Ncr Corporation

Intel Corporation

Edward A McDonald

Thomas F Heil

Arthur F Cochcroft, Jr

James M Ottinger

Robert J Miller

Gene F Young

Byron L Reams

Harry W Scrivener

Craig A Walrath

Daniel C Robbins

Bobby W Batchler

Larry C James

Roy M Stevens

Billy K Taylor

Jimmy D Pike

Richard R Barton

Peter Washington

Jeffrey A Hawkey

P Chris Raeuber

Marti D Miller

James B Kirkland, Jr

There is provided, for use in a processing system containing a plurality of processors coupled to a main memory, a control circuit for perturbing a lock-step sequence of memory requests received from the processors. The control circuit comprises a memory request generator for generating at least one memory request operable to terminate the lock-step sequence of memory requests.

System and method for terminating lock-step sequences in a multiprocessor system

A deadlock-avoidance system for a computer. In a multi-bus, multi-processor computer, one processor may request a lock on a bus, to execute a locked cycle, thereby blocking all other processors, and other agents, from access to the bus. In addition, a conflicting agent may, in effect, lock a resource which is needed by the processor to complete the cycle for which the lock was requested. These two locks can create a deadlock situation which stalls the computer: the processor and the conflicting agent have each locked a resource needed by the other. Under the invention, when a locked cycle is requested by a processor, all other operations are suspended in the computer. Then queues standing in memory controllers are emptied. If a process requested by an agent occupies a resource, such as a bridge, required by the requested locked cycle, that resource is freed. Then the locked cycle is executed.

Method for preventing deadlock by suspending operation of processors, bridges, and devices

A controller is provided, for use in a processing system containing a plurality of processors operable to communicate with a plurality of I/O devices, for directing a first I/O request issued by a first selected one of the plurality of processors to a targeted one of the I/O devices. The controller device comprises a counter for counting a number of retries associated with the first I/O request and comparison circuitry for comparing a count value in the counter with a first predetermined limit, wherein the controller, in response to a determination that the count value in the counter exceeds the first predetermined limit, blocks all other I/O requests issued by the plurality of processors from being directed to the targeted I/O device.

System and method for improved transfer of data between multiple

Multiple subsystem I/O (Input/Output) buses are coupled to one or more system buses of a computer system by interface circuits which perform necessary decoding of memory space and I/O (Input/Output) space for allocation of portions of the memory space and the I/O (Input/Output) space to each I/O (Input/Output) bus. The interface circuits also translate fixed addresses within each I/O (Input/Output) bus to permit proper operation of the I/O (Input/Output) buses with the computer system. The interface circuits are programmed by the computer system to define the allocated memory spaces and I/O (Input/Output) spaces for the corresponding I/O (Input/Output) buses. Programming of the I/O (Input/Output) buses is performed at the time of system configuration by writing appropriate values into configuration registers incorporated into each of the interface circuits.

Apparatus and method for address translation and allocation for a

There is disclosed a memory control circuit for use in a processing system containing a plurality of processors coupled to a main memory by a common bus. The memory control circuit is adapted for implementing directory-based coherency in the processing system according to a selected coherency algorithm and comprises: 1) monitoring circuitry for detecting coherency corruption in a coherency directory associated with the main memory; and 2) coherency control circuitry responsive to a detection of coherency corruption in the coherency directory for dynamically modifying the selected coherency algorithm, thereby enabling the processing system to shut down in a controlled manner. In some embodiments, the monitoring circuitry further detects possible system coherency failure conditions external to the coherency directory and the coherency control circuitry responds to the detection of a possible system coherency failure condition by dynamically modifying the selected coherency algorithm, thereby enabling the processing system to shut down in a controlled manner.

System and method for reliable system shutdown after coherency corruption

A system for booting a multi-processor computer. If a normal boot attempt fails, different processors are selected, one-at-a-time, for performing the boot routine. During the boot routing, all other processors are held inactive. After boot, processors are tested for health. Non-healthy processors are held inactive, and healthy processors are activated as usual.

Booting method for multi-processor computer

A multiprocessor computing apparatus that includes a mechanism for favoring at least one processor over another processor to achieve more equitable access to cached data. Logic for detecting when, for example, a remote and a local processor are attempting to access data from the cache of another local processor is disclosed. Logic that provides an advantage to the remote processor in a manner that achieves fairer access among the various processors is also disclosed.

Multiprocessor computing apparatus having spin lock fairness

A hardware method to concurrently obtain memory access locality information for a large number of contiguous sections of system memory (pages) for the purposes of optimizing memory and process assignments in a multiple-node NUMA architecture computer system including a distributed system memory. Page access monitoring logic is included within each processing node which contains a portion of shared system memory. This page access monitoring logic maintains a plurality of page access counters, each page access counter corresponding to a different memory page address within the shared system memory. Whenever the processing node generates a transaction requiring access to a memory address within system memory, the page access monitoring logic increments a count value contained within the page access counter corresponding to the memory address to which access is sought. Thus, a record of memory access patterns is created which can be used to optimize memory and process assignments in the computer system.

Method and apparatus for determining memory page access information in a

A multiprocessor computing apparatus including a plurality of processors each having a cache memory and preferably arranged in nodes on a system bus. A first cache coherency providing mechanism coupled to the processors for achieving system level cache coherency. A second cache coherency providing mechanism is also provided. When an error is detected in the first cache coherency providing mechanism, this mechanism is disabled and cache coherency is achieved by the second cache coherency providing mechanism. In a preferred embodiment, the first mechanism includes coherency directories and the second mechanism includes bus snooping.

Multiprocessor computing apparatus with optional coherency directory

Logic for determining the average latency of pending pipelined and split bus transactions within a computer system including a bus, such as an Intel Pentium Pro or P6 bus, which supports pipelined and split bus transactions. The logic includes a first counter connected to the bus, and containing a TOTAL QUALIFIED CYCLES count value which is incremented on the start of every qualified bus cycle placed on the bus; logic for determining a cycle COUNT-BY-VALUE representing the number of outstanding or pending qualified bus cycles during any bus cycle; and a second counter which is incremented at the start of every qualified bus cycle occurring during the sample period by the number of outstanding qualified bus cycles to provide a TOTAL LATENCY CLOCKS count value. Divider logic is connected to receive the TOTAL QUALIFIED CYCLES count value from the first counter and the TOTAL LATENCY CLOCKS value from the second counter and divide the TOTAL QUALIFIED CYCLES count value into the TOTAL LATENCY CLOCK value to determine the average number of clocks of latency, or average number of pending bus cycles, per qualified bus cycle.

System for determining the average latency of pending pipelined or split

A mechanism for distributing interrupts to processors within a multi-processing system including a cache memory corresponding to each processor, a main memory, a bus structure connecting the processors and their associated cache memories with the main memory, and a cache coherency mechanism to maintain data consistency between the cache memories and the main memory. An address within the main memory is assigned to each processor within the system, the assigned address being associated with an interrupt for the processor to which it is assigned. For each processor, a copy of the contents of its assigned address is thereafter read into its corresponding cache memory. Thereafter when a cache coherency operation to update the contents or status of the cache memory address occurs, a comparison is made between the cache memory address presented to the cache memory through the system bus structure and a stored interrupt base address. An interrupt signal for the processor is generated when the comparison determines a match between the cache memory address presented to the cache memory through the system bus structure and a stored interrupt base address.

Method for distributing interprocessor interrupt requests via cache

A system and method of establishing error precedence in a computer system which determine a first error to occur. The system determines the order of occurrence of errors in a computer system and includes error precedence modules which record and order errors that occur on a first bus with error that occur on a second bus. Diagnostic processing circuitry reads the errors stored within the error precedence modules and their order of occurrence and determines which bus the first error occurred on.

System and method of establishing error precedence in a computer system

A mechanism for ensuring coherency between a system memory and a cache memory within a processing system including a first split transaction bus, the system memory being connected to the first split transaction bus; a second split transaction bus; a bus agent including the cache memory connected to the second split transaction bus, and a bus interface unit connecting the first and second split transaction busses for transferring bus cycles between the first and second split-transaction busses. The mechanism records bus cycles, such as read cycles, write cycles and cache line invalidate cycles, directed from the first split transaction bus to the second split transaction bus into a transaction queue within the bus interface unit, and sequentially transfers these cycles to the second split transaction bus in the order in which these cycles are recorded into the queue. In another embodiment of the invention, implemented in a system which utilizes a write post negation procedure for to indicate write completion, only read and invalidate cycles received from the first split transaction bus are placed within an ordering queue, while write cycles received from the first split transaction bus are immediately passed through to the second split transaction bus. However, when an invalidate cycle has been recorded to the queue, all write cycles received from the first split transaction bus are held until the queued invalidate cycle has started on the second split transaction bus.

Method for ensuring cycle ordering requirements within a hierarchical

A retry scheme for optimizing use of a first bus in a computer system which includes a plurality of bus masters connected through the first bus to an interface circuit and second bus. The interface circuit includes logic for generating a busy signal when the second bus is in a busy state and logic for generating a retry signal when the interface circuit is addressed by a bus master while the second bus is in a busy state. Each bus master includes logic for receiving the retry signal and relinquishing control of the common bus upon receipt of the retry signal from the interface circuit. A bus arbiter includes logic for receiving the busy signal and preventing any bus master seeking access to the second bus from participating in arbitration for control of the common first bus until the busy signal has been negated. Thus, during the term of the busy signal the first bus may be controlled by any bus master not requiring access to the shared resource. Upon negation of the busy signal, all bus masters will be permitted to compete for ownership of the bus.

Growing community of inventors

Baton Rouge, LA, United States of America

Edward A McDonald