JPS63137349A - Buffer control system - Google Patents

Abstract

PURPOSE:To improve the efficiency of a buffer and to reduce overhead accompanied with data transfer by properly selecting an initialized buffer capacity. CONSTITUTION:A master processor 1 is connected to a slave processor 2 through a data transfer line 3 to transfer data from the slave processor 2 to the master processor 1. A buffer control part 4 is formed in the master processor 1 and a buffer control part 5 is formed in the slave processor 2. The buffer control part 4 secures the prescribed capacity of the buffer at the time of starting the operation of the system, informs the secured capacity to the slave processor 2 as the initial buffer capacity and then informs a returned buffer capacity to the slave processor 2 every return of the used buffer. The buffer control part 5 stores the informed initial buffer capacity, subtracts the stored buffer capacity every use of the buffer and adds the returned buffer capacity every reception of the returned information to update the buffer capacity. The slave processor 2 transfers data within the stored buffer capacity based on an instruction from the buffer control part 5.

Description

[Detailed Description of the Invention] [Summary] In a system in which a main processing unit and a slave processing unit are connected by a data transfer path, when data of the slave processing unit is uncertain in terms of occurrence time and amount of data, is transferred to the main processing unit. , the transfer buffer size of the main processing unit is kept to a certain amount, this buffer size is notified to the sub-processing unit in advance, and the sub-processing unit is configured to transfer data within this buffer size, in order to avoid untransferred data due to buffer shortage. If this occurs, the remaining data is transferred after receiving a notification that the used buffer has been returned from the main processing unit, thereby improving buffer efficiency and reducing overhead caused by interrupts caused by data transfer. Dynamic buffer control method.

[Industrial application field]

With the development of data transfer between computers and similar devices,
A system consisting of a main processing unit and a sub-processing unit with a data transfer path between them, where the sub-processing unit processes various monitoring information or input information to generate data and transfers it to the main processing unit. There is a system.

In such systems, there are cases where the data to be transferred from the slave processing device to the main processing device is uncertain both in generation time and data amount.The present invention relates to control of data transfer buffers in such cases. .

[Conventional technology]

Conventionally, there are two main types of control methods for controlling the data transfer buffer between the main processing unit and the slave processing unit when the generation and amount of data from the slave processing unit change irregularly. The data transfer and buffer control system will be explained below using an example of the prior art shown in FIG.

In addition, the following figures 3(a), (b), and 4(a).

In (b), numbers with circles are attached to the frames indicating each operation, but the numbers start from 1 in each of the four figures above, so drawing numbers are only used when it is likely to be confused with other figures. I will explain it by attaching it.

FIG. 4(a) shows a method in which the slave processor instructs the amount of data to be transferred, waits for the main processor to secure a buffer, and then transfers the data. When the slave processor is ready to transfer the data, the slave processor Instructs the main processing unit to transfer data amount and requests buffer reservation.The main processing unit receives this request, secures the buffer, and returns a buffer reservation notification to the slave processing unit.The slave processing unit receives this notification and The data transfer (2) is performed, and after confirming the receipt (2) of the data transfer, the main processing unit returns a data receipt notification to the slave processing unit, completing the data transfer (3).

In this method, the main processing unit secures a buffer each time it is needed, so the buffer efficiency is high, but when the amount of data to be transferred is large, it is necessary to wait until the required amount of buffer is secured. If this occurs frequently, communication between the main and slave processing units will increase, and the overhead (hereinafter simply abbreviated as overhead) associated with the occurrence of interrupts for data transfer of the main processing unit will tend to increase. be.

Figure 4(b) shows a method in which the main processor prepares a buffer in advance and receives data transfer from the slave processor.After securing the buffer, the main processor notifies the slave processor that the data can be transferred. , the slave processing device receives this, confirms that the transfer is possible, and stores it.■
If the data is ready to be transferred, and if the transfer is OK according to the memory mentioned above, then the data transfer will be carried out.When the master device confirms the receipt of the data, the slave processing device will be notified that the data transfer is complete, and at the same time the master The processing device starts securing a buffer for the next transfer (same operation as ■).

In this method, the main processing unit does not know the amount of data to be transferred, generally requires a large buffer at one time, and there are periods when it is not used, resulting in low buffer efficiency and the need for frequent transfers of small amounts of data. For example, there is little inheritance between the main and slave processing units, and there is little overhead.

[Problem/point that the invention attempts to solve]

As mentioned above, the two typical conventional methods are characterized in that the higher the buffer efficiency, the higher the overhead, and the lower the buffer efficiency, the lower the overhead. There are both advantages and disadvantages. The purpose of the present invention is to solve this problem and provide a buffer control method that has good buffer efficiency and low overhead for the main processing unit.

[Means for solving problems]

The purpose is to provide a system in which a main processor 1 and a slave processor 2 are connected by a data transfer path 3, and data is transferred from the slave processor 2 to the main processor 1, as shown in FIG. This is achieved by providing a buffer management section 4 in the main processing device 1 and a buffer management section 5 in the slave processing device 2 as a control method for the data transfer buffer, and controlling the buffers as follows.

The buffer management unit 4 of the main processing device 1 secures a predetermined amount of buffer at the start of system operation, and notifies the slave processing device 2 as the initial buffer amount. Notify the amount of buffer to be returned.

On the other hand, the buffer management unit 5 of the slave processing device 2 stores the initial buffer amount that has been notified, and thereafter subtracts the stored buffer amount each time it is used and adds it each time it receives the return notification. shall be updated.

When data transfer occurs, the slave processing device 2 is controlled to transfer data within the storage buffer amount according to instructions from the buffer management unit 5, and to transfer the remaining data after waiting for the storage buffer amount ifl to recover. do.

[Effect]

After the initial setting, as long as there is a buffer amount remaining in the buffer storage of the buffer management section of the slave processing unit, the slave processing unit will only be able to communicate with the main processing unit regarding the presence or absence of the buffer. In addition, even if the storage buffer size is insufficient compared to the amount of data to be transferred, the data can be divided and transferred only on the slave processor side, and the remaining data is transferred to the main processor. Upon receiving notification of the returned amount of used buffers on the processing device side, it is possible to perform control such that the data is transferred when the storage buffer amount of the buffer management unit of the slave processing device is updated and increased.

Therefore, if you choose the initial buffer size appropriately,
Without setting aside an unreasonably large buffer,
In addition, since there is little communication left behind between the main and slave processing units, the overhead is small, and divisional transfer can be performed automatically even when it is necessary to transfer not only a relatively small amount of data but also a large amount of data. This makes it possible to both improve buffer efficiency and reduce overhead associated with data transfer.

〔Example〕

An example will be explained using figures.

FIG. 2 is a block diagram of an embodiment of the present invention, and FIG. 3 (
3(a) is an explanatory diagram of the operation of the embodiment of the present invention (in case initial settings and batch transfer is possible), and FIG. ) is shown.

Figure 2 shows the main parts related to data transfer.
, Components that are the same as those in the principle diagram of FIG. 1 are given the same numbers. The configuration will be explained with reference to the figure.

10.1) is a transfer control section of the main processing unit 1 and the slave processing unit 2, respectively, and is a section that controls data transfer.

In the main processing device 1, 4 is a buffer management section according to the present invention, 6 is a processing section that processes the received data from the slave processing device 2, and 8 is a transfer buffer that is secured and managed by the buffer management section 4. For the reception operation, the transfer control unit 10 performs
The data is delivered to the processing section 6 by receiving the data in the transfer buffer and transferring control of this buffer to the processing section 6.

In the slave processing device 2, reference numeral 5 denotes a buffer management section according to the present invention, which is provided with a buffer amount storage section 12 and is configured to store and update the transfer buffer amount of the main processing device 1. 7 is a processing unit that prepares the transfer data, and 9 is a processing buffer that is secured within the slave processing device 2 when the processing unit 7 prepares the transfer data, and the control of this buffer is given to the transfer control unit 1) during transfer. It is now possible to move and transfer data. .

Next, the operation of this embodiment will be explained with reference to FIGS. 3(a) and 3(b). Both figures show the main processing unit 1. Slave processing device 2
The columns are drawn facing each other, and both devices are divided into columns for the buffer management sections 4 and 5 of the present invention and processing sections representing other processing, so that the operation of each section changes downward over time. It is in the form of an action flow diagram in which each action is enclosed in a frame and the flow of the action is connected by lines with arrows.

The processing units that represent the other processes mentioned above are the processing unit 6 and the transfer control unit 10 in the main processing device, and the processing unit 7 in the slave processing device 2.
and transfer control unit 1), and since the operations of processing unit 6.7 and transfer control unit 10.1) themselves are not a feature of the present invention, they will be simplified and written in the respective processing unit columns. It is.

First, the initial setting operation will be explained with reference to FIG. 3(a).

At the start of system operation, the main processing unit 1 secures a predetermined buffer by its buffer management unit 4, and notifies the buffer management unit of the slave processing unit 2 of the buffer amount.
The buffer management section 5 of the slave processing unit 2 stores the buffer amount in the buffer amount storage section 12, thereby completing the initial setting.

Continuing, the case where batch transfer is possible will be explained with reference to FIG. 3(a).

When the processing section of the slave processing device 2 completes preparation of the data to be transferred (2), it notifies the buffer management section 5 of the slave processing device 2 of the amount of data to be transferred and requests transfer.

The buffer management section 5 of the slave processing device 2 compares this transfer amount with the buffer amount storage contents, and if the buffer amount is sufficient, it determines that the transfer is OK and instructs the processing section of the slave processing device 2 to transfer, and also removes the storage buffer. Perform quantity subtraction■. The processing unit that receives the transfer instruction performs data transfer (2).

1 buffer management unit 4 of the main processing unit that received the data transfer ■
After confirming the completion of the transfer, the main processing unit 1 instructs the processing unit 1 to receive and process each buffer. Once the processing by the main processing unit 1 is completed, the buffer is transferred to the buffer management unit 4 of the main processing unit 1. , the buffer management unit 4 checks the returned amount, this returned amount is notified to the buffer management unit of the slave processing device 2, and the buffer management unit 5 of the slave processing device 2 adds the storage buffer amount [phase]. to prepare for the next transfer request. This completes the batch data transfer. Note that the buffer amount subtraction ■ and the buffer amount addition @ in the main processing unit 1 column are not essential to the present invention, but are for confirmation within the main processing unit buffer management section 4.

When the buffer management unit 5 of the slave processing device 2 compares this transfer amount with the contents of the buffer amount storage (■), if the buffer amount is insufficient, it determines that the transfer is NO and transfers the contents of the buffer amount storage to the slave. An instruction is given to the processing section of the processing device 2. If the buffer amount is O at this time, the processing section performs an operation similar to the case of split transfer described later, that is, by having the buffer management section memorize the transfer request amount, the main processing unit 1 uses the buffer. The buffer is returned, and upon receiving the notification, there is a waiting period of 0 until the storage buffer amount of the buffer management unit 5 of the slave processing device 2 is added.

On the other hand, if the buffer amount is not 0 (if the buffer amount is measured by determining a unit, if the buffer amount is one unit or more), division transfer 0 due to buffer shortage, which will be explained next, will be performed.

The case of division transfer ◎ in Fig. 3(a) is shown in Fig. 3(b).
This will be explained in detail.

The judgment of ■ in Figure 3 (a) [also described in Figure 3 (b)] is N.
At O, the processing unit of the slave processing device 2, which received the buffer amount instruction, divides the data to be transferred into the data amount within the specified buffer amount, and divides the data into the data amount of the beginning part. Instructs the remaining capacity to the buffer management section and transfers the first portion of data. The buffer management unit 5 of the slave processing device 2 subtracts the storage buffer amount and stores the remaining amount of data division.

The operation after transferring the first data is to perform the operations from ① to [phase] in FIG. When the storage buffer amount of 5 is added and updated, the contents of the divided remaining amount storage (3) are compared with the contents of the buffer amount storage updated by the buffer amount addition (2) to determine whether the transfer is OK.

In this judgment, if the buffer capacity is sufficient and the transfer is OK, the subsequent operation proceeds in the same way as the transfer OK case in Figure 3 (a) comparison (■) in the case of batch transfer, and the remaining divided data is transferred. ends.

On the other hand, if the buffer size is small and you want to transfer it in parts,
Comparison of FIG. 3(b) in the case of divisional transfer The divisional transfer is continued by repeating the following steps in the same manner as in and after (2) in the case of the transfer NO determination in (2). By repeating this process, if the buffer becomes sufficient for the amount of divided remaining data, the operation proceeds in the same way as in the case of transfer OK in comparison ① in Figure 3 (a) in the case of bulk transfer, and the data transfer of the remaining divided data is completed. It ends.

〔Effect of the invention〕

According to the present invention, by appropriately selecting the initial buffer amount, an unreasonably large buffer is not secured unnecessarily, and communication between the main and slave processing devices is maintained. Since the amount of data is small, overhead can be reduced, and buffer efficiency can be improved because the buffer can be managed to automatically perform divided transfer not only when a relatively small amount of data is transferred, but also when a large amount of data needs to be transferred. This also makes it possible to reduce the overhead associated with data transfer.

[Brief explanation of the drawing]

Fig. 1 is a principle diagram of the present invention, Fig. 2 is a block configuration diagram of an embodiment of the present invention, and Fig. 3 (a) is an explanatory diagram of the operation of the embodiment of the present invention (in case initial settings and batch transfer are possible). , Figure 3 (b
) is an explanatory diagram of the operation of the embodiment of the present invention (in the case of split transfer due to buffer shortage), and FIG. FIG. 4(b) shows an example of the prior art: a method in which the main processing unit prepares a buffer in advance and receives data transfer from the slave processing unit. In the figure, 1 is a main processing unit, 2 is a slave processing unit, 3 is a data transfer path, 4.5 is a buffer management unit, 6.7 is a processing unit, 8 is a transfer buffer, 9 is a processing buffer, 10.1) 12 is the transfer control unit, and 12 is the buffer amount storage unit '14i diagram.

Claims (1)

[Claims] The main processing unit (1) and the slave processing unit (2) are connected to a data transfer path (
3) in a system in which data is transferred from a slave processor (2) to a main processor (1), the main processor (1) has a buffer A buffer management unit (5) is provided in the sub-processing unit (2), and the buffer management unit (4) of the main processing unit (1) secures a predetermined amount of buffer at the start of system operation. Then, the initial buffer amount is notified to the slave processing device (2), and thereafter, each time a used buffer is returned, the returned buffer amount is notified to the slave processing device (2), and the buffer of the slave processing device (2) is The management unit (5) stores the notified initial buffer amount, and thereafter updates the stored buffer amount by subtracting it each time it is used and adding it each time it receives the return notification. When data transfer occurs, the processing device (2) transfers data within the storage buffer amount according to instructions from the buffer management unit (5), and transfers the remaining data after waiting for the storage buffer amount to recover. A buffer control method featuring: