JPH04320537A - Synchronous processing method for parallel processing system - Google Patents

Abstract

PURPOSE:To provide a synchronous processing method for a parallel processing system capable of executing parallel processing while guaranteeing the execution order of plural procedures. CONSTITUTION:After executing a certain procedure, a 1st process 101 adds information necessary and sufficient for executing the succeeding procedure tightly relating to the procedure concerned to data 104 to be processed and inserts the data 104 into a queue 103. A 2nd process 102 extracts the data 104 from the queue 103 and indirectly the succeeding procedure tightly relating to the procedure makes access to concerned based upon the information added to the data 104 to execute the succeeding procedure.

Description

[Detailed description of the invention]

0001

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a synchronous processing method for a parallel processing system that sequentially executes a plurality of procedures in which processing contents for data to be processed are described in accordance with a preset execution order.

0002

2. Description of the Related Art Conventionally, a desired process can generally be achieved by executing a plurality of procedures.

[0003] Here, if it is necessary to process procedure A, procedure B, and procedure C in order for certain data d,
Moreover, the content of procedure C is closely related to procedure A, and procedure B cannot know about its content (that is, it is directly related to procedure C).
).

[0004] When executing each of the above procedures based on these conditions, a single process (sequential execution type)
There is execution by multiple processes (completion waiting type) and execution by multiple processes (completion waiting type).

FIG. 7 shows a diagram for explaining the execution processing of each of the above procedures by a single process. In this figure, process 1 executes various procedures on data d. In this case, if each procedure is in the same context, process 1 executes procedure A, procedure B, and procedure C on data d in this order, or when calling procedure B from procedure A, process 1 uses Pass the entry address etc. of procedure C and call back procedure C within procedure B (indirect procedure call)
By doing so, you can achieve your purpose.

In any of the above cases, as shown in FIG. 8, procedure A is executed on data d (step 11), and then procedure B is executed on data d (step 11). By executing step 12) and procedure C on data d (step 13), the procedures can be executed in the order of procedure A, procedure B, and procedure C.

[0007] Next, a diagram for explaining the execution processing of each of the above procedures by a plurality of processes is shown in FIG. 9, and a flowchart showing the execution processing operation is shown in FIG.

In FIG. 9, a synchronization mechanism such as a queue 4 is provided between a context (or process) 2 and a context (or process) 3. Here, it is assumed that procedure A belongs to process 2 and procedure B belongs to process 3.

Now, as shown in FIG. 10(a), in process 2, procedure A is executed on data d (step 21), and this data d is inserted into queue 4 (step 22). After that, it waits until a notification from process 3 that execution of procedure B has finished (step 2
3) When the execution of procedure B by process 3 is completed, procedure C is executed (step 24).

On the other hand, in process 3, FIG.
), take out data d from queue 4 and write (
Step 31), and procedure B is executed on the extracted data d (step 32). Then, when the execution of procedure B is finished, process 2 is notified of this fact (step 33).

[0011] In the above-described processing using multiple processes, in order to guarantee the order of execution among the multiple processes, process 2 needs to wait for process 3 to finish executing procedure B. In other words, step 2 above
Step 3 must be performed. This means that process 2 must wait until procedure B completes execution, and cannot perform processing in parallel with process 3 during this time.

0012

As described above, in the above-mentioned conventional processing by multiple processes, after process 2 finishes executing procedure A, it waits for the completion of execution of procedure B within the context of procedure A. Further, after process 3 finishes executing procedure B, procedure C must be executed, and while process 3 is executing procedure B, process 2 is in a standby state and cannot perform any processing. Therefore, there is a problem in that parallel processing, which is the original purpose of dividing contexts, is not substantially performed, and no improvement in processing efficiency can be expected.

The present invention has been made in view of the above circumstances, and an object thereof is to provide a synchronization processing method for a parallel processing system that can perform parallel processing while guaranteeing the execution order of a plurality of procedures. .

[0014]

[Means for Solving the Problems] In order to achieve the above object, the present invention provides a parallel processing system that sequentially executes a plurality of procedures in which processing contents for data to be processed are described in accordance with a preset execution order. In the synchronous processing method, a first process that executes a certain procedure;
and a second process that executes a procedure subsequent to the certain procedure, and after executing the certain procedure, the first process executes a subsequent procedure that is closely related to the certain procedure in the data. The second process indirectly calls and executes a subsequent procedure that is closely related to the certain procedure based on the information added to the data. It is characterized by

[0015]

[Operation] According to the synchronous processing method for a parallel processing system according to the present invention, a certain procedure is executed by a first process, and a subsequent procedure closely related to the certain procedure is executed by a second process. Ru. Therefore, after executing a certain procedure, the first process does not need to be involved in the execution of subsequent procedures that are closely related to the certain procedure, so it can perform other processes.

[0016]

DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of a synchronous processing method for a parallel processing system according to the present invention will be described below with reference to FIGS. 1 to 6.

FIG. 1 shows a schematic configuration of an embodiment of a parallel processing system to which the synchronous processing method according to the present invention is applied, and the system of the embodiment shown in this figure is applied to a general-purpose computer system. be able to.

In the figure, the parallel processing system includes a first
process 101 and second process 102, and a queue (hereinafter referred to as "Queue") provided between each process.
(referred to as a queue) 103.

In this embodiment, as explained in the above-mentioned conventional technology, there is a case where it is necessary to process procedure A, procedure B, and procedure C in the order of procedure A, procedure B, and procedure C for certain data.
The content of is closely related to procedure A, and procedure B cannot know about the content (that is, cannot directly call procedure C).

Now, the first process 101 has a function of setting necessary and sufficient information to execute procedure C on the data 104 to be processed, while the second
The process 102 refers to the information for procedure C set in the data 104, and executes procedure C based on the reference result.
It has a function to call indirectly.

Note that the first process 101 and the second process 102 are executed by a control means such as a processor or a central processing unit (not shown). Further, each of the above-mentioned procedures and data 104 are allocated to a storage means (not shown).

Note that in this embodiment, procedure A belongs to the first process 101, and procedure B belongs to the second process 101.
2, and Procedure A, Procedure B, and Procedure C.
It is assumed that both exist within the same context.

Next, the basic parallel processing operation of the parallel processing system will be explained with reference to FIGS. 2 and 3.

Note that FIG. 2 is a flowchart showing the processing operation of the first process 101, and FIG. 3 is a flowchart showing the processing operation of the second process 102.

In the first process 101, as shown in FIG. 2, procedure A is executed (step 201), and then,
information necessary and sufficient to execute procedure C on the data 104;
In this embodiment, for example, the address value of the area where procedure C is stored is stored (step 202), and the data 104 is inserted into the queue 103 (step 2).
03). After step 203, other processing can be executed (step 204). Note that in the data 104 shown in FIG. 1, the shaded portion is
This indicates that sufficient information necessary to execute procedure C is stored.

On the other hand, in the second process 102, as shown in FIG.
(Step 302). And that procedure B
After the execution of the procedure C set in the data 104
The result of this reference (
Call back procedure C (indirect procedure call) based on the address value of the area where procedure C is stored,
Procedure C is executed (step 303).

In this way, when procedure C is being executed by the second process 102, the first process 10
1 need not be involved in the execution of procedure C in any way. Further, since the processing of procedure C itself is based on the method defined by the first process 101, the second process 102 does not need to know what is being executed. In other words, procedure C
There is no need to know what the process is.

[0028] In step 202, the information necessary and sufficient to execute procedure C is set in the data 104 after the execution of procedure A is completed, but this is set within procedure A. (In other words, procedure A
The setting process for the data 104 may be described in the data 104.

Furthermore, in this embodiment, the address value of the area where procedure C is stored is used as information necessary and sufficient to execute procedure C, but this is because procedure A, procedure B, and procedure C are the same. Since it exists within the context of , it is possible to use it as an address value. Therefore, if each procedure exists in a different context, the information necessary and sufficient to execute procedure C can be an offset value or index value to the area where procedure C is stored. .

Next, an embodiment of the synchronization processing method according to the present invention will be described using a communication protocol processing system as an example.

FIG. 4 shows a schematic configuration of an embodiment of a parallel processing system that implements a communication protocol processing system.
A first process 401 and a second process 402 corresponding to the first process 101 and second process 102, respectively, and a queue 40 corresponding to the queue 103
3.

[0032] Here, in the communication protocol processing system, procedure A is transferred to upper layers (layers 3 to 7 of the OSI reference model).
The procedure B is the physical interface driver (i.e., the process of sending packets to the line), the procedure C is the post-processing after the transmission of the packet buffer is completed, and the data 104 is the packet buffer 404. . It is also assumed that procedure A belongs to the first process 401 and procedure B belongs to the second process 402. Furthermore, it is assumed here that each procedure exists within the same context. Note that for the queue 403, other general inter-process communication methods may be used.

Next, the parallel processing operation of the parallel processing system that implements the communication protocol processing system will be explained with reference to FIGS. 5 and 6.

Note that FIG. 5 is a flowchart showing the processing operation of the first process 401, and FIG. 6 is a flowchart showing the processing operation of the second process 402.

Now, as shown in FIG. 5, the first process 401 performs packet transmission preparation processing such as a data set and a destination address set (step 501).
Thereafter, after storing the entry address of the buffer release routine in a predetermined area of the packet buffer 404 (step 502), the packet buffer 404
is inserted into the queue 403 (step 503). After completing step 503, other processing is executed (step 504), and then it is determined whether all processing has been completed (step 505). Here, if all the processing has not been completed, the process returns to step 501 and the steps after this step are executed, whereas if all the processing has been completed, the processing is ended. Note that in the packet buffer 404 shown in FIG. 4, the diagonally shaded portion indicates that the entry address of the buffer release routine is stored.

On the other hand, the second process 402 retrieves the packet buffer 404 from the queue 403, as shown in FIG.
(step 601) and performs packet transmission processing (step 602). As a result, the packet will be sent out onto a line (not shown). Step 60
2, the entry address set in the packet buffer 404 is called (CALL) and a buffer release routine is executed (step 603). As a result, packet buffer 404 will be freed. After completing step 603, it is determined whether all processing has been completed (step 604). Here, if all the processing has not been completed, the process returns to step 601 and the steps from this step onwards are executed, whereas if all the processing has been completed, the processing is ended.

As mentioned above, the second process 402 includes:
After completing the packet transmission process, the buffer release process (routine) will be continued. Note that the buffer release process itself is based on the method defined by the first process 401, so there is no need for the second process 402 to know what it is executing.

Therefore, since the processing by the first process 401 and the processing by the second process 402 can be performed in parallel, the first process 401 can perform the following processing regardless of the processing by the second process 402. Processing can be performed for the transmission destination, and therefore, the same data can be quickly transmitted to a plurality of different transmission destinations.

In addition, in a processing system that implements a virtual circuit (virtual line), procedure C in the processing system of the above-mentioned communication protocol is used as a queue for retransmitting the packet buffer (hereinafter referred to as a retransmission queue). By inserting the packet into the queue (referred to as a queue), the packets sent by the second process 402 are inserted into the queue (referred to as a queue), without changing the flows shown in FIGS. It becomes possible to hold the data in a retransmission queue.

As explained above, according to this embodiment, even if an asynchronous element such as a queue is inserted between procedure A and procedure B, procedure C is always executed after procedure B is completed. This allows parallel processing while guaranteeing the execution order of multiple procedures.

[0041] Also, in the past, procedure C was
However, according to the present invention, procedure C is executed within the context of procedure B, so that the process that was executing procedure A is actually executed in the same context as procedure C. There is no need for any involvement in this process.

[0042]

Effects of the Invention As explained above, according to the present invention, when executing a plurality of procedures including a certain procedure and a subsequent procedure that is closely related to the certain procedure, Adding information necessary and sufficient to execute a subsequent procedure to the data to be processed, and another process different from the process executing the procedure refers to the information added to the data, Based on the reference result, a subsequent procedure that is closely related to the certain procedure is indirectly called and executed, so that a certain procedure and the subsequent procedure that is closely related to the certain procedure are It is possible to perform parallel processing while guaranteeing the execution order of multiple procedures including

[Brief explanation of the drawing]

FIG. 1 is a schematic configuration diagram showing an embodiment of a parallel processing system to which a synchronous processing method according to the present invention is applied.

FIG. 2 is a flowchart showing processing operations of a first process in this embodiment.

FIG. 3 is a flowchart showing the processing operation of a second process in this embodiment.

FIG. 4 is a schematic configuration diagram showing an embodiment of a parallel processing system that implements a communication protocol processing system.

FIG. 5 is a flowchart showing processing operations of a first process in the embodiment shown in FIG. 4;

FIG. 6 is a flowchart showing processing operations of a second process in the embodiment shown in FIG. 4;

FIG. 7 is a diagram for explaining conventional processing by a single process.

FIG. 8 is a flowchart showing the processing operation of a single process.

FIG. 9 is a diagram for explaining conventional processing using multiple processes.

FIG. 10 is a flowchart showing processing operations by multiple processes.

[Explanation of symbols]

101, 401...first process, 102, 402...second process, 103, 403...queue.

Claims (1)

[Claims] 1. A synchronous processing method for a parallel processing system in which a plurality of procedures in which processing contents for data to be processed are described are sequentially executed according to a preset execution order, wherein a first process that executes a certain procedure; and a second process that executes a subsequent procedure of the certain procedure, and after executing the certain procedure, the first process includes a subsequent process that is closely related to the certain procedure in the data. Add sufficient information necessary to carry out the procedure,
The second process indirectly calls and executes a subsequent procedure that is closely related to the certain procedure based on the information added to the data. Method.