Skip to main content

Table 2 Time- and memory-performance results for constructing sas (and lcp-array in case of caps-sa) with 32 threads

From: Fast, parallel, and cache-friendly suffix array construction

Dataset

Divsufsort

Parallel-divsufsort

Parallel-dc3

Caps-sa

Human (GRCh38)

556 (25)

273 (32)

113 (110)

93 (45)

Human (T2T)

575 (26)

279 (34)

119 (116)

106 (48)

CdBG (Human reads)

722 (33)

379 (44)

149 (172)

114 (61)

Great white shark

771 (36)

410 (48)

176 (186)

121 (64)

Axolotl

10489 (236)

3424 (311)

–

1341 (848)

Bacteria (1K)

726 (33)

437 (43)

164 (172)

259 (60)

  1. divsufsort is shown as a serial benchmark. Time is reported in seconds, and the memory usages are reported in GBs in parentheses. Best performances among the parallel algorithms in each instance are highlighted. parallel-dc3 could not be run on Axolotl because we could not modify the PBBS code-base to accommodate the large numerical size