Performance improvement for nvtext::minhash

[davidwendt]

Description

Improves performance of nvtext::minhash by minimizing character counting in the internal logic. The MinHash strings are expected to be very long ( > 1KB). Improvement is measure to be up to 2x.

Checklist

• I am familiar with the Contributing Guidelines.
• New or existing tests cover these changes.
• The documentation is up to date with these changes.

[davidwendt]

From nvbench results

|  num_rows  |  row_width  |  hash_width  |  seed_count  |   Ref Time |   Cmp Time |          Diff |   %Diff |
|------------|-------------|--------------|--------------|------------|------------|---------------|---------|
|    1024    |     128     |      5       |      2       | 193.856 us | 181.411 us |    -12.445 us |  -6.42% |
|    4096    |     128     |      5       |      2       | 227.873 us | 182.397 us |    -45.476 us | -19.96% |
|    8192    |     128     |      5       |      2       | 280.639 us | 197.543 us |    -83.096 us | -29.61% |
|   16364    |     128     |      5       |      2       | 389.130 us | 223.790 us |   -165.340 us | -42.49% |
|   32768    |     128     |      5       |      2       | 610.014 us | 282.108 us |   -327.906 us | -53.75% |
|   262144   |     128     |      5       |      2       |   3.729 ms |   1.113 ms |  -2615.250 us | -70.14% |
|    1024    |     512     |      5       |      2       | 281.461 us | 206.475 us |    -74.986 us | -26.64% |
|    4096    |     512     |      5       |      2       | 362.247 us | 238.903 us |   -123.344 us | -34.05% |
|    8192    |     512     |      5       |      2       | 511.638 us | 292.133 us |   -219.505 us | -42.90% |
|   16364    |     512     |      5       |      2       | 814.275 us | 387.908 us |   -426.367 us | -52.36% |
|   32768    |     512     |      5       |      2       |   1.418 ms | 587.679 us |   -829.977 us | -58.55% |
|   262144   |     512     |      5       |      2       |   9.899 ms |   3.402 ms |  -6496.691 us | -65.63% |
|    1024    |    2048     |      5       |      2       | 624.283 us | 346.543 us |   -277.740 us | -44.49% |
|    4096    |    2048     |      5       |      2       | 900.445 us | 452.052 us |   -448.393 us | -49.80% |
|    8192    |    2048     |      5       |      2       |   1.417 ms | 665.112 us |   -751.827 us | -53.06% |
|   16364    |    2048     |      5       |      2       |   2.465 ms |   1.036 ms |  -1428.609 us | -57.96% |
|   32768    |    2048     |      5       |      2       |   4.518 ms |   1.812 ms |  -2706.182 us | -59.90% |
|   262144   |    2048     |      5       |      2       |  33.360 ms |  12.553 ms | -20807.037 us | -62.37% |
|    1024    |     128     |      10      |      2       | 196.147 us | 170.822 us |    -25.325 us | -12.91% |
|    4096    |     128     |      10      |      2       | 234.330 us | 186.371 us |    -47.959 us | -20.47% |
|    8192    |     128     |      10      |      2       | 292.730 us | 209.369 us |    -83.360 us | -28.48% |
|   16364    |     128     |      10      |      2       | 416.179 us | 241.465 us |   -174.713 us | -41.98% |
|   32768    |     128     |      10      |      2       | 661.388 us | 303.404 us |   -357.983 us | -54.13% |
|   262144   |     128     |      10      |      2       |   4.133 ms |   1.273 ms |  -2859.794 us | -69.19% |
|    1024    |     512     |      10      |      2       | 291.149 us | 219.325 us |    -71.824 us | -24.67% |
|    4096    |     512     |      10      |      2       | 385.945 us | 262.891 us |   -123.054 us | -31.88% |
|    8192    |     512     |      10      |      2       | 554.882 us | 330.057 us |   -224.825 us | -40.52% |
|   16364    |     512     |      10      |      2       | 898.362 us | 439.449 us |   -458.912 us | -51.08% |
|   32768    |     512     |      10      |      2       |   1.575 ms | 675.748 us |   -899.270 us | -57.10% |
|   262144   |     512     |      10      |      2       |  11.165 ms |   3.992 ms |  -7172.857 us | -64.24% |
|    1024    |    2048     |      10      |      2       | 678.483 us | 409.528 us |   -268.955 us | -39.64% |
|    4096    |    2048     |      10      |      2       | 997.831 us | 543.615 us |   -454.216 us | -45.52% |
|    8192    |    2048     |      10      |      2       |   1.582 ms | 822.555 us |   -759.219 us | -48.00% |
|   16364    |    2048     |      10      |      2       |   2.772 ms |   1.234 ms |  -1537.355 us | -55.46% |
|   32768    |    2048     |      10      |      2       |   5.093 ms |   2.149 ms |  -2944.546 us | -57.81% |
|   262144   |    2048     |      10      |      2       |  37.772 ms |  14.850 ms | -22921.371 us | -60.68% |
|    1024    |     128     |      25      |      2       | 207.323 us | 179.497 us |    -27.826 us | -13.42% |
|    4096    |     128     |      25      |      2       | 249.153 us | 204.989 us |    -44.164 us | -17.73% |
|    8192    |     128     |      25      |      2       | 316.093 us | 230.204 us |    -85.890 us | -27.17% |
|   16364    |     128     |      25      |      2       | 459.841 us | 279.298 us |   -180.543 us | -39.26% |
|   32768    |     128     |      25      |      2       | 748.611 us | 379.865 us |   -368.746 us | -49.26% |
|   262144   |     128     |      25      |      2       |   4.789 ms |   1.822 ms |  -2966.954 us | -61.95% |
|    1024    |     512     |      25      |      2       | 331.493 us | 269.785 us |    -61.707 us | -18.62% |
|    4096    |     512     |      25      |      2       | 446.998 us | 329.370 us |   -117.628 us | -26.32% |
|    8192    |     512     |      25      |      2       | 667.508 us | 456.063 us |   -211.445 us | -31.68% |
|   16364    |     512     |      25      |      2       |   1.109 ms | 643.706 us |   -464.991 us | -41.94% |
|   32768    |     512     |      25      |      2       |   1.996 ms |   1.051 ms |   -945.149 us | -47.36% |
|   262144   |     512     |      25      |      2       |  14.409 ms |   6.706 ms |  -7703.241 us | -53.46% |
|    1024    |    2048     |      25      |      2       | 828.190 us | 591.819 us |   -236.371 us | -28.54% |
|    4096    |    2048     |      25      |      2       |   1.252 ms | 836.633 us |   -414.874 us | -33.15% |
|    8192    |    2048     |      25      |      2       |   2.017 ms |   1.339 ms |   -677.821 us | -33.60% |
|   16364    |    2048     |      25      |      2       |   3.594 ms |   2.082 ms |  -1511.921 us | -42.06% |
|   32768    |    2048     |      25      |      2       |   6.667 ms |   3.701 ms |  -2966.655 us | -44.50% |
|   262144   |    2048     |      25      |      2       |  49.920 ms |  26.234 ms | -23686.493 us | -47.45% |
|    1024    |     128     |      5       |      26      | 266.604 us | 243.072 us |    -23.532 us |  -8.83% |
|    4096    |     128     |      5       |      26      | 369.774 us | 357.338 us |    -12.435 us |  -3.36% |
|    8192    |     128     |      5       |      26      | 546.339 us | 524.452 us |    -21.887 us |  -4.01% |
|   16364    |     128     |      5       |      26      | 865.386 us | 806.152 us |    -59.234 us |  -6.84% |
|   32768    |     128     |      5       |      26      |   1.491 ms |   1.402 ms |    -88.768 us |  -5.95% |
|   262144   |     128     |      5       |      26      |  10.441 ms |   9.727 ms |   -714.349 us |  -6.84% |
|    1024    |     512     |      5       |      26      | 567.700 us | 521.053 us |    -46.647 us |  -8.22% |
|    4096    |     512     |      5       |      26      | 959.818 us | 903.960 us |    -55.857 us |  -5.82% |
|    8192    |     512     |      5       |      26      |   1.588 ms |   1.579 ms |     -8.881 us |  -0.56% |
|   16364    |     512     |      5       |      26      |   2.732 ms |   2.632 ms |    -99.895 us |  -3.66% |
|   32768    |     512     |      5       |      26      |   4.990 ms |   4.842 ms |   -147.820 us |  -2.96% |
|   262144   |     512     |      5       |      26      |  36.764 ms |  35.780 ms |   -984.386 us |  -2.68% |
|    1024    |    2048     |      5       |      26      |   1.758 ms |   1.544 ms |   -214.086 us | -12.18% |
|    4096    |    2048     |      5       |      26      |   3.312 ms |   3.150 ms |   -162.507 us |  -4.91% |
|    8192    |    2048     |      5       |      26      |   5.698 ms |   5.581 ms |   -117.682 us |  -2.07% |
|   16364    |    2048     |      5       |      26      |  10.175 ms |   9.892 ms |   -283.592 us |  -2.79% |
|   32768    |    2048     |      5       |      26      |  19.019 ms |  18.590 ms |   -428.750 us |  -2.25% |
|   262144   |    2048     |      5       |      26      | 142.517 ms | 139.764 ms |  -2753.184 us |  -1.93% |
|    1024    |     128     |      10      |      26      | 277.937 us | 261.317 us |    -16.620 us |  -5.98% |
|    4096    |     128     |      10      |      26      | 384.665 us | 368.642 us |    -16.023 us |  -4.17% |
|    8192    |     128     |      10      |      26      | 566.374 us | 546.126 us |    -20.248 us |  -3.57% |
|   16364    |     128     |      10      |      26      | 907.391 us | 798.656 us |   -108.734 us | -11.98% |
|   32768    |     128     |      10      |      26      |   1.552 ms |   1.368 ms |   -183.853 us | -11.84% |
|   262144   |     128     |      10      |      26      |  10.882 ms |   9.374 ms |  -1507.620 us | -13.85% |
|    1024    |     512     |      10      |      26      | 616.924 us | 564.238 us |    -52.686 us |  -8.54% |
|    4096    |     512     |      10      |      26      |   1.042 ms | 970.627 us |    -70.978 us |  -6.81% |
|    8192    |     512     |      10      |      26      |   1.682 ms |   1.653 ms |    -29.064 us |  -1.73% |
|   16364    |     512     |      10      |      26      |   2.895 ms |   2.717 ms |   -177.754 us |  -6.14% |
|   32768    |     512     |      10      |      26      |   5.254 ms |   4.934 ms |   -319.588 us |  -6.08% |
|   262144   |     512     |      10      |      26      |  38.492 ms |  36.068 ms |  -2424.049 us |  -6.30% |
|    1024    |    2048     |      10      |      26      |   2.010 ms |   1.811 ms |   -198.338 us |  -9.87% |
|    4096    |    2048     |      10      |      26      |   3.630 ms |   3.447 ms |   -183.313 us |  -5.05% |
|    8192    |    2048     |      10      |      26      |   6.141 ms |   6.040 ms |   -101.531 us |  -1.65% |
|   16364    |    2048     |      10      |      26      |  10.837 ms |  10.367 ms |   -469.505 us |  -4.33% |
|   32768    |    2048     |      10      |      26      |  20.125 ms |  19.191 ms |   -933.884 us |  -4.64% |
|   262144   |    2048     |      10      |      26      | 149.540 ms | 142.846 ms |  -6693.143 us |  -4.48% |
|    1024    |     128     |      25      |      26      | 304.311 us | 277.693 us |    -26.618 us |  -8.75% |
|    4096    |     128     |      25      |      26      | 415.681 us | 385.801 us |    -29.880 us |  -7.19% |
|    8192    |     128     |      25      |      26      | 599.293 us | 551.448 us |    -47.844 us |  -7.98% |
|   16364    |     128     |      25      |      26      | 952.504 us | 837.496 us |   -115.008 us | -12.07% |
|   32768    |     128     |      25      |      26      |   1.664 ms |   1.422 ms |   -242.341 us | -14.56% |
|   262144   |     128     |      25      |      26      |  11.723 ms |   9.641 ms |  -2081.667 us | -17.76% |
|    1024    |     512     |      25      |      26      | 751.137 us | 706.851 us |    -44.286 us |  -5.90% |
|    4096    |     512     |      25      |      26      |   1.241 ms |   1.216 ms |    -24.526 us |  -1.98% |
|    8192    |     512     |      25      |      26      |   2.031 ms |   2.043 ms |     11.428 us |   0.56% |
|   16364    |     512     |      25      |      26      |   3.489 ms |   3.320 ms |   -168.836 us |  -4.84% |
|   32768    |     512     |      25      |      26      |   6.388 ms |   5.992 ms |   -396.116 us |  -6.20% |
|   262144   |     512     |      25      |      26      |  46.953 ms |  43.520 ms |  -3432.680 us |  -7.31% |
|    1024    |    2048     |      25      |      26      |   2.564 ms |   2.432 ms |   -132.059 us |  -5.15% |
|    4096    |    2048     |      25      |      26      |   4.600 ms |   4.524 ms |    -76.239 us |  -1.66% |
|    8192    |    2048     |      25      |      26      |   7.747 ms |   8.473 ms |    725.970 us |   9.37% |
|   16364    |    2048     |      25      |      26      |  13.620 ms |  13.193 ms |   -426.827 us |  -3.13% |
|   32768    |    2048     |      25      |      26      |  25.247 ms |  24.278 ms |   -968.509 us |  -3.84% |
|   262144   |    2048     |      25      |      26      | 187.823 ms | 179.514 ms |  -8309.640 us |  -4.42% |

[davidwendt]

/merge