Browse Prior Art Database

Area and Power Efficient Multi Priority Level Round Robin Arbitration Scheme for Shared Resource

IP.com Disclosure Number: IPCOM000204386D
Publication Date: 2011-Feb-22
Document File: 3 page(s) / 122K

Publishing Venue

The IP.com Prior Art Database

Abstract

As more and more miniaturization occurs in the VLSI industry, more functionality is placed on chip. This drives the requirement for support of high data intensive applications and puts pressure on shared resources e.g., system bus, memory, etc. For Quality of Service (QOS) for shared resources and fair play, there is a need for an arbitration scheme that is efficient in terms of area, power and scalability. This paper discusses fixed priority arbitration for a multi-priority level system for shared resources. The scheme presented is different from the traditional approach of distributing a token in a round robin scheme. In the traditional approach, the number of token storages is proportional to the number of requests at each priority level, which increase linearly with the number of master increases. The arbitrator discussed here arbitrates at two levels; it first decides a winner based on priority of the requests and then utilizes a round robin for the requests in the winning priority. The area increase is sub-linear by modifying the round robin token storage approach.

This text was extracted from a Microsoft Word document.
At least one non-text object (such as an image or picture) has been suppressed.
This is the abbreviated version, containing approximately 54% of the total text.

Area and Power Efficient Multi Priority Level Round Robin Arbitration Scheme for Shared Resource

Abstract:

As more and more miniaturization occurs in the VLSI industry, more functionality is placed on chip. This drives the requirement for support of high data intensive applications and puts pressure on shared resources e.g., system bus, memory, etc.  For Quality of Service (QOS) for shared resources and fair play, there is a need for an arbitration scheme that is efficient in terms of area, power and scalability. This paper discusses fixed priority arbitration for a multi-priority level system for shared resources. The scheme presented is different from the traditional approach of distributing a token in a round robin scheme.  In the traditional approach, the number of token storages is proportional to the number of requests at each priority level, which increase linearly with the number of master increases.  The arbitrator discussed here arbitrates at two levels; it first decides a winner based on priority of the requests and then utilizes a round robin for the requests in the winning priority.  The area increase is sub-linear by modifying the round robin token storage approach.

Details:

Traditional Technique:          

In the traditional arbitration scheme, the number of token for each priority level is the same as the number of masters supported. Any master can request at any priority level based on user programming. A token is used to enable or disable a request to participate in round robin arbitration only once. One example is taken where number of masters requesting a shared resource are 32 and there are 8 priority levels increasing from level 0 to level 7.

Traditional Round Robin token storage in multi priority arbitration- Fig-1

 

Implication of Above Technique:

·         Area usage is high due to maintaining history of the token distributed at each priority level to enable the round robin process.

·         Power increases in proportion to the area.

·         Scalability is linear with the number of masters supported. In this case if the number of masters increases from 32 to 64 then area is almost double.

The arbitration scheme discussed next addresses these issues.

Area and Power efficient Arbitration scheme:

This scheme handles token distribution in a round robin scheme a little differently. In round robin the token moves from one requesting master to another requesting master every cycle.  Only needed information as per round robin definition processed by the last master number that is basically a token needs to be disabled for all the masters serviced and needs...