SU717790A1 - Device for calculation of large networks - Google Patents

Description

   - -; /. . .:

This invention relates to the field of computing and may be used; Try to build digital spe cialized devices to determine the optimal paths in heterogeneous

 continuous environments.

A device is known for determining the optimal path in a continuous inhomogeneous medium as the shortest path defined on a graph of an arbitrary grid K ph | information with information (in nodes covering this medium fl.

The closest in technical essence to the present invention is a device for calculating large networks, comprising a branch modeling unit, the first output of which is connected by the do input as well. the first input is with the output of the control unit, the first input of which is connected to the first output of the external storage unit, the input and the second output of which are connected respectively to the water supply.

an output and a second input of a branch modeling unit, and a block of memory elements f 23.

The main disadvantage of the known device is the increase in the solution time at the constant lattice size due to the large number of fragments, as well as due to the restructuring of the computing device from the fragment to the frame;

The essence of this invention is fast-paced. .

Claims (2)

In the cauldron, the goal is achieved by additionally input and output logic to the device. switches and a block for searching for fragments of the medium; the outputs of the block of memory elements are connected respectively to the inputs 1 of the input logic switch, the outputs of which are connected respectively to the group of inputs of the simulation of branches, the first group of outputs of which are connected respectively with. The search inputs for fragments I of the medium and the second group of outputs are connected respectively to the group of inputs of the output of the logic switch, the outputs of which are connected respectively to the inputs of the block of memory elements, and the input is connected to the search pbixeay. fragments of the medium. The media fragment search block contains a directional storage matrix, a directional transmission node, a counter / K (Ai ychestvstv shifts and reversible shift register, and the first group of inputs of the directional storage matrix enters the g | euppoy of the block inputs, the second group of inputs of the memory matrix the failures are respectively connected to the inputs of the reverse shift register, the outputs of the directional storage matrix are connected respectively to the transfer transmission unit, the first output of which is connected to the input of the reverse registration The second gate is connected to the input of the shift amount counter, and the third one is at the output of the block, the output of the counterdifferentiate shifts is connected to the input of the node YachIHanjpiaBrieHHjiii. search for fragments of the medium, in Fig. 3 is an example illustrating the operation of the device; v.; -.. ;; .. -; - / v - -; The block diagram of the device (see Fig. 1) contains a block modeling branches 1, in the external memory block 2, block, ele. . memory 3, block up | equals 4, output logic switch 5, input logic switch 6, and search engine environment 7 blocks of memory elements 3 in combination with output 5 and input 6 switches represent a block: in 6 matches: whom collection of fragments 8,.,., ..- ,; . . The block of searching for fragments of environment 7 (see Fig. 2) consists of the matrix of storing the directions of the path 9, the reversing shift register 10, the counter if 1 of the shifts 11 and the sending node of directions 12. To illustrate the operation of the device, consider an example (see Fig. 3 ), where in a continuous inhomogeneous medium it is necessary to determine the optimal path, for example; between two points 1 and 11.; Investigated Medium Covered by a grating, for example, of a rectangular configuration with nodes 1-25 (the connection between the nodes of the graph-lattice in Fig. 3 is omitted from 1.t -. -3 --. - ... / ...: , -., -,., .- i, -;:., -. НЫ /. - .... - l. The principle of the device operation the following Control unit 4 (see Fig. 1) ensures the operation of the device in two stages In the first stage, from the external storage unit 2 into the block for modeling the branches 1, the weight is weighted, which are integral characteristics determined from the corresponding rectangular area, and the shortest path from node 1 to the nodes is determined He ate 11 (see Fig. 3). The distinguished shortest path indicates that a zoon is separated from the continuous non-uniform medium under study, the integral characteristics of which are minimal, therefore the desired optimal path is assumed. Information about the shortest path, namely the number of nodes through which this path passes, in this case 1.8, 12, 16, 17, 14, 11,; and the direction of the exit of the path from each node (for a rectangular lattice with diagonals, this number is eight) is entered in block search medium fragments 7. They use It is in the second phase of the task: node numbers - for fragments job number and directions - for specifying the boundary nodes. In the block of searching for fragments of environment 7 (see FIG. 2), the information is stored in the directional storage matrix 9, which is a logic control trigger matrix. The number of rows of matrix 9 is equal to the number of elements of nodes in the graph, and the number of columns is equal to the number of directions of the path from the node. Information is written into matrix 9 in parallel when the shortest path is extracted from the resulting tree, i.e., when the logical zero signal is transmitted from the end node to the initial: zero. In this case, information about the direction of the path from it is recorded in the line corresponding to the node number. The number of nodes lying on the critical path determines the number of fragments that must be considered to find the desired solution, in this example, seven fragments (nodes 1, 8, 12, 16, 17, 14, 11). At the second stage, the control unit 4 provides successively in a given n: or: d, i.e., in order to traverse the shortest path from the initial node to the final node, select fragments using the medium fragment search block 7.- and determine the shortest paths from the initial boundary nodes, defined by the direction of the shortest path from the previous node, to the final boundary nodes, determined by the direction of the shortest path exit from the node under consideration with the help of a fragment conjugation block the fragment is stored in the external storage unit 2, and then the shortest path from the initial node 1 to the final node, node 11 is allocated. If the resulting path satisfies the conditions of the problem, the device stops, if not, then the second step is repeated for new inset zone defined by the shortest path obtained. Consider separately both modes of operation of the device. The selection of a fragment is carried out using a fragment search block 7 (see FIG. 2). The first fragment is specified by the initial node, and the shortest path comes from the YoOtbFd. Subsequent fragments are determined using the reverse shift register 10. In this case, information about the direction of exit and output of the SU node is selected from matrix 9 and copied to the transmission direction transfer 12. Based on this information, the node has transmitted the corresponding node numbering of the network model directions 12 is determined by the number of the next node, i.e. the number of shifts is formed and the direction of the shift of the reverse shift is 10, the number of shifts of the register 10 is controlled by count COK 11. - .. / -.; -. In order to move, for example, from the first fragment defined by the initial node 1 (see Fig. 3), to the second fragment defined by the direction of the shortest path from this bridle, i.e. to the fragment defined by node 8, it is necessary shift register 10 shift, right to 7 bits. This shift is performed as follows. From matrix 9 (see Fig. 2), the direction of the exit of the path from node 1 is transmitted to the node of direction 12 transmission. An analysis is performed there, on the basis of which the number of shifts is set in the counter code 11 in the reverse code and then shifted with register shift 10. Pulse overflow from counter 11 turns the shift register to shift 10. If the number of the subsequent node has a smaller number, as it happens, for example. With the passage from the fragment defined by wap 17 to the fragment defined by node 14, the shift register 10 is shifted to the left by three bits. Conjugation of the results obtained from the sequential examination of fragments is carried out; jc in conjunction with the block of conjugation of fragments 8 (see Fig. 1). Block of elements pamd-; TI 3, INCLUDED in the block of conjugation of fragments 8, consists of two groups of counters l - KI l, -K, which are alternately connected to the initial and boundary nodes of the block 1 of modeling branches on this fragment. When analyzing the first fragment, any one of the groups of counters of the block of memory elements 3 is connected to the end boundary nodes of this fragment. The par | 1 analysis of the second / fragment that group, which was connected to the end boundary nodes of the first fragment, is connected to the initial, and the other to the end boundary nodes, the subsequent fragment. The group of counters, which are connected to the terminal nodes of the first fragment, contains the number of pulses characterizing the temporal shift between the paths leading to the boundary line of the secondary boundary line. In the second group of counters connected to the end boundary nodes of the second fragment, the information obtained determines the time shift between the paths leading from the initial node of the first fragment to the end nodes of the second IT fragment, d. The length of the coat -. The path to the boundary angles of the second fragment will be fixed in the metering counter located in the control unit. Analbgitao calculates the paths on the subsequent fragments, switching the end and initial boundary nodes with the corresponding group of counters of the memory cell is performed respectively with the help of output 5 and input 6 logic switches. Thus, the performance of the device in. According to the invention, the time to solve a problem can be significantly reduced by reducing the number of fragments, as well as by automating device reorganization from fragment to fragment. Claim 1, A device for calculating large networks, comprising a branch modeling unit, the first output of which is connected to the input, and the first input is connected to the output of the control unit, the first input of which is connected to the first output of the external storage unit, the input and the second output the stroke of which is connected with the second output and the second input of the branch modeling, and a block of memory elements, characterized in that, in order to increase speed, it is additionally introduced into the input and output logic switches and the block for searching for fragments of the medium, and the outputs of the block of elements memory connected respectively to the inputs input logic switch, you ..,. „.„., Fer the moves of which are connected respectively to the input group of the branch modeling unit, the first output group is connected respectively to the inputs of the media fragment search unit, and the second output array is connected to the input switch group of the output switch, respectively, the output port 6 is connected respectively to the inputs a block of memory elements, and tioiftaito a scanner to the output of a block for searching fragments of a medium .- / 2. The device according to claim 1, which means that the search unit of the fragment media in the environment contains a matrix of direct directions, the node transmitting directions,. a shift number estimator and a reversible shift register, the first group of inputs of the directional storage memory matrix is a group of block inputs, the second group of directions storage matrix inputs are respectively connected to the reversal shift register outputs, outputs of the direction storage matrix, respectively, are connected to the input node of the directional transmission node, first output which is connected to the input of the reverse shift register, the second output is connected to the input of the counter of the number of shifts, and the third output is m block, output Counter for the number of shifts soy and en with the input yai / ia transfer directions. Information sources, accepted during examination 1. Vashev V.V. Approximate Mbt, solving some variational problems with local functionals. Bulletin of the Ukrainian SSR Academy of Sciences, 1974, fc 11. , 2. Dodonov, AG, Khadzhinov, V, V, On one / method of solution: large networks on cj (} analogs. Collection. Hybrid computers and complexes, Kiev, Naukova Dumka 1975 (prototype). ig f . , ig, 2 ig.Z