fpga device realizing on-chip network transmission bandwidth expansion function

Abstract

This application discloses an FPGA device for implementing the function of expanding network transmission bandwidth on a chip, and relates to the field of FPGA technology. When the FPGA bare chip is integrated with a predetermined function module built with a hard core IP node, the logic resource module inside the FPGA bare chip is used Configured to form a soft-core IP node connected to a hard-core IP node to form a NOC network structure, realize the increase of nodes, and expand the transmission bandwidth of the predetermined functional module. On the other hand, the soft-core IP node can be additionally connected to the input and output of the predetermined functional module Signals can also expand the transmission bandwidth of predetermined functional modules. Further, the FPGA device also includes a silicon connection layer with a silicon connection layer NOC network inside, so that the nodes inside the FPGA die can be connected to the silicon connection layer to form a larger NOC structure, further realizing the increase of nodes and expanding predetermined functional modules transmission bandwidth.

Description

technical field [0001] The invention relates to the field of FPGA technology, in particular to an FPGA device for realizing the function of expanding network transmission bandwidth on a chip. Background technique [0002] With the continuous improvement of integrated circuit manufacturing technology and the sharp increase in the amount of data, artificial intelligence has achieved rapid development in recent years, especially in March 2016 AlphaGo and the world champion of Go, professional nine-dan Go player Li Shishi played Go man-machine battle, Winning with a total score of 4 to 1 pushed artificial intelligence to unprecedented heights. The rapid development of human intelligence is inseparable from the improvement of the efficiency of big data, high-performance computing and machine learning algorithms. In the stage where machine learning algorithms are still being iteratively optimized, FPGAs have unique advantages in the field of artificial intelligence due to their p...

AI Technical Summary

Problems solved by technology

In the stage where machine learning algorithms are still being iteratively optimized, FPGAs have unique advantages in the field of artificial intelligence due to their programmability and powerful vector computing capabilities. The FPGA field is also optimizing the FPGA architecture and adding dedicated Artificial intelligence engine, so now there are some ways to design and integrate a dedicated AI module inside the FPGA die, but after the AI ​​module is integrated inside the FPGA die, it is limited by the chip structure, and the transmission bandwidth and computing efficiency are relatively limited. It is difficult to achieve the expected work effect

Images