Intel Corporation

Other

David G Carson

P K Nizar

David B Johnson

David L Budde

Craig B Peterson

Anthony L Cornish

Nitin V Sarangdhar

Brent S Baxter

George R Hayek

Brad W Hosler

Doran K Wilde

Colyn Case

Brian K Langendorf

Gary A Solomon

George W Cox

Kevin C Kahn

Norman J Rasmussen

Robert C Duzett

Kim A Meinerth

Herman D D'Hooge

Stephen R Colley

John L Wipfli

David Finzi

Yoav Hochberg

Adi Golbert

William L Bain, Jr

Dave Papworth

Robert P Voll

Scott A Ogilvie

A high throughput memory access port is provided. The port includes features which provide higher data transfer rates between system memory and video/graphics or audio adapters than is possible using standard local bus architectures, such as PCI or ISA. The port allows memory read and write requests to be pipelined in order to hide the effects of memory access latency. In particular, the port allows bus transactions to be performed in either a non-pipelined mode, such as provided by PCI, or in a pipelined mode. In the pipelined mode, one or more additional memory access requests are permitted to be inserted between a first memory access request and its corresponding data transfer. In contrast, in the non-pipelined mode, an additional memory access request cannot be inserted between a first memory access request and its corresponding data transfer.

High-throughput interconnect having pipelined and non-pipelined bus transaction modes

A high throughput memory access interface is provided. The interface includes features which provide higher data transfer rates between system memory and video/graphics or audio adapters than is possible using standard local bus architectures, such as PCI or ISA. The interface allows memory access requests to be performed in such a manner that only portions of an access request are required to be transmitted to the target device for certain bus transactions. Each access request includes command bits, address bits, and length bits. In the initiating device, each access request is separated into three segments, which are stored in separate registers in both the initiating device and the target device. Only the segment which contains the lowest order address bits and the length bits is required by the target device to initiate the bus transaction. Thus, if either of the other two segments has not changed since the previous access request, then such segment or segments are not transmitted to the target. If such segment or segments have changed since the previous access request, then they are provided to the target only for purposes of updating state in the target. Access requests may optionally be provided to the target on a separate port from the port used to transmit data in response to access requests.

High-throughput interconnect allowing bus transactions based on partial

A multi-processor system includes an interrupt bus used for arbitrating among eligible processors to determine which processor is to service of an interrupt request. The interrupt bus comprises wired-OR connection data lines that are used for arbitration. A local interrupt controller that handles the acceptance of interrupt request messages on the interrupt bus is associated with each processor. To minimize interruption of high priority tasks, interrupts can be accepted by the processor in the system that is currently running the lowest priority task. An arbitration protocol governs the interrupt bus and determines the lowest priority processor. The arbitration protocol includes choosing one among the lowest priority processors by means of a random priority scheme that uses an arbitration ID that is updated with each message.

Protocol for interrupt bus arbitration in a multi-processor system

A multi-processor programmable interrupt controller system that includes: an I/O interrupt controller for receiving interrupt requests from an I/O subsystem; multiple processor interrupt controllers, each associated with a specific processor for dispensing of accepted interrupts; and an interrupt controller bus primarily for the transmission of interrupt requests between interrupt controller units and for bus and priority arbitration, using a standard message format and arbitration protocol. The system is implemented, in part, by incorporating the processor interrupt controller with its associated processor into a single integrated circuit. The common system bus which normally carries all system traffic is not used for interrupt request messages. The interrupt controller bus is used for this purpose and thus results in a more efficient system by relieving the system bus of interrupt service requests and the related interrupt request traffic.

Multi-processor computer system with interrupt controllers providing

Programmable multi-processor interrupt controller system with a

A multiprocessor programmable interrupt controller system has an interrupt bus, distinct from the system (memory) bus, for handling interrupt request (IRQ) related messages. Each processor chip has an on-board interrupt acceptance unit (IAU) coupled to the interrupt bus to accept IRQs and to broadcast IRQs that it generates. I/O device interrupt lines are connected to one or more interrupt delivery units (IDUs) that are each coupled to the interrupt bus to broadcast I/O-generated IRQs. The interrupt bus is a synchronous three-wire bus having one clock wire and two wires for data transmission. Arbitration for control of the interrupt bus by the IAUs and IDUs uses one of the data wires. Lowest priority IRQ delivery mode uses a similar one-wire arbitration procedure for determining which IAU has the lowest current priority task running in its associated on-chip processor. A modification to this procedure also provides uniform distribution of IRQs to eligible processors. The actual servicing of the IRQs is done via the system bus. IAU acceptance logic is minimized by allowing retry of a delivered message when the acceptance latches are full. The increase in interrupt bus traffic due to retry is minimized by controlling the time intervals between rebroadcasts of unaccepted IRQs. Exponential timers control this interval so that each succeeding interval is a multiplicative factor, typically 2, greater than the preceding interval.

Multiprocessor programmable interrupt controller system with separate

A multiprocessor programmable interrupt controller (MPIC) system has an interrupt bus, distinct from the system (memory) bus, for handling interrupt-related messages. I/O device interrupt lines are connected to one or more interrupt delivery units (IDUs) that are each coupled to the interrupt bus for broadcasting of I/O-generated interrupt request messages. Each processor chip has an on-board interrupt acceptance unit (IAU) that can accept interrupt requests from the interrupt bus and can broadcast on the interrupt bus interrupt request messages generated by its associated processor. Each processor can request to read the contents of the IAU control registers that are associated with another target processor. In that case, a remote read request message is generated by the IAU of the local processor and responded to, without software intervention, by the IAU of the target processor. A remote read status field indicates to the local processor the status of the data contained in a remote read register. The remote IAU is expected to respond in a fixed number of interrupt bus cycles. If the remote agent is unable to do so, then the remote read status field becomes 'Invalid.' If successful, the remote read status resolves to 'Valid.' The processor polls this field to determine the completion and success of the remote read request. Remote read requests are always successful (although the data may be valid or invalid) in that they are never retried. Remote read requests are primarily a debug feature, and a 'hung' remote IAU that is unable to respond to a remote read request should not cause the debugging software to hang on the local processor.

Multiprocessor interrupt controller with remote reading of interrupt

A multiprocessor programmable interrupt controller system, for use in a multiprocessor system in which one processor unit is a functional redundant checking (FRC) unit, has a synchronous interrupt bus, distinct from the system (memory) bus, with an interrupt bus clock that has a frequency that is a subharmonic of the FRC unit master CPU clock, for handling interrupt request (IRQ) related messages and maintaining synchronism between the master and checker CPUs of the FRC unit. Additional embodiments provide for the use of D-type flip-flop synchronizers to accommodate FRC units whose internal (core) clock or external bus clock are not harmonically related to the interrupt clock frequency. Each processor unit has an interrupt acceptance unit (IAU) coupled to the interrupt bus for the acceptance of IRQs and for broadcasting of IRQs generated by its associated on-chip processor. I/O device interrupt lines are connected to one or more interrupt delivery units (IDUs) that are each coupled to the interrupt bus for broadcasting of I/O-generated IRQs. The interrupt bus is a synchronous three-wire bus having one clock wire and two data wires for 2-bit parallel-serial data transmission. Arbitration for control of the interrupt bus by the IAUs and IDUs uses one of the data wires. Lowest priority IRQ delivery mode uses a similar one-wire arbitration procedure for determining which IAU has the lowest current priority task running in its associated on-chip processor. A modification to the lowest priority mode arbitration procedure also provides for uniform distribution of IRQs to eligible processors. The actual servicing of the IRQs is done by means of the system bus.

Multiprocessor programmable interrupt controller system adapted to

A multi-processor programmable interrupt controller system which includes: an I/O interrupt controller for receiving interrupt requests from an I/O subsystem; multiple processor interrupt controllers, each associated with a specific processor for dispensing of accepted interrupts; and an interrupt controller bus primarily for the transmission of interrupt requests between interrupt controller units and for priority arbitration, using a standard message format and arbitration protocol.

Multi-processor programmable interrupt controller system

An I/O processor for controlling data transfer between a local bus and an I/O bus. An Execution Unit, an I/O bus sequencer, and a local bus sequencer are connected to a register file. The register file is uniformly addressed and each of the Execution Unit, the local bus sequencer, and the I/O bus sequencer have read/write access to the register file. The register file is comprised of a plurality of register sets. The Execution Unit includes a programmed processor which is programmed to allocate the register sets among tasks running on the processor by passing register-set descriptors between the tasks in the form of messages. The local bus sequencer includes a packet-oriented multiprocessor bus, there being a variable number of bytes in each of the packets. The I/O sequencer includes logic for multibyte sequencing of data at a bus-dependent data rate between the I/O bus and the register file. Each of the tasks includes a task frame, each task frame including register-set pointers. The register-set pointers map between logical addresses used in the instructions of the tasks used to access the pointers and physical register-set addresses used to access the register. Programmed logic in each of the Execution Unit, the local bus sequencer, and the I/O bus sequencer dynamically allocate the register sets to the sending and destination tasks.

Memory-based interagent communication mechanism

A number of intelligent nodes (bus interface units-BIUs and memory control units-MCUs) are provided in a matrix composed of processor buses (105) with corresponding error-reporting and control lines (106); and memory buses (107) with corresponding error-reporting and control lines (108). Error-detection mechanisms deal with information flow occuring across area boundaries. Each node (100, 101, 102, 103) has means for logging errors and reporting errors on the error report lines (106, 108). If an error recurs the node at which the error exists initiates an error message which is received and repropagated on the error report lines by all nodes. The error message identifies the type of error and the node ID at which the error was detected. Confinement area isolation logic in a node isolates a faulty confinement area of which the node is a part, upon the condition that the node ID in an error report message identifies the node as a node which is a part of a faulty confinement area. Logic in the node reconfigures at least part of the system upon the condition that the node ID in the error report message identifies the node as a node which is part of a confinement area which should be recofigured to recover from the error reported in the error report message.

Apparatus for recovery from failures in a multiprocessing system

A number of intelligent nodes (bus-interface units-BIUs and memory-control units-MCUs) are provided in a matrix composed of processor buses (105) with corresponding error-reporting and control lines (106); and memory buses (107) with corresponding error-reporting and control lines (108). Each node (100, 101, 102, 103) has means for logging errors and reporting errors on the error-report lines (106, 108). Processor modules (110) and memory modules (112) are each connected to a node which controls access to a common memory bus (107). Each node includes means (a married bit-170 and a shadow bit-172) for marrying modules in pairs such that each module in the pair tracks the operations directed to the module pair, and each module in the pair alternates with the other module in the handling of requests or replies. Each node registers the ID of the other node in a spouse ID register. Comparison logic (162, 164) in each node resets the married bit upon the condition that the node ID (identifying the node at which the error occurred) in an error-report message is equal to the ID stored in the spouse ID register, thus identifying the spouse node (the partner of the node in which the comparison logic is located) as the source of the error. Resetting the married bit splits apart the primary/shadow pair, so that the error-free module takes over and ceases to alternate with its partner.

Apparatus for redundant operation of modules in a multiprocessing system

A number of intelligent bus interface units (100) are provided in a matrix of orthogonal lines interconnecting processor modules (110) and memory control unit (MCU) modules (112). The matrix is composed of processor buses (105) and corresponding control lines; and memory buses (107) with corresponding control lines (108). At the intersection of these lines is a bus interface unit node (100). The bus interface units function to pass memory requests from a processor module to a memory module attached to an MCU node and to pass any data associated with the requests. The memory bus is a packet-oriented bus. Accesses are handled by means of a series of messages transmitted by message generator (417) in accordance with a specific control protocol. Packets comprising one or more bus transmission slots are issued sequentially and contiguously. Each slot in a packet includes an opcode, address, data, control, and parity-check bits. Write-request packets and read-request packets are issued to the memory-control unit. The memory-control unit responds with reply packets. A message controller (416), bus monitor (413), and pipeline and reply monitor ( 414), run the memory bus in a three-level pipeline mode. There may be three outstanding requests in the bus pipeline. Any further requests must wait for a reply message to free-up a slot in the pipeline before proceeding. Request messages increase the length of the pipeline and reply messages decrease the length of a pipeline. A control message, called a blurb, does not affect the pipeline length and can be issued when the pipeline is not full. The different messages are distinguished by three control signals (405) that parallel the data portion of the bus. The message generator (417) and interface logic (404) drive these control lines to indicate the message type, the start and end of the message, and possible error conditions. The pipeline and reply monitor (414) and the message controller (416) cooperate to insert a reply to a particular request in the pipeline position corresponding to the particular request that invoked the reply.

Interface for use between a memory and components of a module switching

An arbitration mechanism comprising a request FIFO (408) for storing ones and zeros corresponding to received requests in the order that they are made. A one indicates that the request was made by the module in which the FIFO is located, and a zero indicates that the request was made by one of a number of other similar modules. The request status information from the other modules is received over signal lines (411) connected between the modules. This logic separates multiple requests into time-ordered slots, such that all requests in a particular time slot may be serviced before any requests in the next time slot. A store (409) stores a unique logical module number. An arbiter (410) examines this logical number bit-by-bit in successive cycles and places a one in a grant queue (412) upon the condition that the bit examined in a particular cycle is a zero and signals this condition over the signal lines. If the bit examined in a particular cycle is a one, the arbiter drops out of contention and signals this condition over the signal lines (411). This logic orders multiple requests within a single time slot, which requests are made by multiple modules, in accordance with the logical module numbers of the modules making the requests. The grant queue (412) stores status information (ones and zeros) corresponding to granted requests in the order that they are granted--a one indicating that the granted request was granted to the module in which the grant queue is located, and a zero indicating that the granted request was granted to one of the other modules. The granted request status information from the other modules is received over the signal lines (411). This logic separates multiple granted requests such that only one request corresponding to a particular module is at the head of any one grant queue at any one time.

Growing community of inventors

Portland, OR, United States of America

David G Carson