JP3438056B2 - Shared memory management method - Google Patents

Description

Description: BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a multi-task system using a pipeline system in which jobs are sequentially processed by a plurality of tasks. The present invention relates to a shared memory management method used by the user. 2. Description of the Related Art In a conventional memory management system, when a memory area release request is issued because a job is not aware of a job, the memory area is immediately released, so that it is necessary to monitor a subsequent task at a task level. There is. FIG. 3 shows a block diagram of a first example of the conventional shared memory management method. It is composed of a plurality of tasks 1, tasks 2, a job queue 3, a job management unit 4, a memory management unit 5, and a memory area 6. The state of the system shown in FIG. 3 is that the job 1 is being processed by the task 1, the job 1 is being processed by the task 2, and the jobs 2 and 3 perform the processing of the task 2 in the job queue 3. Waiting for you. The memory block A of the memory area 6 is a shared memory secured at the stage of the job 1 in the task 1, and the job 1, the job 2, and the job 3 in the task 2.
It is required when processing. [0004] In the system of Example 1 of the conventional shared memory management system shown in FIG. 3, a task 1 issues a memory area release request by monitoring the job end status of a subsequent task 2 through a job management unit 4. In this case, before issuing the memory area release request to the memory management unit 5, the task 1 inquires the job management unit 4 how far the processing in the task 2 has been completed. In the case of the state shown in FIG. 3, since it is known that the task 2 is currently processing the job 1, the task 2 has to cancel the memory area release. Such monitoring is performed until the job 3 ends, and a release request for the memory block A is issued to the memory management unit 5 only after confirming that the job 3 has ended. Upon receiving this, the memory management unit 5 releases the memory block A. FIG. 4 is a block diagram showing a second example of the conventional shared memory management system. In this method, the memory area shared for the subsequent task 2 is copied, and the copied area is released at the end of the job by the subsequent task 2, and the memory block A is released by the task 1 when it becomes unnecessary. Task 1
Jobs 1, job 2, copy of memory block A for each job during processing of job 3 - to the memory block A _1, the memory block A _2, providing a memory block A _3. Therefore, the task 1 may release the memory block A immediately at the time of the job 4. Memory block A ₁ ,
The memory block A ₂ and the memory block A ₃ are released when each job is completed by the task 2. [0006] However, when the system of the first example of the conventional management system shown in FIG.
Since there is a request for securing and releasing the memory area 6 in the processing units A, B, and C, it is necessary to perform a job monitoring process via the job management unit 4 for each process.
Since there are a plurality of places for issuing a memory area release request in the task 1, it is necessary to put such monitoring processing in a plurality of places in the task 1. For this reason, the system configuration becomes complicated, and the independence of the task 1 is lost because the execution state of the job of the subsequent task 2 is monitored. In addition, there is a problem in that the processing performance of the system is degraded because it is necessary to wait until the end of the job of the subsequent task. In the case of the second example of the conventional management method shown in FIG. 4, the problem as in the above method does not occur, but necessary memory blocks are copied, so that the use efficiency of the memory is reduced. In addition, since each task issues a memory area release request, the processing efficiency of the entire system is reduced. There is a problem in that the use efficiency of the memory is reduced and time is wasted for copying. SUMMARY OF THE INVENTION The present invention has been made in view of the above problems, and has as its object to provide a shared memory management method which eliminates the above problems and efficiently manages a memory area shared by a plurality of tasks. . According to the present invention, when a certain task 1 is executed as shown in FIG. 1, the present invention includes a post-task 2 which is executed later from the task 1. When a request to release the shared memory area is issued, the memory management unit 5 accepts the address (or identifier) of the memory block A and the job ID for identifying the final job using the memory block A. Means for recording and receiving data on the cable 51; According to the present invention, by providing the job ID, the job management unit 4 monitors jobs processed by each task, and transfers the job ID to the memory management unit 5 at the end of each job.
The memory management unit 5 compares the job ID with the job ID recorded in the reception table 51, and can release the memory block A secured up to now by confirming that they match. An embodiment of the present invention will be described below in detail with reference to the drawings. FIG. 1 shows a block diagram of the shared memory management system of the present invention. A plurality of tasks 1, a task 2, a job queue 3, a job management unit 4, a memory management unit 5, a reception table
And a memory area 6. A memory management system for managing limited memory resources and making adjustments among a plurality of processing systems. The memory management unit 5 has a function of managing a memory release request from each task by a job ID. Is a memory management mechanism that detects the end of the last job and releases the secured memory area. When the processing unit A of the task 1 stores the processing result in the memory area and the processing unit B of the other task 2 needs it, the task 1 executes the job processing at the stage when the processing unit A becomes unnecessary. A memory area release request is issued to the memory management unit 5 with the ID added. When the job management unit 4 confirms that the job having the job ID has been completed based on the job completion notification from the job management unit 4, the memory management unit 5 releases the actual memory area. From the viewpoint of the processing unit A, there is no problem if the processing is continued with the memory area 6 being released when the memory area 6 is released with the ID. That is, task 1 can be considered as a process similar to the single task. In addition, since the special job ID is immediately released, the memory area 6 used locally in the processing system can be released. As shown in FIG. 1, there are a task 1 having processing units A, B, and C, and a task 2 for performing processing using data processed by the task 1. An outline of the process will be described using an example in which the queue 3 exists. The state of the system in FIG. 1 is as follows: job 1 is processing job 4; task 2 is processing job 1;
Reference numeral 3 denotes a standby state for performing the processing of the task 2 in the job queue. Memory block A is task 1 and job 1
And is required when processing job 1, job 2, and job 3 in task 2. The task 1 may issue a request to release the memory block A with the job ID = 3 when the memory block A becomes unnecessary, and may proceed to the next processing. Upon receiving the memory area release request with ID, the memory management unit 5 receives the job ID = 3 and the address (or identifier) of the memory block A, which are the release request, in the reception table 51 and completes the task 1 normally. Back. FIG. 2 shows an example of a job ID and a reception table 51 of a memory block to be released. Thereafter, the memory management unit 5 confirms that the job 3 has been completed by the task 2 based on the job completion notification from the job management unit 4, and releases the memory block A for the first time. By adding the job ID in this way, it is not necessary for each processing unit in the task 1 to be aware of the processing state of the task 2, and there is no need to copy the memory block A. It is possible to perform management. As described above, according to the present invention, the following excellent effects can be expected . (1) when a memory area release, maintain independence between tasks because there is no need to be aware of the need or subsequent task status to monitor the job end of the subsequent task in each processing unit of the preceding stage task this
And the system construction is simplified. (2) Since each processing unit can shift to the next processing without waiting for the end of the processing of the subsequent task, system efficiency is improved. (3) Since there is no need to copy data to the subsequent task, the memory use efficiency is improved.

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a block diagram showing a configuration of a shared memory management system according to the present invention. FIG. 2 is a diagram showing an example of a reception table of a job ID and a memory block to be released. FIG. 3 is a block diagram showing a configuration example 1 of a conventional shared memory management system. FIG. 4 is a block diagram showing a configuration example 2 of a conventional shared memory management system. [Description of Signs] 1 Task 2 Task 3 Job queue 4 Job management unit 5 Memory management unit 51 Reception table 6 Memory area

Claims (1)

(57) [Claim 1] Shared memory management used in a shared manner between a plurality of tasks in a multitask system using a pipeline system in which jobs are sequentially processed by a plurality of tasks In the method, when a task is executed in the multi-task system, when a request for securing a shared memory area to be used in common with a subsequent task is issued from the task, the requested memory area is secured, when the reception means for Installing receive <br/> a job ID for identifying the last job an address or identifier of the reserved memory area to be used a memory area is provided, the corresponding task is performed, and the job ID accepting an address or identifier of the memory area to the receiving unit, to release the <br/> accepted only memory domain detects that the job with the job ID has been completed Means for managing a common memory.