mgawan / adept

revamping adept from scratch to make more usable in library form

ADEPT

ADEPT is a GPU accelerated sequence alignment library for short DNA reads and protein sequences. It provides API calls for asychronously performing sequence alignments on GPUs while keeping track of progress on GPUs so that work can be performed on CPUs. Different capabilities of ADEPT API are explored using examples in the examples folder.

Dependencies

ADEPT can be built using CUDA 9.0 or later and a version of GCC compatible with the CUDA version that is used.

Building

To build ADEPT library along with the C++ and Python examples, first move into the top level directory of the repo and follow the below steps:

To build:

mkdir build
cd build
export PYTHONPATH=$PWD:$PYTHONPATH
cmake -DADEPT_USE_PYTHON=ON ../
make

To run

To run the installed examples (examples are also used as test):

make test

If you do not need python module of ADEPT, simply use:

mkdir build
cd build
cmake ../
make

and follow the above steps to run the tests.

Contact

If you need help modifying the library to match your specific use-case or for other issues and bug reports please open an issue or reach out at [email protected]

Citation

Awan, M.G., Deslippe, J., Buluc, A. et al. ADEPT: a domain independent sequence alignment strategy for gpu architectures. BMC Bioinformatics 21, 406 (2020). https://doi.org/10.1186/s12859-020-03720-1

Credits

1. Muaaz Awan
2. Muhammad Haseeb

License

ADEPT: a domain independent sequence alignment strategy for GPU architectures Copyright (c) 2019, The Regents of the University of California, through Lawrence Berkeley National Laboratory (subject to receipt of any required approvals from the U.S. Dept. of Energy). All rights reserved.

If you have questions about your rights to use or distribute this software, please contact Berkeley Lab's Intellectual Property Office at [email protected].

NOTICE. This Software was developed under funding from the U.S. Department of Energy and the U.S. Government consequently retains certain rights. As such, the U.S. Government has been granted for itself and others acting on its behalf a paid-up, nonexclusive, irrevocable, worldwide license in the Software to reproduce, distribute copies to the public, prepare derivative works, and perform publicly and display publicly, and to permit other to do so.

adept's Issues

Segmentation fault during execution

Hello,

Inside driver::kernel_launch at driver.cpp, should this for loop go up to query_seqs.size()? It seems that there are cases we are trying to access a position on offset_que, that doesn't exist. Maybe you need the iterations to go up to batch_size?

ADEPT/adept/driver.cpp

Line 115 in ed08b42

for(int i = 0; i < query_seqs.size(); i++){

simple_sw fails for different maximum lengths of ref/query

Hello,

I try to execute the simple_sw example but with larger sequences. I increased the MAX_REF_LEN and MAX_QUERY_LEN and even for the sample inputs the correctness test fails. For example, I used the following values and it fails:
constexpr int MAX_REF_LEN = 8000; constexpr int MAX_QUERY_LEN = 2000;

Is there a limit on the maximum length of ref and query or the number of sequences? Also is there a relation between these 2 values? Can the maximum value for the reference and query be the same?

Little Mistakes in file readings

Hey, there are some silly mistakes when reading files. In the case of simple_sw.cpp line 65 and 67 are the same comparison, I guess one should be done with lineQ, and on line 88 you are checking the length of Q against largestA (which I guess should be largestB) and updating largestB. The same happens for other examples.
It is a silly mistake but I thought you might want to correct it.