Intel Corporation

Cisco Technology, Inc.

Cisco Systems Inc.

Manoj Khare

Akhilesh Kumar

Lily Pao Looi

Faye A Briggs

Kai Cheng

Kenneth C Creta

Sin Sim Tan

Judge K Arora

Kent G Fielden

Rajesh Patil

Ioannis T Schoinas

Harshvardhan P Sharangpani

Robert T George

Michel P Cekleov

Sudarshan Bala Cadambi

Ken Creta

Harshvardan Sharangpani

Judge Kennedy Arora

Judge Kennedy Cisco Systems Inc Arora

A method for reducing memory latency in a multi-node architecture. In one embodiment, a speculative read request is issued to a home node before results of a cache coherence protocol are determined. The home node initiates a read to memory to complete the speculative read request. Results of a cache coherence protocol may be determined by a coherence agent to resolve cache coherency after the speculative read request is issued.

Method and apparatus for reducing memory latency in a cache coherent multi-node architecture

A rule processor and method for using the same are disclosed. In one embodiment, the rule processor comprises a general purpose register file, an instruction sequencer to provide instructions, a decoder coupled to the general purpose register file to decode a set of instructions specified by the instruction sequencer, and a state machine unit coupled to the decoder and having state machine registers to store one or more state machines and state machine execution hardware coupled to the state machine registers to evaluate the one or more state machines in response to executing one or more of the set of instructions and based on information from one or both of the decoder and the general purpose register file.

Programmable processor apparatus integrating dedicated search registers and dedicated state machine registers with associated execution hardware to support rapid application of rulesets to data

A method and apparatus for a mechanism for handling explicit writeback in a cache coherent multi-node architecture is described. In one embodiment, the invention is a method. The method includes receiving a read request relating to a first line of data in a coherent memory system. The method further includes receiving a write request relating to the first line of data at about the same time as the read request is received. The method further includes detecting that the read request and the write request both relate to the first line. The method also includes determining which request of the read and write request should proceed first. Additionally, the method includes completing the request of the read and write request which should proceed first.

Mechanism for handling explicit writeback in a cache coherent multi-node architecture

An approach for pipelining ordered input/output transactions to coherent memory in a distributed memory, cache coherent, multi-processor system. A prefetch engine prefetches data from the distributed, coherent memory in response to a transaction from an input/output bus directed to the distributed, coherent memory. An input/output coherent cache buffer receives the prefetched data and is kept coherent with the distributed, coherent memory and with other caching agents in the system.

Method and apparatus for pipelining ordered input/output transactions to coherent memory in a distributed memory, cache coherent, multi-processor system

Embodiments of the invention provide a programmable FSA building block, having a number of programmable registers and associated logic implemented therein, that provide the capability of contextually evaluating complex REs of arbitrary size against multiple data streams. Embodiments of the invention provide fully programmable hardware in which all of the states of an RE are instantiated and all of the states are fully connected. For one embodiment, the building blocks have a fixed number of states to facilitate implementation on a chip. For such an embodiment, an RE having an excessive number of states is implemented on two or more FSA building blocks and the FSA building blocks are then stitched together to effect evaluation of the RE. For one embodiment, two or more REs having a number of states less than the fixed number of states of a building block may be implemented with a single building block.

Methods and apparatuses for evaluation of regular expressions of arbitrary size

A method and apparatus for a mechanism for handling i/o transactions with known transaction length to coherent memory in a cache coherent multi-node architecture is described. In one embodiment, the invention is a method. The method includes receiving a request for a current copy of a data line. The method further includes finding the data line within a cache-coherent multi-node system. The method also includes copying the data line without disturbing a state associated with the data line. The method also includes providing a copy of the data line in response to the request. The method also includes determining if the data line is a last data line of a transaction based on a known transaction length of the transaction.

Mechanism for handling I/O transactions with known transaction length to coherent memory in a cache coherent multi-node architecture

Embodiments of the present invention relate to methods and apparatus for managing transaction requests in a multi-node architecture. In one embodiment, a previously received ordered group request may be forwarded to a destination agent. Whether a next received ordered group request belongs to a same ordered group as the previously received ordered group request may be determined. Additionally, it may be determined whether an ordering fork is encountered if the next received ordered group request belongs to the same ordered group as the previously received ordered group request. If an ordering fork is encountered, it may be determined whether a request complete message for the previously received ordered group request has been received. If the request complete message for the previously received ordered group request has not been received and the next received ordered group request in the same ordered group is at least one of a un-ordered request and a forward-ordered request, then the next received ordered group request may be forwarded to the destination agent after the request complete message for the previously received at least one of a forward-ordered request and a sequential-ordered request issued on a different path at the ordering fork has been received.

Method and apparatus for managing transaction requests in a multi-node architecture

A method and apparatus are described for providing an implicit write-back in a distributed shared memory environment implementing a snoop based architecture. A requesting node submits a single read request to a snoop based architecture controller switch. The switch recognizes that another node other than the requesting node and the home node for the desired data has a copy of the data. The switch directs the request to the responding node that is not the home node. The responding node, having modified the data, provides a single response back to the switch that causes the switch to both update the data at the home node and answer the requesting node. The updating of the data at the home node is done without receiving an explicit write instruction from the requesting node.

Mechanism for initiating an implicit write-back in response to a read or snoop of a modified cache line

A method of sending messages from a node to a receiving agent. In one embodiment, if a outbound message that is stored in a buffer in the node is unsuccessfully sent to the receiving agent more than a threshold number of times, outbound messages currently stored in the buffer are sent to the receiving agent. It is determined that these outbound messages have been successfully sent before any other outbound messages are sent to the receiving agent. In a further embodiment, an outbound message is successfully sent if a success confirmation message is received for the outbound message from the receiving agent. In a still further embodiment, a retry response is received from the receiving agent for an outbound message if a buffer in the receiving agent that stores incoming outbound messages does not have room for the outbound message.

Method and apparatus for preventing starvation in a multi-node architecture

An example embodiment of a computer system utilizing a central snoop filter includes several nodes coupled together via a switching device. Each of the nodes may include several processors and caches as well as a block of system memory. All traffic from one node to another takes place through the switching device. The switching device includes a snoop filter that tracks cache line coherency information for all caches in the computer system. The snoop filter has enough entries to track the tags and state information for all entries in all caches in all of the system's nodes. In addition to the tag and state information, the snoop filter stores information indicating which of the nodes has a copy of each cache line. The snoop filter serves in part to keep snoop transactions from being performed at nodes that do not contain a copy of the subject cache line, thereby reducing system overhead, reducing traffic across the system interconnect busses, and reducing the amount of time required to perform snoop transactions.

Method and apparatus for centralized snoop filtering

A method of invalidating a cache line in a system having a plurality of nodes that include a processor and a cache memory. A request to invalidate a cache line that is caching a particular memory block is sent from a first node. The request is a request to invalidate a cache line in another node without returning to the first node the data stored in a cache line to be invalidated. In an embodiment, the data in the cache line to be invalidated is not returned to the first node even if the cache line is in the modified state. In a further embodiment, new data is written to a cache line in the first node that is caching the particular memory block without writing old data that was stored in that cache line back to a memory.

Method and apparatus for invalidating a cache line without data return in a multi-node architecture

According to one embodiment, a method is disclosed. The method includes receiving a first request from a first node in a multi-node computer system to invalidate a first cache line at a second node. The method also includes receiving a second request from the second node to invalidate the first cache line at the first node and detecting the concurrent requests at conflict detection circuitry.

Mechanism for handling conflicts in a multi-node computer architecture

According to one embodiment, a method is disclosed. The method comprises receiving a read request from a first node in a multi-node computer system to read data from a memory at a second node. Subsequently, a write request from a third node is received to write data to the memory at the second node. The read request and write request is detected at conflict detection circuitry. Finally, read data from the memory at the second node is transmitted to the first node.

Growing community of inventors

Saratoga, CA, United States of America

Manoj Khare