Efficient usage strategy of limited shared memory in Graphical Processing Unit (GPU) for accelerate DNA sequence alignment / Ahmad Hasif Azman

DNA sequence alignment is expected to help uncover important information about the human body, disease, genetics and other biological relationships when discovered. In addition, intensive efforts have been made to improve the performance of sequence alignment through hardware-based acceleration using the Graphical Processing Unit (GPU) accelerator. This implementation is becoming increasingly popular due to the flexibility of the accelerator design, parallel computational solutions and the ability to simultaneously increase the performance of the alignment. The performance of the DNA sequence alignment system is highly dependent on the algorithm, GPU designed architecture and accelerator performance. In this study, the focus is on utilizing the memory capabilities of GPUs to accelerate the Smith-Waterman algorithm has been proposed. Three new approaches based on global memory, shared memory and a combination of global and shared memory are used in this design. Moreover, the execution time proves that the design is able to speed up the computational process by about 90% compared to the Central Processing Unit (CPU). Again, the result proves that the acceleration of the GPU is able to speed up the processing of the DNA sequence alignment without affecting the result. Finally, the results obtained have shown that the proposed system offers better performance and design than previous work on accelerating SWA DNA sequence alignment using GPU accelerators.