Dismiss
IP.com applications will be updated on Sunday, March 5, from 11 am to 2 pm ET, to add new functionality and content. You may experience brief service interruptions during this period. We apologize for any inconvenience.
Browse Prior Art Database

A Method of Resolving Collision between DMA and PCI Devices with Shared Address/Data Path to Memory

IP.com Disclosure Number: IPCOM000007630D
Original Publication Date: 2002-Apr-10
Included in the Prior Art Database: 2002-Apr-10
Document File: 3 page(s) / 46K

Publishing Venue

Motorola

Related People

Thang Q. Nguyen: AUTHOR [+2]

Abstract

For architectures with a single port access to local memory and accesses to local memory is arbitrated by an internal arbiter between internal Direct Memory Access (DMA) and external Peripheral Component Interconnect (PCI) devices, fairness between the devices can be achieved by limiting the DMAs access bandwidth to local memory; thus increasing the chance of successful PCI devices accesses to local memory. However, this is generally done at the expense of DMA performance and overall fairness is not achieved optimally. There should be an algorithm in ensuring that all devices (external PCI and internal DMA devices) get a fair chance to access local memory and ensuring forward progress in the system without sacrificing the DMA or PCI performance. This document presents a method of arbitration that ensures fairness between internal DMA and external PCI devices while maintaining maximum performances among them.

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 50% of the total text.

A Method of Resolving Collision between DMA and PCI Devices with Shared Address/Data Path to Memory

by

Thang Q. Nguyen

Chau-Shing Hui

I. Abstract:

For architectures with a single port access to local memory and accesses to local memory is arbitrated by an internal arbiter between internal Direct Memory Access (DMA) and external Peripheral Component Interconnect (PCI) devices, fairness between the devices can be achieved by limiting the DMA’s access bandwidth to local memory; thus increasing the chance of successful PCI devices’ accesses to local memory. However, this is generally done at the expense of DMA performance and overall fairness is not achieved optimally. There should be an algorithm in ensuringthat all devices (external PCI and internal DMA devices) get a fair chance to access local memory and ensuring forward progress in the system without sacrificing the DMA or PCI performance. This document presents a method of arbitration that ensures fairness between internal DMA and external PCI devices’ while maintaining maximum performances among them.

II. Body:

Arbitration Scheme

Arbitration among the PCI devices is done either by an off-chip PCI arbiter or an on-chip PCI arbiter while an internal DMA arbiter does arbitration among the DMA devices. The winner among the PCI device and the winner among the DMA devices have to go through a second level of logic arbitration. If both DMA and PCI request the local memory access at the same time, then the PCI request will win. If the PCI device is currently accessing local memory and there is an outstanding DMA request, then after the currently PCI transaction is finished, the DMA will have access to local memory. All subsequent PCI accesses to local memory with be retried. If internal DMA controllers are accessing local memory, and a PCI Retry occurs when PCI is trying to access local memory, the current DMA transaction has to be terminated within n cache line transfer where n is a number that is specific to a design. All subsequent DMA requests are halt in favor of subsequent incoming PCI transactions.  A multi-bit internal DMA hold-off counter starts counting while the DMA controllers are off.  If the counter has not expired and anoth...