WorkshopsFast Methods 2: Effective SpMV Implementation on Modern Parallel Architectures
67
reads
Sparse matrix vector multiplication (SpMV) is widely used in numerical calculation applications. Moreover, the execution time of SpMV is often the dominant time expenditure in many applications. Therefore, the acceleration of SpMV is extremely important. Today, two parallel computation hardware, GPU and MIC are attracting great attention. GPUs are used to accelerate various computations at many computer centers, laboratories, and advanced industries. Kepler is NVIDIA’s latest high performance GPU, which has some new features and much higher performance than existing GPUs. MIC is an Intel’s new parallel computation hardware, which has many computation cores and expected to obtain better performance than existing CPUs. While MIC is in an early phase of popularization, a lot of scientists are paying attention to this new hardware. The author has taken up SpMV on GPU in a few years and now trying to take up MIC, too. In this talk, current optimization techniques and performance of GPU and MIC are revealed.
reads
Satoshi Ohshima
2013-03-28
09:55:00 - 10:20:00
101 , Mathematics Research Center Building (ori. New Math. Bldg.)
Sparse matrix vector multiplication (SpMV) is widely used in numerical calculation applications. Moreover, the execution time of SpMV is often the dominant time expenditure in many applications. Therefore, the acceleration of SpMV is extremely important. Today, two parallel computation hardware, GPU and MIC are attracting great attention. GPUs are used to accelerate various computations at many computer centers, laboratories, and advanced industries. Kepler is NVIDIA’s latest high performance GPU, which has some new features and much higher performance than existing GPUs. MIC is an Intel’s new parallel computation hardware, which has many computation cores and expected to obtain better performance than existing CPUs. While MIC is in an early phase of popularization, a lot of scientists are paying attention to this new hardware. The author has taken up SpMV on GPU in a few years and now trying to take up MIC, too. In this talk, current optimization techniques and performance of GPU and MIC are revealed.