PROJECT TITLE :
Boosting the FM-Index on the GPU: Effective Techniques to Mitigate Random Memory Access
The recent advent of high-throughput sequencing machines manufacturing massive amounts of short reads has boosted the interest in efficient string searching techniques. As of today, many mainstream sequence alignment software tools depend on a special data structure, called the FM-index, that allows for quick exact searches in large genomic references. However, such searches translate into a pseudo-random memory access pattern, thus creating memory access the limiting factor of all computation-economical implementations, both on CPUs and GPUs. Here, we have a tendency to show that many strategies can be put in place to get rid of the memory bottleneck on the GPU: additional compact indexes can be implemented by having a lot of threads work cooperatively on larger memory blocks, and a k-step FM-index will be used to additional reduce the quantity of memory accesses. The mix of those and alternative optimisations yields an implementation that is able to process concerning two Gbases of queries per second on our take a look at platform, being concerning eight× faster than a comparable multi-core CPU version, and concerning 3× to 5× faster than the FM-index implementation on the GPU provided by the recently announced Nvidia NVBIO bioinformatics library.
Did you like this research project?
To get this research project Guidelines, Training and Code... Click Here