With PR #595 merged, we now also index N regions of the reference, which adds some unnecessary memory usage (depending on the reference). For example, on GRCh38, the number of randstrobes goes up by 78 million (from 611409875 to 689205916) with that PR. This increases memory usage by ~1 GB.
It should be relatively easy to just keep track of the N regions and ignore them when generating syncmers.
Originally posted by @marcelm in #595 (review)
With PR #595 merged, we now also index N regions of the reference, which adds some unnecessary memory usage (depending on the reference). For example, on GRCh38, the number of randstrobes goes up by 78 million (from 611409875 to 689205916) with that PR. This increases memory usage by ~1 GB.
It should be relatively easy to just keep track of the
Nregions and ignore them when generating syncmers.Originally posted by @marcelm in #595 (review)