JPS62175869A - Processing system for deciding module arrangement - Google Patents

Abstract

PURPOSE:To shorten the processing time with a titled system by performing the calculation of the shift destination and the decision of shift in parallel with each other for all circuit modules in case the arrangement of those modules is decided automatically by a computer. CONSTITUTION:In a module arrangement processing mode the prescribed positions of arrangement are divided into blocks and a block is allocated to each processor 21 in charge. Then the shift destination block is decided by a prescribed logic for each module in charge after the initial arrangement information, etc. are received. Here the name of the module whose shift destination belongs to another block is transmitted to the processor 21 which takes charge of the shift destination block. Thus the processor 21 is replaced to the position of the module in the corresponding block according to the received information. Otherwise a module that can be accepted at an idle position is decided and this decision is informed to the processor 21 at the remote side as necessary.

Description

[Detailed Description of the Invention] [Summary] A processing method that speeds up the determination of module placement movement in order to reduce the wiring length between circuit modules.

A processing device that divides module placement into multiple blocks and processes each block in parallel processes the process of finding movement destination candidates, notifies each other of the results, and moves modules between different blocks based on the notification. decide. With this method, the number of processing steps is reduced in inverse proportion to the square of the number of blocks.

[Industrial application field]

The present invention relates to a processing method for speeding up the determination of changes in module placement for reducing the wiring length between circuit modules in automatic design using a computer system.

In the layout design of large-scale integrated circuits (LSI), etc., there is a problem of optimizing the arrangement of modules so as to shorten the wiring length between circuit modules.

Since the number of modules contributes to the number of steps required to solve this problem in the order of its square, as the degree of integration of LSI increases,
The processing effort increases rapidly, and it is especially necessary to speed up the processing of placement problems and suppress the increase in development time.

[Prior Art and Problems to be Solved by the Invention] FIG. 3 is a diagram illustrating a process for optimizing module arrangement using the center of gravity relaxation method. In the figure, the number 1 indicates an empty position (referred to as a slot) where a module can be placed.

Figure 3 (a) shows an example of the process of determining the center of gravity vector. For example, the center of gravity vector of module A, which requires wiring with modules B and C, is vector 3 pointing from module A to B, and vector 3 pointing from module A to C. Vector 6 obtained by dividing the sum vector 5 with vector 4 by the number of connection destinations 2 is set as the center of gravity vector of module A. If module A is moved to the destination pointed to by this center of gravity vector, the wiring to the module will be indicates the shortest length.

In the module placement process, first, the initial placement of modules is created appropriately based on the functional relationship of each module, and the destination of the desired module from this initial placement is determined, for example, by the center of gravity relaxation method as described above. If the movement is possible considering various conditions, the movement is determined.

In this process, the centroid vector of a certain module is determined as described above, and the centroid vectors of the modules to which the centroid vector points are sequentially determined. For example, the centroid vector of modules D and H in FIG. vector 7
．． As shown in 8, when two modules are in a relationship in which they mutually move to the other's position, the movement of these two modules can be immediately determined.

In general, the centroid vectors 9, l of modules F, G, H
When the gravity center vectors that are sequentially weighted form a loop, as in the case illustrated in 0111, the related modules can be sequentially moved. Further, even when a gravity center vector pointing to an empty position is obtained, as shown in the gravity center vector 12 of module I, it is possible to determine the movement of the module related to the gravity center vector connected thereto.

According to this method of sequentially determining the center of gravity vector for each module and determining the movement destination, the processing time increases rapidly as the number of modules increases.
Problems arise in the development of LSIs that include a huge number of modules, resulting in increased design time and costs.

[Means for solving problems]

FIG. 1 is a block diagram showing the configuration of the present invention and a flowchart of processing.

FIG. 1(a) shows the configuration of a computer system, where 2o is a host computer and 21 is a processing device.

FIG. 1 (bl shows the main processing flow in each processing device 21, and 30 to 35 are processing steps.

[For production]

For example, as shown in the figure, the plurality of processing devices 21 have a communication bus between them and adjacent processing devices 21, and can communicate with any two processing devices via those communication paths or by relaying with the processing device 21. The configuration is such that information can be transferred between 21.

Further, each processing device 21 is connected to the host computer 20 so that information can be transferred thereto.

When processing the module arrangement, a predetermined arrangement position is divided into blocks, and each two processing apparatuses are assigned one block to be handled.

The host computer 20 creates an initial layout and transfers the initial layout information and module connection information to all the processing devices 21 .

Each processing device 21 uses the information to determine the modules arranged in the block it is in charge of, as shown in FIG.
Execute the process shown in b).

That is, upon receiving the initial arrangement information etc. (processing step 30
), in processing step 31, a destination block is determined for each responsible module using a predetermined logic.

In processing step 32, the module name of the module whose destination is another block is transmitted to the processing device 21 in charge of the destination block. Also, in the same way, it receives notifications addressed to itself sent from other processing devices 21.

In processing step 33, based on the received information, a module that can be replaced in the position of the module in the block in charge or that can be received in a vacant position is determined, and the partner processing device 21 is notified if necessary.

In processing step 34, it is determined whether the processing should be terminated based on predetermined conditions, and if the processing termination conditions are satisfied, the processing is terminated.

If the process should be continued, in processing step 35, if there is a module whose move destination is the own block, and if there is a module that has not been decided to be moved, the name of that module is changed to the module name that should be moved from the own block but has not been decided to be moved. In combination with the module name, the notification is sent to the processing device 21 of the latter destination block, and a similar notification addressed to itself is received.

In processing step 33, based on this received information, if there is a module that can be accepted, it is determined to move it in the same manner as described above.

According to the above configuration, the calculation of the movement destination and the movement determination based on the result are processed in parallel across all modules, so that the processing time can be significantly shortened depending on the number of processing devices 21 that perform parallel processing. can.

〔Example〕

FIG. 2 is a diagram illustrating the arrangement of modules such as LSI. In the figure, 40 indicates a slot where a circuit module can be placed, and 41 indicates an example of a module that is placed.

In the module arrangement process of the present invention, all the slots 40 are divided equally into a plurality of blocks 42. In the illustrated example, a block consisting of 12 slots is divided into 9 blocks.
The processing of each block is shown in Figure 1 (a).
) to each processing device 21.

The host computer 20 creates an initial arrangement and transfers initial arrangement information and module connection information (hereinafter collectively referred to as module connection information) to all processing devices 21 .

The module connection information is, for example, a list consisting of a module name, a placement slot number determined in the initial placement of the module, and one or more connection destination module names.

Each processing device 21 uses the information to determine the modules arranged in the block to which it belongs.
The process shown in FIG. 1(b) is executed.

That is, when the module connection information is received (processing step 30), in processing step 31, for example, the above-mentioned gravity center vector is calculated for each module in charge, and the block pointed to by the gravity center vector is set as the movement destination.

In processing step 32, a notification of the module name of the module whose movement destination is another block (hereinafter referred to as movement module notification) is sent to the processing device 21 in charge of the movement destination block. In addition, in the same manner, the mobile module notification sent from another processing device 21 and addressed to itself is received.

In processing step 33, a module that can be moved into the assigned block is determined based on the information of the received movement module notification.

In this case, if there is a pair of movement destinations that mutually specify the other block as the movement destination, both modules can exchange slots, so movement is determined.

Furthermore, if there is an empty slot, it is possible to immediately decide to move.

However, in general, there may be multiple moving module notifications received by the processing device 21 of one block that meet the above conditions and compete with each other, so appropriate priority logic is provided to select and move. In this case, the other party's processing device 21 is notified.

As a material for this priority selection process, for example, the amount of wiring length reduction in the case of movement is evaluated using an appropriate method, added to the movement module notification and transmitted, and priority is given to the one with the largest reduction in wiring length. You may also do so.

As a result of the above-described processing, generally there remain movement module notifications for which movement has not been determined. In the processing step 34, for example, it is determined that there are no undetermined mobile module notifications for at least one of the outgoing and received mobile module notifications, or that a predetermined number of repetitions has been reached, and the processing is terminated, and in other cases, the processing proceeds to the processing step 35. move on.

In the processing step 35, for example, a pair of transmitting and receiving undetermined mobile module notifications is created, and for the pair, the module name in the received notification is added to the outgoing mobile module notification and sent. Receive similar notifications for.

In processing step 33, based on this received information, if there is a module that can be accepted as described above, it is determined to move.
In this case, by adding module names to mobile module notifications as described above, you can monitor the return of notifications that include the name of the module you are responsible for, and by such mobile module notifications, Detects the formation of a center-of-gravity vector loop such as , and determines the movement of those modules.

The movement determined by the above processing is, for example,
1 to the host computer 20, and the above-mentioned placement process is repeated an appropriate number of times, usually with the placement after movement as the initial placement, and the process is completed if, for example, the wiring length is evaluated and the degree of improvement is small. .

Note that the module arrangement within each block is based on each processing device 21.
Each case can be processed in the usual way and movements can be determined and improved if necessary.

[Effects of the Invention] As is clear from the above description, according to the present invention, in processing a problem of determining the placement of circuit modules in automatic design using a computer system, calculation of the destination of the module and determination of the movement are performed as follows. Since processing is performed in parallel across all modules, there is a significant industrial effect of shortening processing time and reducing design costs for LSIs and the like.

[Brief explanation of drawings]

FIG. 1 is a block diagram of the configuration of an embodiment of the present invention and a flowchart of processing, FIG. 2 is an explanatory diagram of the module arrangement, and FIG. 3 is an explanatory diagram of the center of gravity relaxation method. In the figure, 1.41 is a module, 2.40 is a slot, 7~
12 is a center of gravity vector, 20 is a host computer, 21 is a processing device, 30 to 35 are processing steps, and 42 is a block.

Claims (1)

[Claims] A plurality of circuit modules are provided at a plurality of predetermined positions, and each of the modules is moved according to a predetermined logic so as to reduce the required length of wiring connecting the modules. When determining the position to be processed, a plurality of processing devices (21) corresponding to each of the plurality of processing devices (21) are provided in a block formed by dividing the plurality of locations, and each processing device (21) is
21) determines the destination block based on the above logic for each of the modules arranged in the corresponding block (31), and notifies the processing device (21) corresponding to the destination block of the module name. Shi (32, 35)
, each processing device (21) determines the movement of a module between different blocks based on the received notification (33).
) A module placement determination processing method characterized by being configured as follows.