International Business Machines Corporation

Advanced Micro Devices Corporation

Lenovo (singapore) Pte. Ltd.

James O Nicholson

Ravi Kumar Arimilli

Sudhir Dhawan

David W Siegel

Frederick E Strietelmeier

Steven Mark Thurber

John Thomas O'Quin, Ii

John C O'Quin, Iii

Danny Marvin Neal

Chester A Heath

Bruce James Wilkie

Daniel N De Araujo

James T Hanna

George A Lerom

James D Reid

Richard Louis Arndt

Richard Allen Kelley

Edward John Silha

Douglas Craig Bossen

Charles Andrew McLaughlin

Albert Y Chang

Richard G Fogg, Jr

Daniel James Henderson

Joseph R Mathis

Amy M Youngs

John Hughes Rost

A method and device for error detection includes performing error detection for each data word received in a burst access to a memory. When no error is detected, the data words are written to a cache and indicated as valid data. In response to detecting an error in a data word, the error is corrected and the corrected data written to the cache without indicating the data as valid. In addition, the location of the detected error, indicating the data symbol associated with the error, is recorded in an error vector. The error vectors associated with each data word in the burst access are compared to determine whether a detected error was properly corrected.

Data error correction device and methods thereof

Methods, apparatus, and computer program products are disclosed for clock signal synchronization among computers in a network, including designating, as a primary clock signal for all the computers in a network, a clock signal from one of the computers in the network; providing the primary clock signal, simultaneously and in parallel, from the computer whose clock signal is designated as the primary clock signal to all the other computers in the network; and providing the primary clock signal, simultaneously and in parallel, from each computer in the network to all computers in the network through multiplexers and phase locked loops, with the primary clock signal locked in phase across all the computers by a phase locked loop on each computer.

Clock signal synchronization among computers in a network

Memory bank pairs are sorted utilizing variables determined by a scoring criteria. The scores for the variables are based on the number of memory blocks in a memory bank that are filled; the total number of memory bank pairs that are identically populated (same set of blocks per bank) and the number of memory bank pairs that either match or are close to a standard interleave value. Sort schemes are determined by the values of each variable. A first sort scheme is attempted and after the sort scheme is complete, if all possible configurable banks are not configured, the banks are marked un-configured and another sort scheme is tried. Each sort scheme, utilizing a maximum of four schemes, is attempted until a method is found that configures all possible configurable bank pairs. Sorting is done for up to three levels, i.e., all bank pairs are sorted according to a first value, then all bank pairs with equal values are sorted according to a second value. Then the resulting matching bank pairs are sorted according to the third value. This method reduces the time required to determine a minimum and sufficient number of interleave registers required to configure all possible memory configurations in a given mixed interleave memory.

Method for optimally configuring memory in a mixed interleave system

Device selects lines from each I/O device are brought into a PCI host bridge individually so that the device number of a failing device may be logged in an error register when an error is seen on the PCI bus. Until the error register is reset, subsequent load and store operations are delayed until the device number of the subject device may be checked against the error register. If the subject device is a previously failing device, the load/store operation to that device is prevented from completing, either by forcing bad parity or zeroing all byte enables. By forcing bad parity of zero byte enables, the I/O device will respond to the load or store request by activating its device select line, but will not accept store data. Operations to devices which are not logged in the error register are permitted to proceed normally, as are all load store operations when the error register is clear. Normal system operations are thus not impacted, and operations during error recovery are permitted to proceed if no further damage will be caused by such operations.

Enhanced error handling for I/O load/store operations to a PCI device via bad parity or zero byte enables

A method and apparatus are provided in which a control scheme is implemented to enable a PCI bus to operate more than two PCI slots into which PCI devices may be installed. The PCI slots are checked to determine if a PCI device is installed in the slots and the speed at which the installed PCI devices are capable of running. If any of the slots has a 33 MHz device installed in any of the slots, the system is enabled to run more than two slots, and all of the PCI devices will run at 33 MHz. When no 33 MHz cards or devices are installed in the PCI slots, and PCI devices are only installed in the first two slots, then the system is enabled to run only the first two slots at the speed of 66 MHz. In one alternative embodiment, a default configuration routine sets the PCI bus speed at one of the operating frequencies and modifies that default if it is determined during a system configuration cycle that another speed is more appropriate.

Variable slot configuration for multi-speed bus

An interrupt subsystem within a data processing system is scalable from low-end uni-processor systems to high-end multi-processor (MP) systems. This interrupt subsystem provides for queueing of interrupts from many sources, and for queueing of interrupts to the best processor in a multi-processor system. The external interrupt mechanism is separated into two layers, an interrupt routing layer and an interrupt presentation layer. The interrupt routing layer routes the interrupt conditions to the appropriate instance of an interrupt management area within the interrupt presentation layer. The interrupt presentation layer communicates the interrupt source to the system software which is to service/process the interrupt. By providing two layers within the interrupt subsystem, application or system software can be written which is independent from the types or sources of interrupts. The interrupt routing layer hides the details of a particular hardware implementation from the software. The interrupt presentation layer interfaces to the system and/or application software, and provides hardware independent functionality.

Scalable system interrupt structure for a multi-processing system

A computer system having a plurality of devices which transmit and receive information over a channel is presented. The system includes, in the preferred embodiment, a central arbitration control circuit and a local arbiter associated with each device contending for channel access. Each local arbiter, corresponding to a device which desires channel access, generates a channel request signal to the central control circuit. At the appropriate time when the channel becomes available, the central control circuit generates an arbitrate signal. All local arbiters, then contending for channel access, compare the priority level on the arbitration bus with the priority value of the device it is arbitrating on behalf of, with the winning device gaining access to the channel. Each of the local arbiters contains a programmable circuit which enables the arbiter to operate either utilizing a linear priority arbitration technique or a fairness priority arbitration technique. Thus, each device is dynamically programmable to operate in a linear mode or a fairness mode depending on user and/or application needs. Additionally, each local arbiter includes a circuit which detects whether the device is a burst transfer device or a single transfer device.

Computer system having dynamically programmable linear/fairness priority

A cache for use with input/output devices attached to an input/output bus. Requests for access to system memory by an input/output device pass through the cache. Access authority is checked to determine whether an input/output device is authorized to access that particular page. If it is not, access is denied. Each input/output device has access to a portion of the cache, so that activity by one device will not interfere with activity by another.

I/O cache controller containing a buffer memory partitioned into lines

A computer system has a cache located between input/output devices and a main system memory. All system memory accesses by the input/output devices are made through the cache. Memory accesses through the cache are limited to those addresses which are accessible to a central processor and input/output devices. All access to such addresses by the central processor are made through the cache.

Computer system with input/output cache

A DMA controller coupled to two separate buses controls the transfer of data between them. To effect a block data transfer, data is simultaneously read on one bus and written on the other. This allows data to be transferred between buses at the maximum transfer rate supported by the slower bus.

Computer system DMA transfer

A cache for use with input/output devices attached to an input/output bus. Requests for access to system memory by an input/output device pass through the cache. Virtual memory addresses used by the input/output devices are translated into real addresses in the system memory. Virtual memory can be partitioned, with some virtual addresses being mapped to a second memory attached to the input/output bus.

Input/output cache

A computer system can transfer data between a master subsystem and a slave subsystem on bus address lines as well as bus data lines during a high speed data transfer. Data is clocked during the high speed transfer by a high speed clock signal which is separate from a normal bus clock signal. Data is transferred at the maximum rate which can be handled by both the master subsystem and the slave subsystem.

Data transfer using bus address lines

A DMA controller has an attached, dedicated memory. Data objects are stored on the heap and connected by pointers. Each data object contains DMA block transfer control parameters. A single block transfer made up of several separate transfers, with each separate transfer defined by one data object. The single block transfer is defined by linking several data objects into a list. The DMA controller consecutively performs the transfers in a linked list without requiring control by a system central processor.

Growing community of inventors

Austin, TX, United States of America

James O Nicholson