JPH0233650A - Task scheduling system for multiprocessor system - Google Patents

Abstract

PURPOSE:To improve the performance of a multiprocessor system by measuring the idle time of an unused processor by generating a task schedule via the processor in case the measured idle time exceeds a fixed level. CONSTITUTION:When a task schedule is produced at a time point T1 by a processor 1 of a multiprocessor system, a 1st processor selection means 50 selects a processor and confirms the processors 2-N except the processor 1 to retrieve the unused processors. In case processors 1 and 3 are unused, a processor allocating means 70 updates an idle counter of a task schedule control table 30 and schedules for the processor allocation waiting tasks against the unused processors selected by the means 50. Then the idle counters are set at the initial values to both processors 1 and 3, and the task queues of the table 30 are allocated to the processors 1 and 3.

Description

[Detailed Description of the Invention] [Field of Industrial Application] The present invention relates to a control method for a multiprocessor system.
In particular, it relates to task scheduling methods in multiprocessor systems.

[Conventional technology]

In a multiprocessor system aimed at improving the performance of a computer system, task scheduling is performed to improve system performance by minimizing wasteful idle time of each processor. and,
In this type of task scheduling method, the usage status of each processor is checked at the time of task scheduling, an available processor is selected, and tasks are assigned to that processor.

[Problem to be solved by the invention]

In the conventional task scheduling method mentioned above.

In a processor in which a state requiring a task schedule occurs, the usage status of each processor is checked and a processor is selected so that there is no wasted free time. In this investigation of the usage status of each processor, the status of processors other than the own processor is investigated using methods such as interprocessor communication and shared memory reference that reflects the processor status. Time lag,
Depending on the timing of updates/references to the shared memory, it is not always possible to check the exact state of the processor, and even available processors may not be scheduled. In such a case, there is a drawback that tasks waiting for processor allocation are not scheduled, resulting in a decrease in system performance.

An object of the present invention is to provide a task scheduling method for a multiprocessor system that can eliminate the above drawbacks and improve system performance.

[Means to solve the problem]

According to the invention.

A task scheduling method for a multiprocessor system having a plurality of processors and a main memory that can be commonly referenced by all of the processors.
an idle time measuring means for measuring idle time when the idle time is in an unused state; an area for storing the idle time measured by the idle time measuring means for each processor; and information for managing tasks waiting for processor allocation in the system. a task schedule control table provided on the main memory, including: a first processor selection unit that searches the usage status of each processor and selects an available processor at the time of task scheduling; the task schedule control table; a second processor selection means for selecting an available processor by referring to an area indicating idle time on the table; a processor selected by the first processor selection means and the second processor selection means; and processor allocation means for allocating a task waiting for allocation.

At the time of task scheduling, the 10-processor allocating means selects available resources by the first processor selecting means, allocates a processor to the task waiting for processor allocation, and furthermore, the first processor selecting means Among the available processors that are not selected, the idle time measuring means for measuring unused time and the second processor selecting means select the processor and allocate the processor to the task waiting for processor allocation. A task scheduling method for a multiprocessor system is obtained.

〔Example〕

Next, the present invention will be described in detail with reference to the drawings.

FIG. 1 is an embodiment showing an example of a multiprocessor system to which the task scheduling method of the present invention is applied, and shows a case where there are N processors.

The multiprocessor system of this embodiment includes processors l, 2. . . . N, a main memory section side having a main memory I, a bus 10 connecting them, and an idle counter provided on the main memory 30 and indicating the time when each processor is in an unused state. , a task schedule control table (below) in which tasks waiting for processor allocation are registered.

This is also indicated by 30. ), an idle time measuring means 40 that subtracts the idle counter on the task schedule control table (9) at fixed time intervals when each processor is in an unused state, and a task schedule control table 30.
A first processor selection means selects which processor a task is assigned to based on the usage status of each processor, and a first processor selector selects which processor a task is assigned to based on the idle count on the task schedule control table 30. 2 processor selection means (30 and
Processor allocation means 70 that allocates tasks to processors after setting the idle count on the processor selected by the first gross selection means mark and the second processor selection means (a) to the initial value; Consisting of Said means 40°5
0,60. and 70 are provided in the main storage section.

Figure 2 is a diagram showing the contents of the task schedule control table, and shows all processors 1 on the system.
,2. ..., for N, the idle time measuring means 40
consists of an idle counter that is decremented at fixed time intervals when each processor is in an unused state, and a task queue waiting for processor assignment, in which tasks waiting for processor assignment are registered.

FIG. 3 is a time chart showing the usage status of each processor 1, 2 to N in chronological order.

FIG. 4 is a flowchart of a processing example of the idle time measuring means 40,
5 is a flowchart of an example of processing by the first processor selection means 50, FIG. 6 is a flowchart of an example of processing by the second processor selection means (a), and FIG. 7 is a flowchart of an example of processing by the processor allocation means 70. be.

The operation of this embodiment will be explained below with reference to each figure.

First, the operation of the task schedule will be explained by taking as an example the case where a task schedule occurs in processor 1 at time TI shown in the time chart of FIG. 3, with reference to FIGS. 2, 3, 5, and 7. I will explain while doing so.

When a task schedule occurs in the processor 1 at time TI, the first processor selection means 50 selects a processor. Referring to steps 51 and 52 in FIG. 5, this processor selection is performed based on the state of the own processor 1 and the processor 2 other than the own processor 1.・
. . , N is checked, and unused processors are searched for. At time Tl in FIG. 2, processor 1 and processor 3 are unused. Next, the processor allocation means 70 updates the idle counter of the task schedule control table (9) and causes the first processor selection means 50 to schedule a task waiting for processor allocation to the unused processor selected by the processor selection means 50. This processor allocation is performed in step 7 of FIG.
].

Referring to 72, the idle counts for processors 1 and 3 are set to initial values, and tasks are taken out of the task queue waiting for processor assignment in task schedule control table I and assigned to processors 1 and 3.

Thereafter, when the processor 2 changes from the used state to the unused state after the first processor selection means 50 by the processor 1 is completed, the processor 1 does not know that the processor 2 has become unused, and Unable to schedule tasks.

Even if there is a task waiting for processor allocation, processor 2 remains unused.

Next, the operation after the flosser 2 becomes unused at time T2 in FIG. 3 will be described with reference to FIGS. 2, 3, and 4.

At time T2, the processor 2 is in an unused state, and the idle time measuring means 40 updates the idle counter on the task schedule control table (g). This idle counter update.

Referring to step 41 in FIG. 4, the idle counter of the processor 2 on the task schedule control table I is referenced and subtracted at regular time intervals.

Next, with reference to FIGS. 2, 3, 4, 6, and 7, we will explain the operation in which the idle counter becomes zero in the processor 2 at time T3 in FIG. 3 and a task schedule occurs. explain.

At time T3, when the idle counter for processor 2 becomes zero, the task schedule changes to processor 2.
Occurs in Referring to Figure 4 steps 41.42,
The idle counter for the processor 2 in the task schedule control table 30 becomes zero, and the second processor selection means 60 is activated. Here, task scheduling is performed in the processor 2, and the task is scheduled in the processor 2. Referring to step 61.62 in FIG. 6 and step 71.72 in FIG. Selected for task scheduling. Processor 2 is selected here. Thereafter, the idle counter in the task schedule control table for the processor 2 selected by the second processor selection means mark is set to an initial value, and the task queue waiting for processor allocation is referred to, and the task waiting for processor allocation is transferred to the processor 2. Assign to.

A task waiting for processor allocation is scheduled for the selected processor according to the procedure described above.

〔Effect of the invention〕

According to the present invention as described in detail above.

In a task scheduling method in a multiprocessor system, the idle time of an unused processor is measured, and if the idle state continues for a certain period of time, a task schedule is generated for that processor, thereby eliminating wasted idle time of each processor. This makes it possible to reduce the amount of noise and improve system performance.

(Koki l:ta), 40... Idle time measuring means,
50...first processor selection means, 60...second
processor selection means, 70...processor allocation means.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing an example of a multiprocessor system to which the present invention is applied, FIG. 2 is a diagram showing an example of the contents of a task schedule control table, and FIG. 3 is a block diagram showing an example of a multiprocessor system to which the present invention is applied.
2. . . . A time chart showing the usage status and operation of N in chronological order. Figure 4 shows the idle time measuring means 4.
5 is a flowchart of a processing example of the first processor selection means 50, FIG. 6 is a flowchart of a processing example of the second processor selection means 60, and FIG. 7 is a flowchart of a processing example of the second processor selection means 60. It is a flowchart of a processing example. In Figure 1. I,...,N...Processor, 10...Bus, 20...Main memory, 30...Task schedule control table agent (7783) Patent attorney Noriyasu Ikeda (・
4gu', ・Figure 4 Figure 4 ■■ Figure Figure Time Figure ■10

Claims (1)

[Claims] 1. In a task scheduling method for a multiprocessor system including a plurality of processors and a main memory that can be commonly referenced by all of the processors, each of the processors is in an unused state. an idle time measuring means for measuring the idle time; and the main memory including an area for storing the idle time measured by the idle time measuring means for each processor and information for managing tasks waiting for processor allocation in the system. a task schedule control table provided on the task schedule control table; first processor selection means for searching the usage status of each processor and selecting an available processor at the time of task scheduling; and indicating idle time on the task schedule control table; a second processor selection means for selecting an available processor by referring to an area; means for allocating a task waiting for processor allocation to the processor selected by the first processor selection means and the second processor selection means; ; the processor allocation means selects an available processor by the first processor selection means at the time of task scheduling, allocates the processor to the task waiting for processor allocation, and further selects the processor from the first processor selection means. In the available processors that are not selected by the means, the idle time measuring means for measuring unused time and the second processor selecting means select the processors and allocate the processors to the tasks waiting for processor allocation. A task scheduling method for multiprocessor systems featuring: