Dong Chen

Room 737, Computer Studies Building
Department of Computer Science
University of Rochester
Rochester, NY, 14627



A collection of OpenCL benchmarks for performance evaluation: NVIDIA SDK, Parboil, SHOC, SNU NPB.
Memory Access Analysis
A LLVM based tool which can extract memory access pattern of global memory accesses from OpenCL kernel code. A linear abstract representation is designed to describe the global memory accesses.
OpenCL Runtime
A Distributed Shared OpenCL Memory (DSOM) is designed in OpenCL runtime, which relieves users of having to manage data transfer explicitly among multiple devices. DSOM allocates shared buffers in the system memory and treats the on-device memory as a software managed virtual cache buffer. The management is based on the global memory access pattern.
OpenCL Performance Portability
Targeting on different computing devices with different architectures (CPU, GPU, MIC), optimizations to improve performance of OpenCL programs are summarized. A performance concerned code transformation (from fine-grained program to coarse-grained program) is being studied.



Fall 2014: CSC 400(proposal), CSC 453, CSC 456, CSC 486

UR Library | UR HRMS | Gym Hours | Campus Calendar | UR Registrar | URCS People | URCS Home Page