|
1 | 1 | """
|
2 |
| -An example script to benchmark neighbor-finding algorithms in CustomizeLattice. |
3 |
| -
|
4 |
| -This script demonstrates the performance difference between the KDTree-based |
5 |
| -neighbor search and a baseline all-to-all distance matrix method. |
6 |
| -As shown by the results, the KDTree approach offers a significant speedup, |
7 |
| -especially when calculating for a large number of neighbor shells (large max_k). |
| 2 | +Benchmark: Compare neighbor-building time between KDTree and distance-matrix |
| 3 | +methods in CustomizeLattice for varying lattice sizes. |
8 | 4 | """
|
9 | 5 |
|
10 |
| -import timeit |
11 |
| -from typing import Any, Dict, List |
| 6 | +import argparse |
| 7 | +import csv |
| 8 | +import time |
| 9 | +from typing import Iterable, List, Tuple, Optional |
| 10 | +import logging |
| 11 | + |
| 12 | +import numpy as np |
| 13 | + |
| 14 | +# Silence verbose infos from the library during benchmarks |
| 15 | + |
| 16 | +logging.basicConfig(level=logging.WARNING) |
| 17 | + |
| 18 | +from tensorcircuit.templates.lattice import CustomizeLattice |
| 19 | + |
| 20 | + |
| 21 | +def run_once( |
| 22 | + n: int, d: int, max_k: int, repeats: int, seed: int |
| 23 | +) -> Tuple[float, float]: |
| 24 | + """Run one size point and return (time_kdtree, time_matrix).""" |
| 25 | + rng = np.random.default_rng(seed) |
| 26 | + ids = list(range(n)) |
| 27 | + |
| 28 | + # Collect times for each repeat with different random coordinates |
| 29 | + kdtree_times: List[float] = [] |
| 30 | + matrix_times: List[float] = [] |
| 31 | + |
| 32 | + for _ in range(repeats): |
| 33 | + # Generate different coordinates for each repeat |
| 34 | + coords = rng.random((n, d), dtype=float) |
| 35 | + lat = CustomizeLattice(dimensionality=d, identifiers=ids, coordinates=coords) |
| 36 | + |
| 37 | + # KDTree path - single measurement |
| 38 | + t0 = time.perf_counter() |
| 39 | + lat._build_neighbors(max_k=max_k, use_kdtree=True) |
| 40 | + kdtree_times.append(time.perf_counter() - t0) |
12 | 41 |
|
| 42 | + # Distance-matrix path - single measurement |
| 43 | + t0 = time.perf_counter() |
| 44 | + lat._build_neighbors(max_k=max_k, use_kdtree=False) |
| 45 | + matrix_times.append(time.perf_counter() - t0) |
13 | 46 |
|
14 |
| -def run_benchmark() -> None: |
15 |
| - """ |
16 |
| - Executes the benchmark test and prints the results in a formatted table. |
17 |
| - """ |
18 |
| - # --- Benchmark Parameters --- |
19 |
| - # A list of lattice sizes (N = number of sites) to test |
20 |
| - site_counts: List[int] = [10, 50, 100, 200, 500, 1000, 1500, 2000] |
| 47 | + return float(np.mean(kdtree_times)), float(np.mean(matrix_times)) |
21 | 48 |
|
22 |
| - # Use a large k to better showcase the performance of KDTree in |
23 |
| - # finding multiple neighbor shells, as suggested by the maintainer. |
24 |
| - max_k: int = 2000 |
25 | 49 |
|
26 |
| - # Reduce the number of runs to keep the total benchmark time reasonable, |
27 |
| - # especially with a large max_k. |
28 |
| - number_of_runs: int = 3 |
29 |
| - # -------------------------- |
| 50 | +def parse_sizes(s: str) -> List[int]: |
| 51 | + return [int(x) for x in s.split(",") if x.strip()] |
30 | 52 |
|
31 |
| - results: List[Dict[str, Any]] = [] |
32 | 53 |
|
33 |
| - print("=" * 75) |
34 |
| - print("Starting neighbor finding benchmark for CustomizeLattice...") |
35 |
| - print(f"Parameters: max_k={max_k}, number_of_runs={number_of_runs}") |
36 |
| - print("=" * 75) |
| 54 | +def format_row(n: int, t_kdtree: float, t_matrix: float) -> str: |
| 55 | + speedup = (t_matrix / t_kdtree) if t_kdtree > 0 else float("inf") |
| 56 | + return f"{n:>8} | {t_kdtree:>12.6f} | {t_matrix:>14.6f} | {speedup:>7.2f}x" |
| 57 | + |
| 58 | + |
| 59 | +def main(argv: Optional[Iterable[str]] = None) -> int: |
| 60 | + p = argparse.ArgumentParser(description="Neighbor-building time comparison") |
| 61 | + p.add_argument( |
| 62 | + "--sizes", |
| 63 | + type=parse_sizes, |
| 64 | + default=[128, 256, 512, 1024, 2048], |
| 65 | + help="Comma-separated site counts to benchmark (default: 128,256,512,1024,2048)", |
| 66 | + ) |
| 67 | + p.add_argument( |
| 68 | + "--dims", type=int, default=2, help="Lattice dimensionality (default: 2)" |
| 69 | + ) |
| 70 | + p.add_argument( |
| 71 | + "--max-k", type=int, default=6, help="Max neighbor shells k (default: 6)" |
| 72 | + ) |
| 73 | + p.add_argument( |
| 74 | + "--repeats", type=int, default=5, help="Repeats per measurement (default: 5)" |
| 75 | + ) |
| 76 | + p.add_argument("--seed", type=int, default=42, help="PRNG seed (default: 42)") |
| 77 | + p.add_argument("--csv", type=str, default="", help="Optional CSV output path") |
| 78 | + args = p.parse_args(list(argv) if argv is not None else None) |
| 79 | + |
| 80 | + print("=" * 74) |
37 | 81 | print(
|
38 |
| - f"{'Sites (N)':>10} | {'KDTree Time (s)':>18} | {'Baseline Time (s)':>20} | {'Speedup':>10}" |
| 82 | + f"Benchmark CustomizeLattice neighbor-building | dims={args.dims} max_k={args.max_k} repeats={args.repeats}" |
39 | 83 | )
|
40 |
| - print("-" * 75) |
| 84 | + print("=" * 74) |
| 85 | + print(f"{'N':>8} | {'KDTree(s)':>12} | {'DistMatrix(s)':>14} | {'Speedup':>7}") |
| 86 | + print("-" * 74) |
41 | 87 |
|
42 |
| - for n_sites in site_counts: |
43 |
| - # Prepare the setup code for timeit. |
44 |
| - # This code generates a random lattice and is executed before timing begins. |
45 |
| - # We use a fixed seed to ensure the coordinates are the same for both tests. |
46 |
| - setup_code = f""" |
47 |
| -import numpy as np |
48 |
| -from tensorcircuit.templates.lattice import CustomizeLattice |
| 88 | + rows: List[Tuple[int, float, float]] = [] |
| 89 | + for n in args.sizes: |
| 90 | + t_kdtree, t_matrix = run_once(n, args.dims, args.max_k, args.repeats, args.seed) |
| 91 | + rows.append((n, t_kdtree, t_matrix)) |
| 92 | + print(format_row(n, t_kdtree, t_matrix)) |
49 | 93 |
|
50 |
| -np.random.seed(42) |
51 |
| -coords = np.random.rand({n_sites}, 2) |
52 |
| -ids = list(range({n_sites})) |
53 |
| -lat = CustomizeLattice(dimensionality=2, identifiers=ids, coordinates=coords) |
54 |
| -""" |
55 |
| - # Define the Python statements to be timed. |
56 |
| - stmt_kdtree = f"lat._build_neighbors(max_k={max_k})" |
57 |
| - stmt_baseline = f"lat._build_neighbors_by_distance_matrix(max_k={max_k})" |
58 |
| - |
59 |
| - try: |
60 |
| - # Execute the timing. timeit returns the total time for all runs. |
61 |
| - time_kdtree = ( |
62 |
| - timeit.timeit(stmt=stmt_kdtree, setup=setup_code, number=number_of_runs) |
63 |
| - / number_of_runs |
64 |
| - ) |
65 |
| - time_baseline = ( |
66 |
| - timeit.timeit( |
67 |
| - stmt=stmt_baseline, setup=setup_code, number=number_of_runs |
68 |
| - ) |
69 |
| - / number_of_runs |
70 |
| - ) |
71 |
| - |
72 |
| - # Calculate and store results, handling potential division by zero. |
73 |
| - speedup = time_baseline / time_kdtree if time_kdtree > 0 else float("inf") |
74 |
| - results.append( |
75 |
| - { |
76 |
| - "n_sites": n_sites, |
77 |
| - "time_kdtree": time_kdtree, |
78 |
| - "time_baseline": time_baseline, |
79 |
| - "speedup": speedup, |
80 |
| - } |
81 |
| - ) |
82 |
| - print( |
83 |
| - f"{n_sites:>10} | {time_kdtree:>18.6f} | {time_baseline:>20.6f} | {speedup:>9.2f}x" |
84 |
| - ) |
85 |
| - |
86 |
| - except Exception as e: |
87 |
| - print(f"An error occurred at N={n_sites}: {e}") |
88 |
| - break |
89 |
| - |
90 |
| - print("-" * 75) |
91 |
| - print("Benchmark complete.") |
| 94 | + if args.csv: |
| 95 | + with open(args.csv, "w", newline="", encoding="utf-8") as f: |
| 96 | + w = csv.writer(f) |
| 97 | + w.writerow(["N", "time_kdtree_s", "time_distance_matrix_s", "speedup"]) |
| 98 | + for n, t_kdtree, t_matrix in rows: |
| 99 | + speedup = (t_matrix / t_kdtree) if t_kdtree > 0 else float("inf") |
| 100 | + w.writerow([n, f"{t_kdtree:.6f}", f"{t_matrix:.6f}", f"{speedup:.2f}"]) |
| 101 | + |
| 102 | + print("-" * 74) |
| 103 | + print(f"Saved CSV to: {args.csv}") |
| 104 | + |
| 105 | + print("-" * 74) |
| 106 | + print("Done.") |
| 107 | + return 0 |
92 | 108 |
|
93 | 109 |
|
94 | 110 | if __name__ == "__main__":
|
95 |
| - run_benchmark() |
| 111 | + raise SystemExit(main()) |
0 commit comments