Academia-Industry ML
Home > Research > AP

The Application Processor SIG is primarily responsible for studying GPU processor architecture and heterogeneous multicore systems. For the study of GPU processor architecture, the focus will be on GPU architecture. The development of the “general-purpose computing on graphics processing units” (GPGPU) technology provides GPUs with the ability to perform multimedia computing such as image processing and computer vision computing. The establishment of the HSA Foundation accelerated the integration of CPU and GPU into a single chip, increasing the possibility of close collaboration between the CPU and the GPU. The scheduled development themes of the Application Processor SIG are as follows: (a) study the graphics architecture of advanced graphics algorithms; (b) investigate and use heterogeneous system architecture-based programmed graphics pipeline; and (c) consider the architectural design of graphic chips used in advanced processing. For the study of heterogeneous multicore systems, the focus will be on full-system simulations, which are crucial for conducting performance evaluations and design space exploration. For multicore platforms, which are characterized by extremely challenging software and hardware development, performance monitoring is the basis for optimizing system performance. The Application Processor SIG will develop a tool set to simulate and analyze heterogeneous multicore CPU+GPU systems software and hardware to help users explore and optimize their systems. Quick emulator (QEMU) and Multi2Sim will be combined to serve as the platform for analyzing heterogeneous processors. The said platform supports Linux/Android and OpenCL applications and enable users to analyze and explore the workload of system architecture as well as perform software debugging and optimization. In addition, users will be allowed to adjust crucial hardware configurations.

Circuit technology
Communication system
System applications