File size: 5,800 Bytes
9fe063f
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
"""
Condensate: Rust vs Python Benchmark

Runs the same workloads through both backends and compares:
- Build time
- Prediction latency
- Accuracy
- Scoring time

Run: python3 test_rust_vs_python.py
"""

import numpy as np
import time
import os
import sys

sys.path.insert(0, os.path.dirname(__file__))

from membrane import Membrane
from graph_builder import GraphBuilder
from predictor import Predictor
from rust_predictor import is_rust_available, RustPredictorWrapper


def generate_workload(num_layers, num_hot, iterations):
    """Generate access log from a simulated workload."""
    Membrane.clear()

    state = {f"layer_{i}": np.random.randn(64, 64).astype(np.float32)
             for i in range(num_layers)}
    wrapped = Membrane.wrap(state, "model")

    hot_set = set(range(num_hot))
    for _ in range(iterations):
        for i in range(num_layers):
            if i in hot_set:
                _ = wrapped[f"layer_{i}"]
            elif np.random.random() < 0.03:
                _ = wrapped[f"layer_{i}"]
        time.sleep(0.001)

    return Membrane.get_log()


def benchmark_python(log):
    """Benchmark the Python predictor."""
    # Build graph
    start = time.monotonic()
    graph = GraphBuilder(causal_window_ns=5_000_000)
    graph.build(log)
    build_time = time.monotonic() - start

    # Learn predictor
    predictor = Predictor()
    start = time.monotonic()
    predictor.learn(graph)
    learn_time = time.monotonic() - start

    # Score
    start = time.monotonic()
    result = predictor.score(log)
    score_time = time.monotonic() - start

    # Single prediction latency
    start = time.monotonic()
    for _ in range(1000):
        predictor.predict("model.layer_0")
    predict_time = (time.monotonic() - start) / 1000

    return {
        "build_ms": build_time * 1000,
        "learn_ms": learn_time * 1000,
        "score_ms": score_time * 1000,
        "predict_us": predict_time * 1_000_000,
        "accuracy": result["accuracy"],
        "predictions": result["predictions_made"],
        "hits": result["hits"],
    }


def benchmark_rust(log):
    """Benchmark the Rust predictor."""
    if not is_rust_available():
        return None

    wrapper = RustPredictorWrapper()

    # Build graph + learn (combined in Rust path)
    start = time.monotonic()
    wrapper.learn_from_log(log)
    build_learn_time = time.monotonic() - start

    # Score
    start = time.monotonic()
    result = wrapper.score(log)
    score_time = time.monotonic() - start

    # Single prediction latency
    start = time.monotonic()
    for _ in range(1000):
        wrapper.predict("model.layer_0")
    predict_time = (time.monotonic() - start) / 1000

    return {
        "build_ms": build_learn_time * 1000,
        "learn_ms": 0,  # combined with build
        "score_ms": score_time * 1000,
        "predict_us": predict_time * 1_000_000,
        "accuracy": result["accuracy"],
        "predictions": result["predictions_made"],
        "hits": result["hits"],
    }


def run_comparison(name, num_layers, num_hot, iterations):
    """Run both backends on the same workload and compare."""
    print(f"\n{'='*65}")
    print(f"  {name}")
    print(f"  {num_layers} layers, {num_hot} hot, {iterations} iterations")
    print(f"{'='*65}")

    log = generate_workload(num_layers, num_hot, iterations)
    print(f"  Events generated: {len(log)}")

    # Python
    py = benchmark_python(log)

    # Rust
    rs = benchmark_rust(log)

    # Print comparison
    print(f"\n  {'Metric':<25} {'Python':>12} {'Rust':>12} {'Speedup':>10}")
    print(f"  {'-'*25} {'-'*12} {'-'*12} {'-'*10}")

    if rs:
        py_build = py['build_ms'] + py['learn_ms']
        rs_build = rs['build_ms']
        build_speedup = py_build / rs_build if rs_build > 0 else float('inf')

        score_speedup = py['score_ms'] / rs['score_ms'] if rs['score_ms'] > 0 else float('inf')
        predict_speedup = py['predict_us'] / rs['predict_us'] if rs['predict_us'] > 0 else float('inf')

        print(f"  {'Build + Learn':<25} {py_build:>10.1f}ms {rs_build:>10.1f}ms {build_speedup:>9.1f}x")
        print(f"  {'Score':<25} {py['score_ms']:>10.1f}ms {rs['score_ms']:>10.1f}ms {score_speedup:>9.1f}x")
        print(f"  {'Single predict()':<25} {py['predict_us']:>10.1f}μs {rs['predict_us']:>10.1f}μs {predict_speedup:>9.1f}x")
        print(f"  {'Accuracy':<25} {py['accuracy']:>11.1f}% {rs['accuracy']:>11.1f}%")
        print(f"  {'Predictions':<25} {py['predictions']:>12} {rs['predictions']:>12}")
        print(f"  {'Hits':<25} {py['hits']:>12} {rs['hits']:>12}")
    else:
        print(f"  Rust backend not available — showing Python only")
        print(f"  {'Build + Learn':<25} {py['build_ms'] + py['learn_ms']:>10.1f}ms")
        print(f"  {'Score':<25} {py['score_ms']:>10.1f}ms")
        print(f"  {'Single predict()':<25} {py['predict_us']:>10.1f}μs")
        print(f"  {'Accuracy':<25} {py['accuracy']:>11.1f}%")

    return py, rs


if __name__ == "__main__":
    print("=" * 65)
    print("  CONDENSATE — Rust vs Python Benchmark")
    print("=" * 65)

    if is_rust_available():
        print("  Rust backend: AVAILABLE")
    else:
        print("  Rust backend: NOT AVAILABLE")
        print("  Build it: cd rust_core && maturin develop --release")
        print("  Showing Python-only results for now.")

    run_comparison("Small (inference-like)", 16, 4, 50)
    run_comparison("Medium (mixed workload)", 32, 8, 50)
    run_comparison("Large (stress test)", 64, 8, 30)

    print(f"\n{'='*65}")
    if is_rust_available():
        print("  Rust core is live. Production-ready prediction engine.")
    else:
        print("  Build the Rust core to see speedup numbers:")
        print("  cd rust_core && maturin develop --release")
    print(f"{'='*65}")