Electrical computers and digital data processing systems: input/ – Access arbitrating
Reexamination Certificate
2001-10-15
2004-11-16
Gaffin, Jeffrey (Department: 2182)
Electrical computers and digital data processing systems: input/
Access arbitrating
C710S005000, C710S006000, C710S040000, C710S052000, C710S053000, C710S309000
Reexamination Certificate
active
06820151
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
This invention relates to computer system input/output (I/O) and, more particularly, to transaction handling in an I/O node.
2. Description of the Related Art
In a typical computer system, one or more processors may communicate with input/output (I/O) devices over one or more buses. The I/O devices may be coupled to the processors through an I/O bridge which manages the transfer of information between a peripheral bus connected to the I/O devices and a shared bus connected to the processors. Additionally, the I/O bridge may manage the transfer of information between a system memory and the I/O devices or the system memory and the processors.
Unfortunately, many bus systems suffer from several drawbacks. For example, multiple devices attached to a bus may present a relatively large electrical capacitance to devices driving signals on the bus. In addition, the multiple attach points on a shared bus produce signal reflections at high signal frequencies which reduce signal integrity. As a result, signal frequencies on the bus are generally kept relatively low in order to maintain signal integrity at an acceptable level. The relatively low signal frequencies reduce signal bandwidth, limiting the performance of devices attached to the bus.
Lack of scalability to larger numbers of devices is another disadvantage of shared bus systems. The available bandwidth of a shared bus is substantially fixed (and may decrease if adding additional devices causes a reduction in signal frequencies upon the bus). Once the bandwidth requirements of the devices attached to the bus (either directly or indirectly) exceeds the available bandwidth of the bus, devices will frequently be stalled when attempting access to the bus, and overall performance of the computer system including the shared bus will most likely be reduced. An example of a shared bus used by I/O devices is a peripheral component interconnect (PCI) bus.
Many I/O bridging devices use a buffering mechanism to buffer a number of pending transactions from the PCI bus to a final destination bus. However buffering may introduce stalls on the PCI bus. Stalls may be caused when a series of transactions are buffered in a queue and awaiting transmission to a destination bus and a stall occurs on the destination bus, which stops forward progress. Then a transaction that will allow those waiting transactions to complete arrives at the queue and is stored behind the other transactions. To break the stall, the transactions in the queue must somehow be reordered to allow the newly arrived transaction to be transmitted ahead of the pending transactions. Thus, to prevent scenarios such as this, the PCI bus specification prescribes a set of reordering rules that govern the handling and ordering of PCI bus transactions.
To overcome some of the drawbacks of a shared bus, some computers systems may use packet-based communications between devices or nodes. In such systems, nodes may communicate with each other by exchanging packets of information. In general, a “node” is a device which is capable of participating in transactions upon an interconnect. For example, the interconnect may be packet-based, and the node may be configured to receive and transmit packets. Generally speaking, a “packet” is a communication between two nodes: an initiating or “source” node which transmits the packet and a destination or “target” node which receives the packet. When a packet reaches the target node, the target node accepts the information conveyed by the packet and processes the information internally. A node located on a communication path between the source and target nodes may relay or forward the packet from the source node to the target node.
Additionally, there are systems that use a combination of packet-based communications and bus-based communications. For example, a system may connect to a PCI bus and a graphics bus such as AGP. The PCI bus may be connected to a packet bus interface that may then translate PCI bus transactions into packet transactions for transmission on a packet bus. Likewise the graphics bus may be connected to an AGP interface that may translate AGP transactions into packet transactions. Each interface may communicate with a host bridge associated with one of the processors or in some cases to another peripheral device.
When PCI devices initiate the transactions, the packet-based transactions may be constrained by the same ordering rules as set forth in the PCI Local Bus specification. The same may be true for packet transactions destined for the PCI bus. These ordering rules are still observed in the packet-based transactions since transaction stalls that may occur at a packet bus interface may cause a deadlock at that packet bus interface. This deadlock may cause further stalls back into the packet bus fabric. In addition, AGP transactions may follow a set of transaction ordering rules to ensure proper delivery of data.
Depending on the configuration of the I/O nodes, transactions may be forwarded through a node to another node either in a direction to the host bridge or away from the host bridge. Alternatively, transactions may be injected into packet traffic at a particular node. In either scenario, an I/O node architecture that may control the transactions as the transactions are sent along the communication path may be desirable.
SUMMARY OF THE INVENTION
Various embodiments of a starvation avoidance mechanism for an input/output node of a computer system are disclosed. In one embodiment, a scheduler unit for an input/output node of a computer system includes a first buffer circuit coupled to receive control commands from a first source and a second buffer circuit coupled to receive control commands from a second source. The first buffer circuit includes a first plurality of buffers for storing selected control commands received from the first source and the second buffer circuit includes a second plurality of buffers for storing selected control commands received from the second source. The scheduler further includes an arbitration circuit coupled to the first buffer circuit and to the second buffer circuit. The arbitration circuit may be configured to arbitrate between the control commands stored in the first buffer circuit and the control commands stored in the second buffer circuit. To avoid starving either the first or the second buffer circuits from sending their respective control commands, the outcome of selected arbitration cycles may be dependent upon a number of times in which a control command from a given one of the buffers is blocked due to an unavailable destination.
In one particular implementation, the arbitration circuit includes a first arbitration unit that may be configured to arbitrate between the selected control commands stored within the first plurality of buffers. Further, the arbitration circuit includes a second arbitration unit which may be configured to arbitrate between the selected control commands stored within the second plurality of buffers. The arbitration circuit further includes a fairness unit which is coupled to the first arbitration unit and the second arbitration unit and may be configured to determine a current transaction request rate for the input/output node and to establish an arbitration priority which is dependent upon the current transaction request rate.
The arbitration circuit may further include a starvation unit that is coupled to the fairness unit and may be configured to count the number of times in which a control command from a given one of the buffers is blocked due to an unavailable destination. The starvation unit may be further configured to store a value corresponding to a maximum allowable number of times in which a control command from a given one of the buffers is blocked due to an unavailable destination. The arbitration circuit may be further configured to select the blocked control command from the given one of the buffers in response to the value corresponding to the maximum allowable number being equal to the cou
Advanced Micro Devices , Inc.
Curran Stephen J.
Gaffin Jeffrey
Kivlin B. Noäl
Knapp Justin
LandOfFree
Starvation avoidance mechanism for an I/O node of a computer... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Starvation avoidance mechanism for an I/O node of a computer..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Starvation avoidance mechanism for an I/O node of a computer... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3286124