Operation device
A computing device and computing result technology, applied in computing, computers, energy-saving computing, etc., can solve problems such as increased power consumption, increased times of memory reading and writing, increased number of times of memory reading and writing, etc.
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
Method used
Image
Examples
no. 1 approach
[0031] (structure)
[0032] Embodiments will be described below with reference to the drawings.
[0033] figure 1 It is a block diagram showing the configuration of the computing system 1 of the first embodiment. In the description of the embodiment and modifications, the features Figure X Represents all or part of the feature map or feature quantity, weight W represents all or part of the weight, operation unit M represents all or part of the operation unit, region B represents all or part of the region, and operation result Y represents the entire operation result or part of it.
[0034] The computing system 1 has a memory 2 and a computing device 3 . The calculation device 3 is connected to the memory 2, and reads various information from the memory 2 to perform calculations.
[0035] Memory 2 stores network structure information S, weights W and feature Figure X , for example constituted by SRAM. Network structure information S, weight W and features Figure X Fo...
no. 2 approach
[0099] In the first embodiment and the modified example, the calculation instructing unit 11 performed the characteristic convolution calculation Cv Figure X After scanning in the xy direction, slide in the z direction to perform the next scan in the xy direction. However, after scanning in the z direction, slide in either the x direction or the y direction to perform the next scan. The scanning in the z direction constitutes. In the description of this embodiment, descriptions of the same configurations as those of other embodiments or modified examples are omitted.
[0100] For example, the calculation device 3 sequentially adds and stores the intermediate calculation result Ic in the area B1 through scanning in the z direction, and outputs the intermediate calculation result Ic from the area B1 to the memory 2 after scanning in the z direction, and stores the intermediate calculation result Ic in the area B1. Area B1 is cleared. Next, slide in the x direction, and sequen...
no. 3 approach
[0107] Next, a third embodiment will be described. In the description of this embodiment, descriptions of the same configurations as those of other embodiments or modified examples are omitted.
[0108] Figure 14 It is a block diagram showing an example of the configuration of the computing system 1 of the third embodiment.
[0109] Such as Figure 14 As shown, in the memory 2, the network structure information S, the weight matrix W and the feature Figure X . The network structure information S has weight matrix identification information as identification information of the weight matrix W. The weight matrix W represents all or part of the weight matrix.
[0110] Figure 15 It is a table for explaining an example of the calculation procedure of the calculation system 1 of the third embodiment.
[0111] Such as Figure 15 As shown, in sequence 1, the computing device 3 reads in the preset feature from the memory 2 Figure X The convolution operation Cv is performed...
PUM
Login to View More Abstract
Description
Claims
Application Information
Login to View More