Wednesday, 7 December 2011

[Bio-bwa-help] enhancing speed of bwa aln and bwa samse/sampe

IMHO BWA is already very fast and individual reads are easily split to parallelise the process. Though I guess having a threaded version simplifies workflow development. 


---------- Forwarded message ----------
From: Mark Kelly
Date: 2011/12/7
Subject: Re: [Bio-bwa-help] enhancing speed of bwa aln and bwa samse/sampe



We have created a threaded parallel bwa samse and sampe

enhancement that also includes overlapped i/o of the seq

and aln files for some performance gains, see below.

Attached is a patch file of this enhancement for bwa-0.5.9.  

We will be working on a 0.6.1 version soon.


Here at Convey, we also have accelerated bwa aln (0.5.9) to

run many times faster on our server and FPGA-based coprocessor

than an x86 multi-core server.  If you are interested, please



sampe example:


% bwa sampe ….


original 0.5.9 version


real  53m5.028s

user  47m16.860s

sys   1m20.615s


% bwa sampe –t 8 ….


latest 0.5.9 patch version


real  10m29.499s

user  53m31.964s

sys   1m16.295s


addl cmd-line options are


   -t INT   number of threads [1]

   -T       dont resize read buffer for threads

   -X       disable async read seq method

   -Y       disable async print results method


Note memory usage will go up some when

more cores are requested.


Let me know if you have any questions.




No comments:

Post a Comment

Datanami, Woe be me