发明名称 |
Scheduling in a multicore architecture |
摘要 |
This invention relates to scheduling threads in a multicore processor. Executable transactions may be scheduled using at least one distribution queue, which lists executable transactions in order of eligibility for execution, and multilevel scheduler which comprises a plurality of linked individual executable transaction schedulers. Each of these includes a scheduling algorithm for determining the most eligible executable transaction for execution. The most eligible executable transaction is outputted from the multilevel scheduler to the at least one distribution queue. |
申请公布号 |
US9164953(B2) |
申请公布日期 |
2015.10.20 |
申请号 |
US201414267830 |
申请日期 |
2014.05.01 |
申请人 |
Synopsys, Inc.;Fujitsu Semiconductor Limited |
发明人 |
Lippett Mark David |
分类号 |
G06F15/80;G06F1/32;G06F9/48;G06F9/46;G06F9/50 |
主分类号 |
G06F15/80 |
代理机构 |
Fenwick & West LLP |
代理人 |
Fenwick & West LLP |
主权项 |
1. A method of scheduling executable transactions in a multicore processor comprising a plurality of processor elements, wherein at least one processor element comprises reconfigurable logic, the method comprising:
providing a first configuration queue of executable transactions corresponding to a first configuration of the reconfigurable logic and a second configuration queue of executable transactions corresponding to a second configuration of the reconfigurable logic, the reconfigurable logic comprising one of a field programmable gate array (FPGA) or a memory, the executable transactions allocated to and ready for execution by the reconfigurable logic; providing a plurality of executable transaction schedulers, each executable transaction scheduler comprising a scheduling algorithm for determining a most eligible executable transaction for execution from a number of candidate executable transactions; linking the executable transaction schedulers together into a multilevel scheduler; outputting the most eligible executable transaction from the multilevel scheduler to one of the first configuration queue and the second configuration queue; outputting the executable transactions of the first configuration queue to the reconfigurable processor element for execution when the reconfigurable logic is configured according to the first configuration; reconfiguring the reconfigurable logic according to a second configuration when a pre-determined threshold is reached; and outputting the executable transactions of the second configuration queue to the configurable processor element for execution when the reconfigurable logic is configured according to the second configuration. |
地址 |
Mountain View CA US |