. If the specified pallette_id is zero,
/// that signifies the init state for both the tile config and the tile data, and the tiles are zeroed.
/// Any invalid configurations will result in #GP fault.
///
/// [Intel's documentation](https://www.intel.com/content/www/us/en/docs/intrinsics-guide/index.html#text=_tile_loadconfig&ig_expand=6875)
#[inline]
#[target_feature(enable = "amx-tile")]
#[unstable(feature = "x86_amx_intrinsics", issue = "126622")]
pub unsafe fn _tile_loadconfig(mem_addr: *const u8) {
    ldtilecfg(mem_addr);
}

/// Stores the current tile configuration to a 64-byte memory location specified by mem_addr.
/// The tile configuration format is specified below, and includes the tile type pallette,
/// the number of bytes per row, and the number of rows. If tiles are not configured, all zeroes will be stored to memory.
///
/// [Intel's documentation](https://www.intel.com/content/www/us/en/docs/intrinsics-guide/index.html#text=_tile_storeconfig&ig_expand=6879)
#[inline]
#[target_feature(enable = "amx-tile")]
#[unstable(feature = "x86_amx_intrinsics", issue = "126622")]
pub unsafe fn _tile_storeconfig(mem_addr: *mut u8) {
    sttilecfg(mem_addr);
}

/// Load tile rows from memory specifieid by base address and stride into destination tile dst using the tile configuration previously configured via _tile_loadconfig.
///
/// [Intel's documentation](https://www.intel.com/content/www/us/en/docs/intrinsics-guide/index.html#text=_tile_loadd&ig_expand=6877)
#[inline]
#[rustc_legacy_const_generics(0)]
#[target_feature(enable = "amx-tile")]
#[unstable(feature = "x86_amx_intrinsics", issue = "126622")]
pub unsafe fn _tile_loadd<const DST: i32>(base: *const u8, stride: usize) {
    static_assert_uimm_bits!(DST, 3);
    tileloadd64(DST as i8, base, stride);
}

/// Release the tile configuration to return to the init state, which releases all storage it currently holds.
///
/// [Intel's documentation](https://www.intel.com/content/www/us/en/docs/intrinsics-guide/index.html#text=_tile_release&ig_expand=6878)
#[inline]
#[target_feature(enable = "amx-tile")]
#[unstable(feature = "x86_amx_intrinsics", issue = "126622")]
pub unsafe fn _tile_release() {
    tilerelease();
}

/// Store the tile specified by src to memory specifieid by base address and stride using the tile configuration previously configured via _tile_loadconfig.
///
/// [Intel's documentation](https://www.intel.com/content/www/us/en/docs/intrinsics-guide/index.html#text=_tile_stored&ig_expand=6881)
#[inline]
#[rustc_legacy_const_generics(0)]
#[target_feature(enable = "amx-tile")]
#[unstable(feature = "x86_amx_intrinsics", issue = "126622")]
pub unsafe fn _tile_stored<const DST: i32>(base: *mut u8, stride: usize) {
    static_assert_uimm_bits!(DST, 3);
    tilestored64(DST as i8, base, stride);
}

/// Load tile rows from memory specifieid by base address and stride into destination tile dst using the tile configuration
/// previously configured via _tile_loadconfig. This intrinsic provides a hint to the implementation that the data will
/// likely not be reused in the near future and the data caching can be optimized accordingly.
///
/// [Intel's documentation](https://www.intel.com/content/www/us/en/docs/intrinsics-guide/index.html#text=_tile_stream_loadd&ig_expand=6883)
#[inline]
#[rustc_legacy_const_generics(0)]
#[target_feature(enable = "amx-tile")]
#[unstable(feature = "x86_amx_intrinsics", issue = "126622")]
pub unsafe fn _tile_stream_loadd<const DST: i32>(base: *const u8, stride: usize) {
    static_assert_uimm_bits!(DST, 3);
    tileloaddt164(DST as i8, base, stride);
}

/// Zero the tile specified by tdest.
///
/// [Intel's documentation](https://www.intel.com/content/www/us/en/docs/intrinsics-guide/index.html#text=_tile_zero&ig_expand=6885)
#[inline]
#[rustc_legacy_const_generics(0)]
#[target_feature(enable = "amx-tile")]
#[unstable(feature = "x86_amx_intrinsics", issue = "126622")]
pub unsafe fn _tile_zero<const DST: i32>() {
    static_assert_uimm_bits!(DST, 3);
    tilezero(DST as i8);
}

/// Compute dot-product of BF16 (16-bit) floating-point pairs in tiles a and b,
/// accumulating the intermediate single-precision (32-bit) floating-point elements
/// with elements in dst, and store the 32-bit result back to tile dst.
///
/// [Intel's documentation](https://www.intel.com/content/www/us/en/docs/intrinsics-guide/index.html#text=_tile_dpbf16ps&ig_expand=6864)
#[inline]
#[rustc_legacy_const_generics(0, 1, 2)]
#[target_feature(enable = "amx-bf16")]
#[unstable(feature = "x86_amx_intrinsics", issue = "126622")]
pub unsafe fn _tile_dpbf16ps<const DST: i32, const A: i32, const B: i32>() {
    static_assert_uimm_bits!(DST, 3);
    static_assert_uimm_bits!(A, 3);
    static_assert_uimm_bits!(B, 3);
    tdpbf16ps(DST as i8, A as i8, B as i8);
}

/// Compute dot-product of bytes in tiles with a source/destination accumulator.
/// Multiply groups of 4 adjacent pairs of signed 8-bit integers in a with corresponding
/// signed 8-bit integers in b, producing 4 intermediate 32-bit results.
/// Sum these 4 results with the corresponding 32-bit integer in dst, and store the 32-bit result back to tile dst.
///
/// [Intel's documentation](https://www.intel.com/content/www/us/en/docs/intrinsics-guide/index.html#text=_tile_dpbssd&ig_expand=6866)
#[inline]
#[rustc_legacy_const_generics(0, 1, 2)]
#[target_feature(enable = "amx-int8")]
#[unstable(feature = "x86_amx_intrinsics", issue = "126622")]
pub unsafe fn _tile_dpbssd<const DST: i32, const A: i32, const B: i32>() {
    static_assert_uimm_bits!(DST, 3);
    static_assert_uimm_bits!(A, 3);
    static_assert_uimm_bits!(B, 3);
    tdpbssd(DST as i8, A as i8, B as i8);
}

/// Compute dot-product of bytes in tiles with a source/destination accumulator.
/// Multiply groups of 4 adjacent pairs of signed 8-bit integers in a with corresponding
/// unsigned 8-bit integers in b, producing 4 intermediate 32-bit results.
/// Sum these 4 results with the corresponding 32-bit integer in dst, and store the 32-bit result back to tile dst.
///
/// [Intel's documentation](https://www.intel.com/content/www/us/en/docs/intrinsics-guide/index.html#text=_tile_dpbsud&ig_expand=6868)
#[inline]
#[rustc_legacy_const_generics(0, 1, 2)]
#[target_feature(enable = "amx-int8")]
#[unstable(feature = "x86_amx_intrinsics", issue = "126622")]
pub unsafe fn _tile_dpbsud<const DST: i32, const A: i32, const B: i32>() {
    static_assert_uimm_bits!(DST, 3);
    static_assert_uimm_bits!(A, 3);
    static_assert_uimm_bits!(B, 3);
    tdpbsud(DST as i8, A as i8, B as i8);
}

/// Compute dot-product of bytes in tiles with a source/destination accumulator.
/// Multiply groups of 4 adjacent pairs of unsigned 8-bit integers in a with corresponding
/// signed 8-bit integers in b, producing 4 intermediate 32-bit results.
/// Sum these 4 results with the corresponding 32-bit integer in dst, and store the 32-bit result back to tile dst.
///
/// [Intel's documentation](https://www.intel.com/content/www/us/en/docs/intrinsics-guide/index.html#text=_tile_dpbusd&ig_expand=6870)
#[inline]
#[rustc_legacy_const_generics(0, 1, 2)]
#[target_feature(enable = "amx-int8")]
#[unstable(feature = "x86_amx_intrinsics", issue = "126622")]
pub unsafe fn _tile_dpbusd<const DST: i32, const A: i32, const B: i32>() {
    static_assert_uimm_bits!(DST, 3);
    static_assert_uimm_bits!(A, 3);
    static_assert_uimm_bits!(B, 3);
    tdpbusd(DST as i8, A as i8, B as i8);
}

/// Compute dot-product of bytes in tiles with a source/destination accumulator.
/// Multiply groups of 4 adjacent pairs of unsigned 8-bit integers in a with corresponding
/// unsigned 8-bit integers in b, producing 4 intermediate 32-bit results.
/// Sum these 4 results with the corresponding 32-bit integer in dst, and store the 32-bit result back to tile dst.
///
/// [Intel's documentation](https://www.intel.com/content/www/us/en/docs/intrinsics-guide/index.html#text=_tile_dpbuud&ig_expand=6872)
#[inline]
#[rustc_legacy_const_generics(0, 1, 2)]
#[target_feature(enable = "amx-int8")]
#[unstable(feature = "x86_amx_intrinsics", issue = "126622")]
pub unsafe fn _tile_dpbuud<const DST: i32, const A: i32, const B: i32>() {
    static_assert_uimm_bits!(DST, 3);
    static_assert_uimm_bits!(A, 3);
    static_assert_uimm_bits!(B, 3);
    tdpbuud(DST as i8, A as i8, B as i8);
}

/// Compute dot-product of FP16 (16-bit) floating-point pairs in tiles a and b,
/// accumulating the intermediate single-precision (32-bit) floating-point elements
///  with elements in dst, and store the 32-bit result back to tile dst.
///
/// [Intel's documentation](https://www.intel.com/content/www/us/en/docs/intrinsics-guide/index.html#text=_tile_dpfp16ps&ig_expand=6874)
#[inline]
#[rustc_legacy_const_generics(0, 1, 2)]
#[target_feature(enable = "amx-fp16")]
#[unstable(feature = "x86_amx_intrinsics", issue = "126622")]
pub unsafe fn _tile_dpfp16ps<const DST: i32, const A: i32, const B: i32>() {
    static_assert_uimm_bits!(DST, 3);
    static_assert_uimm_bits!(A, 3);
    static_assert_uimm_bits!(B, 3);
    tdpfp16ps(DST as i8, A as i8, B as i8);
}

/// Perform matrix multiplication of two tiles containing complex elements and accumulate the results into a packed single precision tile.
/// Each dword element in input tiles a and b is interpreted as a complex number with FP16 real part and FP16 imaginary part.
/// Calculates the imaginary part of the result. For each possible combination of (row of a, column of b),
/// it performs a set of multiplication and accumulations on all corresponding complex numbers (one from a and one from b).
/// The imaginary part of the a element is multiplied with the real part of the corresponding b element, and the real part of
/// the a element is multiplied with the imaginary part of the corresponding b elements. The two accumulated results are added,
/// and then accumulated into the corresponding row and column of dst.
///
/// [Intel's documentation](https://www.intel.com/content/www/us/en/docs/intrinsics-guide/index.html#text=_tile_cmmimfp16ps&ig_expand=6860)
#[inline]
#[rustc_legacy_const_generics(0, 1, 2)]
#[target_feature(enable = "amx-complex")]
#[unstable(feature = "x86_amx_intrinsics", issue = "126622")]
pub unsafe fn _tile_cmmimfp16ps<const DST: i32, const A: i32, const B: i32>() {
    static_assert_uimm_bits!(DST, 3);
    static_assert_uimm_bits!(A, 3);
    static_assert_uimm_bits!(B, 3);
    tcmmimfp16ps(DST as i8, A as i8, B as i8);
}

/// Perform matrix multiplication of two tiles containing complex elements and accumulate the results into a packed single precision tile.
/// Each dword element in input tiles a and b is interpreted as a complex number with FP16 real part and FP16 imaginary part.
/// Calculates the real part of the result. For each possible combination of (row of a, column of b),
/// it performs a set of multiplication and accumulations on all corresponding complex numbers (one from a and one from b).
/// The real part of the a element is multiplied with the real part of the corresponding b element, and the negated imaginary part of
/// the a element is multiplied with the imaginary part of the corresponding b elements.
/// The two accumulated results are added, and then accumulated into the corresponding row and column of dst.
///
/// [Intel's documentation](https://www.intel.com/content/www/us/en/docs/intrinsics-guide/index.html#text=_tile_cmmrlfp16ps&ig_expand=6862)
#[inline]
#[rustc_legacy_const_generics(0, 1, 2)]
#[target_feature(enable = "amx-complex")]
#[unstable(feature = "x86_amx_intrinsics", issue = "126622")]
pub unsafe fn _tile_cmmrlfp16ps<const DST: i32, const A: i32, const B: i32>() {
    static_assert_uimm_bits!(DST, 3);
    static_assert_uimm_bits!(A, 3);
    static_assert_uimm_bits!(B, 3);
    tcmmrlfp16ps(DST as i8, A as i8, B as i8);
}

#[allow(improper_ctypes)]
extern "C" {
    #[link_name = "llvm.x86.ldtilecfg"]
    fn ldtilecfg(mem_addr: *const u8);
    #[link_name = "llvm.x86.sttilecfg"]
    fn sttilecfg(mem_addr: *mut u8);
    #[link_name = "llvm.x86.tileloadd64"]
    fn tileloadd64(dst: i8, base: *const u8, stride: usize);
    #[link_name = "llvm.x86.tileloaddt164"]
    fn tileloaddt164(dst: i8, base: *const u8, stride: usize);
    #[link_name = "llvm.x86.tilerelease"]
    fn tilerelease();
    #[link_name = "llvm.x86.tilestored64"]
    fn tilestored64(dst: i8, base: *mut u8, stride: usize);
    #[link_name = "llvm.x86.tilezero"]
    fn tilezero(dst: i8);
    #[link_name = "llvm.x86.tdpbf16ps"]
    fn tdpbf16ps(dst: i8, a: i8, b: i8);
    #[link_name = "llvm.x86.tdpbuud"]
    fn tdpbuud(dst: i8, a: i8, b: i8);
    #[link_name = "llvm.x86.tdpbusd"]
    fn tdpbusd(dst: i8, a: i8, b: i8);
    #[link_name = "llvm.x86.tdpbsud"]
    fn tdpbsud(dst: i8, a: i8, b: i8);
    #[link_name = "llvm.x86.tdpbssd"]
    fn tdpbssd(dst: i8, a: i8, b: i8);
    #[link_name = "llvm.x86.tdpfp16ps"]
    fn tdpfp16ps(dst: i8, a: i8, b: i8);
    #[link_name = "llvm.x86.tcmmimfp16ps"]
    fn tcmmimfp16ps(dst: i8, a: i8, b: i8);
    #[link_name = "llvm.x86.tcmmrlfp16ps"]
    fn tcmmrlfp16ps(dst: i8, a: i8, b: i8);
}

#[cfg(test)]
mod tests {
    use crate::core_arch::x86::_mm_cvtness_sbh;
    use crate::core_arch::x86_64::*;
    use core::mem::transmute;
    use stdarch_test::simd_test;
    #[cfg(target_os = "linux")]
    use syscalls::{syscall, Sysno};

    #[allow(non_camel_case_types)]
    #[repr(packed)]
    #[derive(Copy, Clone, Default, Debug, PartialEq)]
    struct __tilecfg {
        /// 0 `or` 1
        palette: u8,
        start_row: u8,
        /// reserved, must be zero
        reserved_a0: [u8; 14],
        /// number of bytes of one row in each tile
        colsb: [u16; 8],
        /// reserved, must be zero
        reserved_b0: [u16; 8],
        /// number of rows in each tile
        rows: [u8; 8],
        /// reserved, must be zero
        reserved_c0: [u8; 8],
    }

    impl __tilecfg {
        fn new(palette: u8, start_row: u8, colsb: [u16; 8], rows: [u8; 8]) -> Self {
            Self {
                palette,
                start_row,
                reserved_a0: [0u8; 14],
                colsb,
                reserved_b0: [0u16; 8],
                rows,
                reserved_c0: [0u8; 8],
            }
        }

        const fn as_ptr(&self) -> *const u8 {
            self as *const Self as *const u8
        }

        fn as_mut_ptr(&mut self) -> *mut u8 {
            self as *mut Self as *mut u8
        }
    }

    #[cfg(not(target_os = "linux"))]
    #[target_feature(enable = "amx-tile")]
    fn _init_amx() {}

    #[cfg(target_os = "linux")]
    #[target_feature(enable = "amx-tile")]
    #[inline]
    unsafe fn _init_amx() {
        let mut ret: usize;
        let mut xfeatures: usize = 0;
        ret = syscall!(Sysno::arch_prctl, 0x1022, &mut xfeatures as *mut usize)
            .expect("arch_prctl ARCH_GET_XCOMP_PERM syscall failed");
        if ret != 0 {
            panic!("Failed to get XFEATURES");
        } else {
            match 0b11 & (xfeatures >> 17) {
                0 => panic!("AMX is not available"),
                1 => {
                    ret = syscall!(Sysno::arch_prctl, 0x1023, 18)
                        .expect("arch_prctl ARCH_REQ_XCOMP_PERM syscall failed");
                    if ret != 0 {
                        panic!("Failed to enable AMX");
                    }
                }
                3 => {}
                _ => unreachable!(),
            }
        }
    }

    #[simd_test(enable = "amx-tile")]
    unsafe fn test_tile_loadconfig() {
        let config = __tilecfg::default();
        _tile_loadconfig(config.as_ptr());
        _tile_release();
    }

    #[simd_test(enable = "amx-tile")]
    unsafe fn test_tile_storeconfig() {
        let config = __tilecfg::new(1, 0, [32; 8], [8; 8]);
        _tile_loadconfig(config.as_ptr());
        let mut _config = __tilecfg::default();
        _tile_storeconfig(_config.as_mut_ptr());
        _tile_release();
        assert_eq!(config, _config);
    }

    #[simd_test(enable = "amx-tile")]
    unsafe fn test_tile_zero() {
        _init_amx();
        let mut config = __tilecfg::default();
        config.palette = 1;
        config.colsb[0] = 64;
        config.rows[0] = 16;
        _tile_loadconfig(config.as_ptr());
        _tile_zero::<0>();
        let mut out = [[1_i8; 64]; 16];
        _tile_stored::<0>(&mut out as *mut [i8; 64] as *mut u8, 64);
        _tile_release();
        assert_eq!(out, [[0; 64]; 16]);
    }

    #[simd_test(enable = "amx-tile")]
    unsafe fn test_tile_stored() {
        _init_amx();
        let mut config = __tilecfg::default();
        config.palette = 1;
        config.colsb[0] = 64;
        config.rows[0] = 16;
        _tile_loadconfig(config.as_ptr());
        _tile_zero::<0>();
        let mut out = [[1_i8; 64]; 16];
        _tile_stored::<0>(&mut out as *mut [i8; 64] as *mut u8, 64);
        _tile_release();
        assert_eq!(out, [[0; 64]; 16]);
    }

    #[simd_test(enable = "amx-tile")]
    unsafe fn test_tile_loadd() {
        _init_amx();
        let mut config = __tilecfg::default();
        config.palette = 1;
        config.colsb[0] = 64;
        config.rows[0] = 16;
        _tile_loadconfig(config.as_ptr());
        _tile_zero::<0>();
        let mat = [1_i8; 1024];
        _tile_loadd::<0>(&mat as *const i8 as *const u8, 64);
        let mut out = [[0_i8; 64]; 16];
        _tile_stored::<0>(&mut out as *mut [i8; 64] as *mut u8, 64);
        _tile_release();
        assert_eq!(out, [[1; 64]; 16]);
    }

    #[simd_test(enable = "amx-tile")]
    unsafe fn test_tile_stream_loadd() {
        _init_amx();
        let mut config = __tilecfg::default();
        config.palette = 1;
        config.colsb[0] = 64;
        config.rows[0] = 16;
        _tile_loadconfig(config.as_ptr());
        _tile_zero::<0>();
        let mat = [1_i8; 1024];
        _tile_stream_loadd::<0>(&mat as *const i8 as *const u8, 64);
        let mut out = [[0_i8; 64]; 16];
        _tile_stored::<0>(&mut out as *mut [i8; 64] as *mut u8, 64);
        _tile_release();
        assert_eq!(out, [[1; 64]; 16]);
    }

    #[simd_test(enable = "amx-tile")]
    unsafe fn test_tile_release() {
        _tile_release();
    }

    #[simd_test(enable = "amx-bf16,avx512f")]
    unsafe fn test_tile_dpbf16ps() {
        _init_amx();
        let bf16_1: u16 = _mm_cvtness_sbh(1.0).to_bits();
        let bf16_2: u16 = _mm_cvtness_sbh(2.0).to_bits();
        let ones: [u8; 1024] = transmute([bf16_1; 512]);
        let twos: [u8; 1024] = transmute([bf16_2; 512]);
        let mut res = [[0f32; 16]; 16];
        let mut config = __tilecfg::default();
        config.palette = 1;
        (0..=2).for_each(|i| {
            config.colsb[i] = 64;
            config.rows[i] = 16;
        });
        _tile_loadconfig(config.as_ptr());
        _tile_zero::<0>();
        _tile_loadd::<1>(&ones as *const u8, 64);
        _tile_loadd::<2>(&twos as *const u8, 64);
        _tile_dpbf16ps::<0, 1, 2>();
        _tile_stored::<0>(&mut res as *mut [f32; 16] as *mut u8, 64);
        _tile_release();
        assert_eq!(res, [[64f32; 16]; 16]);
    }

    #[simd_test(enable = "amx-int8")]
    unsafe fn test_tile_dpbssd() {
        _init_amx();
        let ones = [-1_i8; 1024];
        let twos = [-2_i8; 1024];
        let mut res = [[0_i32; 16]; 16];
        let mut config = __tilecfg::default();
        config.palette = 1;
        (0..=2).for_each(|i| {
            config.colsb[i] = 64;
            config.rows[i] = 16;
        });
        _tile_loadconfig(config.as_ptr());
        _tile_zero::<0>();
        _tile_loadd::<1>(&ones as *const i8 as *const u8, 64);
        _tile_loadd::<2>(&twos as *const i8 as *const u8, 64);
        _tile_dpbssd::<0, 1, 2>();
        _tile_stored::<0>(&mut res as *mut [i32; 16] as *mut u8, 64);
        _tile_release();
        assert_eq!(res, [[128_i32; 16]; 16]);
    }

    #[simd_test(enable = "amx-int8")]
    unsafe fn test_tile_dpbsud() {
        _init_amx();
        let ones = [-1_i8; 1024];
        let twos = [2_u8; 1024];
        let mut res = [[0_i32; 16]; 16];
        let mut config = __tilecfg::default();
        config.palette = 1;
        (0..=2).for_each(|i| {
            config.colsb[i] = 64;
            config.rows[i] = 16;
        });
        _tile_loadconfig(config.as_ptr());
        _tile_zero::<0>();
        _tile_loadd::<1>(&ones as *const i8 as *const u8, 64);
        _tile_loadd::<2>(&twos as *const u8, 64);
        _tile_dpbsud::<0, 1, 2>();
        _tile_stored::<0>(&mut res as *mut [i32; 16] as *mut u8, 64);
        _tile_release();
        assert_eq!(res, [[-128_i32; 16]; 16]);
    }

    #[simd_test(enable = "amx-int8")]
    unsafe fn test_tile_dpbusd() {
        _init_amx();
        let ones = [1_u8; 1024];
        let twos = [-2_i8; 1024];
        let mut res = [[0_i32; 16]; 16];
        let mut config = __tilecfg::default();
        config.palette = 1;
        (0..=2).for_each(|i| {
            config.colsb[i] = 64;
            config.rows[i] = 16;
        });
        _tile_loadconfig(config.as_ptr());
        _tile_zero::<0>();
        _tile_loadd::<1>(&ones as *const u8, 64);
        _tile_loadd::<2>(&twos as *const i8 as *const u8, 64);
        _tile_dpbusd::<0, 1, 2>();
        _tile_stored::<0>(&mut res as *mut [i32; 16] as *mut u8, 64);
        _tile_release();
        assert_eq!(res, [[-128_i32; 16]; 16]);
    }

    #[simd_test(enable = "amx-int8")]
    unsafe fn test_tile_dpbuud() {
        _init_amx();
        let ones = [1_u8; 1024];
        let twos = [2_u8; 1024];
        let mut res = [[0_i32; 16]; 16];
        let mut config = __tilecfg::default();
        config.palette = 1;
        (0..=2).for_each(|i| {
            config.colsb[i] = 64;
            config.rows[i] = 16;
        });
        _tile_loadconfig(config.as_ptr());
        _tile_zero::<0>();
        _tile_loadd::<1>(&ones as *const u8, 64);
        _tile_loadd::<2>(&twos as *const u8, 64);
        _tile_dpbuud::<0, 1, 2>();
        _tile_stored::<0>(&mut res as *mut [i32; 16] as *mut u8, 64);
        _tile_release();
        assert_eq!(res, [[128_i32; 16]; 16]);
    }

    #[simd_test(enable = "amx-fp16")]
    unsafe fn test_tile_dpfp16ps() {
        _init_amx();
        let ones = [1f16; 512];
        let twos = [2f16; 512];
        let mut res = [[0f32; 16]; 16];
        let mut config = __tilecfg::default();
        config.palette = 1;
        (0..=2).for_each(|i| {
            config.colsb[i] = 64;
            config.rows[i] = 16;
        });
        _tile_loadconfig(config.as_ptr());
        _tile_zero::<0>();
        _tile_loadd::<1>(&ones as *const f16 as *const u8, 64);
        _tile_loadd::<2>(&twos as *const f16 as *const u8, 64);
        _tile_dpfp16ps::<0, 1, 2>();
        _tile_stored::<0>(&mut res as *mut [f32; 16] as *mut u8, 64);
        _tile_release();
        assert_eq!(res, [[64f32; 16]; 16]);
    }

    #[simd_test(enable = "amx-complex")]
    unsafe fn test_tile_cmmimfp16ps() {
        _init_amx();
        let ones = [1f16; 512];
        let twos = [2f16; 512];
        let mut res = [[0f32; 16]; 16];
        let mut config = __tilecfg::default();
        config.palette = 1;
        (0..=2).for_each(|i| {
            config.colsb[i] = 64;
            config.rows[i] = 16;
        });
        _tile_loadconfig(config.as_ptr());
        _tile_zero::<0>();
        _tile_loadd::<1>(&ones as *const f16 as *const u8, 64);
        _tile_loadd::<2>(&twos as *const f16 as *const u8, 64);
        _tile_cmmimfp16ps::<0, 1, 2>();
        _tile_stored::<0>(&mut res as *mut [f32; 16] as *mut u8, 64);
        _tile_release();
        assert_eq!(res, [[64f32; 16]; 16]);
    }

    #[simd_test(enable = "amx-complex")]
    unsafe fn test_tile_cmmrlfp16ps() {
        _init_amx();
        let ones = [1f16; 512];
        let twos = [2f16; 512];
        let mut res = [[0f32; 16]; 16];
        let mut config = __tilecfg::default();
        config.palette = 1;
        (0..=2).for_each(|i| {
            config.colsb[i] = 64;
            config.rows[i] = 16;
        });
        _tile_loadconfig(config.as_ptr());
        _tile_zero::<0>();
        _tile_loadd::<1>(&ones as *const f16 as *const u8, 64);
        _tile_loadd::<2>(&twos as *const f16 as *const u8, 64);
        _tile_cmmrlfp16ps::<0, 1, 2>();
        _tile_stored::<0>(&mut res as *mut [f32; 16] as *mut u8, 64);
        _tile_release();
        assert_eq!(res, [[0f32; 16]; 16]);
    }
}