Parallelize canonization

[AgentElement]

• canonize::is_tree now uses bitsets instead of hashsets when counting the subgraph's nodes.
• Previously, when a tree has multiple children, their canonical strings were combined as something equivalent to (child1)(child2)...(childn). The internal delimiter pairs )( are replaced by a distinct third delimiter akin to a comma, yielding shorter labels: (child1,child2,...,childn). This change was good in theory, but was slower in practice and abandoned.
• When using ParallelMode::Always or ParallelMode:DepthOne, the matches function now bins subgraphs into isomorphism classes in parallel (using a parallel-aware dashmap::DashMap).

[jdaymude]

Benchmarking Results for 60cfc9c (this) vs. 7539cef (main)

The results of cargo bench --bench benchmark -- --baseline 7539cef --load-baseline 60cfc9c:

bench_enumerate/gdb13_1201/extend                        
                        time:   [26.430 ms 26.437 ms 26.444 ms]
                        change: [-0.2968% -0.2563% -0.2165%] (p = 0.00 < 0.05)
                        Change within noise threshold.
bench_enumerate/gdb13_1201/grow-erode                        
                        time:   [12.281 ms 12.284 ms 12.286 ms]
                        change: [+0.5535% +0.5853% +0.6124%] (p = 0.00 < 0.05)
                        Change within noise threshold.
bench_enumerate/gdb17_200/extend                        
                        time:   [372.26 ms 372.34 ms 372.42 ms]
                        change: [+1.1204% +1.1493% +1.1788%] (p = 0.00 < 0.05)
                        Performance has regressed.
bench_enumerate/gdb17_200/grow-erode                        
                        time:   [113.34 ms 113.38 ms 113.42 ms]
                        change: [+0.4600% +0.5015% +0.5467%] (p = 0.00 < 0.05)
                        Change within noise threshold.
bench_enumerate/checks/extend                        
                        time:   [146.57 ms 146.67 ms 146.79 ms]
                        change: [+0.5118% +0.6872% +0.8401%] (p = 0.00 < 0.05)
                        Change within noise threshold.
bench_enumerate/checks/grow-erode                        
                        time:   [31.614 ms 31.624 ms 31.635 ms]
                        change: [-0.4693% -0.4279% -0.3851%] (p = 0.00 < 0.05)
                        Change within noise threshold.
bench_enumerate/coconut_55/extend                        
                        time:   [1.2344 s 1.2347 s 1.2350 s]
                        change: [-0.9497% -0.3562% +0.1354%] (p = 0.24 > 0.05)
                        No change in performance detected.
bench_enumerate/coconut_55/grow-erode                        
                        time:   [272.49 ms 272.56 ms 272.63 ms]
                        change: [+0.9121% +0.9549% +0.9956%] (p = 0.00 < 0.05)
                        Change within noise threshold.

bench_canonize/gdb13_1201/nauty                        
                        time:   [106.23 ms 106.28 ms 106.34 ms]
                        change: [+0.7569% +0.8276% +0.9035%] (p = 0.00 < 0.05)
                        Change within noise threshold.
bench_canonize/gdb13_1201/tree-nauty                        
                        time:   [55.936 ms 55.962 ms 55.987 ms]
                        change: [-9.8061% -9.7010% -9.5862%] (p = 0.00 < 0.05)
                        Performance has improved.
bench_canonize/gdb17_200/nauty                        
                        time:   [2.6299 s 2.6315 s 2.6332 s]
                        change: [-0.2784% -0.2072% -0.1337%] (p = 0.00 < 0.05)
                        Change within noise threshold.
bench_canonize/gdb17_200/tree-nauty                        
                        time:   [1.1628 s 1.1635 s 1.1642 s]
                        change: [-4.9351% -4.8376% -4.7398%] (p = 0.00 < 0.05)
                        Performance has improved.
bench_canonize/checks/nauty                        
                        time:   [1.2371 s 1.2403 s 1.2431 s]
                        change: [+1.1575% +1.4465% +1.6836%] (p = 0.00 < 0.05)
                        Performance has regressed.
bench_canonize/checks/tree-nauty                        
                        time:   [444.06 ms 444.90 ms 445.73 ms]
                        change: [-1.4436% -1.1516% -0.8590%] (p = 0.00 < 0.05)
                        Change within noise threshold.
bench_canonize/coconut_55/nauty                        
                        time:   [11.654 s 11.749 s 11.849 s]
                        change: [-3.1066% -2.2170% -1.4172%] (p = 0.00 < 0.05)
                        Performance has improved.
bench_canonize/coconut_55/tree-nauty                        
                        time:   [4.3415 s 4.3484 s 4.3550 s]
                        change: [+2.3180% +2.5432% +2.7653%] (p = 0.00 < 0.05)
                        Performance has regressed.

bench_bounds/gdb13_1201/no-bounds                        
                        time:   [63.187 ms 63.505 ms 63.818 ms]
                        change: [+5.7547% +6.4152% +7.0704%] (p = 0.00 < 0.05)
                        Performance has regressed.
bench_bounds/gdb13_1201/log                        
                        time:   [58.366 ms 58.644 ms 58.924 ms]
                        change: [+3.9892% +4.6793% +5.4122%] (p = 0.00 < 0.05)
                        Performance has regressed.
bench_bounds/gdb13_1201/int                        
                        time:   [48.720 ms 48.981 ms 49.244 ms]
                        change: [+5.8524% +6.5710% +7.2963%] (p = 0.00 < 0.05)
                        Performance has regressed.
bench_bounds/gdb13_1201/int-vec                        
                        time:   [46.788 ms 47.071 ms 47.361 ms]
                        change: [+4.2874% +5.2914% +6.3024%] (p = 0.00 < 0.05)
                        Performance has regressed.
bench_bounds/gdb17_200/no-bounds                        
                        time:   [1.2363 s 1.2478 s 1.2595 s]
                        change: [-3.5566% -2.0265% -0.4742%] (p = 0.02 < 0.05)
                        Change within noise threshold.
bench_bounds/gdb17_200/log                        
                        time:   [783.92 ms 789.39 ms 795.00 ms]
                        change: [-3.0747% -2.1475% -1.3117%] (p = 0.00 < 0.05)
                        Performance has improved.
bench_bounds/gdb17_200/int                        
                        time:   [174.07 ms 174.72 ms 175.43 ms]
                        change: [-12.428% -11.920% -11.387%] (p = 0.00 < 0.05)
                        Performance has improved.
bench_bounds/gdb17_200/int-vec                        
                        time:   [76.214 ms 76.555 ms 76.877 ms]
                        change: [-28.631% -27.947% -27.254%] (p = 0.00 < 0.05)
                        Performance has improved.
bench_bounds/checks/no-bounds                        
                        time:   [1.1020 s 1.1356 s 1.1721 s]
                        change: [-4.9606% -0.9356% +2.7356%] (p = 0.63 > 0.05)
                        No change in performance detected.
bench_bounds/checks/log time:   [475.20 ms 488.11 ms 502.81 ms]
                        change: [-3.1365% +0.4069% +4.2252%] (p = 0.84 > 0.05)
                        No change in performance detected.
bench_bounds/checks/int time:   [90.709 ms 92.265 ms 93.862 ms]
                        change: [-5.6455% -2.9099% +0.0618%] (p = 0.07 > 0.05)
                        No change in performance detected.
bench_bounds/checks/int-vec                        
                        time:   [15.968 ms 16.115 ms 16.279 ms]
                        change: [-3.2919% -1.4946% +0.1872%] (p = 0.12 > 0.05)
                        No change in performance detected.
bench_bounds/coconut_55/no-bounds                        
                        time:   [16.867 s 17.189 s 17.544 s]
                        change: [-4.9439% -2.4353% +0.2764%] (p = 0.09 > 0.05)
                        No change in performance detected.
bench_bounds/coconut_55/log                        
                        time:   [13.861 s 14.111 s 14.379 s]
                        change: [-2.2578% -0.3164% +1.7106%] (p = 0.78 > 0.05)
                        No change in performance detected.
bench_bounds/coconut_55/int                        
                        time:   [784.03 ms 796.46 ms 808.98 ms]
                        change: [-5.9655% -3.1128% -0.4027%] (p = 0.05 < 0.05)
                        Change within noise threshold.
bench_bounds/coconut_55/int-vec                        
                        time:   [234.33 ms 237.22 ms 239.86 ms]
                        change: [-6.0805% -4.7966% -3.6285%] (p = 0.00 < 0.05)
                        Performance has improved.

bench_memoize/gdb13_1201/no-memoize                        
                        time:   [22.327 ms 22.724 ms 23.388 ms]
                        change: [-3.0192% +5.3551% +14.669%] (p = 0.26 > 0.05)
                        No change in performance detected.
bench_memoize/gdb13_1201/frags-index                        
                        time:   [25.596 ms 25.906 ms 26.456 ms]
                        change: [-2.4206% +3.4592% +9.4365%] (p = 0.33 > 0.05)
                        No change in performance detected.
bench_memoize/gdb13_1201/nauty-index                        
                        time:   [49.883 ms 50.180 ms 50.488 ms]
                        change: [+3.6248% +4.7283% +5.7843%] (p = 0.00 < 0.05)
                        Performance has regressed.
bench_memoize/gdb13_1201/tree-nauty-index                        
                        time:   [47.280 ms 47.590 ms 47.910 ms]
                        change: [+6.0997% +7.1288% +8.1485%] (p = 0.00 < 0.05)
                        Performance has regressed.
bench_memoize/gdb17_200/no-memoize                        
                        time:   [51.625 ms 51.999 ms 52.348 ms]
                        change: [-36.733% -35.490% -34.235%] (p = 0.00 < 0.05)
                        Performance has improved.
bench_memoize/gdb17_200/frags-index                        
                        time:   [55.219 ms 55.541 ms 55.915 ms]
                        change: [-33.405% -32.121% -30.815%] (p = 0.00 < 0.05)
                        Performance has improved.
bench_memoize/gdb17_200/nauty-index                        
                        time:   [88.178 ms 88.592 ms 88.994 ms]
                        change: [-24.692% -23.838% -23.026%] (p = 0.00 < 0.05)
                        Performance has improved.
bench_memoize/gdb17_200/tree-nauty-index                        
                        time:   [76.348 ms 76.703 ms 77.047 ms]
                        change: [-28.509% -27.639% -26.762%] (p = 0.00 < 0.05)
                        Performance has improved.
bench_memoize/checks/no-memoize                        
                        time:   [11.679 ms 11.978 ms 12.274 ms]
                        change: [-2.0990% -0.0432% +2.0696%] (p = 0.97 > 0.05)
                        No change in performance detected.
bench_memoize/checks/frags-index                        
                        time:   [11.327 ms 11.412 ms 11.483 ms]
                        change: [-1.0101% +0.3051% +1.6749%] (p = 0.66 > 0.05)
                        No change in performance detected.
bench_memoize/checks/nauty-index                        
                        time:   [21.899 ms 22.180 ms 22.460 ms]
                        change: [-1.8757% +1.0243% +4.1871%] (p = 0.54 > 0.05)
                        No change in performance detected.
bench_memoize/checks/tree-nauty-index                        
                        time:   [16.007 ms 16.163 ms 16.324 ms]
                        change: [-1.0252% +0.8112% +2.6985%] (p = 0.42 > 0.05)
                        No change in performance detected.
bench_memoize/coconut_55/no-memoize                        
                        time:   [300.53 ms 310.92 ms 320.03 ms]
                        change: [-6.6146% -2.4958% +1.4217%] (p = 0.24 > 0.05)
                        No change in performance detected.
bench_memoize/coconut_55/frags-index                        
                        time:   [254.89 ms 261.93 ms 268.36 ms]
                        change: [-7.8595% -4.8567% -1.7930%] (p = 0.00 < 0.05)
                        Performance has improved.
bench_memoize/coconut_55/nauty-index                        
                        time:   [273.24 ms 275.89 ms 278.34 ms]
                        change: [-4.7965% -3.1433% -1.5129%] (p = 0.00 < 0.05)
                        Performance has improved.
bench_memoize/coconut_55/tree-nauty-index                        
                        time:   [232.61 ms 236.49 ms 240.08 ms]
                        change: [-7.0343% -5.2680% -3.4177%] (p = 0.00 < 0.05)
                        Performance has improved.

[jdaymude]

Benchmarking Results for f134492 (this) vs. 7539cef (main)

The results of cargo bench --bench benchmark -- bench_canonize --baseline 7539cef --load-baseline f134492:

bench_canonize/gdb13_1201/nauty                                                                          
                        time:   [104.64 ms 104.68 ms 104.72 ms]
                        change: [-0.7522% -0.6907% -0.6252%] (p = 0.00 < 0.05)
                        Change within noise threshold.
bench_canonize/gdb13_1201/tree-nauty                                                                           
                        time:   [52.627 ms 52.705 ms 52.785 ms]
                        change: [-15.112% -14.956% -14.786%] (p = 0.00 < 0.05)
                        Performance has improved.
bench_canonize/gdb17_200/nauty                                                                          
                        time:   [2.7782 s 2.7819 s 2.7857 s]
                        change: [+5.3668% +5.4965% +5.6508%] (p = 0.00 < 0.05)
                        Performance has regressed.
bench_canonize/gdb17_200/tree-nauty                                                                          
                        time:   [1.0386 s 1.0402 s 1.0417 s]
                        change: [-15.078% -14.925% -14.788%] (p = 0.00 < 0.05)
                        Performance has improved.
bench_canonize/checks/nauty                                                                          
                        time:   [1.2724 s 1.2766 s 1.2810 s]
                        change: [+4.0984% +4.4158% +4.8083%] (p = 0.00 < 0.05)
                        Performance has regressed.
bench_canonize/checks/tree-nauty                                                                          
                        time:   [406.69 ms 406.99 ms 407.27 ms]
                        change: [-9.8070% -9.5757% -9.3654%] (p = 0.00 < 0.05)
                        Performance has improved.
bench_canonize/coconut_55/nauty                                                                          
                        time:   [11.978 s 11.988 s 11.997 s]
                        change: [-0.3465% -0.2357% -0.1383%] (p = 0.00 < 0.05)
                        Change within noise threshold.
bench_canonize/coconut_55/tree-nauty                                                                          
                        time:   [3.8617 s 3.8859 s 3.9119 s]
                        change: [-8.9870% -8.3631% -7.7245%] (p = 0.00 < 0.05)
                        Performance has improved.

CanonizeMode::TreeNauty is now roughly 8–15% faster depending on the dataset, as desired. Somewhat inexplicably, CanonizeMode::Nauty is benchmarking around 5% slower on gdb17_200 and checks even though nothing affecting that mode was changed?

[jdaymude]

LGTM.