Method and device for implementing fast RR scheduler with non-fixed interval jump

By adopting a fast RR scheduler method with non-fixed interval jump in the RR scheduler, the low latency problem caused by traditional RR scheduling is solved, and efficient low-latency transmission of data channels in the 5G era is achieved.

CN115766531BActive Publication Date: 2025-02-25苏州特思恩科技有限公司
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
CN202211391153.4
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-11-11
Publication Date
2025-02-25
Estimated Expiration
2042-11-11

AI Technical Summary

Technical Problem

Traditional RR scheduling methods lead to low latency problems in the 5G era, especially in industrial Ethernet, because it polls on data channels without data requests, resulting in unnecessary latency.

Method used

Using a fast RR scheduler with non-fixed interval jump, it can create data path request signals, request signal sample space and request sample space to perform logic and operation and high priority path selection, skip channels without data requests, and achieve low Delayed data transfer.

Benefits of technology

The delay caused by traditional polling scheduling is reduced, and the operation skipping on channels without data requests is realized, the low latency performance of data transmission is improved, and there is no limit on the number of data paths and clock frequency.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN115766531B_ABST
    Figure CN115766531B_ABST
Patent Text Reader

Abstract

A method and device for implementing a fast RR scheduler with non-fixed interval jumps, the present invention relates to the field of communication chip technology. In the field of communication chip design, traditional RR scheduling will poll the data channel according to the clock cycle, regardless of whether there is a data transmission request on the data channel; this will cause unnecessary delays, because the data on some channels will be sent with a delay due to the access of its previous channel (this channel has no data request), thereby causing delays. The present invention uses a reasonable algorithm to skip the data path without data request, effectively shorten the RR polling cycle in real time, and effectively reduce the delay caused by using the traditional RR scheduling method at the communication chip interface traffic convergence point.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of communication technology, and in particular to a method and device for implementing a fast RR scheduler with non-fixed interval jumps. Background Art

[0002] In the design of communication chips, it is inevitable to encounter the problem of scheduling data messages. Especially in places where interface traffic converges, a round-robin scheduling method is generally used to extract traffic from different data paths.

[0003] This conventional RR technology can access all data paths one by one according to the data path bandwidth allocation to determine whether the data paths have data transmission requests.

[0004] In the 5G era, communication equipment is required to have lower latency, so low latency has become a higher priority indicator for communication chips, especially in industrial Ethernet.

[0005] Traditional RR scheduling polls the data channel according to the clock cycle, regardless of whether there is a data request on the data channel; this brings unnecessary delays, because the data on some channels will be sent later due to the access of the previous channel (this channel has no data request), resulting in delays. Summary of the invention

[0006] The purpose of the present invention is to overcome the shortcomings of the above-mentioned technology and to provide a method and device for implementing a fast RR scheduler with non-fixed interval jumps. The method and device are easy to implement, have simple logic, have no clock rate bottleneck, can skip data paths without data requests, and realize low-latency transmission of data paths with transmission data requests.

[0007] The present invention provides a method and device for implementing a fast RR scheduler with non-fixed interval jump, comprising the following steps:

[0008] S1. Create a data path request signal. The signal bit width corresponds to the number of data path entries. Each bit represents whether the path has a request action. A logic high level indicates that the path has a request, and a logic low level indicates that the path has no request.

[0009] S2, creating a data path request signal sample space, that is, copying the data path request signal N times, where the value N is the number of data paths;

[0010] S3, create a data path request to select a sample space, that is, there is a sample space, the number of elements in the sample space is the number of data paths, the bit width of each element is the number of data paths, and only one bit in each element is set, and the set bit corresponds to the number of the data path;

[0011] S4, performing a logical AND operation on the data path request selection sample space and the data path request signal sample space to obtain a data path valid request sample space, wherein the number of logical 1s in the sample space represents the total number of requests;

[0012] S5, analyze the result in S4 and select the high priority path, negate the result of this element, keep other elements unchanged, and perform a logical AND operation with the result in S4 to remove the requested path to be processed;

[0013] S6. Repeat S4 and S5 until there is no logic 1 in the result, and POP the request data group out of the queue.

[0014] Based on the above technical solution, a round of operations can be completed in the fastest clock cycle of RR polling scheduling.

[0015] Compared with the prior art, the advantages of the present invention are as follows:

[0016] 1. The present invention reduces the delay caused by the traditional polling scheduling of periodic access to the data path;

[0017] 2. The present invention has no limitation on the number of data paths and design clock frequency; BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 4-channel fast RR dispatch system

[0019] Figure 2 Fast RR scheduling algorithm for 4 channels DETAILED DESCRIPTION

[0020] The present invention will be further explained in detail below by taking 4-channel data request as an example in conjunction with the accompanying drawings.

[0021] The present invention provides a method and device for implementing a fast RR scheduler with non-fixed interval jump, comprising the following steps:

[0022] S1, chnl0, chnl1, chnl2, chnl3 are the conditions for request occurrence; if in a certain clock cycle, channel 0 and channel 2 have requests, then the data path request signal reqs = 1010 is constructed;

[0023] S2, create a data path request sample space,

[0024] reqsSamp={1010,1010,1010,1010};

[0025] S3. Create a data path request to select a sample space. The number of primary colors in this sample space is the number of data paths. The bit width of each primary color is the number of data paths. There is only one bit set in each element. This set bit corresponds to the data path number:

[0026] reqSelG = {1000,0100,0010,0001};

[0027] S4. Perform a logical AND operation on the data path request selection sample space and the data path request signal sample space to obtain the data path valid request sample space. The number of logical 1s in this sample space represents the total number of requests. The valid sample space value is:

[0028] result1 = {1000 0000 0010 0000};

[0029] Through the selector, we know that channel 0 has a request, and the logic can process the request of channel 1 first;

[0030] S5, analyze the result in S4 and select the high priority path, negate the result of this element, keep other elements unchanged, and perform a logical AND operation with the result in S4 to remove the requested path to be processed; the result after processing is:

[0031] result2 = {0000 0000 0010 0000}; It can be seen that channel 2 has a request, and the logic processes the request of channel 2;

[0032] This skips channel 1;

[0033] S6. Repeat S5, and the result is {0000 0000 0000 0000};

[0034] S7. POP the path request signal from the request FIFO.

Claims

1. A method for implementing a fast RR scheduler with non-fixed interval jumps, characterized in that The following steps are involved: S1. Create a data path request signal. The signal bit width corresponds to the number of data path entries. Each bit represents whether the path has a request action. A logic high level indicates that the path has a request, and a logic low level indicates that the path has no request. S2, creating a data path request signal sample space, that is, copying the data path request signal N times, where the value N is the number of data paths; S3, create a data path request to select a sample space, that is, there is a sample space, the number of elements in the sample space is the number of data paths, the bit width of each element is the number of data paths, and only one bit in each element is set, and the set bit corresponds to the number of the data path; S4, performing a logical AND operation on the data path request selection sample space and the data path request signal sample space to obtain a data path valid request sample space, wherein the number of logical 1s in the sample space represents the total number of requests; S5, analyze the result in S4 and select the high priority path, negate the result of this element, keep other elements unchanged, and perform a logical AND operation with the result in S4 to remove the requested path to be processed; S6. Repeat S4 and S5 until there is no logical 1 in the result, and POP the request data group from the queue.

Citation Information

Patent Citations

  • Novel bus polling scheduling method and system based on FPGA

    CN113722069A

  • Network-on-chip environment and method for reduction of latency

    WO2006106475A1