What is TFM-CUDA ?
TFM-CUDA is a CUDA implementation of parallel algorithms able to:
- scan a matrix or a set of matrices against a sequence (see also TFM-Scan),
- compute the P-value corresponding to a score, or the score corresponding to a Pvalue (see also TFM-Pvalue),
- compare two matrices
The method is described in the paper referenced below, and uses parallelism on graphics processing units (GPUs). TFM-CUDA was designed to speed-up some commonly computations on PWMs. If you prefer full-featured tools, please use TFM-Scan and TFM-Pvalue.
Download
TFM-CUDA needs the NVIDIA CUDA SDK. TFM-CUDA is licensed under GPLv3.
TFM-CUDA-0.5.tar.gz
(2010/06/28)
TFM-CUDA-0.4.tar.gz
(2009/11/30)
TFM-CUDA-0.3.tar.gz
(2009/04/10)
TFM-CUDA-0.2.tar.gz
(2009/01/27)
TFM-CUDA-0.1.tar.gz
(2008/12/01)
See the Changelog.
References
- Mathieu Giraud and Jean-Stéphane Varré, Parallel Position Weight Matrices Algorithms, Parallel Computing, 10.1016/j.parco.2010.10.001
- An earlier version appeared in International Symposium on Parallel and Distributed Computing (ISPDC 2009), 65-69, 10.1109/ISPDC.2009.31, 2009