You are cordially invited by the COE/ICS, to attend a Graduate Seminar, on the above given title, by Mr. Mohammed Omer Assayony, Ph.D Student, College of Computer Science & Engineering, on Tuesday, 16th April, 2013, at 02:30 PM, in Building 22, Room 132.
Abstract: The advances of Graphic Processing Units (GPU) technology and the introduction of CUDA programming model facilitate developing new solutions for sparse and dense linear algebra solvers. Matrix Transpose is one important linear algebra procedure that has deep impact in various computational science and engineering applications. Several implementations have been proposed in the literature to enhance the computation speed. However, several factors hinder the expected performance of large matrix transpose on GPU devices. The degradation in performance involves the memory access pattern such as coalesce access in the global memory and bank conflict in the shared memory of streaming multiprocessors within the GPU. Two matrix transpose algorithms are proposed to alleviate the aforementioned issues of ensuring coalesced access and conflict free bank access. Evaluation of proposed have comparable execution times to optimized algorithms in the literature. The main advantage of proposed algorithms is that they eliminate bank conflicts while allocating exactly the tile size of the memory space. However, to the best of our knowledge an extra space of T×T tile N× (N+1) needs to be allocated in the published research.
About the speaker: Mr. Mohammed Omer Haj Assayony, is a Ph.D. candidate at KFUPM (Computer Science & Engineering). He received MS (Computer Science) from Universiti Sains Malaysia (USM) in 2009 and B.Sc. (Computer Science) from Al-Ahgaff University, Yemen in 2003. He is working as a lecture in Alandalus University, Yemen since 2004. His area of interest is in designing parallel algorithms for scientific applications.