`samtools merge` multi-thread mode

[nimarafati]

It seems one can run samtools merge. Have you tested it? or samtools set the threads corresponding to number of files to merge?

[cnluzon]

Yes, that is a good suggestion, it could potentially improve the speed at which individual pooled BAM files are produced. However I have not tested this. I imagine it could make more of a difference in the case the BAM files are large (if the BAM files are quite sparse it might even be counter-productive). On the other hand if a run has many pools, Snakemake will spawn multiple merging processes at once if the number of availalbe threads is significantly larger than the amount of threads requested by a single pooling rule, parallelizing the processing at a different level. There is probably an optimal value for threads that depends on the size of the minute run, the sequencing depth and the number of available CPUs.

It is possible to override the specific threads value once it's set, but I would still need to add a -@ {threads} parameter to the rule to make it available and make these tests.