vyre-self-substrate 0.6.3

Vyre self-substrate: vyre using its own primitives on its own scheduler problems. The recursion-thesis layer between vyre-primitives and vyre-driver.
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
use super::*;
use crate::dispatch_buffers::u32_slice_to_le_bytes;
use crate::optimizer::dispatcher::{DispatchError, OptimizerDispatcher};
use std::sync::Mutex;
use vyre_foundation::ir::Program;

mod reference_contracts;
mod release_path_contracts;

struct CsrChangedDispatcher {
    outputs: Vec<Vec<u8>>,
}

impl OptimizerDispatcher for CsrChangedDispatcher {
    fn dispatch(
        &self,
        _program: &Program,
        inputs: &[Vec<u8>],
        _grid_override: Option<[u32; 3]>,
    ) -> Result<Vec<Vec<u8>>, DispatchError> {
        if inputs.len() != 7 && inputs.len() != 8 {
            return Err(DispatchError::BadInputs(format!(
                "Fix: csr_forward_or_changed test dispatcher expected 7 legacy inputs or 8 changed-history inputs, got {}.",
                inputs.len()
            )));
        }
        Ok(self.outputs.clone())
    }
}

struct RecordingCsrChangedDispatcher {
    outputs: Vec<Vec<u8>>,
    frontier_inputs: Mutex<Vec<Vec<u32>>>,
}

impl OptimizerDispatcher for RecordingCsrChangedDispatcher {
    fn dispatch(
        &self,
        _program: &Program,
        inputs: &[Vec<u8>],
        _grid_override: Option<[u32; 3]>,
    ) -> Result<Vec<Vec<u8>>, DispatchError> {
        self.frontier_inputs
            .lock()
            .expect("Fix: frontier recording mutex should not be poisoned")
            .push(crate::hardware::dispatch_buffers::read_u32s(&inputs[5]));
        Ok(self.outputs.clone())
    }
}

struct StaticCsrInputRecordingDispatcher {
    outputs: Vec<Vec<u8>>,
    edge_targets: Mutex<Vec<Vec<u32>>>,
}

impl OptimizerDispatcher for StaticCsrInputRecordingDispatcher {
    fn dispatch(
        &self,
        _program: &Program,
        inputs: &[Vec<u8>],
        _grid_override: Option<[u32; 3]>,
    ) -> Result<Vec<Vec<u8>>, DispatchError> {
        self.edge_targets
            .lock()
            .expect("Fix: static input recording mutex should not be poisoned")
            .push(crate::hardware::dispatch_buffers::read_u32s(&inputs[2]));
        Ok(self.outputs.clone())
    }
}

fn linear_graph() -> (Vec<u32>, Vec<u32>, Vec<u32>) {
    // 0 -> 1 -> 2 -> 3
    (vec![0, 1, 2, 3, 3], vec![1, 2, 3], vec![1, 1, 1])
}

#[test]
fn gpu_into_decodes_exact_outputs_into_reused_frontier() {
    let dispatcher = CsrChangedDispatcher {
        outputs: vec![
            u32_slice_to_le_bytes(&[0b1111]),
            u32_slice_to_le_bytes(&[0, 0, 0, 0]),
        ],
    };
    let (off, tgt, msk) = linear_graph();
    let mut frontier = Vec::with_capacity(4);
    let ptr = frontier.as_ptr();
    forward_closure_via_change_flag_gpu_into(
        &dispatcher,
        4,
        &off,
        &tgt,
        &msk,
        &[0b0001],
        0xFFFF_FFFF,
        4,
        &mut frontier,
    )
    .expect("Fix: dispatch succeeds");
    assert_eq!(frontier, vec![0b1111]);
    assert_eq!(frontier.as_ptr(), ptr);
}

#[test]
fn gpu_rejects_extra_outputs() {
    let dispatcher = CsrChangedDispatcher {
        outputs: vec![
            u32_slice_to_le_bytes(&[0b1111]),
            u32_slice_to_le_bytes(&[0, 0, 0, 0]),
            u32_slice_to_le_bytes(&[99]),
        ],
    };
    let (off, tgt, msk) = linear_graph();
    let err = forward_closure_via_change_flag_gpu(
        &dispatcher,
        4,
        &off,
        &tgt,
        &msk,
        &[0b0001],
        0xFFFF_FFFF,
        4,
    )
    .expect_err("extra outputs must be rejected");
    assert!(
        matches!(err, DispatchError::BackendError(_)),
        "unexpected error: {err:?}"
    );
}

#[test]
fn gpu_rejects_trailing_changed_bytes() {
    let dispatcher = CsrChangedDispatcher {
        outputs: vec![u32_slice_to_le_bytes(&[0b1111]), vec![0, 0, 0, 0, 1]],
    };
    let (off, tgt, msk) = linear_graph();
    let err = forward_closure_via_change_flag_gpu(
        &dispatcher,
        4,
        &off,
        &tgt,
        &msk,
        &[0b0001],
        0xFFFF_FFFF,
        4,
    )
    .expect_err("trailing changed bytes must be rejected");
    assert!(
        matches!(err, DispatchError::BackendError(_)),
        "unexpected error: {err:?}"
    );
}

#[test]
fn gpu_rejects_non_boolean_changed_flag() {
    let dispatcher = CsrChangedDispatcher {
        outputs: vec![
            u32_slice_to_le_bytes(&[0b1111]),
            u32_slice_to_le_bytes(&[2]),
        ],
    };
    let (off, tgt, msk) = linear_graph();
    let err = forward_closure_via_change_flag_gpu(
        &dispatcher,
        4,
        &off,
        &tgt,
        &msk,
        &[0b0001],
        0xFFFF_FFFF,
        1,
    )
    .expect_err("non-boolean changed flag must be rejected");
    assert!(
        matches!(err, DispatchError::BackendError(_)),
        "unexpected error: {err:?}"
    );
}

#[test]
fn gpu_rejects_bad_seed_width_without_clobbering_frontier() {
    struct NoDispatch;

    impl OptimizerDispatcher for NoDispatch {
        fn dispatch(
            &self,
            _program: &Program,
            _inputs: &[Vec<u8>],
            _grid_override: Option<[u32; 3]>,
        ) -> Result<Vec<Vec<u8>>, DispatchError> {
            panic!("bad seed width must be rejected before dispatch");
        }
    }

    let (off, tgt, msk) = linear_graph();
    let mut scratch = ForwardChangedGpuScratch::default();
    let mut frontier = vec![0xCAFE_BABEu32];
    let capacity = frontier.capacity();

    let err = forward_closure_via_change_flag_gpu_with_scratch_into(
        &NoDispatch,
        4,
        &off,
        &tgt,
        &msk,
        &[],
        0xFFFF_FFFF,
        5,
        &mut scratch,
        &mut frontier,
    )
    .expect_err("bad seed width must be rejected before mutating reusable frontier storage");

    assert!(matches!(err, DispatchError::BadInputs(_)));
    assert_eq!(frontier, vec![0xCAFE_BABEu32]);
    assert_eq!(frontier.capacity(), capacity);
    assert!(scratch.inputs.is_empty());
    assert_eq!(scratch.program_builds(), 0);
}

#[test]
fn gpu_reuses_dispatch_input_buffers() {
    let dispatcher = CsrChangedDispatcher {
        outputs: vec![
            u32_slice_to_le_bytes(&[0b1111]),
            u32_slice_to_le_bytes(&[0, 0, 0, 0]),
        ],
    };
    let (off, tgt, msk) = linear_graph();
    let mut scratch =
        ForwardChangedGpuScratch::with_input_capacities(&[32, 32, 32, 32, 32, 32, 32, 8], 1);
    let mut frontier = Vec::with_capacity(4);
    let input_caps = scratch.inputs.iter().map(Vec::capacity).collect::<Vec<_>>();
    let frontier_ptr = frontier.as_ptr();
    forward_closure_via_change_flag_gpu_with_scratch_into(
        &dispatcher,
        4,
        &off,
        &tgt,
        &msk,
        &[0b0001],
        0xFFFF_FFFF,
        4,
        &mut scratch,
        &mut frontier,
    )
    .unwrap();
    assert_eq!(
        scratch.inputs.iter().map(Vec::capacity).collect::<Vec<_>>(),
        input_caps
    );
    assert_eq!(frontier.as_ptr(), frontier_ptr);
    assert_eq!(frontier, vec![0b1111]);
}

#[test]
fn gpu_refreshes_static_inputs_when_same_shape_graph_content_changes() {
    let dispatcher = StaticCsrInputRecordingDispatcher {
        outputs: vec![
            u32_slice_to_le_bytes(&[0b0001]),
            u32_slice_to_le_bytes(&[0]),
        ],
        edge_targets: Mutex::new(Vec::new()),
    };
    let edge_offsets = vec![0, 1, 2, 3, 3];
    let first_targets = vec![1, 2, 3];
    let second_targets = vec![2, 3, 0];
    let edge_kind_mask = vec![1, 1, 1];
    let mut scratch = ForwardChangedGpuScratch::default();
    let mut frontier = Vec::new();

    forward_closure_via_change_flag_gpu_with_scratch_into(
        &dispatcher,
        4,
        &edge_offsets,
        &first_targets,
        &edge_kind_mask,
        &[0b0001],
        0xFFFF_FFFF,
        1,
        &mut scratch,
        &mut frontier,
    )
    .expect("Fix: first same-shape dispatch should succeed");
    forward_closure_via_change_flag_gpu_with_scratch_into(
        &dispatcher,
        4,
        &edge_offsets,
        &second_targets,
        &edge_kind_mask,
        &[0b0001],
        0xFFFF_FFFF,
        1,
        &mut scratch,
        &mut frontier,
    )
    .expect("Fix: second same-shape dispatch should refresh static CSR inputs");

    let recorded_targets = dispatcher
        .edge_targets
        .lock()
        .expect("Fix: static input recording mutex should not be poisoned");
    assert_eq!(
        recorded_targets.as_slice(),
        &[first_targets, second_targets]
    );
    assert_eq!(
        scratch.program_builds(),
        1,
        "Fix: same-shape graph content changes should refresh staged static inputs without rebuilding the primitive program."
    );
}

#[test]
fn gpu_reuses_cached_program_by_primitive_key() {
    let history_dispatcher = CsrChangedDispatcher {
        outputs: vec![
            u32_slice_to_le_bytes(&[0b1111]),
            u32_slice_to_le_bytes(&[0, 0, 0, 0]),
        ],
    };
    let legacy_dispatcher = CsrChangedDispatcher {
        outputs: vec![
            u32_slice_to_le_bytes(&[0b1111]),
            u32_slice_to_le_bytes(&[0]),
        ],
    };
    let (off, tgt, msk) = linear_graph();
    let mut scratch = ForwardChangedGpuScratch::default();
    let mut frontier = Vec::new();

    forward_closure_via_change_flag_gpu_with_scratch_into(
        &history_dispatcher,
        4,
        &off,
        &tgt,
        &msk,
        &[0b0001],
        0xFFFF_FFFF,
        4,
        &mut scratch,
        &mut frontier,
    )
    .expect("Fix: first changed-history dispatch should build one program");
    assert_eq!(scratch.program_builds(), 1);

    forward_closure_via_change_flag_gpu_with_scratch_into(
        &history_dispatcher,
        4,
        &off,
        &tgt,
        &msk,
        &[0b0011],
        0xFFFF_FFFF,
        4,
        &mut scratch,
        &mut frontier,
    )
    .expect("Fix: identical primitive key should reuse the cached program");
    assert_eq!(scratch.program_builds(), 1);

    forward_closure_via_change_flag_gpu_with_scratch_into(
        &history_dispatcher,
        4,
        &off,
        &tgt,
        &msk,
        &[0b0001],
        0b0001,
        4,
        &mut scratch,
        &mut frontier,
    )
    .expect("Fix: changed allow mask should rebuild the primitive program");
    assert_eq!(scratch.program_builds(), 2);

    forward_closure_via_change_flag_gpu_with_scratch_into(
        &legacy_dispatcher,
        4,
        &off,
        &tgt,
        &msk,
        &[0b0001],
        0b0001,
        65,
        &mut scratch,
        &mut frontier,
    )
    .expect("Fix: switching changed-history policy should rebuild the program");
    assert_eq!(scratch.program_builds(), 3);
}

#[test]

fn gpu_rejects_mismatched_edge_arrays() {
    let dispatcher = CsrChangedDispatcher {
        outputs: vec![
            u32_slice_to_le_bytes(&[0b1111]),
            u32_slice_to_le_bytes(&[0, 0, 0, 0]),
        ],
    };
    let err = forward_closure_via_change_flag_gpu(
        &dispatcher,
        2,
        &[0, 1, 1],
        &[1],
        &[],
        &[0b01],
        0xFFFF_FFFF,
        1,
    )
    .expect_err("mismatched edge arrays must be rejected");
    assert!(matches!(err, DispatchError::BadInputs(_)));
}

#[test]
fn generated_gpu_seed_copy_bounds_to_primitive_frontier_words() {
    for node_count in 1u32..=512 {
        let frontier_words = node_count.div_ceil(32) as usize;
        let edge_offsets = vec![0; node_count as usize + 1];
        for extra_words in 0..8usize {
            let seed_len = frontier_words + extra_words;
            let seed = (0..seed_len)
                .map(|idx| 0xA5A5_0000u32 ^ idx as u32 ^ node_count)
                .collect::<Vec<_>>();
            let dispatcher = RecordingCsrChangedDispatcher {
                outputs: vec![
                    u32_slice_to_le_bytes(&vec![0; frontier_words]),
                    u32_slice_to_le_bytes(&[0]),
                ],
                frontier_inputs: Mutex::new(Vec::new()),
            };
            let mut frontier = Vec::new();

            let result = forward_closure_via_change_flag_gpu_into(
                &dispatcher,
                node_count,
                &edge_offsets,
                &[],
                &[],
                &seed,
                0xFFFF_FFFF,
                1,
                &mut frontier,
            );

            if extra_words == 0 {
                result.expect("Fix: exact-width empty-edge generated CSR closure should dispatch");
                let observed = dispatcher
                    .frontier_inputs
                    .lock()
                    .expect("Fix: frontier recording mutex should not be poisoned");
                assert_eq!(
                    observed.len(),
                    1,
                    "node_count={node_count} extra_words={extra_words}"
                );
                assert_eq!(
                    observed[0],
                    seed[..frontier_words],
                    "node_count={node_count} extra_words={extra_words}"
                );
            } else {
                let err = result.expect_err(
                    "Fix: oversized generated seed must be rejected instead of silently truncated",
                );
                assert!(
                    matches!(err, DispatchError::BadInputs(_)),
                    "node_count={node_count} extra_words={extra_words} err={err:?}"
                );
                let observed = dispatcher
                    .frontier_inputs
                    .lock()
                    .expect("Fix: frontier recording mutex should not be poisoned");
                assert!(
                    observed.is_empty(),
                    "node_count={node_count} extra_words={extra_words}"
                );
            }
        }
    }
}