Back to results
US 12,613,743 B2Granted
Method and terminal for scheduling processors
AssigneeSeoul National University R&Db Foundation
InventorsYoungki Lee, Changjin Jeong, Jingyu Lee, Changmin Jeon, Joo Seong Jeong, Byung-Gon Chun, Donghyun Kim
Life of this patent
Estimating the term…
Abstract
A single terminal performs scheduling of processing a request from a plurality of applications by using heterogeneous processors. The single terminal may include an analysis unit partitioning a request from an application in units and generating at least one subgraph, a profiling unit predicting an operation execution time for at least one frequency of at least one processor capable of processing the subgraph, and a scheduler performing scheduling based on a request from the application and the operation execution time.
Claims9
A terminal performing scheduling comprising:
- a plurality of processors;
- a memory storing instructions, wherein the instructions, when executed by one of the plurality of processors, cause the terminal to:
- execute an analysis unit to partition a request received from at least one application into units and generate a plurality of subgraphs, each subgraph including one or more of the units;
- execute a profiling unit to receive the plurality of subgraphs, and determine an operation execution time for each of the plurality of processors to process each of the received subgraphs;
- and
- execute a scheduler to assign at least one subgraph of the received subgraphs to at least one processor based on the operation execution time, wherein the scheduler is further executed to:
- receive a current frequency of each of the at least one processor and transmit the current frequency of each of the at least one processor to the profiling unit;
- request, from the profiling unit, an operation execution time corresponding to the current frequency, among available frequencies of the at least one processor, capable of processing the at least one subgraph, and receive, from the profiling unit, the operation execution time;
- and
- select subgraphs and processors based on predicted operation execution time corresponding to the current frequency,
- wherein the analysis unit is further executed to derive a plurality of subgraph sequences based on types of the processors capable of processing each of the units, and generate the subgraphs by combining consecutive units in each of the subgraph sequences, and
- wherein, when the consecutive units are processable by a same processor, the consecutive units are combined into the subgraphs.
9 more elements
A method of performing scheduling in a terminal, the method comprising:
- generating, by an analysis unit, a plurality of subgraphs by partitioning a request received from at least one application into a plurality of units;
- transmitting, by the analysis unit, the plurality of subgraphs to a profiling unit;
- transmitting, by the analysis unit, the plurality of units to a scheduler;
- determining, by the profiling unit, an operation execution time for each of a plurality of processors to process each of the subgraphs;
- performing, by the scheduler, scheduling by inserting the plurality of units into a queue having a same policy as the request;
- and
- assigning, by the scheduler, at least one subgraph of the subgraphs to at least one processor of the plurality of processors for processing a pending unit in the queue based on the operation execution time, wherein the assigning comprises:
- receiving a current frequency of each of the at least one processor and
- transmitting the current frequency of each of the at least one processor to the profiling unit;
- requesting, from the profiling unit, an operation execution time corresponding to the current frequency, among available frequencies of the at least one processor, capable of processing the at least one subgraph, and receiving, from the profiling unit, the operation execution time;
- and
- selecting subgraphs and processors based on predicted operation execution time corresponding to the current frequency,
- wherein the analysis unit is executed to derive a plurality of subgraph sequences based on types of the processors capable of processing each of the units, and generate the subgraphs by combining consecutive units in each of the subgraph sequences, and
- wherein, when the consecutive units are processable by a same processor, the consecutive units are combined into the subgraphs.
11 more elements
Dependencies are inferred from the claim text.