aprender-compute 0.71.4

High-performance SIMD compute library with GPU support, LLM inference engine, and GGUF model loading (was: trueno)
Documentation
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
use super::super::microkernels::*;
use super::super::*;

// ========================================================================
// Phase 2: Microkernel Tests
// ========================================================================

#[test]
fn test_microkernel_scalar_single_k() {
    // MR=8, NR=6, K=1
    let a = vec![1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0]; // 8x1
    let b = vec![1.0, 2.0, 3.0, 4.0, 5.0, 6.0]; // 1x6
    let mut c = vec![0.0; MR * NR]; // 8x6 column-major

    microkernel_scalar(1, &a, &b, &mut c, MR);

    // c[j,i] = a[i] * b[j]
    for j in 0..NR {
        for i in 0..MR {
            let expected = a[i] * b[j];
            assert!(
                (c[j * MR + i] - expected).abs() < 1e-6,
                "Mismatch at ({}, {}): {} vs {}",
                i,
                j,
                c[j * MR + i],
                expected
            );
        }
    }
}

#[test]
fn test_microkernel_scalar_accumulation() {
    let a = vec![1.0; MR * 4]; // 8x4
    let b = vec![1.0; 4 * NR]; // 4x6
    let mut c = vec![0.0; MR * NR];

    microkernel_scalar(4, &a, &b, &mut c, MR);

    // Each output should be 4.0 (sum of 4 ones)
    for val in &c {
        assert!((val - 4.0).abs() < 1e-6);
    }
}

#[test]
#[cfg(target_arch = "x86_64")]
fn test_microkernel_avx2_matches_scalar() {
    if !is_x86_feature_detected!("avx2") || !is_x86_feature_detected!("fma") {
        return;
    }

    let k = 64;
    let a: Vec<f32> = (0..MR * k).map(|i| (i as f32) * 0.1).collect();
    let b: Vec<f32> = (0..k * NR).map(|i| (i as f32) * 0.01).collect();

    let mut c_scalar = vec![0.0; MR * NR];
    let mut c_avx2 = vec![0.0; MR * NR];

    microkernel_scalar(k, &a, &b, &mut c_scalar, MR);

    // SAFETY: test-only usage with controlled inputs
    unsafe {
        microkernel_8x6_avx2(k, a.as_ptr(), b.as_ptr(), c_avx2.as_mut_ptr(), MR);
    }

    for i in 0..MR * NR {
        let diff = (c_scalar[i] - c_avx2[i]).abs();
        let rel_diff = diff / c_scalar[i].abs().max(1e-10);
        assert!(
            rel_diff < 1e-5,
            "Mismatch at {}: scalar={}, avx2={}, rel_diff={}",
            i,
            c_scalar[i],
            c_avx2[i],
            rel_diff
        );
    }
}

// ========================================================================
// AVX2 ASM Microkernel Tests (coverage for microkernel_8x6_avx2_asm)
// ========================================================================

#[test]
#[cfg(target_arch = "x86_64")]
fn test_microkernel_avx2_asm_matches_scalar() {
    if !is_x86_feature_detected!("avx2") || !is_x86_feature_detected!("fma") {
        return;
    }

    let k = 64;
    let a: Vec<f32> = (0..MR * k).map(|i| (i as f32) * 0.1).collect();
    let b: Vec<f32> = (0..k * NR).map(|i| (i as f32) * 0.01).collect();

    let mut c_scalar = vec![0.0; MR * NR];
    let mut c_asm = vec![0.0; MR * NR];

    microkernel_scalar(k, &a, &b, &mut c_scalar, MR);

    // SAFETY: test-only usage with controlled inputs
    unsafe {
        microkernel_8x6_avx2_asm(k, a.as_ptr(), b.as_ptr(), c_asm.as_mut_ptr(), MR);
    }

    for i in 0..MR * NR {
        let diff = (c_scalar[i] - c_asm[i]).abs();
        let rel_diff = diff / c_scalar[i].abs().max(1e-10);
        assert!(
            rel_diff < 1e-4,
            "avx2_asm mismatch at {}: scalar={}, asm={}, rel_diff={}",
            i,
            c_scalar[i],
            c_asm[i],
            rel_diff
        );
    }
}

#[test]
#[cfg(target_arch = "x86_64")]
fn test_microkernel_avx2_asm_k_less_than_4() {
    if !is_x86_feature_detected!("avx2") || !is_x86_feature_detected!("fma") {
        return;
    }

    for k in [1, 2, 3] {
        let a: Vec<f32> = (0..MR * k).map(|i| (i as f32 + 1.0) * 0.5).collect();
        let b: Vec<f32> = (0..k * NR).map(|i| (i as f32 + 1.0) * 0.3).collect();

        let mut c_scalar = vec![0.0; MR * NR];
        let mut c_asm = vec![0.0; MR * NR];

        microkernel_scalar(k, &a, &b, &mut c_scalar, MR);

        // SAFETY: test-only usage with controlled inputs
        unsafe {
            microkernel_8x6_avx2_asm(k, a.as_ptr(), b.as_ptr(), c_asm.as_mut_ptr(), MR);
        }

        for i in 0..MR * NR {
            let diff = (c_scalar[i] - c_asm[i]).abs();
            assert!(
                diff < 1e-4,
                "avx2_asm k={} mismatch at {}: scalar={}, asm={}, diff={}",
                k,
                i,
                c_scalar[i],
                c_asm[i],
                diff
            );
        }
    }
}

#[test]
#[cfg(target_arch = "x86_64")]
fn test_microkernel_avx2_asm_k_with_remainder() {
    if !is_x86_feature_detected!("avx2") || !is_x86_feature_detected!("fma") {
        return;
    }

    // k=5 → 1 unrolled iteration + 1 remainder
    // k=7 → 1 unrolled iteration + 3 remainder
    for k in [5, 6, 7, 9, 13, 17] {
        let a: Vec<f32> = (0..MR * k).map(|i| ((i % 10) as f32) * 0.1).collect();
        let b: Vec<f32> = (0..k * NR).map(|i| ((i % 7) as f32) * 0.2).collect();

        let mut c_scalar = vec![0.0; MR * NR];
        let mut c_asm = vec![0.0; MR * NR];

        microkernel_scalar(k, &a, &b, &mut c_scalar, MR);

        // SAFETY: test-only usage with controlled inputs
        unsafe {
            microkernel_8x6_avx2_asm(k, a.as_ptr(), b.as_ptr(), c_asm.as_mut_ptr(), MR);
        }

        for i in 0..MR * NR {
            let diff = (c_scalar[i] - c_asm[i]).abs();
            let rel_diff = diff / c_scalar[i].abs().max(1e-10);
            assert!(
                rel_diff < 1e-4,
                "avx2_asm k={} mismatch at {}: scalar={}, asm={}, rel_diff={}",
                k,
                i,
                c_scalar[i],
                c_asm[i],
                rel_diff
            );
        }
    }
}

#[test]
#[cfg(target_arch = "x86_64")]
fn test_microkernel_avx2_asm_k_exact_multiple_of_4() {
    if !is_x86_feature_detected!("avx2") || !is_x86_feature_detected!("fma") {
        return;
    }

    for k in [4, 8, 16, 32] {
        let a: Vec<f32> = (0..MR * k).map(|i| (i as f32) * 0.05).collect();
        let b: Vec<f32> = (0..k * NR).map(|i| (i as f32) * 0.02).collect();

        let mut c_scalar = vec![0.0; MR * NR];
        let mut c_asm = vec![0.0; MR * NR];

        microkernel_scalar(k, &a, &b, &mut c_scalar, MR);

        // SAFETY: test-only usage with controlled inputs
        unsafe {
            microkernel_8x6_avx2_asm(k, a.as_ptr(), b.as_ptr(), c_asm.as_mut_ptr(), MR);
        }

        for i in 0..MR * NR {
            let diff = (c_scalar[i] - c_asm[i]).abs();
            let rel_diff = diff / c_scalar[i].abs().max(1e-10);
            assert!(
                rel_diff < 1e-4,
                "avx2_asm k={} mismatch at {}: scalar={}, asm={}",
                k,
                i,
                c_scalar[i],
                c_asm[i]
            );
        }
    }
}

// ========================================================================
// True ASM Microkernel Tests (coverage for microkernel_8x6_true_asm)
// ========================================================================

#[test]
#[cfg(target_arch = "x86_64")]
fn test_microkernel_true_asm_matches_scalar() {
    if !is_x86_feature_detected!("avx2") || !is_x86_feature_detected!("fma") {
        return;
    }

    let k = 64;
    let a: Vec<f32> = (0..MR * k).map(|i| (i as f32) * 0.1).collect();
    let b: Vec<f32> = (0..k * NR).map(|i| (i as f32) * 0.01).collect();

    let mut c_scalar = vec![0.0; MR * NR];
    let mut c_asm = vec![0.0; MR * NR];

    microkernel_scalar(k, &a, &b, &mut c_scalar, MR);

    // SAFETY: test-only usage with controlled inputs
    unsafe {
        microkernel_8x6_true_asm(k, a.as_ptr(), b.as_ptr(), c_asm.as_mut_ptr(), MR);
    }

    for i in 0..MR * NR {
        let diff = (c_scalar[i] - c_asm[i]).abs();
        let rel_diff = diff / c_scalar[i].abs().max(1e-10);
        assert!(
            rel_diff < 1e-4,
            "true_asm mismatch at {}: scalar={}, asm={}, rel_diff={}",
            i,
            c_scalar[i],
            c_asm[i],
            rel_diff
        );
    }
}

#[test]
#[cfg(target_arch = "x86_64")]
fn test_microkernel_true_asm_k_less_than_4() {
    if !is_x86_feature_detected!("avx2") || !is_x86_feature_detected!("fma") {
        return;
    }

    for k in [1, 2, 3] {
        let a: Vec<f32> = (0..MR * k).map(|i| (i as f32 + 1.0) * 0.5).collect();
        let b: Vec<f32> = (0..k * NR).map(|i| (i as f32 + 1.0) * 0.3).collect();

        let mut c_scalar = vec![0.0; MR * NR];
        let mut c_asm = vec![0.0; MR * NR];

        microkernel_scalar(k, &a, &b, &mut c_scalar, MR);

        // SAFETY: test-only usage with controlled inputs
        unsafe {
            microkernel_8x6_true_asm(k, a.as_ptr(), b.as_ptr(), c_asm.as_mut_ptr(), MR);
        }

        for i in 0..MR * NR {
            let diff = (c_scalar[i] - c_asm[i]).abs();
            assert!(
                diff < 1e-4,
                "true_asm k={} mismatch at {}: scalar={}, asm={}, diff={}",
                k,
                i,
                c_scalar[i],
                c_asm[i],
                diff
            );
        }
    }
}

#[test]
#[cfg(target_arch = "x86_64")]
fn test_microkernel_true_asm_k_with_remainder() {
    if !is_x86_feature_detected!("avx2") || !is_x86_feature_detected!("fma") {
        return;
    }

    // Test k values that exercise both main loop and remainder
    for k in [5, 6, 7, 9, 11, 15] {
        let a: Vec<f32> = (0..MR * k).map(|i| ((i % 10) as f32) * 0.1).collect();
        let b: Vec<f32> = (0..k * NR).map(|i| ((i % 7) as f32) * 0.2).collect();

        let mut c_scalar = vec![0.0; MR * NR];
        let mut c_asm = vec![0.0; MR * NR];

        microkernel_scalar(k, &a, &b, &mut c_scalar, MR);

        // SAFETY: test-only usage with controlled inputs
        unsafe {
            microkernel_8x6_true_asm(k, a.as_ptr(), b.as_ptr(), c_asm.as_mut_ptr(), MR);
        }

        for i in 0..MR * NR {
            let diff = (c_scalar[i] - c_asm[i]).abs();
            let rel_diff = diff / c_scalar[i].abs().max(1e-10);
            assert!(
                rel_diff < 1e-4,
                "true_asm k={} mismatch at {}: scalar={}, asm={}, rel_diff={}",
                k,
                i,
                c_scalar[i],
                c_asm[i],
                rel_diff
            );
        }
    }
}

#[test]
#[cfg(target_arch = "x86_64")]
fn test_microkernel_true_asm_k_exact_multiple_of_4() {
    if !is_x86_feature_detected!("avx2") || !is_x86_feature_detected!("fma") {
        return;
    }

    for k in [4, 8, 12, 16, 32, 64] {
        let a: Vec<f32> = (0..MR * k).map(|i| (i as f32) * 0.05).collect();
        let b: Vec<f32> = (0..k * NR).map(|i| (i as f32) * 0.02).collect();

        let mut c_scalar = vec![0.0; MR * NR];
        let mut c_asm = vec![0.0; MR * NR];

        microkernel_scalar(k, &a, &b, &mut c_scalar, MR);

        // SAFETY: test-only usage with controlled inputs
        unsafe {
            microkernel_8x6_true_asm(k, a.as_ptr(), b.as_ptr(), c_asm.as_mut_ptr(), MR);
        }

        for i in 0..MR * NR {
            let diff = (c_scalar[i] - c_asm[i]).abs();
            let rel_diff = diff / c_scalar[i].abs().max(1e-10);
            assert!(
                rel_diff < 1e-4,
                "true_asm k={} mismatch at {}: scalar={}, asm={}",
                k,
                i,
                c_scalar[i],
                c_asm[i]
            );
        }
    }
}

#[test]
#[cfg(target_arch = "x86_64")]
fn test_microkernel_true_asm_accumulates_into_c() {
    if !is_x86_feature_detected!("avx2") || !is_x86_feature_detected!("fma") {
        return;
    }

    let k = 8;
    let a: Vec<f32> = vec![1.0; MR * k];
    let b: Vec<f32> = vec![1.0; k * NR];

    // Pre-fill C with non-zero values
    let mut c = vec![10.0f32; MR * NR];

    // SAFETY: test-only usage with controlled inputs
    unsafe {
        microkernel_8x6_true_asm(k, a.as_ptr(), b.as_ptr(), c.as_mut_ptr(), MR);
    }

    // Each output should be 10.0 + k = 18.0
    for (i, val) in c.iter().enumerate() {
        assert!(
            (*val - 18.0).abs() < 1e-3,
            "true_asm accumulation at {}: expected 18.0, got {}",
            i,
            val
        );
    }
}

// ========================================================================
// FALSIFY-BLIS-ASM-OOB: the asm kernel must not read past its A or B panel
// ========================================================================
//
// The last ir panel of a packed A buffer ends exactly at the end of its
// allocation. A kernel that reads even one byte past the panel can cross
// into an unmapped page and fault, and only when the allocator happens to
// put the buffer there. These tests put the panel end on a PROT_NONE page,
// so any read past it faults on every run instead of one run in many.

#[cfg(all(target_arch = "x86_64", target_os = "linux"))]
mod guard_page {
    use std::ffi::c_void;

    const PROT_NONE: i32 = 0;
    const PROT_READ: i32 = 1;
    const PROT_WRITE: i32 = 2;
    const MAP_PRIVATE: i32 = 0x02;
    const MAP_ANONYMOUS: i32 = 0x20;
    const SC_PAGESIZE: i32 = 30;

    extern "C" {
        fn mmap(
            addr: *mut c_void,
            len: usize,
            prot: i32,
            flags: i32,
            fd: i32,
            off: i64,
        ) -> *mut c_void;
        fn mprotect(addr: *mut c_void, len: usize, prot: i32) -> i32;
        fn munmap(addr: *mut c_void, len: usize) -> i32;
        fn sysconf(name: i32) -> i64;
    }

    /// `len` f32s whose last element is the last readable byte before a
    /// PROT_NONE page.
    pub struct GuardedF32 {
        base: *mut c_void,
        map_len: usize,
        data: *mut f32,
        len: usize,
    }

    impl GuardedF32 {
        pub fn new(values: &[f32]) -> Self {
            // SAFETY: plain libc calls; every result is checked before use.
            unsafe {
                let page = usize::try_from(sysconf(SC_PAGESIZE)).expect("page size");
                let bytes = values.len() * 4;
                let data_pages = bytes.div_ceil(page).max(1);
                let map_len = (data_pages + 1) * page;
                let base = mmap(
                    std::ptr::null_mut(),
                    map_len,
                    PROT_READ | PROT_WRITE,
                    MAP_PRIVATE | MAP_ANONYMOUS,
                    -1,
                    0,
                );
                assert!(base as isize != -1, "mmap failed");
                let guard = base.cast::<u8>().add(data_pages * page);
                assert_eq!(mprotect(guard.cast(), page, PROT_NONE), 0, "mprotect failed");
                let data = guard.sub(bytes).cast::<f32>();
                std::ptr::copy_nonoverlapping(values.as_ptr(), data, values.len());
                Self { base, map_len, data, len: values.len() }
            }
        }

        pub fn as_ptr(&self) -> *const f32 {
            self.data
        }

        pub fn len(&self) -> usize {
            self.len
        }
    }

    impl Drop for GuardedF32 {
        fn drop(&mut self) {
            // SAFETY: base/map_len are exactly what mmap returned.
            unsafe {
                munmap(self.base, self.map_len);
            }
        }
    }
}

#[test]
#[cfg(all(target_arch = "x86_64", target_os = "linux"))]
fn falsify_blis_asm_oob_001_true_asm_stays_inside_panels() {
    if !is_x86_feature_detected!("avx2") || !is_x86_feature_detected!("fma") {
        return;
    }

    // Multiples of 4 (the unroll), each remainder, and the KC block size.
    for k in [4usize, 5, 6, 7, 8, 12, 64, 255, 256] {
        let a_vals: Vec<f32> = (0..MR * k).map(|i| ((i % 17) as f32) * 0.25 - 2.0).collect();
        let b_vals: Vec<f32> = (0..k * NR).map(|i| ((i % 13) as f32) * 0.5 - 3.0).collect();
        let a = guard_page::GuardedF32::new(&a_vals);
        let b = guard_page::GuardedF32::new(&b_vals);
        assert_eq!(a.len(), MR * k);
        assert_eq!(b.len(), k * NR);

        let mut c_ref = vec![1.0f32; MR * NR];
        let mut c_asm = vec![1.0f32; MR * NR];
        microkernel_scalar(k, &a_vals, &b_vals, &mut c_ref, MR);

        // SAFETY: A holds MR*k and B holds k*NR f32s; C holds MR*NR with ldc = MR.
        unsafe {
            microkernel_8x6_true_asm(k, a.as_ptr(), b.as_ptr(), c_asm.as_mut_ptr(), MR);
        }

        for i in 0..MR * NR {
            let tol = 1e-4 * c_ref[i].abs().max(1.0);
            assert!(
                (c_ref[i] - c_asm[i]).abs() <= tol,
                "FALSIFY-BLIS-ASM-OOB-001 k={k} i={i}: scalar {} vs asm {}",
                c_ref[i],
                c_asm[i]
            );
        }
    }
}