trueno 0.17.2

High-performance SIMD compute library with GPU support for matrix operations
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

#!/usr/bin/env python3
"""
Benchmark Regression Checker

Parses Criterion benchmark output and compares against baseline.
Fails CI if any benchmark shows >5% regression.

Usage:
    ./scripts/check_regression.py [--baseline FILE] [--current FILE]

    cat bench.txt | ./scripts/check_regression.py --baseline baseline.txt
"""

import argparse
import re
import sys
from typing import Dict, Tuple, Optional


def parse_time(time_str: str) -> float:
    """Convert time string to nanoseconds.

    Examples:
        "52.46 ns" -> 52.46
        "856.3 ns" -> 856.3
        "9.99 µs" -> 9990.0
        "31.95 ms" -> 31950000.0
        "0.65 s" -> 650000000.0
    """
    match = re.match(r'([\d.]+)\s*(ns|µs|us|ms|s)', time_str.strip())
    if not match:
        raise ValueError(f"Cannot parse time: {time_str}")

    value = float(match.group(1))
    unit = match.group(2)

    multipliers = {
        'ns': 1,
        'µs': 1_000,
        'us': 1_000,
        'ms': 1_000_000,
        's': 1_000_000_000
    }

    return value * multipliers[unit]


def parse_criterion_output(text: str) -> Dict[str, float]:
    """Parse Criterion benchmark output and extract mean times.

    Returns dict mapping benchmark name to mean time in nanoseconds.
    """
    results = {}

    # Pattern to match benchmark results
    # Examples:
    # gpu_vec_add/GPU/1000    time:   [31.707 ms 31.950 ms 32.184 ms]
    # gpu_vec_add/Scalar/1000 time:   [52.142 ns 52.462 ns 52.823 ns]
    pattern = r'^(\S+)\s+time:\s+\[([\d.]+\s*(?:ns|µs|us|ms|s))\s+([\d.]+\s*(?:ns|µs|us|ms|s))\s+([\d.]+\s*(?:ns|µs|us|ms|s))\]'

    for line in text.split('\n'):
        line = line.strip()
        match = re.match(pattern, line)
        if match:
            name = match.group(1)
            # Group 3 is the mean (middle value)
            mean_time = parse_time(match.group(3))
            results[name] = mean_time

    return results


def format_time(ns: float) -> str:
    """Format nanoseconds as human-readable string."""
    if ns >= 1_000_000_000:
        return f"{ns / 1_000_000_000:.2f}s"
    elif ns >= 1_000_000:
        return f"{ns / 1_000_000:.2f}ms"
    elif ns >= 1_000:
        return f"{ns / 1_000:.2f}µs"
    else:
        return f"{ns:.2f}ns"


def compare_benchmarks(
    baseline: Dict[str, float],
    current: Dict[str, float],
    regression_threshold: float = 0.05,
    warning_threshold: float = 0.02
) -> Tuple[bool, str]:
    """Compare current benchmarks against baseline.

    Returns:
        (passed, report): Whether check passed and detailed report
    """
    regressions = []
    warnings = []
    improvements = []
    unchanged = []
    missing = []

    for name, baseline_time in sorted(baseline.items()):
        if name not in current:
            missing.append(name)
            continue

        current_time = current[name]
        change = (current_time - baseline_time) / baseline_time

        if change > regression_threshold:
            regressions.append((name, baseline_time, current_time, change))
        elif change > warning_threshold:
            warnings.append((name, baseline_time, current_time, change))
        elif change < -warning_threshold:
            improvements.append((name, baseline_time, current_time, change))
        else:
            unchanged.append(name)

    # New benchmarks (in current but not in baseline)
    new_benchmarks = [name for name in current if name not in baseline]

    # Generate report
    lines = []
    lines.append("=" * 60)
    lines.append("BENCHMARK REGRESSION REPORT")
    lines.append("=" * 60)
    lines.append("")

    if regressions:
        lines.append(f"REGRESSIONS (>{regression_threshold*100:.0f}% slower): {len(regressions)}")
        lines.append("-" * 60)
        for name, base, curr, change in regressions:
            lines.append(f"  {name}")
            lines.append(f"    Baseline: {format_time(base)}")
            lines.append(f"    Current:  {format_time(curr)}")
            lines.append(f"    Change:   +{change*100:.1f}% (FAIL)")
            lines.append("")

    if warnings:
        lines.append(f"WARNINGS ({warning_threshold*100:.0f}%-{regression_threshold*100:.0f}% slower): {len(warnings)}")
        lines.append("-" * 60)
        for name, base, curr, change in warnings:
            lines.append(f"  {name}")
            lines.append(f"    Baseline: {format_time(base)}")
            lines.append(f"    Current:  {format_time(curr)}")
            lines.append(f"    Change:   +{change*100:.1f}%")
            lines.append("")

    if improvements:
        lines.append(f"IMPROVEMENTS (>{warning_threshold*100:.0f}% faster): {len(improvements)}")
        lines.append("-" * 60)
        for name, base, curr, change in improvements:
            lines.append(f"  {name}")
            lines.append(f"    Baseline: {format_time(base)}")
            lines.append(f"    Current:  {format_time(curr)}")
            lines.append(f"    Change:   {change*100:.1f}%")
            lines.append("")

    if missing:
        lines.append(f"MISSING (in baseline, not in current): {len(missing)}")
        lines.append("-" * 60)
        for name in missing:
            lines.append(f"  {name}")
        lines.append("")

    if new_benchmarks:
        lines.append(f"NEW (in current, not in baseline): {len(new_benchmarks)}")
        lines.append("-" * 60)
        for name in new_benchmarks:
            lines.append(f"  {name}")
        lines.append("")

    # Summary
    lines.append("=" * 60)
    lines.append("SUMMARY")
    lines.append("=" * 60)
    total = len(baseline)
    lines.append(f"  Total benchmarks: {total}")
    lines.append(f"  Regressions:      {len(regressions)}")
    lines.append(f"  Warnings:         {len(warnings)}")
    lines.append(f"  Improvements:     {len(improvements)}")
    lines.append(f"  Unchanged:        {len(unchanged)}")
    lines.append(f"  Missing:          {len(missing)}")
    lines.append(f"  New:              {len(new_benchmarks)}")
    lines.append("")

    if regressions:
        lines.append(f"RESULT: FAIL - {len(regressions)} regression(s) detected")
    elif warnings:
        lines.append(f"RESULT: PASS (with warnings) - {len(warnings)} warning(s)")
    else:
        lines.append("RESULT: PASS")

    lines.append("=" * 60)

    passed = len(regressions) == 0
    report = '\n'.join(lines)

    return passed, report


def build_arg_parser() -> argparse.ArgumentParser:
    """Build and return the argument parser for the regression checker."""
    parser = argparse.ArgumentParser(
        description='Check benchmark performance regressions'
    )
    parser.add_argument(
        '--baseline', '-b',
        required=True,
        help='Baseline benchmark output file'
    )
    parser.add_argument(
        '--current', '-c',
        help='Current benchmark output file (default: stdin)'
    )
    parser.add_argument(
        '--regression-threshold', '-r',
        type=float,
        default=0.05,
        help='Regression threshold (default: 0.05 = 5%%)'
    )
    parser.add_argument(
        '--warning-threshold', '-w',
        type=float,
        default=0.02,
        help='Warning threshold (default: 0.02 = 2%%)'
    )
    return parser


def read_file_or_exit(file_path: str, label: str) -> str:
    """Read a file and return its contents, or exit with an error message."""
    try:
        with open(file_path, 'r') as f:
            return f.read()
    except FileNotFoundError:
        print(f"Error: {label} file not found: {file_path}", file=sys.stderr)
        sys.exit(1)


def read_current_input(current_path: Optional[str]) -> str:
    """Read current benchmark data from a file or stdin."""
    if current_path:
        return read_file_or_exit(current_path, "Current")
    return sys.stdin.read()


def parse_and_validate(text: str, label: str) -> Dict[str, float]:
    """Parse criterion output and exit if no benchmarks are found."""
    results = parse_criterion_output(text)
    if not results:
        print(f"Error: No benchmarks found in {label}", file=sys.stderr)
        sys.exit(1)
    return results


def main():
    parser = build_arg_parser()
    args = parser.parse_args()

    baseline_text = read_file_or_exit(args.baseline, "Baseline")
    current_text = read_current_input(args.current)

    baseline = parse_and_validate(baseline_text, "baseline")
    current = parse_and_validate(current_text, "current")

    passed, report = compare_benchmarks(
        baseline,
        current,
        args.regression_threshold,
        args.warning_threshold
    )

    print(report)
    sys.exit(0 if passed else 1)


if __name__ == '__main__':
    main()