JPH04288660A - Parallel processing system for plural processors - Google Patents

Abstract

PURPOSE:To use a single program to execute the processing while operating plural processors in parallel in relation to one another. CONSTITUTION:A single program where the name (discrimination information) of each processor is added to the head of each instruction word is inputted to a program input part 4 and is stored in a program storage part 5. A system control part 7 sets operation start and end information of each processor to a timing register 3 in accordance with the program taken out from the part 5 by start indicating information, and the register 3 indicates the start to each processor to start a corresponding processor 1 or 2 in accordance with operation start and end position information and indicates the end to terminate the processing of the corresponding processor. During this processing, each processor successively takes out instructions corresponding to operation start and end position information from the storage part 5 to execute the processing of instructions.

Description

[Detailed description of the invention]

0001

[Industrial Field of Application] The present invention relates to a parallel processing method for multiple processing units in an electronic computer system, and in particular, a system that enables parallel processing operations for multiple processing units (central processing units) using a single program. Concerning parallel processing methods.

0002

[Prior Art] In the conventional technology, when the operation of a processing device is described in a compiler class language, for example, FORTR
One program unit, such as AN or C language, usually has the main purpose of describing the operation of one processing device, and one program unit usually corresponds to one processing device. It was hot. Regarding this technique, see, for example, pages 10-11 and pages 23-30 of "Introduction to Computing Systems (Iwanami Course Software Science 1), published by Iwanami Shoten, October 6, 1988."

[0003]Also, a distributed processing system is conventionally known in which a plurality of processing devices distributed in various locations operate while being associated with each other to execute a series of processing in parallel. Requires a program for each processing device to operate each processing device and a program for controlling the processing devices in association with each other,
A single program does not operate multiple processing devices in parallel.

0004

[Problems to be Solved by the Invention] In each of the above-mentioned conventional techniques, one program is operated in correspondence with one processing device, and the parallel operation of a plurality of processing devices in a system consisting of a plurality of processing devices is There was a problem in that it was not possible to write a compiler class as one program unit and cause multiple processing units in a system consisting of multiple processing units to operate in parallel.

[0005] Accordingly, an object of the present invention is to solve the problems of the prior art described above and to make it possible to operate a plurality of processing devices in parallel while correlating them with each other using a single program (user program). The object of the present invention is to provide a parallel processing method for multiple processing devices.

[0006]

[Means for Solving the Problems] In order to achieve the above object, the parallel processing system of the present invention includes a plurality of processing devices and identification information for identifying each of the plurality of processing devices at the beginning of each instruction word. a program input section for inputting a single assigned program; and a program input unit that selects and starts each of the plurality of processing devices at a predetermined timing while associating them with each other by the single program, and issues instructions to each processing device. The present invention is characterized by comprising a selection control means for a processing device that executes the processing in parallel.

[0007]

[Operation] The operation based on the above configuration will be explained.

According to the present invention, identification information for identifying each of the plurality of processing devices is attached to the beginning of each instruction word of a single program, and the single program allows the processing of the plurality of processing devices. When each of the processing devices is automatically selectively activated in relation to each other, each processing device executes the processing of the instructions assigned to it in parallel.

Means for selecting and activating each processing device according to the identification information attached to the program and distributing instructions to these processing devices for execution (processing device selection control means) is a system control unit. It includes a timing register. When the system control unit obtains an instruction related to the start and end of operation of the processing unit from a single program, if this instruction is an instruction to set the operation start location or operation end location of the processing unit, the system control unit stores these location information in the timing register. Set to . Further, if this instruction is a processing device startup instruction, a start instruction is set in the timing register, and if this instruction is a processing device termination instruction, an end instruction is set in the timing register. Furthermore, when the timing register cannot be set, these settings are performed after waiting for the timing register to become settable.

[0010] The timing register stores location information, start instruction information, and end information of the operation start location and operation end location from the system control unit and the processing device for each processing device.
For each processing device, if a start instruction is given, the corresponding processing device is selected and activated, and if a termination instruction is given, the corresponding processing device is terminated.

In this way, each instruction (instructions for various processing devices are mixed) constituting a single program is selected by the selection control means (system control unit and timing register) of the processing device. The information is then distributed to the processing device addressed to the addressee indicated by the identification information, and processed in parallel at a predetermined timing relationship.

0012

[Embodiments] Hereinafter, embodiments of the present invention will be explained with reference to the drawings.

FIG. 1 shows the configuration of a system and data flow related to the present invention. As an example, a system including two processing devices will be explained. The processing device 1 and the processing device 2 in FIG. 1 are examples thereof.

<Principle of the present invention> First, the principle of the present invention will be explained. Conventionally, the expression of computer operation was
A program is generally said to be something that is expressed according to certain predetermined rules (usually called a computer language). In addition, when a computer language (for example, FORTRAN, C, COBOL, etc.) is used within one program unit, one processing device (
It is also generally known that the term "processing device" is used hereinafter to refer to what is generally referred to as a central processing unit of a computer. Further, the instructions of the processing device are for executing the above-mentioned computer language, and are necessarily 1
Within a program unit, there is only one processing unit (rather than one that targets one processing unit), so this type of instruction is generally a set of instructions that are not aware of the processing unit. It is believed that.

In contrast to the conventional concept described above, the present invention newly introduces the following concept into the system in order to be aware of a plurality of processing devices. Basic concept: All "commands" that operate "processing devices and means that function equivalently" have a name that identifies the "processing device or means that functions equivalently" on which the "commands" operate. shall be. This name is hereinafter referred to as "execution means/identification name" (for example, "utilization/processing device number"). An example of the structure of the "instruction" is shown in FIG. When this basic concept is introduced, first, it becomes possible to express the operations of a plurality of processing devices within one program unit.

Next, the range of motion expression is determined by the functional elements that make up the system. FIG. 1 shows an example of the configuration of functional elements of the system used in the present invention. As shown in FIG. 1, in order to realize the above basic concept, at least a processing device 1, a processing device 2, a timing register 3, a program input section 4, a program storage section 5, and a startup instruction input section 6 are required. and a system control section 7 are required. When there are three or more processing devices, only the processing device increases as a component, and the other components remain the same. The meaning and movement of each will be explained using the movement of the entire system below.

<Operation of the entire system> (1) Program configuration First, an example of a program input by the user is shown in FIG. It is assumed that the "instructions" constituting the program each have the "execution means/identification name" described above. For example, if the processing device 1 has “1
FIG. 3 shows an example in which a number "2" is given to the processing device 2, and a number "0" is given to the system control unit 7.

[0018] The "instructions" constituting the "program" include:
Prepare the following two types. - "Control Instruction" - "General Instruction" The former "control instruction" is an "instruction" for operating multiple processing devices in parallel. Parallel operation is used to mean that multiple processing units within one system operate independently of each other.

The latter ``general instruction'' is the conventionally used ``instruction'' with an ``execution means/identification name'' added. The conventionally used "instruction" is, for example, in FORTRAN, the "instruction" that makes it possible to execute A1=B1+C1 A2=B2+3.0 GO TO 1030 .

The "execution means/identification name" is used to distinguish each of the processing devices (for example, processing device 1 and processing device 2) and system control unit 7 that exist in one system. Therefore, for example, "0" is set to the system control unit 7, "1" is set to the processing device 1, and "2" is set to the processing device 2, which will be described later.
The "execution means/identification name" is assigned in the following manner.

FIG. 3 shows an example of a "program" configured under the above conditions.

(2) Input and storage of "program" First,
The system inputs the "program" created by the user, adds a name to the input "program", and stores it.
The added name is hereinafter referred to as the "program name." The "program" has "information identifying the executing means or processing device" (hereinafter referred to as "executing means/identification name") for each command. The "program" is input through the program input section 4, and is stored in the program storage section 5.

(3) Starting the "program" Next, the system calls the "program" specified by the user,
Execute. This is "startup instruction information" (including "program name") input by the user via the startup instruction input section 6.
When the system control unit 7 obtains the above, the system control unit 7 calls the above-mentioned stored “program” corresponding to the “program name” of the “startup instruction information”, and uses the “program” to
The system starts working.

(4) Operation of the system by the called "program" First, the system control unit 7 starts the "program" (hereinafter referred to as "designated program") corresponding to the "program name" of the "startup instruction information". From there, instructions for the system control unit 7 (instructions with usage/processing device number=0) are extracted as control instructions and executed in order. In the example of FIG. 3, 〇1, 〇2, 〇6, 〇7 (■ is expressed as 〇1. The same applies hereinafter) are applicable commands. For example, when the system control unit 7 obtains 〇1 and 〇2, the system control unit 7 receives “operation start location information” (for example, start address) of the processing device 1.
and "operation end location information" (for example, end address) are set in the timing register 3, and then an instruction to start the operation of the processing device 1 is set in the timing register 3.

When an instruction to start an operation is set, the timing register 3 starts up the instructed processing device, in this example, starts up the processing device 1.

When the processing device 1 is started, it receives “operation start location information” (for example, a start address) and “operation end location information” (for example, an end address) from the area for the processing device 1 in the timing register 3. , and the corresponding portions of the "program" in the program storage section 5 (in the example, instructions 03 to 05) are read out and executed in order.

Further, as soon as the system control unit 7 executes 01 and 02 above, that is, sets an instruction to start the operation of the processing device 1 in the timing register 3, it immediately reads out the next control command, Execute. In this example, 〇6 in Figure 3
Since 〇7 is the command for the next system control unit 7,
Read and execute 〇6 and 〇7. When the system control unit 7 obtains 06 and 07, the system control unit 7
"Operation start location information" (for example, start address) and "operation end location information" (for example, end address) are set in the timing register 3, and then an instruction to start the operation of the processing device 2 is set in the timing register 3. Set to .

When the timing register 3 is set with an instruction to start an operation, it starts the instructed processing device, in this example, the processing device 2. Since the processing device 1 is operating at this point, the processing device 1 and the processing device 2 perform parallel operations (processing in substantially the same time period).

When the processing device 2 is started, it receives “operation start location information” (for example, a start address) and “operation end location information” (for example, an end address) from the area for the processing device 2 in the timing register 3. , and the corresponding portions of the "program" in the program storage unit 5 (in the example, instructions from 08 to 010) are read out and executed in order.

The above is an explanation of how a plurality of processing devices start and perform parallel operations. However, in FIG. 3, processing devices 1 and 2 are not necessarily synchronized and operate independently. Therefore, next, the problem of timing alignment (synchronization) between each processing device when performing this parallel operation will be explained.

(5) Operation for synchronizing a plurality of processing devices Next, a description will be given of an operation for synchronizing a plurality of processing devices. As an example, a case will be described in which a new operation of the processing device 1 is performed after the operations of both the processing device 1 and the processing device 2 are completed. In this case, configuring the program will result in
For example, the program has the form shown in FIG. The program in FIG. 7 first includes a group of instructions (1) for the system control unit 7 that starts the processing device 1, and secondly, instructions for the processing device 1 that are executed by starting the group of instructions (1). There is group (2),
Third, there is an instruction group (3) for the system control unit 7 that starts the processing device 2, and fourth, there is an instruction group (4) for the processing device 2 that is executed by starting the instruction group (3). Fifth, there is an instruction group (5) for the system control unit 7 for starting the processing device 1, and finally, there is an instruction group (5) for the processing device 1 that is executed by starting the instruction group (5). 6). This example is the same as instructions 01 to 010 in FIG.
This is the same as explained in . In other words, the instruction group (1) in FIG. 7 is the instructions 01 and 02 in FIG.
) are 03 to 05, instruction group (3) is 06 and 07, instruction group (
4) is equivalent to 〇8 to 〇10. Also, instruction group (5)
correspond to 〇11 and 〇12, and instruction group (6) corresponds to 〇13 to 〇15. Here, the fifth and last (sixth) instruction group (5) will be explained.

To match the timing, use the fact that the movement of the timing register 3, for example, the timing register 3 can be set from the system control unit 7 only when the operation of the corresponding processing device is completed. . For example, as shown in FIG. 8 as a modification of instructions (5) and (6) in FIG. 3 or FIG.
First, a setting command for "action end location" and a setting command for "action start location" and "action end location" for the processing device 2 are arranged, followed by a command to start the processing device 1. In this way, the processing device 1 will be activated when both the processing device 1 and the processing device 2 are finished, and synchronization can be performed.

The above is an explanation of the operation for synchronizing a plurality of processing devices.

<Description of System Control Unit 7> The operation of the system control unit 7 will be explained below using the flowchart of FIG. 4.

[0035] First, the "startup instruction information" input by the user
(including the "program name") is obtained via the startup instruction input section 6 (step 21).

Next, the specified "program" is called from the "startup instruction information", and it is first checked whether there is an instruction for the system control section 7 (step 22). If there is a command, the process goes to step 23; if not, the process goes to step 31 (skipping other than control commands).

Next, in step 23, instructions for the system control section 7 are read out from the program, which are instructions (a type of control instruction) regarding the start and end of operation of the processing device, and the process proceeds to the next step 24. go.

In step 24, the timing register 3
The "setability information" for the corresponding processing device is checked, and if the corresponding processing device is in the settable state, the process goes to step 25, and if the timing register is in the settable state, it waits until it becomes settable.

Next, from step 25, the following operations are performed.

- First, the control command determines the "operation start location" (in the above-mentioned storage "program") regarding each processing device.
If it is a command to set the “operation end location” (step 2
5) Set the information of "operation start location" and "operation end location" in the timing register described below (step 26),
Go to step 22, - If the instruction is a “startup” instruction for the processing unit (step 2
7) Set the "start instruction" in the timing register (step 28), and go to step 22. - If the instruction is the "end" instruction of the processing device (step 2
9) Set the "end instruction" in the timing register (step 30) and go to step 22. Also, in step 31, check whether all processing devices have finished operating, and if all processing devices have finished operating, proceed to step 32, and if the operation is not completed, wait until it is completed.

Finally, in step 32, the entire system is terminated.

<Description of timing register 3> Below,
The operation of the timing register 3 will be explained using the flowchart of FIG. First, specify the first processing unit (
Step 41). Next, the designated processing device issues a "start instruction"
If it is a "start instruction", it starts the corresponding processing device (step 43), displays that the processing device is in operation (step 44), and proceeds to step 47. If it is not a "start instruction", go to step 45.

In step 45, it is checked whether the designated processing device has a "termination instruction" or not. If it is a "termination instruction", the designated processing device is terminated (step 46), and if it is not a "termination instruction", Go to the next step 47 without doing anything. Next, in step 47, it is checked whether the processing device in question is displayed as being in operation, and if it is not displayed in operation, the process goes to step 51, and if it is displayed in operation, the “completion report” of the processing device in question is checked. Then, check whether the processing device has finished (step 48). If the processing device has not finished, proceed to step 5.
1, and if the processing device is finished, it displays that the processing device can be set (step 49), displays that the processing device is stopped (step 50), and goes to the next step 51. .

Next, in step 51, it is determined whether all the processing devices have been seen. If not, the process goes to step 52, and if all the processing devices have been seen, the process returns to the beginning of step 41. In step 52, the next processing device is specified, and then the process goes to step 42. This is the operation of the timing register 3.

<Description of One Processing Apparatus> The operation of one processing apparatus will be explained below using the flowchart of FIG. 6.

First, when the processing device is started by the “start instruction” in the timing register 3, the information indicating the “operation start location” in the area for the processing device in the timing register 3 is read out. The information thus obtained is set (moved) to the "program start location" of the processing device (step 61).

Next, the information indicating the "operation end location" in the area for the processing device in question in the timing register 3 is read out, and the read information is set in the "program end location" of the processing device in question ( (Step 6)
2).

Next, an instruction is retrieved from the "program start location" of the stored "program" and brought to the processing device (step 63).

Next, it is checked whether the "execution means/identification name" of the retrieved instruction indicates the processing device in question (step 64). If the "execution means/identification name" indicates the processing device in question, it is determined that the The instruction is transferred to the processing device and executed (step 65), and the "execution means/identification name" of the retrieved instruction is
If does not indicate the processing device in question, the command is ignored and the process goes to the next step 66.

Next, in step 66, the contents of the "program start location" are advanced to the location of the next instruction, and the process proceeds to step 67.

Next, in step 67, it is checked whether the "program start location" matches the "program end location". If they match, go to step 69; if not, go to step 68.

In step 68, the timing register 3
Check whether there is an "end instruction" in the area for the processing device. If there is an "end instruction", the process goes to step 69; if not, the process goes to step 63 to process the next command.

Next, in step 69, a "completion of operation report" is sent to the timing register 3, and finally, the processing device is terminated (step 70).

The above is the operation of one processing device.

Finally, the configuration of the above embodiment will be summarized. The data processing system according to this embodiment consists of the following six components.

(1) Program input section: The program input here is created by the user, and each instruction has "information identifying the means of execution or processing device" (execution means/identification name). shall be taken as a thing.

(2) A program storage unit that stores the input program.

(3) A start-up instruction input section for inputting "start-up instruction information" for starting the above-mentioned stored "program".

(4) System control section: This system control section obtains the above-mentioned "start-up instruction information", calls the above-mentioned stored "program", and issues instructions regarding the start and end of operation of the processing device from the program. In addition, if the timing register is in a settable state by obtaining setting enable/disable information from the timing register (described later), then: If the command sets the location and end location of the operation, set the information of the "start location" and "end location" in the timing register described below, and
If the instruction is a "start" instruction for the processing device, set the "start instruction" in the timing register; - If the instruction is the "end" instruction for the processing device, set the "end instruction" in the timing register. , If the timing register is in the unsettable state, it waits until it becomes settable, and when it becomes settable, performs the setting operation to the timing register described above.

(5) Timing register: This timing register is shared by a plurality of processing devices, and contains information on the “operation start location” and “operation end location” obtained from either the system control unit or the processing device. , "start instruction" and "end instruction" are stored for each corresponding processing device, and for each processing device, if it is a “start instruction”, the corresponding processing device is started,
If it is a "termination instruction", the corresponding processing device is terminated.

(6) Multiple processing devices: When one or more of these processing devices is activated by the “start instruction” of the timing register, Instructions are sequentially fetched from the "operation start point" of the "program", and if the "execution means/identification name" of the fetched instructions indicates the processing device, the instructions are executed, and the "execution means/identification name" of the fetched instructions is executed. If the "name" does not indicate the processing device, the command is ignored (not executed, skipped), and the execution location (
Repeat the operation until the executed instruction part) matches the "operation end location" in the timing register, and when the repetition is finished, "report the end of the operation" to the timing register, and if there is a "termination instruction" from the timing register. , terminates the processing device and sends a "report of completion of operation" to the timing register.

[0062]

As described above in detail, according to the present invention, identification information for identifying each of a plurality of processing devices is attached to each instruction of a single program, and each instruction is Since the system distributes and executes programs to each processing device according to the identification information, it is no longer necessary to run multiple programs to operate multiple processing devices, as was the case in the past. This has the effect that a plurality of processing devices can be interconnected and execute processing in parallel.

Therefore, the present invention is effective in (1) performing parallel operations while associating a plurality of processing procedures within one program unit, such as in technical calculations. Also,(
2) The program used in the present invention is the same as the conventional program if the control command and identification information (utilization/processing device number) are ignored, so it is effective even when connected to a single processing device. .

[Brief explanation of the drawing]

FIG. 1 is a configuration diagram showing the configuration of a processing system to which an embodiment of the present invention is applied, together with the flow of data.

FIG. 2 is a configuration diagram of an example of an instruction that takes into account multiple processing devices.

FIG. 3 is a configuration diagram showing an example of a program.

FIG. 4 is a flowchart showing the operation of the system control unit.

FIG. 5 is a flowchart showing the operation of a timing register.

FIG. 6 is a flowchart showing the operation of one processing device.

FIG. 7 is a configuration diagram showing an example of a single program.

FIG. 8 is a configuration diagram showing an example of a control command group.

[Explanation of symbols]

1, 2 Processing equipment 3 Timing register 4 Program input section 5 Program storage section 6　Start-up instruction input section 7 System control section

Claims (3)

[Claims] 1. A program input unit that inputs a plurality of processing devices and a single program in which identification information for identifying each of the plurality of processing devices is added to the beginning of each instruction word;
Processing device selection control means that selects and activates each of the plurality of processing devices at a predetermined timing while correlating them with each other by the single program, and causes each processing device to execute processing of each instruction in parallel. A parallel processing method for multiple processing devices, characterized by comprising: 2. The selection control means for the processing device includes a system control section and a timing register, and the system control section obtains instructions regarding the start and end of operation of the processing device from the single program. , if the instruction is an instruction to set the operation start location or operation end location for each processing device, information on these locations is set in the timing register, and if the instruction is a startup instruction for the processing device, setting a start instruction in the timing register;
If the instruction is a termination instruction of the processing device, the termination instruction is set in the timing register, and if the timing register is in a setting disabled state, the setting operation to the timing register is performed after waiting until setting becomes possible. 2. The parallel processing method for multiple processing devices according to claim 1, wherein the parallel processing method is configured to perform the following steps. 3. The timing register stores information on an operation start location and an operation end location, start instruction information, and end instruction information obtained from either the system control unit or the processing device for each corresponding processing device. Claim 2 characterized in that the system is configured such that for each processing device, if a start instruction is given, the corresponding processing device is selected and activated, and if a termination instruction is given, the corresponding processing device is terminated. Parallel processing method of multiple processing devices described.