Blame bench/matlab/bench_pacman.m at e7743b1d608e130a37c4a92c35f69bf09749babc

6292f12 A benchmark chapter, and offline benchmarks to compare it againstJeremy Magland 2 months ago1% Offline benchmark of the Pac-Man problem in real MATLAB.

3% Runs the tutorial's own MATLAB layer, unchanged: the chapter driver and the

4% editable script are concatenated into main.m exactly as the page does, and

5% executed here instead of in numbl. Only the two best methods from the

6% in-browser benchmark are swept (the harmonic basis and the graded-panel

7% double layer), with a deeper panel ladder than the page uses, since a native

8% LAPACK has no trouble with the larger dense systems.

1a8bb6a Make the cross-language benchmark measure the languagesJeremy Magland 2 months ago10% Because this runs the chapter's own script, whatever that script does badly

11% it does badly here too, which is a virtue for comparing MATLAB against numbl

12% and a hazard for comparing MATLAB against the Python and Julia ports. The

13% shape the script uses, a scalar double loop for the kernel matrix, happens to

14% be the fastest of the three shapes available in both MATLAB and numbl, so

15% nothing is being given away; if that ever stops being true, the honest fix is

16% to change the chapter, not to keep a faster copy here.

17%

6292f12 A benchmark chapter, and offline benchmarks to compare it againstJeremy Magland 2 months ago18% matlab -batch "run('bench/matlab/bench_pacman.m')"

19%

20% from the repository root writes bench-pacman-matlab.json into the

21% repository root (run() changes the working directory while it executes, so

22% "the current directory" would be a moving target). Load that file from the

23% benchmark chapter's "load offline results" button; because it ran on the

24% same machine as your browser, the timings share a clock and can share an

25% axis.

27here = fileparts(mfilename('fullpath'));

28root = fileparts(fileparts(here));

29outfile = fullfile(root, 'bench-pacman-matlab.json');

31% the deep basis ladders are legitimately rank deficient at the 1e-12 level;

32% Chapter 9 discusses why, and the warning would repeat once per solve

33warning('off', 'MATLAB:rankDeficientMatrix');

35kind = 'both';

36% ladders: zeros skip a method (rows: SW+SOR, basis, panels, Chebyshev)

37nsm = [0, 0, 0, 0, 0, 0;

38 4, 8, 16, 32, 48, 0;

39 4, 8, 12, 16, 20, 24;

40 0, 0, 0, 0, 0, 0];

42work = tempname;

43mkdir(work);

44libdir = fullfile(root, 'src', 'matlab');

45libfiles = dir(fullfile(libdir, '*.m'));

46for k = 1:numel(libfiles)

47 copyfile(fullfile(libdir, libfiles(k).name), work);

48end

49driver = fileread(fullfile(root, 'src', 'chapters', 'ch11', 'driver.m'));

50script = fileread(fullfile(root, 'src', 'chapters', 'ch11', 'pacman.m'));

51fid = fopen(fullfile(work, 'main.m'), 'w');

52fwrite(fid, sprintf('%s\n%s', driver, script));

53fclose(fid);

55params = struct('kind', kind, 'mode', 'sweep', 'nsm', nsm);

56fid = fopen(fullfile(work, 'params.json'), 'w');

57fwrite(fid, jsonencode(params));

58fclose(fid);

60old = cd(work);

61cleanup = onCleanup(@() cd(old));

62run('main.m');

63res = jsondecode(fileread('out.json'));

64cd(old);

66% out.json holds one row per method; keep the two that actually ran

67slots = {'fd2', 'mps', 'bie', 'spectral'};

68keep = [2, 3];

69methods = cell(1, numel(keep));

70for q = 1:numel(keep)

71 k = keep(q);

72 ns = pick_row(res.NS, k);

73 ok = isfinite(ns) & ns > 0;

74 m = struct();

75 m.name = res.names{k};

76 m.slot = slots{k};

77 m.ns = ns(ok);

78 m.E = pick_row(res.E, k);

79 m.E = m.E(ok);

80 m.T = pick_row(res.T, k);

81 m.T = m.T(ok);

82 methods{q} = m;

83end

85out = struct();

86out.schema = 'laplace-dirichlet-bench-1';

1a8bb6a Make the cross-language benchmark measure the languagesJeremy Magland 2 months ago87out.label = sprintf('MATLAB %s, %d threads', version('-release'), maxNumCompThreads);

6292f12 A benchmark chapter, and offline benchmarks to compare it againstJeremy Magland 2 months ago88out.kind = kind;

89out.date = char(datetime('now', 'Format', 'yyyy-MM-dd'));

90out.methods = methods;

92fid = fopen(outfile, 'w');

93fwrite(fid, jsonencode(out));

94fclose(fid);

95fprintf('wrote %s\n', outfile);

97function r = pick_row(M, k)

98% jsondecode gives a matrix when the rows are rectangular and a cell of

99% columns when they are not; accept both.

100if iscell(M)

101 r = M{k}(:)';

102else

103 r = M(k, :);

104end

105end