#!/usr/bin/env python # Copyright 2018-2023 David Robillard # SPDX-License-Identifier: ISC """"Benchmark RDF reading and writing commands.""" # pylint: disable=consider-using-f-string # pylint: disable=consider-using-with # pylint: disable=import-outside-toplevel # pylint: disable=invalid-name # pylint: disable=redefined-outer-name # pylint: disable=too-many-locals # pylint: disable=unspecified-encoding import argparse import csv import itertools import math import os import subprocess import sys import matplotlib class WorkingDirectory: "Scoped context for changing working directory" def __init__(self, working_dir): self.original_dir = os.getcwd() self.working_dir = working_dir def __enter__(self): sys.stderr.write("Entering directory `%s'\n" % self.working_dir) os.chdir(self.working_dir) return self def __exit__(self, exc_type, value, traceback): sys.stderr.write("Leaving directory `%s'\n" % self.working_dir) os.chdir(self.original_dir) def order_of_magnitude(values): "Return the order of magnitude to use for an axis with the given values" if len(values) <= 0: return 0 # Calculate the "best" order of magnitude like ScalarFormatter does val = max(values) oom = math.floor(math.log10(max(1.0, val))) if -3 <= oom <= 3: return 0 # Round down to a sensible (thousand, millions, billions, etc) order remainder = oom % 3 oom = oom - remainder return oom def filename(num): "Filename for a generated file with n statements" return "gen%d.ttl" % num def gen(sp2b_dir, n_min, n_max, step): "Generate files with n_min ... n_max statements if they are not present" with WorkingDirectory(sp2b_dir) as directory: for n in range(n_min, n_max + step, step): out_path = os.path.join( directory.original_dir, "build", filename(n) ) if not os.path.exists(out_path): subprocess.call(["./sp2b_gen", "-t", str(n), out_path]) def write_header(results, progs): "Write the header line for TSV output" results.write("n") for prog in progs: results.write("\t" + os.path.basename(prog.split()[0])) results.write("\n") def parse_time(report): "Return user time and max RSS from a /usr/bin/time -v report" time = memory = None for line in report.split("\n"): after_colon = line.find(":") + 1 if line.startswith("\tUser time"): time = float(line[after_colon:]) elif line.startswith("\tMaximum resident set"): memory = int(float(line[after_colon:]) * 1024) return (time, memory) def get_dashes(): "Generator for plot line dash patterns" dash = 2.0 space = dot = 0.75 yield [] # Solid yield [dash, space] # Dashed yield [dot, space] # Dotted # Dash-dots, with increasing number of dots for each line for i in itertools.count(2): yield [dash, space] + [dot, space] * (i - 1) def plot(in_file, out_filename, x_label, y_label, y_max=None): "Plot a TSV file as SVG" matplotlib.use("agg") import matplotlib.pyplot as plt plt.rcParams.update({"font.size": 7}) fig_height = 1.8 dashes = get_dashes() markers = itertools.cycle(["o", "s", "v", "D", "*", "p", "P", "h", "X"]) reader = csv.reader(in_file, delimiter="\t") header = next(reader) cols = list(zip(*list(reader))) # Create a figure with a grid plt.clf() fig = plt.figure(figsize=(fig_height * math.sqrt(2), fig_height)) ax = fig.add_subplot(111) ax.set_xlabel(x_label) ax.set_ylabel(y_label) ax.grid(linewidth=0.25, linestyle=":", color="0", dashes=[0.2, 1.6]) ax.tick_params(axis="both", width=0.75) x = list(map(float, cols[0])) actual_y_max = 0.0 for i, y in enumerate(cols[1::]): y_floats = list(map(float, y)) y_floats_max = max(y_floats) actual_y_max = max(actual_y_max, y_floats_max) ax.plot( x, y_floats, label=header[i + 1], marker=next(markers), dashes=next(dashes), markersize=3.0, linewidth=1.0, ) # Set Y axis limits to go from zero to the maximum value with a small pad y_max = (1.025 * actual_y_max) if y_max is None else y_max ax.set_ylim([0.0, y_max]) # Set axis magnitudes x_m = (order_of_magnitude(x),) * 2 y_m = (order_of_magnitude([y_max]),) * 2 ax.ticklabel_format(axis="x", style="sci", scilimits=x_m, useMathText=True) ax.ticklabel_format(axis="y", style="sci", scilimits=y_m, useMathText=True) # Save plot plt.legend(labelspacing=0.25) plt.savefig(out_filename, bbox_inches="tight", pad_inches=0.125) plt.close() sys.stderr.write("wrote {}\n".format(out_filename)) def run(progs, n_min, n_max, step): "Benchmark each program with n_min ... n_max statements" with WorkingDirectory("build"): results = { "time": open("serd-time.txt", "w"), "throughput": open("serd-throughput.txt", "w"), "memory": open("serd-memory.txt", "w"), } # Write TSV header for all output files for name, f in results.items(): write_header(f, progs) for n in range(n_min, n_max + step, step): # Add first column (n) to rows rows = {} for name, _ in results.items(): rows[name] = [str(n)] # Run each program and fill rows with measurements for prog in progs: cmd = "/usr/bin/time -v " + prog + " " + filename(n) with open(filename(n) + ".out", "w") as out: sys.stderr.write(cmd + "\n") proc = subprocess.Popen( cmd.split(), stdout=out, stderr=subprocess.PIPE ) time, memory = parse_time(proc.communicate()[1].decode()) rows["time"] += ["%.07f" % time] rows["throughput"] += ["%d" % (n / time)] rows["memory"] += [str(memory)] # Write rows to output files for name, f in results.items(): f.write("\t".join(rows[name]) + "\n") for name, f in results.items(): tsv_filename = "serd-pipe-%s.txt" % name sys.stderr.write("wrote %s\n" % tsv_filename) def plot_results(): "Plot all benchmark results" with WorkingDirectory("build"): plot( open("serd-time.txt", "r"), "serd-time.svg", "Statements", "Time (s)", ) plot( open("serd-throughput.txt", "r"), "serd-throughput.svg", "Statements", "Statements / s", ) plot( open("serd-memory.txt", "r"), "serd-memory.svg", "Statements", "Bytes", ) if __name__ == "__main__": ap = argparse.ArgumentParser( usage="%(prog)s [OPTION]... SP2B_DIR", description=__doc__, formatter_class=argparse.RawDescriptionHelpFormatter, epilog=""" example: %(prog)s --max 300000 \\ --run 'rapper -i turtle -o turtle' \\ --run 'riot --output=ttl' \\ --run 'rdfpipe -i turtle -o turtle' /path/to/sp2b/src/ """, ) ap.add_argument( "--max", type=int, default=3000000, help="maximum triple count" ) ap.add_argument( "--run", type=str, action="append", default=[], help="additional command to run (input file is appended)", ) ap.add_argument( "--no-generate", action="store_true", help="do not generate data" ) ap.add_argument( "--no-execute", action="store_true", help="do not run benchmarks" ) ap.add_argument( "--no-plot", action="store_true", help="do not plot benchmarks" ) ap.add_argument("--steps", type=int, default=6, help="number of steps") ap.add_argument("sp2b_dir", help="path to sp2b test data generator") args = ap.parse_args(sys.argv[1:]) progs = ["serd-pipe -f -i turtle -o turtle"] + args.run min_n = int(args.max / args.steps) max_n = args.max step = min_n if not args.no_generate: gen(args.sp2b_dir, min_n, max_n, step) if not args.no_execute: run(progs, min_n, max_n, step) if not args.no_plot: plot_results()