aprender-explain 0.70.1

PTX/SIMD/wgpu visualization and tracing CLI for 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

use super::super::analyzer::*;
use super::super::types::*;

/// Test `RedundantMoves` detection - consecutive mov chain
#[test]
fn test_redundant_moves_chain() {
    let ptx = r"
.visible .entry test() {
    mov.u32 %r1, %r0;
    mov.u32 %r2, %r1;
    ret;
}
";
    let result = PtxBugAnalyzer::new().analyze(ptx);
    assert!(result.has_bug(&PtxBugClass::RedundantMoves));
}

/// Test `RedundantMoves` - no chain (valid)
#[test]
fn test_redundant_moves_no_chain() {
    let ptx = r"
.visible .entry test() {
    mov.u32 %r1, %r0;
    add.u32 %r2, %r1, 1;
    mov.u32 %r3, %r2;
    ret;
}
";
    let result = PtxBugAnalyzer::new().analyze(ptx);
    assert!(!result.has_bug(&PtxBugClass::RedundantMoves));
}

/// Test `UnoptimizedMemoryPattern` - multiple single loads
#[test]
fn test_unoptimized_memory_single_loads() {
    let ptx = r"
.visible .entry test() {
    ld.global.f32 %f0, [%rd0];
    ld.global.f32 %f1, [%rd1];
    ld.global.f32 %f2, [%rd2];
    ld.global.f32 %f3, [%rd3];
    ret;
}
";
    let result = PtxBugAnalyzer::new().analyze(ptx);
    assert!(result.has_bug(&PtxBugClass::UnoptimizedMemoryPattern));
}

/// Test `UnoptimizedMemoryPattern` - vector loads (valid)
#[test]
fn test_unoptimized_memory_vector_loads() {
    let ptx = r"
.visible .entry test() {
    ld.global.v4.f32 {%f0, %f1, %f2, %f3}, [%rd0];
    ret;
}
";
    let result = PtxBugAnalyzer::new().analyze(ptx);
    assert!(!result.has_bug(&PtxBugClass::UnoptimizedMemoryPattern));
}

/// Test `UnoptimizedMemoryPattern` - few single loads (acceptable)
#[test]
fn test_unoptimized_memory_few_loads() {
    let ptx = r"
.visible .entry test() {
    ld.global.f32 %f0, [%rd0];
    ld.global.f32 %f1, [%rd1];
    ret;
}
";
    let result = PtxBugAnalyzer::new().analyze(ptx);
    // Only 2 single loads - below threshold of 4, should not flag
    assert!(!result.has_bug(&PtxBugClass::UnoptimizedMemoryPattern));
}

/// Test suspicious stride detection in strict mode
#[test]
fn test_unoptimized_memory_suspicious_stride() {
    let ptx = r"
.visible .entry test() {
    mul.wide.u32 %rd0, %r0, 17;
    ld.global.f32 %f0, [%rd0];
    ret;
}
";
    let result = PtxBugAnalyzer::strict().analyze(ptx);
    assert!(result.has_bug(&PtxBugClass::UnoptimizedMemoryPattern));
}

/// Test normal strides are not flagged
#[test]
fn test_unoptimized_memory_normal_stride() {
    let ptx = r"
.visible .entry test() {
    mul.wide.u32 %rd0, %r0, 4;
    ld.global.f32 %f0, [%rd0];
    ret;
}
";
    let result = PtxBugAnalyzer::strict().analyze(ptx);
    // Stride 4 is normal for f32
    assert!(!result.has_bug(&PtxBugClass::UnoptimizedMemoryPattern));
}

/// Test high register pressure detection
#[test]
fn test_high_register_pressure() {
    let ptx = r"
.visible .entry test() {
    .reg .b32 %r<64>;
    .reg .b64 %rd<16>;
    .reg .f32 %f<32>;
    .reg .pred %p<4>;
    ret;
}
";
    // 64 + 16 + 32 + 4 = 116 registers > 64 threshold
    let result = PtxBugAnalyzer::new().analyze(ptx);
    assert!(result.has_bug(&PtxBugClass::HighRegisterPressure));
}

/// Test acceptable register pressure (no bug)
#[test]
fn test_normal_register_pressure() {
    let ptx = r"
.visible .entry test() {
    .reg .b32 %r<16>;
    .reg .b64 %rd<8>;
    .reg .f32 %f<8>;
    .reg .pred %p<4>;
    ret;
}
";
    // 16 + 8 + 8 + 4 = 36 registers < 64 threshold
    let result = PtxBugAnalyzer::new().analyze(ptx);
    assert!(!result.has_bug(&PtxBugClass::HighRegisterPressure));
}

/// Test predicate overflow detection
#[test]
fn test_predicate_overflow() {
    let ptx = r"
.visible .entry test() {
    .reg .pred %p<12>;
    .reg .b32 %r<4>;
    ret;
}
";
    // 12 predicates > 8 limit
    let result = PtxBugAnalyzer::new().analyze(ptx);
    assert!(result.has_bug(&PtxBugClass::PredicateOverflow));
}

/// Test acceptable predicate count (no bug)
#[test]
fn test_normal_predicate_count() {
    let ptx = r"
.visible .entry test() {
    .reg .pred %p<8>;
    .reg .b32 %r<4>;
    ret;
}
";
    // 8 predicates = limit, should not flag
    let result = PtxBugAnalyzer::new().analyze(ptx);
    assert!(!result.has_bug(&PtxBugClass::PredicateOverflow));
}