Skip to content

Robust Predicates Speed Comparision

Jeremy edited this page Aug 4, 2019 · 1 revision

This page will show speed comparisons on linux without and with robust geometric predicates compiled into Polylidar

WITHOUT Robust Geometric Predicates

Name (time in ms)                 Min                 Max                Mean            StdDev              Median               IQR            Outliers       OPS            Rounds  Iterations
-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
test_bad_convex_hull           1.0257 (1.0)        3.6538 (1.30)       1.0682 (1.0)      0.1961 (3.57)       1.0352 (1.0)      0.0290 (1.0)         15;37  936.1199 (1.0)         567           1
test_building1                 2.1618 (2.11)       2.8034 (1.0)        2.1972 (2.06)     0.0550 (1.0)        2.1743 (2.10)     0.0410 (1.41)        24;16  455.1320 (0.49)        370           1
test_building2                 2.5240 (2.46)       3.1666 (1.13)       2.5653 (2.40)     0.0584 (1.06)       2.5443 (2.46)     0.0480 (1.65)        28;15  389.8156 (0.42)        366           1
test_100k_array_lmax         127.8992 (124.69)   131.6918 (46.98)    129.5448 (121.27)   1.0665 (19.40)    129.5521 (125.15)   0.6138 (21.13)         2;2    7.7193 (0.01)          8           1
test_100k_array_alpha        128.0727 (124.86)   132.4049 (47.23)    130.6749 (122.33)   1.4112 (25.68)    130.6245 (126.18)   1.7312 (59.60)         3;0    7.6526 (0.01)          8           1
test_100k_array_xyThresh     130.2316 (126.97)   135.3389 (48.28)    132.0185 (123.59)   1.7027 (30.98)    131.6535 (127.18)   2.1603 (74.37)         2;0    7.5747 (0.01)          8           1
-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

WITH Robust Geometric Predicates (slower)

--------------------------------------------------------------------------------------- benchmark: 6 tests --------------------------------------------------------------------------------------
Name (time in ms)                 Min                 Max                Mean            StdDev              Median               IQR            Outliers       OPS            Rounds  Iterations
-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
test_bad_convex_hull           1.2998 (1.0)        2.0526 (1.0)        1.3561 (1.0)      0.0699 (1.0)        1.3167 (1.0)      0.0563 (1.0)        120;72  737.4264 (1.0)         747           1
test_building1                 2.7249 (2.10)       3.7294 (1.82)       2.8399 (2.09)     0.1120 (1.60)       2.8053 (2.13)     0.1580 (2.81)         47;4  352.1233 (0.48)        294           1
test_building2                 3.1900 (2.45)       4.6644 (2.27)       3.3202 (2.45)     0.1418 (2.03)       3.2857 (2.50)     0.1339 (2.38)        34;12  301.1848 (0.41)        304           1
test_100k_array_xyThresh     154.2823 (118.70)   160.0046 (77.95)    158.4303 (116.83)   2.0101 (28.74)    159.2607 (120.96)   1.8330 (32.56)         1;1    6.3119 (0.01)          7           1
test_100k_array_lmax         156.7735 (120.61)   160.5695 (78.23)    159.2696 (117.45)   1.2911 (18.46)    159.6585 (121.26)   1.4057 (24.97)         2;0    6.2787 (0.01)          7           1
test_100k_array_alpha        158.2368 (121.74)   162.3062 (79.07)    160.5723 (118.41)   1.3218 (18.90)    160.7966 (122.13)   1.5479 (27.50)         2;0    6.2277 (0.01)          7           1
-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------


Summary

Its about 20% slower with robust geometric predicates.

Clone this wiki locally