Doppelmark too slow

[fgvieira]

Dear all,

I recently decided to check doppelmark speed on a BAM file, given claims that it could process a ~250Gb file in little over 30min using 40 cpus (can't find the blog now). However, on a file with 1'709'696'127 reads it took over 10 hours and the average cpu usage was only 3 (even though I set it to 40)!

I am using the command:

doppelmark --parallelism 40 --logtostderr -v 2 --optical-distance -1 --bam input.bam --index input.bai --clear-existing --disk-mate-shards 0 -tag-duplicates -clip-padding 200 -scratch-dir /tmp --metrics output.metrics --output output.bam

Am I doing anything wrong? Is there any options that can be used to increase the speed?

thanks,