With the rapid development of design for specific fields accelerator
heterogeneous system based on accelerator is a new trend of computing architecture design. However
the complex heterogeneous system challenges the programming mode and the efficient interaction between processor and accelerator. How to describe the computing tasks between the main processor and the accelerator
and how to reduce the cost of data transmission between them
and how to make the processor efficiently complete the task management and scheduling of the accelerator are the key technology to ensure the performance of heterogeneous systems. Based on a reconfigurable array driven by data flow
this paper proposes a task management mechanism and interconnection mode with the master controller by abstracting its driving mode
data flow direction
input and output
including the host interface of hardware
task management system of software
etc.
which is implemented and verified on Rocket Core based on risc-v instruction.
关键词
Keywords
references
NAJAFI N, CLAYTON K W, BAER W, et al. An architecture and interface for VLSI sensors[C]// IEEE Solid-state Sensor % Actuator Workshop. Hilton Head Island, SC, USA, IEEE, 1988.
ASANOVIC K, AVIZIENIS R, BACHRACH J, et al. The rocket chip generator[J]. EECS Department, University of California. Berkeley, Tech. Rep. UCB/EECS-2016-17, 2016.
XIONG P X. Programming Guide For RT Thread Real Time Operating System[R]. Shanghai: RT-Thread Studio. 2010.
MARONGIU A, CAPOTONDI A, TAGLIAVINI G, et al. Simplifying Many-Core-Based Heterogeneous SoC Programming With Offload Directives[J]. IEEE Transactions on Industrial Informatics, 2015, 11(4):957-967. DOI: 957
FIOLHAIS L, GONÇALVES F, DUARTE R P, et al. Low Energy Heterogeneous Computing with Multiple RISC-V and CGRA Cores[C]// 2019 IEEE International Symposium on Circuits and Systems (ISCAS). Sapporo, Japan, 2019.
SUN Y X. Hardware Design And Realization Of Semaphore Management In RTOS[D]. Qingdao: Qingdao University, 2015.
ZHANG Y N, MA M M, LI K N. Brief History And Prospect Of Coprocessors[J]. China Science and Technology Information, 2008(13):115-117. DOI: 115
LI Y Y, PANG J M, LI Y B. Multi-dimensional Parallelism Recognition Method Of Nested Loop For Many-core Processors[J]. Application Research of Computers, 2018, 35(11):117-120. DOI: 117