Skip to main content

Table 1 Comparison of computational efficiency of Mash, BinDash, and Dashing at k=31 and various sketch sizes

From: Dashing: fast and accurate genomic distances with HyperLogLog

      DashingDashingDashing
PhaseMeasureklog2(size)MashBinDashOriginalErtl-MLEErtl-JMLE
SketchWall clock (s)311013451157273271277
   1213491157273274270
   1413561159286289278
   1614001226359367299
 Peak mem (MB)311017,72014112,68312,72112,644
   1218,29639912,72312,43012,726
   1419,706142612,63012,87712,853
   1625,127554212,88812,41212,933
DistanceWall clock (s)311019017480100601
   1223681882863082139
   143446672111311378308
   16877736036172425130,506
 Peak mem (MB)31101120409116116116
   121380673371371372
   1427851,709139213921392
   1610,7765,816547654765476
BothWall clock (s)311032461,231345365870
   1237171,3455575792407
   1448011,83113901,4088574
   1610,1774,82943944,45330,433
 Peak mem (MB)311017,72040912,46812,95012,988
   1218,29667312,95813,04213,020
   1419,706170913,95113,78214,205
   1625,127581618,32018,08118,011
  1. The log2(size) column reports the log2 of the sketch size in bytes. “Both” results obtained either by using a combined Sketch+Distance mode (for Dashing) or by combining results from separate sketching and distance-calculation invocations (for Mash and BinDash). Dashing was assessed using three estimation methods: Flajolet’s method using the harmonic mean (“Original”) and Ertl’s MLE and JMLE methods. Italicized entries indicate the lowest space or time for a given experiment