primal-barracuda 0.4.0

barraCuda — BARrier-free Rust Abstracted Cross-platform Unified Dimensional Algebra. Vendor-agnostic GPU compute via WGSL.
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
// SPDX-License-Identifier: AGPL-3.0-or-later
use super::*;

fn rewrite_and_resolve(wgsl: &str) -> String {
    let rewritten = rewrite_f64_infix_to_df64(wgsl).expect("should parse");

    resolve_spans(&rewritten, wgsl)
}

#[test]
fn test_count_f64_infix_simple_add() {
    let wgsl = r"
@group(0) @binding(0) var<storage, read> input: array<f64>;
@group(0) @binding(1) var<storage, read_write> output: array<f64>;

@compute @workgroup_size(64)
fn main(@builtin(global_invocation_id) gid: vec3<u32>) {
let i = gid.x;
let a = input[i];
let b = input[i + 1u];
output[i] = a + b;
}
";
    let count = count_f64_infix_ops(wgsl).expect("should parse");
    assert!(count >= 1, "expected at least 1 f64 infix op, got {count}");
}

#[test]
fn test_count_f64_infix_no_f64_ops() {
    let wgsl = r"
@group(0) @binding(0) var<storage, read> input: array<f32>;
@group(0) @binding(1) var<storage, read_write> output: array<f32>;

@compute @workgroup_size(64)
fn main(@builtin(global_invocation_id) gid: vec3<u32>) {
let i = gid.x;
output[i] = input[i] + 1.0;
}
";
    let count = count_f64_infix_ops(wgsl).expect("should parse");
    assert_eq!(count, 0, "f32 ops should not be counted");
}

#[test]
fn test_rewrite_simple_add() {
    let wgsl = r"
@group(0) @binding(0) var<storage, read> input: array<f64>;
@group(0) @binding(1) var<storage, read_write> output: array<f64>;

@compute @workgroup_size(64)
fn main(@builtin(global_invocation_id) gid: vec3<u32>) {
let i = gid.x;
let a = input[i];
let b = input[i + 1u];
output[i] = a + b;
}
";
    let result = rewrite_and_resolve(wgsl);
    assert!(
        result.contains("_df64_add_f64("),
        "should contain bridge call, got:\n{result}"
    );
}

#[test]
fn test_rewrite_nested_ops() {
    let wgsl = r"
@group(0) @binding(0) var<storage, read> input: array<f64>;
@group(0) @binding(1) var<storage, read_write> output: array<f64>;

@compute @workgroup_size(64)
fn main(@builtin(global_invocation_id) gid: vec3<u32>) {
let i = gid.x;
let a = input[i];
let b = input[i + 1u];
let c = input[i + 2u];
output[i] = (a + b) * c;
}
";
    let result = rewrite_and_resolve(wgsl);
    assert!(
        result.contains("_df64_mul_f64("),
        "should contain mul bridge, got:\n{result}"
    );
    assert!(
        result.contains("_df64_add_f64("),
        "should contain add bridge for nested op, got:\n{result}"
    );
}

#[test]
fn test_rewrite_preserves_u32_ops() {
    let wgsl = r"
@group(0) @binding(0) var<storage, read> input: array<f64>;
@group(0) @binding(1) var<storage, read_write> output: array<f64>;

@compute @workgroup_size(64)
fn main(@builtin(global_invocation_id) gid: vec3<u32>) {
let i = gid.x;
let j = i + 1u;
let a = input[i];
let b = input[j];
output[i] = a * b;
}
";
    let result = rewrite_and_resolve(wgsl);
    assert!(
        result.contains("_df64_mul_f64("),
        "f64 mul should be rewritten, got:\n{result}"
    );
}

#[test]
fn test_rewrite_negation() {
    let wgsl = r"
@group(0) @binding(0) var<storage, read> input: array<f64>;
@group(0) @binding(1) var<storage, read_write> output: array<f64>;

@compute @workgroup_size(64)
fn main(@builtin(global_invocation_id) gid: vec3<u32>) {
let i = gid.x;
let a = input[i];
output[i] = -a;
}
";
    let result = rewrite_and_resolve(wgsl);
    assert!(
        result.contains("_df64_neg_f64("),
        "negation should be rewritten, got:\n{result}"
    );
}

/// End-to-end: full pipeline produces valid WGSL when compiled with df64 core.
#[test]
fn test_full_pipeline_validates() {
    let f64_source = r"
@group(0) @binding(0) var<storage, read> input: array<f64>;
@group(0) @binding(1) var<storage, read_write> output: array<f64>;

@compute @workgroup_size(64)
fn main(@builtin(global_invocation_id) gid: vec3<u32>) {
let i = gid.x;
let a = input[i];
let b = input[i + 1u];
let c = input[i + 2u];
let sum = a + b;
let product = sum * c;
let result = product - a;
output[i] = result;
}
";
    let rewritten = rewrite_f64_infix_full(f64_source).expect("rewrite");

    const DF64_CORE: &str = include_str!("../../../shaders/math/df64_core.wgsl");
    const DF64_TRANS: &str = include_str!("../../../shaders/math/df64_transcendentals.wgsl");
    let full = format!("{DF64_CORE}\n{DF64_TRANS}\n{rewritten}");

    let module = naga::front::wgsl::parse_str(&full)
        .unwrap_or_else(|e| panic!("should parse: {e}\n\nSource:\n{full}"));
    let mut v = naga::valid::Validator::new(
        naga::valid::ValidationFlags::all(),
        naga::valid::Capabilities::all(),
    );
    v.validate(&module)
        .unwrap_or_else(|e| panic!("should validate: {e}\n\nSource:\n{full}"));
}

#[test]
fn test_fma_pattern() {
    let wgsl = r"
@group(0) @binding(0) var<storage, read> a_buf: array<f64>;
@group(0) @binding(1) var<storage, read> b_buf: array<f64>;
@group(0) @binding(2) var<storage, read> c_buf: array<f64>;
@group(0) @binding(3) var<storage, read_write> out: array<f64>;

@compute @workgroup_size(64)
fn main(@builtin(global_invocation_id) gid: vec3<u32>) {
let i = gid.x;
let a = a_buf[i];
let b = b_buf[i];
let c = c_buf[i];
out[i] = a * b + c;
}
";
    let result = rewrite_and_resolve(wgsl);
    assert!(
        result.contains("_df64_add_f64(") && result.contains("_df64_mul_f64("),
        "FMA pattern should produce both, got:\n{result}"
    );
}

// ══════════════════════════════════════════════════════════════
// Chaos tests for the naga rewriter
// ══════════════════════════════════════════════════════════════

#[test]
fn test_chaos_naga_f32_only_shader() {
    let wgsl = r"
@group(0) @binding(0) var<storage, read> input: array<f32>;
@group(0) @binding(1) var<storage, read_write> output: array<f32>;

@compute @workgroup_size(64)
fn main(@builtin(global_invocation_id) gid: vec3<u32>) {
output[gid.x] = input[gid.x] + 1.0;
}
";
    let result = rewrite_f64_infix_to_df64(wgsl).expect("should not fail on f32 shader");
    assert_eq!(result, wgsl, "f32-only shader should be returned unchanged");
}

#[test]
fn test_chaos_naga_invalid_wgsl() {
    let result = rewrite_f64_infix_to_df64("this is not valid wgsl");
    assert!(result.is_err(), "invalid WGSL should return Err");
}

#[test]
fn test_chaos_naga_empty_shader() {
    // Empty source may or may not parse/validate in naga.
    // Either way, the function should not panic.
    let result = rewrite_f64_infix_to_df64("");
    // If naga accepts it, the result is the source unchanged.
    // If naga rejects it, we get an Err. Both are acceptable.
    if let Ok(s) = result {
        assert_eq!(s, "", "empty in, empty out");
    }
}

#[test]
fn test_chaos_count_mixed_f32_f64_ops() {
    let wgsl = r"
@group(0) @binding(0) var<storage, read> input: array<f64>;
@group(0) @binding(1) var<storage, read_write> output: array<f64>;

@compute @workgroup_size(64)
fn main(@builtin(global_invocation_id) gid: vec3<u32>) {
let i = gid.x;
let j = i + 1u;
let k = j * 2u;
let a = input[i];
let b = input[k];
output[i] = a + b;
}
";
    let count = count_f64_infix_ops(wgsl).expect("should parse");
    assert!(
        count >= 1,
        "should count f64 add but not u32 ops, got {count}"
    );
}

#[test]
fn test_chaos_naga_subtraction_chain() {
    let wgsl = r"
@group(0) @binding(0) var<storage, read> input: array<f64>;
@group(0) @binding(1) var<storage, read_write> output: array<f64>;

@compute @workgroup_size(64)
fn main(@builtin(global_invocation_id) gid: vec3<u32>) {
let i = gid.x;
let a = input[i];
let b = input[i + 1u];
let c = input[i + 2u];
let d = input[i + 3u];
output[i] = a - b - c - d;
}
";
    let result = rewrite_and_resolve(wgsl);
    assert!(
        result.contains("_df64_sub_f64("),
        "subtraction chain should use sub bridges, got:\n{result}"
    );
}

#[test]
fn test_chaos_naga_division() {
    let wgsl = r"
@group(0) @binding(0) var<storage, read> input: array<f64>;
@group(0) @binding(1) var<storage, read_write> output: array<f64>;

@compute @workgroup_size(64)
fn main(@builtin(global_invocation_id) gid: vec3<u32>) {
let i = gid.x;
output[i] = input[i] / input[i + 1u];
}
";
    let result = rewrite_and_resolve(wgsl);
    assert!(
        result.contains("_df64_div_f64("),
        "division should be rewritten, got:\n{result}"
    );
}

// ══════════════════════════════════════════════════════════════
// Fault tests for the naga rewriter
// ══════════════════════════════════════════════════════════════

#[test]
fn test_fault_resolve_spans_empty() {
    let result = resolve_spans("no spans here", "original");
    assert_eq!(result, "no spans here");
}

#[test]
fn test_fault_resolve_spans_valid() {
    let original = "hello world";
    let rewritten = "prefix __SPAN__0__5 suffix";
    let result = resolve_spans(rewritten, original);
    assert_eq!(result, "prefix hello suffix");
}

#[test]
fn test_fault_resolve_spans_out_of_bounds() {
    let original = "short";
    let rewritten = "__SPAN__0__999";
    let result = resolve_spans(rewritten, original);
    // Should not panic, marker stays as-is
    assert!(
        result.contains("__SPAN__"),
        "out of bounds should leave marker"
    );
}

#[test]
fn test_fault_resolve_spans_inverted_range() {
    let original = "hello world";
    let rewritten = "__SPAN__5__0";
    let result = resolve_spans(rewritten, original);
    assert!(
        result.contains("__SPAN__"),
        "inverted range should leave marker"
    );
}

#[test]
fn test_fault_bridge_functions_defined() {
    let bf = bridge_functions();
    assert!(bf.contains("_df64_add_f64"), "add bridge");
    assert!(bf.contains("_df64_sub_f64"), "sub bridge");
    assert!(bf.contains("_df64_mul_f64"), "mul bridge");
    assert!(bf.contains("_df64_div_f64"), "div bridge");
    assert!(bf.contains("_df64_neg_f64"), "neg bridge");
    assert!(bf.contains("_df64_gt_f64"), "gt bridge");
    assert!(bf.contains("_df64_lt_f64"), "lt bridge");
    assert!(bf.contains("_df64_gte_f64"), "gte bridge");
    assert!(bf.contains("_df64_lte_f64"), "lte bridge");
}

#[test]
fn test_fault_full_pipeline_no_f64_ops_passthrough() {
    let wgsl = r"
@group(0) @binding(0) var<storage, read> input: array<f32>;
@group(0) @binding(1) var<storage, read_write> output: array<f32>;

@compute @workgroup_size(64)
fn main(@builtin(global_invocation_id) gid: vec3<u32>) {
output[gid.x] = input[gid.x] * 2.0;
}
";
    let result = rewrite_f64_infix_full(wgsl).expect("should succeed");
    assert_eq!(result, wgsl, "no f64 ops = passthrough unchanged");
}

#[test]
fn test_fault_dedup_preserves_non_overlapping() {
    let mut replacements = vec![
        Replacement {
            span_start: 100,
            span_end: 110,
            text: "A".into(),
        },
        Replacement {
            span_start: 50,
            span_end: 60,
            text: "B".into(),
        },
        Replacement {
            span_start: 10,
            span_end: 20,
            text: "C".into(),
        },
    ];
    dedup_overlapping(&mut replacements);
    assert_eq!(replacements.len(), 3, "non-overlapping should all survive");
}

#[test]
fn test_fault_dedup_removes_nested() {
    let mut replacements = vec![
        Replacement {
            span_start: 15,
            span_end: 25,
            text: "inner".into(),
        },
        Replacement {
            span_start: 10,
            span_end: 30,
            text: "outer".into(),
        },
    ];
    // sorted by span_start descending
    replacements.sort_by_key(|r| std::cmp::Reverse(r.span_start));
    dedup_overlapping(&mut replacements);
    assert_eq!(replacements.len(), 1, "nested should be deduped");
    assert_eq!(replacements[0].text, "outer", "outermost should survive");
}