The MASA-CUDAlign extension is used with the MASA architecture to align DNA sequences of unrestricted size with the Smith-Waterman/Needleman-Wunsch algorithm combined with Myers-Miller. It uses the NVIDIA CUDA GPU platform for accelerating the computation time.
Switch branches/tags
Nothing to show
Clone or download
Latest commit b817de6 Aug 28, 2015
Permalink
Failed to load latest commit information.
masa-cudalign-3.9.1.1024 v1.3.9.1024 May 12, 2015
releases v1.3.9.1024 May 12, 2015
LICENSE Initial commit Apr 14, 2015
README.md Update README.md Aug 28, 2015

README.md

MASA-CUDAlign

The MASA-CUDAlign extension is used with the MASA architecture to align DNA sequences of unrestricted size with the Smith-Waterman and Needleman-Wunsch algorithms combined with Myers-Miller. It uses the NVIDIA CUDA platform to accelerate the computation time. This extension is able to align huge DNA sequences with more than 200 million base pairs (MBP).

Download

Latest Version: masa-cudalign-3.9.1.1024.tar.gz

Compiling (cudalign)

tar -xvzf masa-cudalign-3.9.1.1024.tar.gz
cd masa-cudalign-3.9.1.1024
./configure
make

Executing CUDAlign

./cudalign [options] seq1.fasta seq2.fasta

All the command line arguments can be retrieved using the --help parameter. See the wiki for a list of command line examples.

Performance Benchmarks

We have executed MASA-CUDAlign in many different environments. Here we will present the best results in different scenarios. We recommend to read the reference papers in order to understand the feature improvements in each test.

Test Environment (Homogeneous GPU cluster): [CCGRID2014]
Minotauro Cluster - 64 x NVidia Tesla M2090.

Sequence 1 Sequence 2 Len1 Len2 Time GCUPS
NC_000001.10 NC_006468.3 249M 228M 9h09m25s 1726.47

Test Environment (Heterogeneous GPU Nodes): [PPOPP2014]
Laico Labs - 3 Hosts - 1 x NVidia GTX 580 + 2 x NVidia GTX 680.

Sequence 1 Sequence 2 Len1 Len2 Time GCUPS
NC_000019.9 NC_006486.3 59M 64M 7h29m17s 139.60
NC_000020.10 NC_006487.3 63M 62M 7h42m08s 140.31
NC_000021.8 NC_006488.2 48M 46M 4h27m04s 139.63
NC_000022.10 NC_006489.3 51M 50M 5h03m00s 140.36

Test Environment (Heterogeneous GPUs in Single Node): [PPOPP2014]
Panoramix Host - 1 x NVidia Tesla K20c + 2 x NVidia Tesla C2050.

Sequence 1 Sequence 2 Len1 Len2 Time GCUPS
NC_000019.9 NC_006486.3 59M 64M 10h20m52s 101.02
NC_000020.10 NC_006487.3 63M 62M 7h42m08s 100.96
NC_000021.8 NC_006488.2 48M 46M 6h10m38s 100.62
NC_000022.10 NC_006489.3 51M 50M 7h03m36s 101.38

Test Environment (Single GPU): [TPDS2013]
NVIDIA GeForce GTX 560 Ti

Sequence 1 Sequence 2 Len1 Len2 Time GCUPS
CP000051.1 CP000051.1 1M 1M 43s 25.82
BA000035.2 BX927147.1 3M 3M 6m07s 28.15
AE016879.1 AE017225.1 5M 5M 9m18s 48.98
NC_005027.1 NC_003997.3 7M 5M 22m01s 28.28
NT_033779.4 NT_037436.3 23M 25M 5h29m17s 28.59
NC_000024.9 NC_006492.2 59M 24M 13h05m23s 30.18
BA000046.3 NC_000021.7 33M 47M 8h26m09s 50.70

License:

MASA-CUDAlign is an open source project with public license (GPLv3). A copy of the LICENSE is maintained in this repository.

External Links

CUDAlign 3.0: Parallel Biological Sequence Comparison in Large GPU Clusters

Retrieving Smith-Waterman Alignments with Optimizations for Megabase Biological Sequences Using GPU