A dynamic and static noc mapping method for power consumption and temperature

Abstract

The invention discloses a NoC mapping method oriented to the dynamic and static combination of power consumption temperature, which is characterized by the following steps: 1. Establishing a mapping library through a discrete firefly algorithm oriented to power consumption temperature; 2. Putting the mapping library into a host; 3. The host sends the initial mapping scheme to the NoC; 4. The NoC configures the mapping scheme; 5. The NoC monitors and records the traffic of each communication node in real time; 6. The NoC sends the traffic of each node to the host regularly; 8. The host matches the temperature mode according to the maximum temperature node position, and calls the corresponding mapping scheme in the mapping library; 9. The host sends the new mapping scheme to the NoC; step 10, the NoC configures the NoC according to the new mapping scheme, and returns Step 5. The invention can prevent mapping hotspots, balance network loads, and simultaneously adopt a dynamic and static method to reduce online resource occupation and improve mapping optimization efficiency.

Description

technical field [0001] The invention belongs to the technical field of network communication, and in particular relates to a NoC mapping method for dynamic and static combination of power consumption temperature. Background technique [0002] Although the change of network-on-chip (NoC) replacing the traditional bus structure overcomes various problems caused by the traditional structure to a certain extent, with the increase of IP (Intellectual Property) cores and the increase of communication traffic, the The power consumption of the interconnection network has become the focus of NoC design. With the reduction of the chip size and the improvement of the overall performance, the power consumption per unit area also increases. This makes the task mapped to the resource node for processing. The rapid rise even leads to the generation of local hot spots (Hot Spot); the high temperature affects the performance of the chip and may even damage the device, which greatly weakens t...

AI Technical Summary

Problems solved by technology

[0002] Although the change of network-on-chip (NoC) replacing the traditional bus structure overcomes various problems caused by the traditional structure to a certain extent, with the increase of IP (Intellectual Property) cores and the increase of communication traffic, the The power consumption of the interconnection network has become the focus of NoC design. With the reduction of the chip size and the improvement of the overall performance, the power consumption per unit area also increases. This makes the task mapped to the resource node for processing. Rapidly rising, even leading to the generation of local hot spots (Hot Spot); high temperature affects the performance of the chip and may even damage the device, which greatly weakens the advantages of NoC and directly affects the performance of the network system on chip

[0003] NoC mapping is divided into static mapping and dynamic mapping. Static mapping is an offline mapping method. Before the NoC system works, the mapping scheme is obtained through the mapping algorithm, which does not occupy online resources. However, static mapping cannot be adjusted according to the status of the NoC system. The solution optimizes system performance, and the optimization time is long

[0004] Dynamic mapping is an online mapping method. When the NoC system is working, the system working status is obtained, the mapping algorithm is run according to the system status, and the mapping scheme is replaced online. Although the optimization time is fast, dynamic mapping requires a lot of online resources and may not be able to get the optimal solution

[0005] Most of the existing mapping schemes are researches on the choice of one of the two. Optimizing one aspect may degrade the performance of the other, and the traditional NoC mapping algorithm has a single optimization goal, while the optimization goal of most mapping algorithms mainly focuses on communication loss. In terms of performance, such as power consumption and delay, less attention has been paid to the influence of temperature on the mapping system. At the same time, the calculation of node temperature is mainly based on the temperature model tool Hot spot. There is no more effective and direct method to calculate the temperature of each node of the network on chip; For example, the Chinese invention patent CN104079439A "An On-Chip Network Mapping Method Based on Discrete Firefly Algorithm" published on October 1, 2014 improves the traditional firefly algorithm to solve the on-chip network mapping problem, and can quickly and effectively find the optimal mapping solution

The disadvantage of this method is that the mapping optimization target is concentrated on the total communication loss, which may cause the task mapping to be concentrated while reducing the overall communication loss, resulting in higher local node temperature, resulting in hot spots

Images