Sambamba vs samtools performance

[Pjotr Prins]

Emilio Palumbo has been doing some performance tests on sambamba. He
discovered a speed regression in 0.6.7 (2x slower for certain
conditions). This was due to a change in compilation procedure. I had
split all compilations steps to make development easier, but LLVM
manages quite a bit of inline optimizations when you compile into a
single object. Bit late for me to realize!

As a bonus, upcoming 0.6.8 is 10-20% faster than 0.6.6. This is due to
D-compiler and LLVM improvements.

Other findings are that sambamba is still faster than samtools for
index, view and markdup.

sambamba lost out on sort, though I find that with enough RAM and
cores it can still be faster.

If you want to try latest build:

wget http://files.genenetwork.org/software/contrib/sambamba-0.6.8-pre1
md5sum sambamba-0.6.8-pre1
3d6a286947daebb44ac87b4ff2f16c66

Pj.

[Emilio]

Hi,

the code of the tests I ran and a sample report can be found on GitHub at the following link: 

https://github.com/guigolab/sambamBench-nf

It would be great if anybody else other than me could run the benchmark.

Best,

Emilio

--
You received this message because you are subscribed to the Google Groups "sambamba-discussion" group.
To unsubscribe from this group and stop receiving emails from it, send an email to sambamba-discussion+unsub...@googlegroups.com.
To post to this group, send an email to sambamba-discussion@googlegroups.com.
To view this discussion on the web, visit https://groups.google.com/d/msgid/sambamba-discussion/20180214100505.GA15599%40thebird.nl.
For more options, visit https://groups.google.com/d/optout.