From 6b28db8eb2d8a2761f8bde0dcfb9d5ceefe5827b Mon Sep 17 00:00:00 2001 From: flu0r1ne Date: Mon, 31 Oct 2022 17:32:08 -0500 Subject: Update readme --- README.md | 10 +++++++--- 1 file changed, 7 insertions(+), 3 deletions(-) diff --git a/README.md b/README.md index c789e09..7d9b441 100644 --- a/README.md +++ b/README.md @@ -15,8 +15,12 @@ which is impractical for most human genome data. Notes: +- Currently, `qidx` is very inefficient in terms of disk space. When indexing a 33GiB BAM file + (Illumina 35x), it takes up 22GiB on disk when using STD compression. It initially maps 1.2TiB + into memory. This is reduced to ~120GiB due to file holes. ZSTD block compression again reduces + this to 22GiB. When I get a chance, I hope to look into this further. - `qidx` creates a disk-backed hashset using a sparse memory-mapped file. The underlying -operating system must support `mmap` and file holes +operating system must support `mmap` and file holes. - `qidx` doesn't currently support compression. it is currently recommended to -use block-level compression (such as `zfs` `zstd` compression) -- the bamfile must be sorted by query name before the index is built `samtools sort -n` +use block-level compression (such as `zfs` `zstd` compression.) +- The bamfile must be sorted by query name before the index is built `samtools sort -n`. -- cgit v1.2.3