Abstract: In a computing system, cache coherency is performed by selecting one of a plurality of coherency protocols for a first memory transaction. Cache coherency is performed on appropriate caches in the computing system in accordance with the selected one of the plurality of coherency protocols. For a second memory transaction, another selection is made of the plurality of coherency protocols. The selected one of the coherency protocols for the second memory transaction may be the same as or different from the selected one of the plurality of coherency protocols for the first memory transaction.

Abstract: A system and method for interconnecting a plurality of processing element nodes within a scalable multiprocessor system is provided. Each processing element node includes at least one processor and memory. A scalable interconnect network includes physical communication links interconnecting the processing element nodes in a cluster. A first set of routers in the scalable interconnect network route messages between the plurality of processing element nodes. One or more metarouters in the scalable interconnect network route messages between the first set of routers so that each one of the routers in a first cluster is connected to all other clusters through one or more metarouters.

Abstract: Improved method and apparatus for parallel processing. One embodiment provides a multiprocessor computer system that includes a first and second node controller, a number of processors being connected to each node controller, a memory connected to each controller, a first input/output system connected to the first node controller, and a communications network connected between the node controllers. The first node controller includes: a crossbar unit to which are connected a memory port, an input/output port, a network port, and a plurality of independent processor ports. A first and a second processor port connected between the crossbar unit and a first subset and a second subset, respectively, of the processors. In some embodiments of the system, the first node controller is fabricated onto a single integrated-circuit chip.

Abstract: Improved method and apparatus for parallel processing. One embodiment provides a multiprocessor computer system that includes a first and second node controller, a number of processors being connected to each node controller, a memory connected to each controller, a first input/output system connected to the first node controller, and a communications network connected between the node controllers. The first node controller includes: a crossbar unit to which are connected a memory port, an input/output port, a network port, and a plurality of independent processor ports. A first and a second processor port connected between the crossbar unit and a first subset and a second subset, respectively, of the processors. In some embodiments of the system, the first node controller is fabricated onto a single integrated-circuit chip.

Abstract: A processor may operate in one of a plurality of operating states. In a Normal operating state, the processor is not involved with a memory transaction. Upon receipt of a transaction instruction to access a memory location, the processor transitions to a Transaction operating state. In the Transaction operating state, the processor performs changes to a cache line and data associated with the memory location. While in the Transaction operating state, any changes to the data and the cache line is not visible to other processors in the computing system. These changes become visible upon the processor entering a Commit operating state in response to receipt of a commit instruction. After changes become visible, the processor returns to the Normal operating state. If an abort event occurs prior to receipt of the commit instruction, the processor transitions to an Abort operating state where any changes to the data and cache line are discarded.

Abstract: A computer system (10) includes a bus controller (12), a bus (14), a plurality of processing devices (16) and a plurality of enabling switches (18). Each enabling switch (18) corresponds to a separate one of the processing devices (16). Each processing device (16) sends an access request (24) to arbitration logic (22) in the bus controller (12), requesting access to the bus (14). The arbitration logic (22) selects one of the access requests (24) according to a priority protocol. The arbitration logic (22) generates a control signal (20) associated with the selected access request (24). The control signal (20) is provided to the enabling switch (18) corresponding to the processing device (16) that sent the selected access request (24). The enabling switch (18) enables access to the bus (14) for the processing device (16) in response to the control signal (20).

Abstract: An address translation unit generates a physical address for access to a memory from a virtual address using either a translation lookaside buffer or a segmentation buffer. If the virtual address falls within a predetermined range, the address translation unit will use the segmentation buffer to generate the physical address. Upon generation of the physical address, the memory will either receive data from or provide data to a processor in accordance with the instructions being processed by the processor.

Abstract: A system and method for interconnecting a plurality of processing element nodes within a scalable multiprocessor system is provided. Each processing element node includes at least one processor and memory. A scalable interconnect network includes physical communication links interconnecting the processing element nodes in a cluster. A first set of routers in the scalable interconnect network route messages between the plurality of processing element nodes. One or more metarouters in the scalable interconnect network route messages between the first set of routers so that each one of the routers in a first cluster is connected to all other clusters through one or more metarouters.

Abstract: The present invention is directed to a method and a system for maintaining cache coherence in a distributed shared memory (DSM) multiprocessor system. The method begins with a receiving of a shared access request by a receiving node, where the receiving node is an arbitrary node having at least one main memory unit containing information desired to be accessed. Then, the method determines whether the shared access request originates from a local node or from a remote node. When the shared access request originates from a local node, the shared access request is processed as a shared access request. If the shared access request is granted, a sharing vector is generated or updated to reflect the sharing local node(s). When the shared access request originates from a remote node, the shared access request is converted to an exclusive access request and the sharing vector is replaced with a pointer to the requesting remote node. This limits the potential size of the sharing vector according to the local nodes.

Abstract: A modular computer system includes at least two processing functional modules each including a processing unit adapted to process data and adapted to input/output data to other functional modules through at least two ports with each port including a plurality of data lines. At least one routing functional module is adapted to route data and adapted to input/output data to other functional modules through at least two ports with each port including a plurality of data lines. At least one input or output functional module is adapted to input or output data and adapted to input/output data to other functional modules through at least one port including a plurality of data lines. Each processing, routing and input or output functional module includes a local controller adapted to control the local operation of the associated functional module, wherein the local controller is adapted to input and output control information over control lines connected to the respective ports of its functional module.

Abstract: A distributed, shared memory computer architecture that is organized into a set of functionally independent processing nodes operating in a global, shared address space. Each node has one or more local processors, local memory and includes a common communication interface for communicating with other modules within the system via a message protocol. The common communication interface provides a single high-speed communications center within each node to operatively couple the node to one or more external processing nodes, an external routing module, an input/output (I/O) module. The common communication interface that facilitates the ability to incrementally add and swap the nodes of the system without disrupting the overall computing resources of the system.

Abstract: Improved method and apparatus for parallel processing. One embodiment provides a multiprocessor computer system that includes a first and second node controller, a number of processors being connected to each node controller, a memory connected to each controller, a first input/output system connected to the first node controller, and a communications network connected between the node controllers. The first node controller includes: a crossbar unit to which are connected a memory port, an input/output port, a network port, and a plurality of independent processor ports. A first and a second processor port connected between the crossbar unit and a first subset and a second subset, respectively, of the processors. In some embodiments of the system, the first node controller is fabricated onto a single integrated-circuit chip.

Abstract: The present invention is directed to a method and a system for maintaining cache coherence in a distributed shared memory (DSM) multiprocessor system. The method begins with a receiving of a shared access request by a receiving node, where the receiving node is an arbitrary node having at least one main memory unit containing information desired to be accessed. Then, the method determines whether the shared access request originates from a local node or from a remote node. When the shared access request originates from a local node, the shared access request is processed as a shared access request. If the shared access request is granted, a sharing vector is generated or updated to reflect the sharing local node(s). When the shared access request originates from a remote node, the shared access request is converted to an exclusive access request and the sharing vector is replaced with a pointer to the requesting remote node. This limits the potential size of the sharing vector according to the local nodes.

Abstract: A node controller (12) in a computer system (10) includes a processor interface unit (24), a memory directory interface unit (22), and a local block unit (28). In response to a memory location in a memory (17) associated with the memory directory interface unit (22) being altered, the processor interface unit (24) generates an invalidation request for transfer to the memory directory interface unit (22). The memory directory interface unit (22) provides the invalidation request and identities of processors (16) affected by the invalidation request to the local block unit (28). The local block unit (28) determines which ones of the identified processors (16) are present in the computer system (10) and generates an invalidation message for each present processor (16) for transfer thereto. Each of the present processors (16) process their invalidation message and generate an acknowledgment message for transfer to the processor interface unit (24) that generated the invalidation request.

Abstract: A distributed shared memory multiprocessor computer system is provided, which has a number of processors and is divided into partitions. Each partition has within it one or more of the processors, and may also have memory or cache and other related hardware. Although each partition works together and communicates with other partitions to share computational load, the partitions each are independently operable and execute an independent copy of the operating system. The partitions comprise additional features to enable removal of a partition from the operating computer system, and to enable insertion of hardware into the operating computer system.

Abstract: A node controller (12) in a computer system (10) includes a processor interface unit (24), a memory directory interface unit (22), and a local block unit (28). In response to a memory location in a memory (17) associated with the memory directory interface unit (22) being altered, the processor interface unit (24) generates an invalidation request for transfer to the memory directory interface unit (22). The memory directory interface unit (22) provides the invalidation request and identities of processors (16) affected by the invalidation request to the local block unit (28). The local block unit (28) determines which ones of the identified processors (16) are present in the computer system (10) and generates an invalidation message for each present processor (16) for transfer thereto. Each of the present processors (16) process their invalidation message and generate an acknowledgment message for transfer to the processor interface unit (24) that generated the invalidation request.

Abstract: A node controller (12) in a computer system (10) includes a processor interface unit (24), a memory directory interface unit (22), and a local block unit (28). In response to a memory location in a memory (17) associated with the memory directory interface unit (22) being altered, the processor interface unit (24) generates an invalidation request for transfer to the memory directory interface unit (22). The memory directory interface unit (22) provides the invalidation request and identities of processors (16) affected by the invalidation request to the local block unit (28). The local block unit (28) determines which ones of the identified processors (16) are present in the computer system (10) and generates an invalidation message for each present processor (16) for transfer thereto. Each of the present processors (16) process their invalidation message and generate an acknowledgment message for transfer to the processor interface unit (24) that generated the invalidation request.

Abstract: A memory system that includes switches for controlling data transfer that are disposed on the motherboard. The switches are selectively coupled to a controller and to connector receptacles that are adapted to receive a memory module. The memory system also includes resistors that are disposed on the motherboard for terminating data signals. In one embodiment, memory modules are accessed in pairs. That is, the data switches are used to control the flow of data signals such that data signals only flow to one pair of memory modules at any particular time. In one embodiment, the memory system of the present invention includes eight memory modules that use DDR SDRAM memory components. When 8 Mbit, 16 Mbit, 32 Mbit or 64 Mbit memory components are used, this configuration gives a range of memory configurations from 128 megabytes (Mbytes) to 1 gigabyte (Gbyte).

Abstract: A memory system that includes switches for controlling data transfer that are disposed on the motherboard. The switches are selectively coupled to a controller and to connector receptacles that are adapted to receive a memory module. The memory system also includes resistors that are disposed on the motherboard for terminating data signals. In one embodiment, memory modules are accessed in pairs. That is, the data switches are used to control the flow of data signals such that data signals only flow to one pair of memory modules at any particular time. In one embodiment, the memory system of the present invention includes eight memory modules that use DDR SDRAM memory components. When 8 Mbit, 16 Mbit, 32 Mbit or 64 Mbit memory components are used, this configuration gives a range of memory configurations from 128 megabytes (Mbytes) to 1 gigabyte (Gbyte).

Abstract: A multiprocessing system that uses read resources to track cache coherent split transactions on its main system bus. Pending reads are tracked by being associated with read resources. When a read request is issued, it occupies the first available read resource. A pending read request will occupy a read resource until a corresponding read response appears on the bus. If all read resources are filled, future read requestors must wail until a read resource becomes available.