Dismiss
InnovationQ will be updated on Sunday, Oct. 22, from 10am ET - noon. You may experience brief service interruptions during that time.
Browse Prior Art Database

Bandwidth Management for System Data Transfers

IP.com Disclosure Number: IPCOM000006510D
Original Publication Date: 2002-Jan-11
Included in the Prior Art Database: 2002-Jan-11
Document File: 2 page(s) / 35K

Publishing Venue

Motorola

Related People

Srinath Audityan: AUTHOR [+2]

Abstract

This invention proposes a mechanism for bandwidth management for stride transfers. It also proposes a dynamic bandwidth management scheme based on transfer priority. The important components of this invention are: a device capable of initiating stride transfers, a bandwidth management scheme for allocating system bandwidth across multiple transfer initiaters. Existing stride transfer schemes have no mechanism for bandwidth management. This invention proposes a mechanism to solve this problem and performs dynamic bandwidth management based on transfer priority.

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 69% of the total text.

Bandwidth Management for System Data Transfers</DIV><DIV>

Srinath Audityan & </DIV><DIV>Jose M Nunez

Abstract

This invention proposes a mechanism for bandwidth management for stride transfers. It also proposes a dynamic bandwidth management scheme based on transfer priority. The important components of this invention are: </DIV><DIV>a device capable of initiating stride transfers, a bandwidth management scheme for allocating system bandwidth across multiple transfer initiaters. </DIV><DIV>Existing stride transfer schemes have no mechanism for bandwidth management. This invention proposes a mechanism to solve this problem and performs dynamic bandwidth management based on transfer priority.

Body 

All DMA transfers are split into sub-transfers (chunks) as specified by the bandwidth control for proper bandwidth allocation. During stride transfers, if the chunk size is less than the stride size, left over stride data remains. When the channel gets its turn again to transfer its next chunk, it has to start from where it left off. Only after finishing the left over stride data, it can jump to the next location to start a new stride. This is achieved by saving the remaining stride size and address back in the channel and restarting with this saved off value when the left over stride history buffer is valid. In other words, the channel state is saved off in the middle of a stride and restored again when the channel get its turn to transfer data to finish the stride.

<DIV></DIV>

Bandw...