Dr.Abhinav Sarje

Postdoctoral Fellow

Computational Research Division

Lawrence Berkeley National Laboratory

Abstract: In the area of computing, the degree of parallelism is continually increasing at various levels, such as SIMD processing, SMT architectures, and multi/many-cores with thread-level shared memory and SIMT parallelism, just to name a few low-level parallelisms available. To develop high-performance parallel algorithms and efficient software tools in scientific computing, it is necessary to harness such raw computational power available.

This talk will cover some of these areas of parallelism, specifically many-core graphics processors, starting with a basic introduction to the GPU architectures, and the CUDA programming model. As case studies targeting large-scale scientific applications, this will be supplemented with some of the architecture-aware parallel techniques on GPUs, which we have developed in the past couple of years. We will look at the problem of scheduling all-pairs computations with applications to fluid dynamics and systems biology, and at the analyses of X-ray scattering data (SAXS and GISAXS), which are key to the design and fabrication of energy-relevant nano-devices. Control theory provides a unique perspective in the study of quantum dynamics, and has broad applications ranging from many-body physics, quantum simulations, to quantum information processing. We will make a review of some recent development of quantum control including both the fundamental theory as well as some potential experimental implementations. In particular, as an application, I will discuss ultraefficient cooling of resonators by quantum control, and then we will explore the relationship between symmetry and controllability and how this is related to specific control designs.

Date&Time: February 23, 2012（Thursday） 14:30 – 15:30
Location: 606 Conference Room