CALCULATOR AND JOB SCHEDULING METHOD THEREOF

• US 20200133722A1
• Filed: 10/25/2019
• Published: 04/30/2020
• Est. Priority Date: 10/26/2018
• Status: Active Grant

• Current Assignee: Samsung SDS (China) Co., Ltd. (Samsung Group)
• Original Assignee: Samsung SDS (China) Co., Ltd. (Samsung Group)
• Inventors: SUH, Man Suk, ROH, Hwan Kyun, PANG, Gi Beom
• Granted Patent: US 11,494,228 B2
• Time in Patent Office: Days
• CPC Class Codes: G06F 9/4881 Scheduling strategies for dispatcher, e.g. round robin, multi-level priority queues
• CPC Class Codes: G06F 9/5038 considering the execution order of a plurality of tasks, e.g. taking priority or time dependency constraints into consideration scheduling strategies G06F9/4881 and subgroups
• CPC Class Codes: G06F 9/505 considering the load
• CPC Class Codes: G06T 1/20 Processor architectures; Processor configuration, e.g. pipelining

Abstract

A method for scheduling jobs for the calculator includes measuring core utilization of the second-type processor, when the measured core utilization is less than a reference value, transmitting, by the first-type processor, a job suspension instruction to suspend a first job, which is currently being executed, to the second-type processor, in response to the job suspension instruction, copying data of a region occupied by the first job in a memory of the second-type processor to a main memory, copying data of a second job stored in the main memory to the memory of the second-type processor, and transmitting, by the first-type processor, an instruction to execute the second job to the second-type processor.

Claims

1. A job scheduling method performed in a computing system including a first-type processor and a second-type processor, the method comprising:

measuring core utilization of the second-type processor;

when the measured core utilization is less than a reference value, transmitting, by the first-type processor, a job suspension instruction to suspend a first job, which is currently being executed, to the second-type processor;

in response to the job suspension instruction, copying data of a region occupied by the first job in a memory of the second-type processor to a main memory;

copying data of a second job stored in the main memory to the memory of the second-type processor; and

transmitting, by the first-type processor, an instruction to execute the second job to the second-type processor.

2. The job scheduling method of claim 1, further comprising measuring free space of the memory of the second-type processor, wherein the transmitting of the job suspension instruction comprises, when the free space of the memory is less than a reference value and the measured core utilization is less than the reference value, transmitting, by the first-type processor, the job suspension instruction to suspend the first job, which is currently being executed, to the second-type processor.

3. The job scheduling method of claim 1, wherein the transmitting of the job suspension instruction comprises, when the measured core utilization is less than the reference value and the first job uses the second-type processor in a stable state, transmitting, by the first-type processor, the job suspension instruction to suspend the first job, which is currently being executed, to the second-type processor.

4. The job scheduling method of claim 3, wherein in the stable state, a ratio of a use time of the first-type processor to a use time of the second-type processor does not deviate from a certain value by a reference error.

5. The job scheduling method of claim 1, wherein the second-type processor is a graphics processing unit (GPU).

6. The job scheduling method of claim 5, wherein the memory of the GPU and the main memory exchange data with each other in a direct memory access (DMA) manner;

the copying of the data of the region occupied by the first job in the memory of the second-type processor to the main memory comprises directly copying the data of the region occupied by the first job in the memory of the second-type processor to the main memory in the DMA manner; and

the copying of the data of the second job stored in the main memory to the memory of the second-type processor comprises directly copying the data of the second job stored in the main memory to the memory of the second-type processor in the DMA manner.

7. The job scheduling method of claim 1, wherein the second job belongs to the same job switching group as the first job;

and

the job switching group is composed of a plurality of jobs which require memory sizes of the second-type processor within a certain range allocated to the job switching group.

8. The job scheduling method of claim 7, wherein the second job is selected from among jobs which wait to use the second-type processor while being executed through the first-type processor.

9. The job scheduling method of claim 7, wherein a plurality of jobs included in the job switching group to which the first job belongs are sequentially selected as the second job.

10. The job scheduling method of claim 1, wherein the second job requires the memory of the second-type processor corresponding to a size of region occupied by the first job in the memory of the second-type processor or less and waits to use the second-type processor while being executed through the first-type processor.

11. The job scheduling method of claim 1, wherein when a total memory requirement of two or more jobs does not exceed a memory capacity of the second-type processor, the second-type processor supports the two or more jobs to simultaneously use the second-type processor;

and

the transmitting of the job suspension instruction comprises transmitting, by the first-type processor, a job suspension instruction to suspend the first job between the first job and a third job which are currently being executed to the second-type processor.

12. The job scheduling method of claim 11, wherein the second-type processor is a graphics processing unit (GPU);

some of the two or more jobs are executed by the GPU in a first computing environment provisioned by a job scheduler; and

at least some among others of the two or more jobs are executed by the GPU in a second computing environment provisioned by the job scheduler.

13. The job scheduling method of claim 12, wherein the second job belongs to the same job switching group as the first job and is sequentially selected from among jobs which wait to use the second-type processor while being executed through the first-type processor provisioned by the job scheduler to a computing environment in which the first job has been generated.

14. The job scheduling method of claim 12, wherein the second job belongs to the same job switching group as the first job and is selected from among jobs, which wait to use the second-type processor while being executed through the first-type processor provisioned by the job scheduler to a computing environment in which the first job has been generated, on the basis of priorities given to the jobs;

and

the priorities given to the jobs are determined on the basis of priorities given to computing environments in which the jobs have been generated.

15. The job scheduling method of claim 1, wherein the copying of the data of the region occupied by the first job in the memory of the second-type processor to the main memory comprises performing a calculation of the first job in the first-type processor;

and

further comprising, after the transmitting of the instruction to execute the second job, performing processor switching between the first job and the second job when the first job executed in the first-type processor requires a calculation by the second-type processor.

16. The job scheduling method of claim 15, wherein the performing of the processor switching between the first job and the second job comprises performing processor switching between the first job and the second job even if the second job is using the second-type processor when the first job executed in the first-type processor requires a calculation by the second-type processor.

17. The job scheduling method of claim 15, wherein the performing of the processor switching between the first job and the second job comprises:

transmitting, by the first-type processor, a job suspension instruction to suspend the second job, which is currently being executed, to the second-type processor;

in response to the job suspension instruction, copying data of a region occupied by the second job in the memory of the second-type processor to the main memory;

copying the data of the first job stored in the main memory to the memory of the second-type processor; and

transmitting, by the first-type processor, an instruction to execute the first job to the second-type processor.

18. A job scheduling method performed in a computing system including a first-type processor and a second-type processor, the method comprising:

repeatedly loading jobs dequeued from a job queue into a memory of the second-type processor until no more job can be loaded;

measuring core utilization of the second-type processor;

when the measured core utilization is less than a reference value, selecting, by the first-type processor, one suspension job from among jobs currently being executed in the second-type processor and transmitting a job suspension instruction to suspend the selected suspension job to the second-type processor;

in response to the job suspension instruction, copying data of a region occupied by the selected suspension job in the memory of the second-type processor to a main memory;

copying data of a new job, which is selected from the job queue as a job constituting a job switching group together with the suspension job, to the memory of the second-type processor;

transmitting, by the first-type processor, an instruction to execute the new job to the second-type processor; and

monitoring the core utilization of the second-type processor and sequentially executing jobs belonging to the job switching group in the second-type processor every time the core utilization is less than the reference value.

19. A calculator comprising:

a main memory to which a plurality of instructions are loaded;

a first-type processor configured to execute the plurality of instructions loaded into the main memory; and

a second-type processor including a dedicated memory,

wherein the plurality of instructions include;

an instruction to measure core utilization of the second-type processor;

an instruction to transmit a job suspension instruction to suspend a first job, which is currently being executed, to the second-type processor when the measured core utilization is less than a reference value;

an instruction to copy data of a second job stored in the main memory to the memory of the second-type processor when data of a region occupied by the first job in the memory of the second-type processor is copied to the main memory in response to the job suspension instruction; and

an instruction to transmit an instruction to execute the second job to the second-type processor.