JPH083794B2 - Production system using multi-processor - Google Patents

Description

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a production system, and more particularly to a production system using a multiprocessor unit.

(Prior Art) In recent years, it has been proposed to perform inference processing using a production system when performing knowledge information processing using a computer, for example, failure diagnosis of various devices.

As shown in Fig. 2 (a), the production system uses a working memory (Wo) that stores information to be inferred.
rking Memory (abbreviated as WM below) 1 and production memory that stores a sequence of operating conditions for inference (Production Memory R below)
2) and a processor unit (hereinafter, abbreviated as PU) 3 that connects both to execute inference, but each element is 1 in this way.
In the structure corresponding to the one-to-one correspondence, the processing becomes sequential, so that the speeding up of the inference processing cannot be expected.

Therefore, as shown in FIG. 2 (b), PM5-1, which has a plurality of PUs 4-1, 4-2, ... 4-n and stores operating condition sequences divided into each PU, 5-2, ... 5-n is added, and only one WM6 is configured to be shared via the path 7, and the inference processing is shared by multiple PUs to perform processing in parallel to speed up processing. I was planning.

However, in the above-mentioned conventional production system, since each PU 4-1, 4-2, ... 4-n shares the only WM6, while one PU accesses WM6, the other PUs have to wait until the end. Therefore, there is a drawback that the speedup is naturally limited.

(Object of the Invention) The present invention has been made to eliminate the drawbacks in the conventional production system as described above, and an object of the present invention is to provide a production system in which the processing time is remarkably shortened.

(Outline of the Invention) The present invention has the following configuration in order to achieve this object.

That is, each of the plurality of processor units is provided with a working memory and a production memory, the same content data is stored in each of the working memories, and only the data required by each processor unit belonging to each of the production memories is distributed. A communication channel for storing and connecting each of the plurality of processor units in a loop is provided, and each processor unit stores a data rewriting content or data necessary for inference processing via the communication channel to a working memory belonging to another processor unit. Alternatively, it is configured to transfer to the production memory and execute the inference processing.

(Examples of the Invention) The present invention will be described in detail based on the illustrated examples.

FIG. 1 is a block diagram showing the principle of a production system using a multiprocessor according to the present invention. In the figure, 8-1, 8-2, ... 8-n are PUs, and the communication of each PU connects the ports in a loop to form one communication channel 9 and Have the same content in PU
WM10, 10, ... PM11 that stores the operating condition sequence divided into 10-
1, 11-2, ... 11-n are connected.

With such a configuration, since all WMs belonging to each PU store the same content data, each PU can execute inference processing independently of each other and at the same time, so that the desired processing can be performed at extremely high speed. It's easy to see what you can do.

FIG. 3 is a block diagram showing an embodiment of the present invention, and a case where a fault diagnosis of a television is performed using this device will be described. In the figure, 12-1, 12-2, 12-3, and 12-4 are four microprocessor units (Micro Processor Units).
(Hereinafter abbreviated as MPU), each communication port is connected to each other to form one closed loop, and the MPU has memories 13-1, 13-2, 13-3 and 13-4. To connect to each other to form a production system. Further, one of the MPUs, such as the MPU 12-1, is provided with an input keyboard via the host computer 14 that controls communication between the production system and the operator.
15 and the display device 16 are connected.

The memories 13-1 to 13-4 have a storage area corresponding to the above-mentioned PM and a storage area corresponding to the WM.
As the information to be stored in the PM, for example, information as shown in Table 1 in which various failure states and causes thereof are associated with each other is divided and stored.

When a failure diagnosis of a television is performed using the device configured as described above, the operator checks the failure situation of the target television and inputs some or all of them into the host computer 14 using the keyboard 15. The host computer writes the input failure status into the storage area WM of the memory 13-1 connected to the MPU 12-1 via the MPU 12-1, and at the same time the MPU 12-1 transmits another information via the closed loop communication channel. Write to the storage area WM of each memory attached to the MPU.

When an inference start command is issued from the host computer 14 in this state, each MPU simultaneously starts collation of the contents of PM _{1 to} PM ₄ and WM stored in 13-1 to 13-4, and causes the failure cause of each. Candidates are selected and the result is MPU12-1
It is sent to the host computer 14 via the display device 16 and displayed by the display device 16.

It can be easily understood that high-speed inference can be performed because each MPU can function in parallel by independently providing WM and PM for each of the multiple MPUs.

The above example is shown to explain the operation and effect of the present invention shown in FIG. 1, that is, to speed up the inference processing when WM and PM are added to each of a plurality of processor units. The reasoning process in the fault diagnosis device is complicated and therefore the algorithm is also complicated, and each WM needs to be rewritten for each reasoning. In the present invention, this WM rewriting and PM All necessary data can be supplied by a communication channel connected in a loop.

(Effects of the Invention) Since the present invention is configured and functions as described above, it is extremely effective in constructing a production system that executes inference at high speed.

[Brief description of drawings]

FIG. 1 is a block diagram for explaining the principle of the production system of the present invention, FIGS. 2 (a) and 2 (b) are configuration diagrams of a conventional production system, and FIG. 3 is a block diagram showing an embodiment of the present invention. Is. 8-1 to 8-n ... PU (processor unit), 9 ...
Communication channels, 10, 10, ...... 10 WM (working memory), 11-1 to 11-n ...... PM (production memory)

Claims (1)

[Claims] 1. A plurality of processor units each include a working memory and a production memory, each working memory stores data of the same content, and only data required by a processor unit to which each production memory belongs is stored. A communication channel that stores the data in a distributed manner and connects each of the plurality of processor units in a loop is provided. A production system using a multiprocessor, wherein the inference processing by the plurality of processor units is simultaneously performed by transferring to a working memory or a production memory to which the processor memory belongs.