Fulgor: a fast and compact k-mer index for large-scale matching and color queries

Table 2 Total index construction time (elapsed time) and GB of memory (max. RSS), as reported by /usr/bin/time with option -v, using 48 processing threads

	Fulgor		Themisto		MetaGraph		COBS
	hh:mm	GB	hh:mm	GB	hh:mm	GB	hh:mm	GB
EC	00:06	16.89	00:19	17.18	00:46	149.38	00:03	6.39
SE-5K	00:04	12.91	00:11	12.97	00:47	190.99	00:09	8.13
SE-10K	00:09	23.60	00:25	23.58	01:50	218.76	00:17	16.15
SE-50K	01:13	43.76	02:32	96.00	14:16	\(^*\)118.95	01:41	82.49
SE-10K	02:56	73.54	06:25	202.42	26:40	\(^*\)103.99	02:37	83.79
SE-150K	04:36	136.94	10:00	323.10	—	—	04:54	159.31
GB	02:27	115.05	06:21	183.56	10:50	\(^*\)99.54	00:22	17.08

The reported time includes the time to serialize the index on disk and, for Fulgor and Themisto, the time taken by GGCAT to build the ccdBG. We did not observe appreciable differences in space and memory usage when building indexes for Themisto with and without \(k\)-mer sampling, except on the Gut Bacteria collection where sampling is very beneficial. For this reason, we report its best time and memory usage, i.e., that for Themisto-d20. MetaGraph instances marked by \(^*\) were capped to use 100 GB of memory because construction otherwise exceeds total available memory (\(> 500\) GB) on our machine

ISSN: 1748-7188