Releases: bioinfo-pf-curie/mpiBWA
version-1.6
version-1.6
- IO preallocation improvement (issue #68)
- tested with a default stripping on Lustre of 8 OST and 1Gb stripe for the fastqs (lfs setstripe -c 8 -S 1G fastqs) before calling mpiBWA
version-1.5.5
version-1.5.4
improve IO
version-1.5.3
version-1.5.3
- fix a memory leak with trimmed reads and single end (#57)
- improved performances with trimmed alignment
version-1.5.2
version-1.5.1
version-1.5.1
- fix bug in trimmed fastq condition (thanks to Jianshu) (issue #45) (issue #51)
- fix load balancing (thanks to Vasim) (issue #47)
- overlap writing and computing (only done for mpiBWA) (issue #48)
- NEW! shared memory reference is no more the defaut placement. By default reference is shared by individual mpi job (issue #43)
- NEW! add NUMA aware option -z for placement in memory of the reference genome
options available with openMPI are shared, socket, numa, l1, l2, l3, core
options available for other MPI are shared, core
from experiences pinning mpi jobs by numa and sharing
the reference by numa has the better throughput (see documentation) (thanks to tahashmi) (issue #43)
version-1.5
version-1.4
version-1.3
- fix memory leaks (issue #36)
version-1.2
- Add an option -f to fixe the mate score, quality and cigar (like samtools fixmate -m does).
Usefull for compatibility with future mpiMarkdup or samtools markdup. When -f is set there is no more discordant.sam.
A typical workflow is mpiBWAByChr (f) + mpiSort (u + b) + samtools index + samtools markdup
or mpiBWAByChr (f) + mpiMarkdup (issue #31) - Add the option -K to set the number of bases to analyse (used for reproducibility) (issue #32).
- Fix a bug for offset read initialization and invalid free (issue #34)
- Fix load balancing (issue #30)