JP2009151375A - Job scheduling method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a job scheduling method for efficiently using the capability of a CPU 13. <P>SOLUTION: In this job scheduling method in which a CPU 13 performs scheduling in a fixed cycle execution system for executing a plurality of processings within a scheduled time based on job management information, the job management information has the priority information of each of the plurality of processings and the execution time information of a plurality of jobs capable of executing each of the plurality of processings, and when the spare time of the CPU 13 is generated, the job whose execution time is longer than the next to the job scheduled to be performed is selected from among a plurality of jobs based on priority. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to a job scheduling method for performing a plurality of processes by a fixed cycle execution method.

  In a multitask process in which a plurality of different processes are performed at the same time using a single processor, a fixed-cycle execution process is known in which a time for performing each process is assigned as a reserved time in advance. In the fixed-cycle execution process, each process is repeatedly executed at a set fixed-cycle time. Also, a process of repeatedly executing a series of predetermined processes within a predetermined sampling time, such as data sampling, is one type of fixed-cycle execution method when the sampling time is considered as a reserved time. In the fixed-cycle execution method, in order to complete the processing within the reserved time, the processor previously stores the time required for executing the job for executing each processing as job management information. In the fixed-cycle execution method, a scheduler that performs scheduling for registering a plurality of jobs as a queue is used so that the processing is completed within a reserved time based on job management information. In addition to completing the processing within the reserved time, the scheduler minimizes the wasted free time based on the job management information in order to reduce the wasted free time due to the processing event waiting, that is, the processor standby state. To perform scheduling.

  However, even when processing exactly the same job, the time until the processor actually completes processing is not always constant. This is due not only to fluctuations in the processing speed of the processor itself, but also to various factors such as the transfer speed of the peripheral bus. For this reason, to complete the processing within the scheduled time, the worst execution time (WCET: Worst-Case Execution Time) is stored as the time required for a specific job stored as job management information. It is necessary to keep it.

  For this reason, when the time until the processor actually completes processing is shorter than the worst execution time, the processor idle time increases, and the processor capacity cannot be used efficiently. was there. The above problem is particularly remarkable in the case of a process with a large variation in execution time.

  For this reason, Patent Document 1 discloses a task scheduling method that alternately schedules a high priority task and a low priority task when the CPU wait time exceeds a specified value.

  Here, in recent television receivers, since a large number of complicated processes are performed in parallel by digitization, a multi-core processor, for example, one general-purpose processor core in one CPU and eight Heterogeneous multi-core combining simple processor cores is used. However, in a television receiver, 60 images per second, that is, an image of 60 images / second in total of 30 images / second for every odd and even number of scanning lines must be displayed. For this reason, job scheduling for performing a plurality of processes by a fixed-cycle execution method with a period of 1/30 second is important.

  In addition, with the increase in the screen size of television receiver monitors, a single display in which a plurality of images taken from a plurality of input image signals input from a plurality of channels and / or input terminals is divided into a plurality of regions. A multi-screen display function for displaying in each area of the screen has been used.

  In a television receiver having a multi-screen display function, each of a plurality of image signals selected from a plurality of input image signals is subjected to image size conversion processing and format conversion processing by a multi-screen processing circuit. It is necessary to compose as a display screen. For this reason, in job scheduling, a reserved time for performing at least image size conversion processing and format conversion processing is set.

  However, the time required for processing by the multi-screen processing circuit varies depending on the number of images to be displayed on the multi-screen, that is, the number of input image signals. In addition, when a program running on another processor core of the heterogeneous multi-core performs a process of compressing the bandwidth, the speed of the CPU that performs the multi-screen process may be affected and become slow.

  For this reason, when the scheduled time based on the worst execution time is used, the processor idle time increases, and the problem that the processor capacity cannot be used efficiently is remarkable.

The present invention has been made in view of the above problems, and provides a job scheduling method capable of efficiently using the capacity of a processor.
JP-A-4-191935

  An object of the present invention is to provide a job scheduling method that can efficiently use the capacity of a processor.

  According to one aspect of the present invention, based on job management information, a CPU performs scheduling in a periodic execution method in which a plurality of processes are executed within a reserved time, and the job management information includes: A plurality of jobs, each having priority information and execution time information of a plurality of jobs capable of executing each of the plurality of processes, and when a CPU idle time occurs, A job scheduling method is provided that selects a job having a long execution time next to a job scheduled to be executed.

  The present invention provides a job scheduling method that can efficiently use the capacity of a processor.

Embodiments of the present invention will be described below with reference to the drawings.
<First Embodiment>
<Scheduler configuration>
First, the configuration of the scheduler 100 according to the first embodiment of the present invention will be described with reference to FIGS. 1 to 5.

  FIG. 1 is a configuration diagram showing a configuration of a television receiver 1 having a multi-screen processing circuit 11 according to the present embodiment, and FIG. 2 is a diagram showing an example of a multi-screen of the television receiver 1. 3 is a block diagram of the scheduler 100, FIG. 4 is a table showing an example of the job management information 105, and FIG. 5 shows a plurality of jobs with different algorithms that can be executed by the image size conversion processing A2. It is explanatory drawing for demonstrating.

  As shown in FIG. 1, the television receiver 1 according to the present embodiment temporarily stores a plurality of image signals selected from a large number of input image signals in a memory 15, and based on the data, The screen processing circuit 11 performs an image size conversion process and a format conversion process for each image signal, and outputs the image signal to the monitor 12 as an image output unit via the memory 15 again. For example, digital television broadcast signals received by the parabolic antenna 2 are processed by different decoders depending on the broadcast station and broadcast specifications. In FIG. 1, a simplified decoder group including four decoders 4, 5, 6, and 7 is illustrated. Further, an image signal recorded in a built-in hard disk drive (HDD) 8 and an image signal from a video camera or a game machine (not shown) connected to the external terminal 9 are also input to the television receiver 1. An image signal selected by the switching circuit 10 from the plurality of input image signals is processed by the multi-screen processing circuit 11. The multi-screen processing circuit 11 is a circuit that combines a plurality of image signals into one display image signal.

  The CPU 13 of the multi-screen processing circuit 11 has a fixed period of 1/30 seconds in order to process 60 images per second, that is, 30 / second image signals for every odd and even scanning lines. Processing is performed according to the queue memory information registered by the scheduler 100 that performs job scheduling of the execution method. Note that the CPU 13 processes not only the process of the multi-screen processing circuit 11 but also another process B by a fixed-cycle execution method with a period of 1/30 second.

  Next, FIG. 2 shows an example of a monitor display screen 20 of a television receiver that displays a multi-screen synthesized by the multi-screen processing circuit 11. 2A shows a normal single screen 20A, and FIG. 2B shows a double window screen in which two screens 20B and 20C having the same size are displayed side by side. FIG. ) Shows a double window screen in which a small child screen 20D and a large parent screen 20E are displayed side by side. FIG. 2D shows a multi-window screen in which four screens 20F, 20G, 20H, and 20I are displayed simultaneously. Show. The normal screen 20A is displayed with (1920 × 1080) pixels, but each screen of the multi-screen is reduced and displayed. That is, the screens 20B, 20C, 20F, 20G, 20H, and 20I are all displayed with (960 × 540) pixels that are half of the normal screen 20A, and 20D is a quarter (480 × 270) of the normal screen 20A. Displayed with pixels, 20E is displayed with (960 × 540) pixels, which is three-fourths of the normal screen 20A.

  Next, FIG. 3 shows a block diagram of the scheduler 100 according to the present embodiment. The scheduler 100 registers the job executed by the CPU 13 as a queue in the queue memory 108, that is, performs scheduling. In order to display a multi-screen as shown in FIG. 2, the multi-screen processing circuit 11 must execute at least two different purposes. That is, a format conversion process (hereinafter referred to as process A1) for converting the format of the original image signal and an image size conversion process (hereinafter referred to as process A2) for converting the image size of the original image signal. The image size of the original image signal includes HD (1920 × 1080), SD (720 × 480), etc. Here, the image size conversion that reduces the processing A2 by setting the image size of the original image signal to HD This will be described as processing. Of course, there is an image size conversion to enlarge.

  Hereinafter, the process A1 and the process A2 are collectively referred to as a process A, and a process that is executed together with the process A by the same CPU is referred to as a process B. Here, the queue registered in the queue memory 108 shown in FIG. 3 represents a fixed cycle time of 1/30 second, that is, about 33 ms, expressed by a vertical length of a rectangle, and a reservation for executing the processing A among them is performed. For example, the time is 70% of the fixed cycle time. Although the reservation times of the processes A and B can be reset, the job scheduling in a state where the reservation time is fixed will be described below. That is, an example in which the process A is executed in about 23 ms, which is 70% of the fixed cycle time of about 33 ms, will be described.

  As illustrated in FIG. 3, the scheduler 100 includes a selector 103 that selects a job based on information stored in the job management information memory 105 and a selector 103 for registering a job in the queue memory 108. The resetting unit 102 registers the selected job in the queue memory 108. The detector 104 collects data related to the usage status of the CPU 13 in a series of processes running on the fixed-cycle scheduler. For example, the detector 104 performs processing when the CPU 13 performs processing according to information in the registered queue memory 108. The actual time required for completion (hereinafter referred to as “actual measurement execution time”) is detected.

  The job management information memory 105 is a memory for storing management information of a plurality of jobs that actually execute processing. As shown in FIG. 4, in the job scheduling method according to the present embodiment, at least when processing is started, Basic information 107 including priority information and worst execution time information of a job is included. The basic information 107 is input as a parameter together with a fixed period time before the processing is started. The job management information memory 105 also has dynamic information 106 that is actually measured execution time information that is input based on information from the detector 104 after the start of processing.

  In the job scheduling of this embodiment, as shown in FIG. 4, one job is selected from a plurality of jobs that can execute a certain process, and scheduling is performed. Here, an example is shown in which three jobs of job A21, job A22, and job A23 are used for the image size conversion processing A2 for reducing the image size of the original image signal. Job A21, job A22, and job A23 are jobs that can individually execute image size conversion processing. However, the three jobs have different algorithms for performing image size conversion processing as described below.

  Here, a job used for image size conversion processing, that is, processing for reducing the image size of the original image signal will be described with reference to FIG. For example, as shown in FIG. 2D, in order to simultaneously display four images on the screen, the size of each image must be reduced to a quarter. Note that various different algorithms can be used for a job for reducing the image size. Here, for simplicity of explanation, an example of a simple algorithm using pixel data will be described.

  That is, as shown in FIG. 5a, when the original full pixel data is composed of 24 pixels of 6 pixels horizontally and 4 pixels vertically, to reduce the image size to 1/4, It is to convert to 6 pixels of 2 pixels. The algorithm a21 is a so-called thinning algorithm in which one pixel is randomly selected from four pixels of 2 pixels in the vertical direction and 2 pixels in the horizontal direction as the full pixel data, and the reduced image data is created by collecting the pixels. On the other hand, the algorithm a22 creates a new pixel by averaging information of 4 pixels of 2 pixels in the vertical direction and 2 pixels in the horizontal direction of the full pixel data, and creates the reduced image data by collecting the pixels. It is a simple averaging algorithm. The algorithm a23 is also an averaging algorithm that averages the information of nine pixels of three vertical pixels and three horizontal pixels. However, the algorithm a23 does not simply average, but contributes to the reduced pixels according to the distance from the central pixel. change. Therefore, this is a so-called composite averaging algorithm in which information of one pixel among the full pixels is reflected by weighting a plurality of reduced pixels.

  Since the algorithm a21 which is a thinning algorithm is the simplest algorithm, the job A21 using the algorithm a21 has a short processing time as shown in FIG. 4, for example, the worst execution time is 15 ms. However, since the algorithm a21 is simply thinning out, the quality of the reduced image is not good, that is, the algorithm a21 has a low quality as a result of processing. On the other hand, since the algorithm a22 which is a simple averaging algorithm is a relatively simple algorithm, the job A22 using the algorithm a22 has a relatively short processing time as shown in FIG. 4, for example, the worst execution time is 25 ms. Further, since the algorithm a22 is averaging, the quality of the reduced image is normal, that is, the algorithm a22 has a medium quality as a result of processing. Since the algorithm a23, which is a complex averaging algorithm, is complicated, the job A23 using the algorithm a23 has a long processing time as shown in FIG. 4, for example, the worst execution time is 35 ms. However, since the algorithm a23 performs averaging processing by weighting a large number of pixels, the quality of the reduced image is good, that is, the algorithm a23 has high processing quality.

  In the above description, for example, even if the quality of the result of processing of the job 21 by the algorithm a21 is low, it is a relative quality compared to the job of the algorithm by which other processing time is long, and the absolute quality. The processing of job 21 also has a sufficient result quality for practical use.

  Also, as shown in FIG. 4, the process A1, which is a format conversion process for converting the format of the original image signal, uses a plurality of jobs A11, A12, and A13 that can execute the process A1. . Since the job A11 is a simple algorithm, the processing time is short, for example, the worst execution time is 7 ms. However, since the job A11 is a simple process, the quality of the converted image is poor, that is, the job A11 has a low quality as a result of the process. On the other hand, since the job A12 is a slightly simple algorithm, the processing time is relatively short. For example, the worst execution time is 10 ms. In job A12, the quality of the converted image is normal, that is, the quality of the result of processing is medium. Since job A13 is a complex algorithm, the processing time is long, for example, the worst execution time is 15 ms. However, the quality of the converted image is good for job A13, that is, the quality of the result of processing is high.

  As shown in the job management information processing of FIG. 4, the priority of the process A1 that is the format conversion process is 2, which is set lower than the priority 1 of the process A2 that is the image size conversion process. The priority is input as basic information 107 before the start of processing. Further, the measured actual execution times TA1 and TA2 of the job are recorded by the detector 104 after the processing is executed, as will be described later.

  In the job management information shown in FIG. 4, the algorithm and result quality are added for explanation, and may not be included in the actual job management information.

<Scheduling method>
Next, the scheduling method according to the present embodiment will be described with reference to FIGS. FIG. 6 is a flowchart for explaining the flow of scheduling processing by the scheduler 100 according to the present embodiment, and FIG. 7 is an explanatory diagram for explaining rescheduling A by the scheduler 100.

  Hereinafter, the scheduling method of the present embodiment will be described with reference to the flowchart of FIG.

In the scheduling method of the present embodiment, when the idle time Ti of the CPU 13 occurs, the job having the longest execution time is selected based on the priority and next to the job scheduled to be executed among the plurality of jobs. Here, the case where the idle time occurs means a case where the idle time Ti actually occurs and / or a case where it is expected to occur. Further, even when the idle time Ti actually occurs, the idle time Ti does not necessarily occur in the next periodic processing.

<Step S10>
Prior to the start of processing, basic information 107 and fixed period time of job management information are input to the job management information memory 105 as processing parameters. As shown in FIG. 4, the basic information 107 stored in the job management information memory 105 includes process names of a plurality of processes, priority information of A1 and A2, and job IDs of a plurality of jobs that can be executed by the process A1. A11, A12, and A13 and worst execution time information of each job, and A21, A22, and A23 that are job IDs of a plurality of jobs that can be executed by the process A2 and worst execution time information. At this time, the dynamic information 106 in the job management information memory 105, that is, the actually measured execution time of the job may not be registered, and may include information at the time of the previous process.

<Step S11>
At the start of processing, the selector 103 selects each job of a plurality of processes based on the priority information and worst execution time information of the basic information 107 so that the free time Ti of the CPU 13 is minimized, Register in the queue memory 108. That is, as shown in the schedule A0 of FIG. 8, the selector 103 can execute a plurality of processes A1 and A2 based on the priority so that the CPU idle time Ti is minimized in the reserved time. One job is selected from each of the jobs.

  For example, when the reservation time is 23 ms, the selector 103 first selects the job A21 having the worst execution time of 15 ms from the processing A2 with the priority level 1, and then selects the job A11 having the worst execution time of 7 ms from the processing A1 with the priority level 2. Select. Then, as shown in the schedule A0 in FIG. 7, the free time Ti becomes 1 ms with respect to the reservation time of 23 ms.

  Note that when the measured execution time information of each job is stored in the dynamic information 106 of the job management information memory 105, the selector 103 uses the measured execution time information instead of the worst execution time information. Select each job. That is, the selector 103 selects a job based on the worst execution time information when each job is not executed.

<Step S12>
The CPU 13 executes the process A1 with the job A11 according to the information in the queue memory 108 registered by the scheduler 100.

<Step S13>
When the process A1 is completed, the detector 104 detects tA11 which is the actually measured execution time TA1 of the job A11 of the process A1. Then, the detector 104 records the actual measurement execution time tA11 of the job A11, for example, 4 ms, in the dynamic information 106 of the job management information memory 105.

<Step S14>
Next, the detector 104 estimates the free time Ti of the CPU 13 from the execution time and the reserved time stored in the job management information memory 105 of the job A21 scheduled to be executed.

  For example, as shown in the execution result 1 in FIG. 7, when the actually measured execution time tA11 of the job A11 is 4 ms, the worst execution time is 7 ms, so a free time Ti of 3 ms occurs. For this reason, the detector 104 estimates that a free time Ti of 4 ms or more occurs when combined with the free time Ti1 ms in the schedule A0.

<Step S15>
The scheduler 100 examines whether the estimated free time Ti of the CPU 13 is generated to the extent that the rescheduling A can be performed. That is, the execution time stored in the job management information memory 105 of the job A22 having the longest execution time after the job A21 scheduled to be executed among the plurality of jobs in the process A2 is the worst execution time of 25 ms, and the job A21 Is longer than 10 ms and longer than the estimated free time Ti4 ms.

  For this reason, if job A21 is replaced with job A22, there is a high possibility that job A22 cannot be executed within the remaining reservation time. Therefore, the scheduler 100 determines that the free time Ti is not sufficient (No), and if the process is not completed (No) in Step S19, the CPU 13 executes the unexecuted job A21 again in Step S12.

  The CPU 13 executes the process B according to the queue information after the process A, but will be omitted in the following description.

<Step S12>
The CPU 13 executes the process A2 in the job A21 according to the information in the queue memory 108 indicated in the schedule A1 registered by the scheduler 100.

<Step S13>
When the process A2 is completed, the detector 104 detects tA21 which is the actually measured execution time TA2 of the job A21 of the process A2. Then, the detector 104 records the actual measurement execution time tA21 of the job A11, for example, 9 ms, in the dynamic information 106 of the job management information memory 105.

<Step S14>
Further, as shown in the execution result 2 of FIG. 7, the detector 104 estimates the free time Ti of the CPU 13 from the execution time of the scheduled job A11 stored in the job management information memory 105 and the reserved time. Here, the estimation is to calculate the vacant time Ti from the actually measured execution time and estimate it as the vacant time Ti in the next processing.

<Step S15>
The scheduler 100 examines whether the estimated free time Ti of the CPU 13 is generated to the extent that the rescheduling A can be performed.

  In this case, since there are a plurality of processes to be executed next, that is, the process A1 and the process A2, the process A2 having a higher priority is first examined based on the priority. Here, since the idle time Ti of 10 ms occurs, the scheduler 100 is stored in the job management information memory 105 of the job A22 having the longest execution time after the job A21 scheduled to be executed among the plurality of jobs of the process A2. The execution time is the worst execution time of 25 ms, but is just 10 ms longer than the job A21. For this reason, even if the job A21 is replaced with the job A22, there is a high possibility that the job A22 can be executed within the reserved time. Therefore, the scheduler 100 determines that the free time Ti for performing the rescheduling A is sufficient (Yes), and proceeds to step S16.

<Step S16>
In the rescheduling A, the selector 103 selects a job having a longest execution time next to a job scheduled to be executed among the plurality of jobs based on the processing priority. This is because a job using a complicated algorithm with a long processing time has high result quality. Further, even if the idle time Ti occurs, the idle time Ti is not guaranteed by the worst execution time, so the selector 103 is selected last time so that the processing is not completed within the reserved time. Then, the job having the longest execution time is selected after the job that has been processed within the reserved time.

  In addition, since the job is selected based on the priority of the process, it is possible to prioritize the quality of the process with a high priority. When the scheduler 100 selects a job based on the processing priority, it is preferable to select the job in consideration of the balance of result quality. That is, when the scheduler 100 performs rescheduling A that first selects a job with a long execution time and a higher result quality from jobs with the highest priority processing, in the next rescheduling A, Select the job with the longer execution time and the higher result quality from the jobs with the second highest priority, and execute the job with the third highest priority in the next rescheduling A Select a job with a longer result time and a higher result quality, and after rescheduling A to select a job with a longer execution time and a higher result quality from among the jobs with the lowest priority processes, Rescheduling A is performed to select a job with a longer execution time and a higher result quality from among the jobs with the highest processing. This is because even if the result quality of only a process having a high priority among a plurality of processes is prominently high, the result quality as a whole of the multi-screen processing circuit 11 may not be high.

  Here, the selector 103 selects the job A22 having the longest execution time after the job A21 selected in step S11 among the jobs of the process A2 having the priority level 1, and the resetter 102 stores the queue memory 108. As in schedule A1 shown in FIG.

<Step S12>
The CPU 13 executes the process A1 in the job A11 according to the information in the queue memory 108 indicated in the schedule A1 registered by the scheduler 100.

<Step S13>
When the process A1 is completed, the detector 104 detects tA11 which is the actually measured execution time TA1 of the job A11 of the process A1. Then, the detector 104 records the actual measurement execution time tA11 of the job A11, for example, 4 ms, in the dynamic information 106 of the job management information memory 105 again.

<Step S14>
Further, the detector 104 estimates the idle time Ti of the CPU 13 from the execution time of the scheduled job A22 stored in the job management information memory 105 and the reserved time.

  For example, as shown in the execution result 3 in FIG. 7, when the actually measured execution time tA11 of the job A11 is 4 ms again, the execution stored in the job management information memory 105 of the job A22 of the process A2 scheduled to be executed next. Since the worst execution time is 10 ms longer than the previously executed job 21, all processing is completed in just 23 ms. That is, the idle time Ti does not occur for the reservation time of 23 ms.

<Step S15>
Since the idle time Ti does not occur, the scheduler 100 determines that the idle time Ti is not sufficient (No), and executes the unexecuted job A22 through step S19 and again in step S12.

<Step S12>
The CPU 13 executes the process A2 in the job A22 according to the information in the queue memory 108 indicated in the schedule A1 registered by the scheduler 100.

<Step S13>
When the process A2 is completed, the detector 104 detects tA22 which is the actually measured execution time TA2 of the job A22 of the process A2. The detector 104 records the actual measurement execution time tA22 of the job A11, for example, 15 ms, in the dynamic information 106 of the job management information memory 105.

<Step S14>
Further, the detector 104 estimates the free time Ti of the CPU 13 from the execution time of the scheduled job A11 stored in the job management information memory 105 and the reserved time.

  For example, as shown in the execution result 4 of FIG. 7, when the measured execution time tA22 of the job A22 is 15 ms, the measured execution time tA22 of the already executed job A11 is 4 ms, so the processing is completed in 19 ms. For this reason, a free time Ti of 4 ms occurs with respect to the scheduled time of 23 ms.

<Step S15>
The scheduler 100 examines whether the estimated free time Ti of the CPU 13 is generated so that the rescheduling A can be performed.

  In this case, among the scheduled processes A1 and A2 to be executed next, the process A1 with the priority 1 is already the medium result quality job A22, whereas the process 2 with the priority 2 is low. Since it is the result quality job A21, the scheduler 100 first considers the processing 2 based on the priority in consideration of the balance. Here, since a free time Ti of 10 ms occurs, the execution time stored in the job management information memory 105 of the job A12 having the longest execution time next to the job A11 scheduled to be executed among the plurality of jobs in the process A2 is The worst execution time is 10 ms, but it is only 3 ms longer than the job A21. Therefore, the scheduler 100 determines that there is a high possibility that the job A22 can be executed within the reserved time even if the job A11 is replaced with the job A12. That is, the scheduler 100 determines that a free time Ti for performing rescheduling A occurs (Yes).

<Step S16>
In rescheduling A, the selector 103 selects the job A12 having the longest execution time after the previously selected job A11 in the job of the process A1, and the resetter 102 stores the queue memory 108 in the schedule shown in FIG. In step S12, the CPU 13 executes the process again.

<Step S12>
The CPU 13 executes the process A1 in the job A12 according to the information in the queue memory 108 indicated in the schedule A2 registered by the scheduler 100.

<Step S13>
When the process A1 is completed, the detector 104 detects tA12 that is the actually measured execution time TA1 of the job A12 of the process A1. Then, the detector 104 records the actual measurement execution time tA12 of the job A11, for example, 7 ms, in the dynamic information 106 of the job management information memory 105.

<Step S14>
The detector 104 estimates the free time Ti of the CPU 13 from the execution time of the scheduled job A22 stored in the job management information memory 105 and the reserved time.

  For example, as shown in the execution result 5 of FIG. 7, when the actually measured execution time tA12 of the job is 7 ms, the execution time stored in the job management information memory 105 of the job A22 of the process A2 scheduled to be executed next is actually measured. Since the execution time is 15 ms, all processing is completed in 22 ms. That is, only a free time Ti of 1 ms occurs for a reservation time of 23 ms.

<Step S15>
The scheduler 100 determines that the free time Ti is not sufficient (No), and after step S19, executes the unexecuted job A22 again in step S12. The execution result is shown as execution result 6 in FIG.
Thereafter, the process execution and rescheduling A are repeated in the same manner.

  As described above, the job scheduling method according to the present embodiment is a job scheduling method in which scheduling of a periodic execution method in which a plurality of processes are repeated within a reserved time is performed based on job management information. When the management information includes processing priority information and execution time information of a plurality of jobs that can execute the respective processing, and a CPU idle time Ti occurs, a plurality of jobs are based on the priority. Select the job with the longest execution time next to the job scheduled to be executed. Further, the job management information includes measured execution time information of a plurality of jobs, and a job is selected based on the measured execution time information. For this reason, the job scheduling method of the present embodiment can efficiently use the capacity of the processor even when the time required for processing by the multi-screen processing circuit changes.

<Modification of the first embodiment>
Hereinafter, a job scheduling method according to a modification of the first embodiment will be described. Note that the scheduling method according to the present modification is similar to the scheduling method by the scheduler 100 according to the first embodiment, and thus the description of the same parts is omitted. The scheduler 101 that performs the job scheduling method of the present modification performs job scheduling based on the worst execution time of each job so that the processing is always completed within the reserved time every fixed-cycle processing. Each time one job is executed, the scheduler 101 estimates the vacant time Ti from the measured execution time information of the job and the measured execution time information of the remaining jobs scheduled to be executed, and rescheduling A can be performed. Only when there is free time Ti, rescheduling A is performed. Then, after the execution of all jobs is completed, that is, after one cycle of periodic processing is completed, the queue information is reset again to job scheduling based on the worst execution time.

  Alternatively, the scheduler 101 may perform job scheduling based on the measured execution time information of each job so that a predetermined idle time Tk is generated at each fixed-cycle process. Here, the predetermined idle time Tk is information input before the start of processing, and is a safety coefficient parameter for reliably completing the processing within the reserved time. The predetermined vacant time Tk is determined according to fluctuations in the processing speed, but is preferably 10 to 30% of the reserved time. If it is less than the above range, the processing may not be completed within the reserved time, and if it exceeds the range, the use efficiency of the processor is not good.

  In the job scheduling method of this modification, in addition to the effects of the job scheduling method of the first embodiment, the processing can be completed reliably within the reserved time, and the processing reliability is improved.

<Second Embodiment>
Next, the scheduler 102 according to the second embodiment of the present invention will be described. Since the scheduling method according to the second embodiment is similar to the scheduling method by the scheduler 100 according to the first embodiment, the same components are denoted by the same reference numerals and description thereof is omitted.

  FIG. 8 is a flowchart for explaining the flow of processing of the scheduling method according to the second embodiment. FIG. 9 is an explanatory diagram for explaining rescheduling B by the scheduler 102.

  In the scheduling method according to the present embodiment, scheduling is performed based on the actually measured execution time, not the worst execution time, in order to efficiently use the capacity of the processor. For this reason, the possibility that the process is not completed within the reserved time is not zero. For this reason, in the scheduling method according to the present embodiment, if the process is not completed within the reserved time, rescheduling B is performed to complete the process promptly. Here, the case where the processing is not completed within the reserved time means a case where the processing cannot actually be completed and / or a case where the processing cannot be completed.

Hereinafter, the scheduling method of the present embodiment will be described with reference to the flowchart of FIG.
As shown in FIG. 8, since the scheduling method according to the present embodiment is the same as the scheduling method according to the first embodiment from step S10 to step S15 and step S16, rescheduling after step S17 is performed. B will be described.

  As described above, in the scheduling method of the present embodiment, in order to efficiently use the capacity of the processor, a job whose processing is not completed within the reserved time is registered in the scheduling based on the worst execution time. For this reason, when the process is executed, the process may not be completed within the reserved time. If the process is not completed within the reserved time, the scheduler 102 determines Yes in step S16 and performs rescheduling B in step S17.

<Step S17>
The rescheduling B will be described with reference to FIG. For example, as shown in the schedule B0 in FIG. 9, since the actually measured execution time of the job A13 is 4 ms and the actually measured execution time of the job A23 is 19 ms, it is scheduled that the scheduler 102 completes the processing with the reserved time 23 ms. When the job 13 and job A23 are registered and executed in the queue memory 108, a change occurs in the input image signal and the like, and the actual measurement execution time may increase.

  For example, as shown in the scheduled execution result 7 of FIG. 9, if the measured execution time tA13 of the job A13 increases rapidly from 4 ms to 10 ms, the remaining reservation time is only 13 ms, so a schedule requiring 19 ms, that is, the measured execution time It is expected that the 19 ms job A23 cannot be executed within the reserved time. As described above, when all the processes are not completed within the reserved time based on the actually measured execution time information, the scheduler 102 determines that the processes are not completed, and performs rescheduling B in step S18.

<Step S18>
In rescheduling B, the scheduler 102 selects an unexecuted job based on the priority and the job with the shortest execution time next to the job scheduled to be executed among the plurality of jobs.

  In this case, since the process scheduled to be executed next is only process A2, rescheduling B is performed for process A2. That is, the selector 103 selects the job A22 having the shortest execution time after the job A23 scheduled to be executed, the resetter 102 resets the queue memory 108 to the schedule B1 shown in FIG. 9, and again in step S12, the CPU 13 Performs processing according to the information in the queue memory 108 shown in the schedule B1. Then, the scheduler 102 performs step S13, step S14, and step S15, and again determines whether the process is completed within the reserved time in step S17.

<Step S17>
As shown in the execution result 8 of FIG. 9, when the job A22 is not completed within the reserved time, the scheduler 102 determines that the process is not completed, and performs rescheduling B in step S18.

<Step S18>
In rescheduling B, the scheduler 102 selects an unexecuted job based on the priority and the job with the shortest execution time next to the job scheduled to be executed among the plurality of jobs. In other words, since rescheduling B is scheduling that lowers the quality of results, based on priority means that a job with low result quality is selected from a process with low priority.

  In this case, since there are a plurality of processes to be executed next, that is, the process A1 and the process A2, the rescheduling B is first performed on the process A1 having a low priority based on the priority. That is, the selector 103 selects the job A12 having the shortest execution time after the job A13 to be executed, the resetter 102 resets the queue memory 108 to the schedule B2 shown in FIG. 9, and again in step S12, the CPU 13 Performs processing according to the information in the queue memory 108 shown in the schedule B2. Then, the scheduler 102 performs step S13, step S14, and step S15, and again determines whether the process is completed within the reserved time in step S17.

<Step S17>
As shown in the scheduled execution result 9 in FIG. 9, when the actually measured execution time tA12 of the job A12 is 8 ms and the actually measured execution time tA22 of the job A22 is 15 ms, the scheduler 102 is expected to be able to complete the processing (Yes). Therefore, the scheduler 102 performs step S19, step S13, step S14, and step S15 from step S19 without performing rescheduling B, and confirms whether the process is completed within the reserved time in step S17. Make a decision.

<Step S17>
As shown in the execution result 10 of FIG. 9, when the job A22 is completed within the reserved time, the scheduler 102 determines that the process is complete (Yes) and continues the process.

  As described above, in the scheduling method according to the present embodiment, in addition to the effects of the scheduling method according to the first embodiment, if the process is not completed within the scheduled time, the rescheduling B is performed to quickly It is possible to complete the processing.

  When rescheduling B is performed, it is preferable not to perform rescheduling A until a predetermined period or the number of input image signals changes. This is because even if rescheduling A is performed to improve the quality of results, processing is not completed again at the time of the next frame processing, and rescheduling B is likely to be necessary.

If the multi-screen processing is not completed, the television receiver cannot display the latest image on the monitor, and therefore displays the image of the previous frame. That is, the monitor display is in a so-called freeze state. However, if the freeze state is only for a few frames, the viewer will not notice. For this reason, in the job scheduling method according to the present embodiment, in addition to the job scheduling method according to the first embodiment, an output image with the highest possible quality is output, and the adverse effect due to the incomplete processing remains minimal. Therefore, the reliability of the process is improved. In the above embodiment, an example is shown in which the job scheduling method of the scheduler 100 is used for the two processes of the image size conversion process and the format conversion process of the multi-image processing circuit 11. However, it can also be used for job scheduling of three or more processes. Although three jobs can be executed for each process, the number of jobs may be four or more. Of course, the number of jobs for each process does not have to be the same. Further, when there are three or more processes for job scheduling, if at least two processes have a plurality of jobs that can execute the respective processes, one job can be executed for each of the other processes. It may be only. In addition, the job scheduling method of the scheduler 100 is not limited to job scheduling for the process A that is executed periodically, but can be used for the process B that is executed periodically in combination with the process A.

  Furthermore, in the rescheduling A and rescheduling B, an example in which one processing job is replaced by one rescheduling is shown, but a plurality of processing jobs may be replaced by one rescheduling.

  Further, job scheduling by the scheduler 100 may be used for, for example, a voice processing circuit or a data error correction circuit. Furthermore, the job scheduling by the scheduler 100 is not limited to the data processing of the television receiver 1, and scheduling for performing a plurality of different purposes is performed based on the job management information in a fixed cycle execution method. Any job scheduling method can be used for general purposes, and in any case, the processor capacity can be used efficiently.

  The present invention is not limited to the above-described embodiments, and various changes and modifications can be made without departing from the scope of the present invention.

1 is a configuration diagram showing a configuration of a television receiver 1 having a multi-screen processing circuit according to a first embodiment. FIG. It is the figure which showed the example of the multiscreen of a television receiver. It is a block diagram of the scheduler concerning a 1st embodiment. 3 is a table showing an example of job management information according to the first embodiment. It is explanatory drawing for demonstrating the job from which the several algorithm which can perform image size conversion process differs. It is a flowchart for demonstrating the flow of the scheduling process of 1st Embodiment. It is explanatory drawing for demonstrating rescheduling A by the scheduler concerning 1st Embodiment. It is a flowchart for demonstrating the flow of the scheduling process of 2nd Embodiment. It is explanatory drawing for demonstrating rescheduling B by the scheduler concerning 2nd Embodiment.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 ... Television receiver, 2 ... Parabolic antenna, 4, 5, 6, 7 ... Decoder, 8 ... HDD, 9 ... External terminal, 10 ... Switching circuit, 11 ... Multi-screen processing circuit, 12 ... Monitor, 13 ... Queue Memory, 15 ... Memory, 20 ... Monitor display screen, 100 ... Scheduler, 102 ... Reset device, 103 ... Selector, 105 ... Job management information memory, 106 ... Dynamic information, 107 ... Basic information, 108 ... Queue memory, TA1 , TA2 ... actual measurement execution time, Ti ... vacant time

Claims (5)

On the basis of job management information, the CPU performs a scheduling of a fixed-cycle execution method in which a plurality of processes are executed within a reserved time.
The job management information is
Priority information of each of the plurality of processes;
Execution time information of a plurality of jobs capable of executing each of the plurality of processes,
A job scheduling method comprising: selecting a job having a longer execution time after a job scheduled to be executed among the plurality of jobs when the CPU idle time occurs. The job management information includes measured execution time information of the plurality of jobs;
The job scheduling method according to claim 1, wherein the job is selected based on the measured execution time information.   If all jobs are not completed within the reserved time based on the measured execution time information, the unexecuted job is executed based on the priority and the execution time next to the job scheduled to be executed among the plurality of jobs. The job scheduling method according to claim 1, wherein a short job is selected.   The job scheduling method according to any one of claims 1 to 3, wherein a job is selected in consideration of a balance of quality as a result of the plurality of processes.   The job scheduling method according to any one of claims 1 to 4, wherein the CPU performs scheduling of multi-screen processing of a television receiver.

Images