WO2004042585A2 - Methods and apparatus for managing probe requests - Google Patents

Methods and apparatus for managing probe requests Download PDF

Info

Publication number
WO2004042585A2
WO2004042585A2 PCT/US2003/034756 US0334756W WO2004042585A2 WO 2004042585 A2 WO2004042585 A2 WO 2004042585A2 US 0334756 W US0334756 W US 0334756W WO 2004042585 A2 WO2004042585 A2 WO 2004042585A2
Authority
WO
WIPO (PCT)
Prior art keywords
cluster
request
probe
home
probe request
Prior art date
Application number
PCT/US2003/034756
Other languages
French (fr)
Other versions
WO2004042585A3 (en
Inventor
David B. Glasco
Original Assignee
Newisys, Inc.
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Priority claimed from US10/288,399 external-priority patent/US7103726B2/en
Priority claimed from US10/288,347 external-priority patent/US7003633B2/en
Application filed by Newisys, Inc. filed Critical Newisys, Inc.
Priority to AU2003287415A priority Critical patent/AU2003287415A1/en
Priority to EP03781650A priority patent/EP1563387A2/en
Priority to JP2004550375A priority patent/JP2006505855A/en
Priority to CA002505016A priority patent/CA2505016A1/en
Publication of WO2004042585A2 publication Critical patent/WO2004042585A2/en
Publication of WO2004042585A3 publication Critical patent/WO2004042585A3/en

Links

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F12/00Accessing, addressing or allocating within memory systems or architectures
    • G06F12/02Addressing or allocation; Relocation
    • G06F12/08Addressing or allocation; Relocation in hierarchically structured memory systems, e.g. virtual memory systems
    • G06F12/0802Addressing of a memory level in which the access to the desired data or data block requires associative addressing means, e.g. caches
    • G06F12/0806Multiuser, multiprocessor or multiprocessing cache systems
    • G06F12/0815Cache consistency protocols
    • G06F12/0817Cache consistency protocols using directory methods

Definitions

  • the present invention generally relates to accessing data in a multiple processor system. More specifically, the present invention provides techniques for improving data access efficiency while maintaining cache coherency in a multiple processor system having a multiple cluster architecture.
  • Data access in multiple processor systems can raise issues relating to cache coherency.
  • Conventional multiple processor computer systems have processors coupled to a system memory through a shared bus.
  • individual processors are typically designed to work with cache memory.
  • each processor has a cache that is loaded with data that the processor frequently accesses. The cache is read or written by a processor.
  • cache coherency problems arise because multiple copies of the same data can co-exist in systems having multiple processors and multiple cache memories. For example, a frequently accessed data block corresponding to a memory line may be loaded into the cache of two different processors.
  • both processors attempt to write new values into the data block at the same time, different data values may result.
  • One value may be written into the first cache while a different value is written into the second cache. A system might then be unable to determine what value to write through to system memory.
  • cache coherency mechanisms have been developed to address such problems in multiprocessor systems.
  • One solution is to simply force all processor writes to go through to memory immediately and bypass the associated cache. The write requests can then be serialized before overwriting a system memory line.
  • bypassing the cache significantly decreases efficiency gained by using a cache.
  • Other cache coherency mechanisms have been developed for specific architectures.
  • each processor checks or snoops on the bus to determine whether it can read or write a shared cache block.
  • a processor only writes an object when it owns or has exclusive access to the object.
  • Each corresponding cache object is then updated to allow processors access to the most recent version of the object.
  • Bus arbitration is used when both processors attempt to write the same shared data block in the same clock cycle.
  • Bus arbitration logic decides which processor gets the bus first.
  • cache coherency mechanisms such as bus arbitration are effective, using a shared bus limits the number of processors that can be implemented in a single system with a single memory space.
  • Other multiprocessor schemes involve individual processor, cache, and memory systems connected to other processors, cache, and memory systems using a network backbone such as Ethernet or Token Ring.
  • Multiprocessor schemes involving separate computer systems each with its own address space can avoid many cache coherency problems because each processor has its own associated memory and cache.
  • communication is explicit. Messages are sent to move data to another processor and messages are received to accept data from another processor using standard network protocols such as TCP/IP.
  • Multiprocessor systems using explicit communication including transactions such as sends and receives are referred to as systems using multiple private memories.
  • multiprocessor system using implicit communication including transactions such as loads and stores are referred to herein as using a single address space.
  • Multiprocessor schemes using separate computer systems allow more processors to be interconnected while minimizing cache coherency problems. However, it would take substantially more time to access data held by a remote processor using a network infrastructure than it would take to access data held by a processor coupled to a system bus. Furthermore, valuable network bandwidth would be consumed moving data to the proper processors. This can negatively impact both processor and network performance.
  • Performance limitations have led to the development of a point-to-point architecture for connecting processors in a system with a single memory space.
  • individual processors can be directly connected to each other through a plurality of point-to-point links to form a cluster of processors. Separate clusters of processors can also be connected.
  • the point-to-point links significantly increase the bandwidth for coprocessing and multiprocessing functions.
  • using a point- to-point architecture to connect multiple processors in a multiple cluster system sharing a single memory space presents its own problems.
  • probe filter information is used to limit the number of probe requests transmitted to request and remote clusters.
  • a computer system in one embodiment, includes a home cluster having a first plurality of processors and a home cache coherence controller. The first plurality of processors and the home cache coherence controller are interconnected in a point-to-point architecture. The home cache coherence controller is configured to receive a probe request and probe one or more selected clusters. The one or more clusters are selected based on the characteristics associated with the probe request.
  • a method for managing probes is provided. A probe request is received at a home cache coherence controller in a home cluster. The home cluster includes a first plurality of processors and the home cache coherence controller. The first plurality of processors and the home cache coherence controller are interconnected in a point-to-point architecture. One or more clusters are selected for probing based on the characteristics associated with the probe request. The one or more clusters are probed.
  • Figure 1A and IB are diagrammatic representation depicting a system having multiple clusters.
  • Figure 2 is a diagrammatic representation of a cluster having a plurality of processors.
  • Figure 3 is a diagrammatic representation of a cache coherence controller.
  • Figure 4 is a diagrammatic representation showing a transaction flow for a data access request from a processor in a single cluster.
  • Figure 5A-5D are diagrammatic representations showing cache coherence controller functionality.
  • Figure 6 is a diagrammatic representation depicting a transaction flow for a probe request with multiple probe responses.
  • Figure 7 is a diagrammatic representation showing a cache coherence directory.
  • Figure 8 is a diagrammatic representation showing probe filter information that can be used to reduce the number of probes transmitted to various clusters.
  • Figure 9 is a diagrammatic representation showing a transaction flow for probing of a home cluster without probing of other clusters.
  • Figure 10 is a diagrammatic representation showing a transaction flow for probing of a single remote cluster.
  • Figure 11 is a flow process diagram showing the handling of a probe request with probe filter information.
  • Figure 12 is a diagrammatic representation showing memory controller filter information.
  • Figure 13 is a diagrammatic representation showing a transaction flow for probing a single remote cluster without probing a home cluster.
  • Figure 14 is a flow process diagram showing the handling of a probe request at a home cluster cache coherence controller using memory controller filter information.
  • DETAILED DESCRIPTION OF SPECIFIC EMBODIMENTS Reference will now be made in detail to some specific embodiments of the invention including the best modes contemplated by the inventors for carrying out the invention. Examples of these specific embodiments are illustrated in the accompanying drawings. While the invention is described in conjunction with these specific embodiments, it will be understood that it is not intended to limit the invention to the described embodiments.
  • Multi-processor architectures having point-to-point communication among their processors are suitable for implementing specific embodiments of the present invention.
  • numerous specific details are set forth in order to provide a thorough understanding of the present invention. The present invention may be practiced without some or all of these specific details. Well-known process operations have not been described in detail in order not to unnecessarily obscure the present invention.
  • the present application's reference to a particular singular entity includes that possibility that the methods and apparatus of the present invention can be implemented using more than one entity, unless the context clearly dictates otherwise.
  • a cluster of processors includes multiple processors directly connected to each other through point-to-point links.
  • point-to-point links instead of a conventional shared bus or external network, multiple processors are used efficiently in a system sharing the same memory space. Processing and network efficiency are also improved by avoiding many of the bandwidth and latency limitations of conventional bus and external network based multiprocessor architectures. According to various embodiments, however, linearly increasing the number of processors in a point-to- point architecture leads to an exponential increase in the number of links used to connect the multiple processors.
  • multiple clusters are used.
  • the multiple processor clusters are interconnected using a point-to-point architecture.
  • Each cluster of processors includes a cache coherence controller used to handle communications between clusters.
  • the point-to-point architecture used to connect processors are used to connect clusters as well.
  • multiple cluster systems can be built using processors that may not necessarily support multiple clusters.
  • Such a multiple cluster system can be built by using a cache coherence controller to represent nonlocal nodes in local transactions so that local nodes do not need to be aware of the existence of nodes outside of the local cluster. More detail on the cache coherence controller will be provided below.
  • serialization point Any mechanism for ordering data access requests is referred to herein as a serialization point.
  • a serialization point is a memory controller.
  • Various processors in the single cluster system send data access requests to the memory controller.
  • the memory controller is configured to serialize or lock the data access requests so that only one data access request for a given memory line is allowed at any particular time. If another processor attempts to access the same memory line, the data access attempt is blocked until the memory line is unlocked.
  • the memory controller allows cache coherency to be maintained in a multiple processor, single cluster system.
  • a serialization point can also be used in a multiple processor, multiple cluster system where the processors in the various clusters share a single address space. By using a single address space, internal point-to-point links can be used to significantly improve intercluster communication over traditional external network based multiple cluster systems.
  • Various processors in various clusters send data access requests to a memory controller associated with a particular cluster such as a home cluster. The memory controller can similarly serialize all data requests from the different clusters.
  • a serialization point in a multiple processor, multiple cluster system may not be as efficient as a serialization point in a multiple processor, single cluster system. That is, delay resulting from factors such as latency from transmitting between clusters can adversely affect the response times for various data access requests.
  • delay also results from the use of probes in a multiple processor environment. Although delay in intercluster transactions in an architecture using a shared memory space is significantly less than the delay in conventional message passing environments using external networks such as Ethernet or Token Ring, even minimal delay is a significant factor. In some applications, there may be millions of data access requests from a processor in a fraction of a second. Any delay can adversely impact processor performance.
  • probe management is used to increase the efficiency of accessing data in a multiple processor, multiple cluster system.
  • a mechanism for eliciting a response from a node to maintain cache coherency in a system is referred to herein as a probe.
  • a mechanism for snooping a cache is referred to as a probe.
  • a response to a probe can be directed to the source or target of the initiating request. Any mechanism for filtering or reducing the number of probes and probe requests transmitted to various nodes is referred to herein as managing probes.
  • managing probe entails characterizing a probe request to determine if a probe can be transmitted to a reduced number of entities.
  • probe requests are sent to a memory controller that broadcasts probes to various nodes in a system.
  • no knowledge of the cache line state is known. All nodes in the system are probed and the request cluster receives a response from each node.
  • state information associated with various memory lines can be used to reduce the number of transactions. Any mechanism for maintaining state information associated with various memory lines is referred to herein as a coherence directory.
  • a coherence directory typically includes information for memory lines in a local cluster that are cached in a remote cluster.
  • a coherence directory is used to reduce the number of probes to remote quads by inferring the state of local caches.
  • a coherence directory is used to eliminate the transmission of a request to a memory controller in a home cluster.
  • FIG 1A is a diagrammatic representation of one example of a multiple cluster, multiple processor system that can use the techniques of the present invention.
  • Each processing cluster 101, 103, 105, and 107 can include a plurality of processors.
  • the processing clusters 101, 103, 105, and 107 are connected to each other through point-to-point links l l la-f.
  • the multiple processors in the multiple cluster architecture shown in Figure 1A share the same memory space.
  • the point-to-point links l l la-f are internal system connections that are used in place of a traditional front-side bus to connect the multiple processors in the multiple clusters 101, 103, 105, and 107.
  • the point-to-point links may support any point-to-point coherence protocol.
  • FIG. IB is a diagrammatic representation of another example of a multiple cluster, multiple processor system that can use the techniques of the present invention.
  • Each processing cluster 121, 123, 125, and 127 can be coupled to a switch 131 through point-to-point links 141a-d. It should be noted that using a switch and point- to-point links allows implementation with fewer point-to-point links when connecting multiple clusters in the system.
  • a switch 131 can include a processor with a coherence protocol interface. According to various implementations, a multicluster system shown in Figure 1A is expanded using a switch 131 as shown in Figure IB.
  • FIG. 2 is a diagrammatic representation of a multiple processor cluster, such as the cluster 101 shown in Figure 1A.
  • Cluster 200 includes processors 202a-202d, one or more Basic I/O systems (BIOS) 204, a memory subsystem comprising memory banks 206a-206d, point-to-point communication links 208a-208e, and a service processor 212.
  • the point-to-point communication links are configured to allow interconnections between processors 202a-202d, I/O switch 210, and cache coherence controller 230.
  • the service processor 212 is configured to allow communications with processors 202a-202d, I/O switch 210, and cache coherence controller 230 via a JTAG interface represented in Fig. 2 by links 214a-214f. It should be noted that other interfaces are supported. It should also be noted that in some implementations, a service processor is not included in multiple processor clusters.
  • I/O switch 210 connects the rest of the system to I/O adapters 216
  • the service processor of the present invention has the intelligence to partition system resources according to a previously specified partitioning schema.
  • the partitioning can be achieved through direct manipulation of routing tables associated with the system processors by the service processor which is made possible by the point-to-point communication infrastructure.
  • the routing tables are used to control and isolate various system resources, the connections between which are defined therein.
  • the processors 202a-d are also coupled to a cache coherence controller 230 through point-to-point links 232a-d. Any mechanism or apparatus that can be used to provide communication between multiple processor clusters while maintaining cache coherence is referred to herein as a cache coherence controller.
  • the cache coherence controller 230 can be coupled to cache coherence controllers associated with other multiprocessor clusters. It should be noted that there can be more than one cache coherence controller in one cluster.
  • the cache coherence controller 230 communicates with both processors 202a-d as well as remote clusters using a point- to-point protocol.
  • memory banks 206a-206d may comprise double data rate (DDR) memory which is physically provided as dual in-line memory modules (DIMMs).
  • I/O adapter 216 may be, for example, an ultra direct memory access (UDMA) controller or a small computer system interface (SCSI) controller which provides access to a permanent storage device.
  • I/O adapter 220 may be an Ethernet card adapted to provide communications with a network such as, for example, a local area network (LAN) or the Internet.
  • both of I/O adapters 216 and 220 provide symmetric I/O access. That is, each provides access to equivalent sets of I/O. As will be understood, such a configuration would facilitate a partitioning scheme in which multiple partitions have access to the same types of I/O. However, it should also be understood that embodiments are envisioned in which partitions without I/O are created. For example, a partition including one or more processors and associated memory resources, i.e., a memory complex, could be created for the purpose of testing the memory complex.
  • service processor 212 is a Motorola MPC855T microprocessor which includes integrated chipset functions.
  • the cache coherence controller 230 is an Application Specific Integrated Circuit (ASIC) supporting the local point-to-point coherence protocol.
  • ASIC Application Specific Integrated Circuit
  • the cache coherence controller 230 can also be configured to handle a non-coherent protocol to allow communication with I/O devices.
  • the cache coherence controller 230 is a specially configured programmable chip such as a programmable logic device or a field programmable gate array.
  • Figure 3 is a diagrammatic representation of one example of a cache coherence controller 230.
  • the cache coherence controller includes a protocol engine 305 configured to handle packets such as probes and requests received from processors in various clusters of a multiprocessor system.
  • the functionality of the protocol engine 305 can be partitioned across several engines to improve performance. In one example, partitioning is done based on packet type (request, probe and response), direction (incoming and outgoing), or transaction flow (request flows, probe flows, etc).
  • the protocol engine 305 has access to a pending buffer 309 that allows the cache coherence controller to track transactions such as recent requests and probes and associate the transactions with specific processors.
  • Transaction information maintained in the pending buffer 309 can include transaction destination nodes, the addresses of requests for subsequent collision detection and protocol optimizations, response information, tags, and state information.
  • the cache coherence controller has an interface such as a coherent protocol interface 307 that allows the cache coherence controller to communicate with other processors in the cluster as well as external processor clusters.
  • each interface 307 and 311 is implemented either as a full crossbar or as separate receive and transmit units using components such as multiplexers and buffers.
  • the cache coherence controller can also include other interfaces such as a non-coherent protocol interface 311 for communicating with I/O devices. It should be noted, however, that the cache coherence controller 230 does not necessarily need to provide both coherent and non-coherent interfaces. It should also be noted that a cache coherence controller in one cluster can communicate with a cache coherence controller in another cluster.
  • FIG. 4 is a diagrammatic representation showing the transactions for a cache request from a processor in a system having a single cluster without using a cache coherence controller.
  • a processor 401-1 sends an access request such as a read memory line request to a memory controller 403-1.
  • the memory controller 403-1 may be associated with this processor, another processor in the single cluster or may be a separate component such as an ASIC or specially configured Programmable Logic Device (PLD).
  • PLD Programmable Logic Device
  • the memory controller 403-1 then sends probes to the local cache memories
  • CPU 401-1, CPU 401-2, and CPU 401-3 refer to the same processor.
  • FIGS 5A-5D are diagrammatic representations depicting cache coherence controller operation.
  • the use of a cache coherence controller in multiprocessor clusters allows the creation of a multiprocessor, multicluster coherent domain without affecting the functionality of local nodes such as processors and memory controllers in each cluster.
  • processors may only support a protocol that allows for a limited number of processors in a single cluster without allowing for multiple clusters.
  • the cache coherence controller can be used to allow multiple clusters by making local processors believe that the non-local nodes are merely a single local node embodied in the cache coherence controller.
  • the processors in a cluster do not need to be aware of processors in other clusters.
  • non-local nodes refer to nodes not in a request cluster generally and includes nodes in both a remote cluster and nodes in a home cluster.
  • a cluster from which a data access or cache access request originates is referred to herein as a request cluster.
  • a cluster containing a serialization point is referred to herein as a home cluster.
  • Other clusters are referred to as remote clusters.
  • the home cluster and the remote cluster are also referred to herein as non-local clusters.
  • FIG. 5A shows the cache coherence controller acting as an aggregate remote cache.
  • the cache coherence controller 509 accepts the probe from the local memory controller 503-1 and forwards it to non-local node portion 511.
  • a coherence protocol can contain several types of messages.
  • a coherence protocol includes four types of messages; data or cache access requests, probes, responses or probe responses, and data packets.
  • Data or cache access requests usually target the home node memory controller.
  • Probes are used to query each cache in the system. The probe packet can carry information that allows the caches to properly transition the cache state for a specified line. Responses are used to carry probe response information and to allow nodes to inform other nodes of the state of a given transaction. Data packets carry request data for both write requests and read responses.
  • the memory address resides at the local memory controller.
  • nodes including processors and cache coherence controllers outside of a local cluster are referred to herein as non-local nodes.
  • the cache coherence controller 509 then accumulates the response from the non-local nodes and sends a single response in the same manner that local nodes associated with cache blocks 505 and 507 send a single response to processor 501-2.
  • Local processors may expect a single probe response for every local node probed.
  • the use of a cache coherence controller allows the local processors to operate without concern as to whether non-local nodes exist.
  • processor 501-1 and processor 501-2 refer herein to the same component at different points in time during a transaction sequence.
  • processor 501-1 can initiate a data access request and the same processor 501-2 can later receive probe responses resulting from the request.
  • Figure 5B shows the cache coherence controller acting as a probing agent pair.
  • the cache coherence controller 521-1 receives a probe from non-local nodes 531
  • the cache coherence controller 521-1 accepts the probe and forwards the probe to local nodes associated with cache blocks 523, 525, and 527.
  • the cache coherence controller 521-2 then forwards a final response to the non-local node portion 531.
  • the cache coherence controller is both the source and the destination of the probes.
  • the local nodes associated with cache blocks 523, 525, and 527 behave as if the cache coherence controller were a local processor with a local memory request.
  • Figure 5C shows the cache coherence controller acting as a remote memory.
  • the cache coherence controller 543-1 forwards the request to the non-local nodes 553.
  • the cache coherence controller 543-1 When the remote request specifies local probing, the cache coherence controller 543-1 generates probes to local nodes and the probed nodes provide responses to the processor 541-2. Once the cache coherence controller 543-1 has received data from the non-local node portion 553, it forwards a read response to the processor 541-3. The cache coherence controller also forwards the final response to the remote memory controller associated with non-local nodes 553.
  • Figure 5D shows the cache coherence controller acting as a remote processor.
  • the cache coherence controller 561-1 at a first cluster receives a request from a processor in a second cluster, the cache coherence controller acts as a first cluster processor on behalf of the second cluster processor.
  • the cache coherence controller 561-1 accepts the request from portion 575 and forwards it to a memory controller 563-1.
  • the cache coherence controller 561-2 then accumulates all probe responses as well as the data fetched and forwards the final response to the memory controller 563- 2 as well as to non-local nodes 575.
  • multiple cluster systems can be built using processors that may not necessarily support multiple clusters.
  • the cache coherence controller can be used to represent non-local nodes in local transactions so that local nodes do not need to be aware of the existence of nodes outside of the local cluster.
  • FIG. 6 is a diagrammatic representation depicting the transactions for a data request from a local processor sent to a non-local cluster using a cache coherence controller.
  • the multicluster system includes a request cluster 600, a home cluster 620, and a remote cluster 640.
  • the home cluster 620 and the remote cluster 640 as well as any other clusters excluding the request cluster 600 are referred to herein as non-local clusters.
  • Processors and cache coherence controllers associated with local and non-local clusters are similarly referred to herein as local processors, local cache coherence controllers, non-local processors, and non-local cache coherence controllers, respectively.
  • processor 601-1 in a local cluster 600 sends a data access request such as a read request to a cache coherence controller 603- 1.
  • the cache coherence controller 603-1 tracks the transaction in the pending buffer of Figure 3 and forwards the request to a cache coherence controller 621-1 in a home cluster 620.
  • the cache coherence controller 621-1 at the home cluster 620 receives the access request and tracks the request in its pending buffer. In one example, information associated with the requests are stored in the pending buffer.
  • the cache coherence controller 621-1 forwards the access request to a memory controller 623-1 also associated with the home cluster 620. At this point, the memory controller 623-1 locks the memory line associated with the request.
  • the memory line is a unique address in the memory space shared by the multiple processors in the request cluster 600, home cluster 620, and the remote cluster 640.
  • the memory controller 623-1 generates a probe associated with the data access request and forwards the probe to local nodes associated with cache blocks 625 and 627 as well as to cache coherence controller 621-2.
  • messages associated with requests, probes, responses, and data are described as forwarded from one node to another, the messages themselves may contain variations. In one example, alterations are made to the messages to allow the multiple cluster architecture to be transparent to various local nodes. It should be noted that write requests can be handled as well. In write requests, the targeted memory controller gathers responses and sends the responses to the processor when gathering is complete.
  • the cache coherence controller 641-1 associated with the remote cluster 640 receives a probe from cache coherence controller 621-2 and probes local nodes associated with cache blocks 645, 647, and 649.
  • the cache coherence controller 603-2 associated with the request cluster 600 receives a probe and forwards the probe to local nodes associated with cache blocks 605, 607, and 609 to probe the cache blocks in the request cluster 600.
  • Processor 601-2 receives probe responses from the local nodes associated with cache blocks 605, 607, and 609.
  • cache coherence controller 621-3 accumulates probe responses and sends the probe responses to cache coherence controller 603-3, which in turn forwards the probe responses to the processor 601-3.
  • Cache coherence controller 621-4 also sends a read response to cache coherence controller 603-4, which forwards the read response to processor 601-4. While probes and probe responses carry information for maintaining cache coherency in the system, read responses can carry actual fetched data.
  • processor 601-4 may send a source done response to cache coherence controller 603- 5.
  • the transaction is now complete at the requesting cluster 600.
  • Cache coherence controller 603-5 forwards the source done message to cache coherence controller 621-5.
  • Cache coherence controller 621-5 in turn sends a source done message to memory controller 623-2.
  • the memory controller 623-2 can unlock the memory line and the transaction at the home cluster 620 is now complete.
  • Another processor can now access the unlocked memory line.
  • probe responses are gathered at cache coherence controller 603-3.
  • both home cluster probe responses and remote cluster probe responses can be delayed at the home cache coherence controller.
  • remote cluster probe responses have to travel an additional hop in order to reach a request cluster.
  • the latency for transmission of a probe response between a remote cluster and a request cluster may be substantially less than the latency for transmission of a probe response between a remote cluster and a request cluster through a home cluster.
  • Home cluster probe responses are also delayed as a result of this added hop.
  • a cache coherence controller 621-3 may wait to receive a read response message from a memory controller 623-1 before transmitting both a probe response message and a read response message to a cache coherence controller 603-3.
  • a cache coherence controller may be the actual processor generating the request. Some processors may operate as both a processor and as a cache coherence controller.
  • various data access request messages, probes, and responses associated with reads and writes are contemplated.
  • any message for snooping a cache can be referred to as a probe.
  • any message for indicating to the memory controller that a memory line should be unlocked can be referred to as a source done message.
  • transactions shown in Figure 6 show examples of cache coherence controllers performing many different functions, including functions of remote processors, aggregate local caches, probing agent pairs, and remote memory as described with reference to Figures 5A-5D.
  • the cache coherence controller 621-1 at the home cluster 620 is acting as a remote processor.
  • the cache coherence controller receives a request from a request cluster processor, the cache coherence controller is directed to act as the requesting processor on behalf of the request cluster processor.
  • the cache coherence controller 621-1 accepts a forwarded request from processor 601-1 and sends it to the memory controller 623-1, accumulates responses from all local nodes and the memory controller 623-1, and forwards the accumulated responses and data back to the requesting processor 601-3.
  • the cache coherence controller 621-5 also forwards a source done to the local memory controller 623-2.
  • the cache coherence controller 603-1 at the request cluster 600 is acting as a remote memory.
  • the cache coherence controller is designed to forward a request from a processor to a proper remote cluster and ensure that local nodes are probed.
  • the cache coherence controller 603-1 forwards a probe to cache coherence controller 621-1 at a home cluster 620.
  • Cache coherence controller 603-2 also probes local nodes 605, 607, and 609.
  • the cache coherence controller 641-1 at the request cluster 640 is acting as a probing agent pair. As noted above, when a cache coherence controller acting as a probing agent pair receives a probe from a remote cluster, the cache coherence controller accepts the probe and forwards it to all local nodes. The cache coherence controller accumulates the responses and sends a final response back to the request cluster. Here, the cache coherence controller 641-1 sends a probe to local nodes associated with cache blocks 645, 647, and 649, gathers probe responses and sends the probe responses to cache coherence controller 621-3 at home cluster 620. Similarly, cache coherence controller 603-2 also acts as a probing agent pair at a request cluster 600. The cache coherence controller 603-2 forwards probe requests to local nodes including local nodes associated with cache blocks 605, 607, and 609.
  • the cache coherence controller 621-2 and 621-3 is also acting as an aggregate remote cache.
  • the cache coherence controller 621-2 is responsible for accepting the probe from the memory controller 623-1 and forwarding the probe to the other processor clusters 600 and 640. More specifically, the cache coherence controller 621-2 forwards the probe to cache coherence controller 603-2 corresponding to request cluster 600 and to cache coherence controller 641-1 corresponding to remote cluster 640.
  • using a multiple cluster architecture may introduce delay as well as other undesirable elements such as increased traffic and processing overhead.
  • Probes are transmitted to all clusters in the multiple cluster system even though not all clusters need to be probed. For example, if a memory line associated with a probe request is invalid or absent from cache, it may not be necessary to probe all of the caches associated with the various clusters. In a system without a coherence directory, it is typically necessary to snoop all clusters. However, by using a coherence directory, the number of transactions in the system can be reduced by probing only a subset of the clusters in a system in order to minimize traffic and processing overhead.
  • coherence directory By using a coherence directory, global memory line state information (with respect to each cluster) can be maintained and accessed by a memory controller or a cache coherence controller in a particular cluster. According to various embodiments, the coherence directory tracks and manages the distribution of probes as well as the receipt of responses. If coherence directory information indicates that probing of a specific cluster is not required, the probe to the specific cluster can be eliminated. In one example, a coherence directory indicates that probing of requesting and remote clusters is not necessary. A cache coherence controller in a home cluster probes local nodes without forwarding probes to the request and remote clusters. The cache coherence controller in the home cluster then sends a response to the request cluster after probe responses are received. However, in typical multiple cluster systems, a requesting cluster expects a predetermined number of responses from the various probed clusters. In one example, if the multiple cluster system includes four clusters, a request cluster would expect probe responses associated with nodes in all four clusters.
  • the techniques of the present invention provide a completion bit associated with a probe responses.
  • the completion bit indicates to the requesting cluster that no other probe responses from other clusters should be expected. Any mechanisms for notifying a request cluster that no other probe responses should be expected from other clusters is referred to herein as a completion indicator.
  • a completion indicator is a completion bit included in the response sent to a request cluster after local nodes are probed.
  • a completion indicator is separate data transmitted to a request cluster.
  • Figure 7 is one example of a coherence directory that can be used to allow management and filtering of probes.
  • Various coherence directories are available.
  • a full directory provides an entry for every memory line in a system.
  • the coherence directory is maintained at the memory controller and is accessible by a cache coherence controller.
  • a sparse directory is provided with a limited number of entries associated with a selected set of memory lines.
  • the coherence directory 701 includes state information 713, dirty data owner information 715, and an occupancy vector 717 associated with the memory lines 711.
  • the memory line states are modified, owned, shared, and invalid.
  • a memory line In the invalid state, a memory line is not currently available in cache associated with any remote cluster.
  • a memory line In the shared state, a memory line may be present in more than one cache, but the memory line has not been modified in any of these caches.
  • an occupancy vector 717 can be checked to determine what caches share the relevant data.
  • An occupancy vector 717 may be implemented as an N-bit string, where each bit represents the availability of the data in the cache of N clusters. Any mechanism for tracking what clusters hold a copy of the relevant memory line in cache is referred to herein as an occupancy vector.
  • the memory line with address 741 is in the shared state, and the occupancy vector 717 indicates that clusters 1 and 3 each have a copy of the shared memory line in cache.
  • dirty data owner information field 715 can be checked to determine the owner of the dirty data. Any mechanism for indicating what cluster owns a modified copy of the memory line in cache is referred to herein as a dirty data owner information field.
  • the memory line associated with address 781 is modified, and the dirty data owner field 715 indicates that cluster 2 owns the memory line.
  • a dirty memory line is owned by a single cache but may be resident in multiple caches. It has been read by the owning cache, but has not been modified. In this case, the copy held in memory is stale.
  • dirty data owner field 715 can be accessed to determine which cluster owns the dirty data.
  • the memory line associated with address 761 is in the owned state and is owned by cluster 4.
  • the occupancy vector 717 can also be checked to determine what other caches may have the relevant data. In this example, the occupancy vector 717 indicates that clusters 2, 3, and 4 each have a copy of the data associated with the memory line in cache.
  • coherence directory 701 includes the four states of modified, owned, shared, and invalid, it should be noted that particular implementations may use a different set of states. In one example, a system may have the five states of modified, exclusive, owned, shared, and invalid. The techniques of the present invention can be used with a variety of different possible memory line states.
  • the coherence directory tracks the various transactions such as probe requests and responses in a multiple cluster system to determine when memory lines are added to the coherence directory, when memory lines are removed from the directory, and when information associated with each memory line is updated.
  • the techniques of the present invention recognize that the number of transactions such as probe requests can be reduced by managing or filtering probes that do not need to be sent to specific clusters.
  • Figure 8 is a diagrammatic representation showing probe filter information that can be used to reduce the number of transactions in a multiple cluster system. Any criterion that can be used to reduce the number of clusters probed from a home cluster is referred to herein as probe filter information. Transactions such as probe requests can have a variety of characteristics. Some characteristics include whether the probe is a read block (read) 823 or a read block modify (read/write) 825. Other characteristics of the probe include the state of the memory line associated with the probe. In some examples, states include invalid 831, shared 833, owned 835, and modified 837.
  • a coherence directory maintains information for memory lines in the local cluster that are cached in non-local clusters, where non-local clusters can include request and remote clusters. If the state of the memory line associated with a probe is invalid 831 as indicated in the coherence directory, no copies of the memory line reside in other clusters. Consequently, only the home cluster needs to be probed and a completion bit can be used to indicate to a request cluster that the request cluster should expect only a single response from home cluster instead of a response from each of the clusters.
  • the memory line associated with the probe is in the shared state 833, and the transaction is a read transaction, only the home cluster needs to be probed and a completion bit can again be used to indicate to the request cluster that only a single response from home cluster should be expected as indicated in entry 803. It should be noted that the completed bit can not be used in entries 811 and 813.
  • FIG. 9 is a diagrammatic representation depicting one example of transactions for probing only a home cluster as indicated in entries 801, 809, and 803 in Figure 8.
  • processor 901-1 in a local cluster 900 sends a data access request such as a read request to a cache coherence controller 903-1.
  • the cache coherence controller 903-1 forwards the request to a cache coherence controller 921-1 in a home cluster 920.
  • the cache coherence controller 921-1 at the home cluster 920 receives the access request and forwards the access request to a memory controller 923-1, which then probes local nodes 925, 927, and cache coherence controller 921-2.
  • a cache coherency controller 921-1 is typically responsible for updating the coherence directory during various transactions.
  • the cache coherence controller 921-2 determines characteristics associated with the probe from the memory controller 923-1 to determine whether remote probes are needed and whether a completion bit can be used.
  • the cache coherence controller 921-2 determines that no remote probes are needed and does not forward probes to the remote cluster 940 or to request cluster 900.
  • cache coherence controller 921-4 After cache coherence controller 921-4 receives the probe responses from local nodes as well as the read response from the memory controller 923-1, the response message with a completion indicator is transmitted to the request cluster. With the completion indicator, the request cluster does not wait for additional responses from other clusters.
  • the coherence controller 903-4 forwards the response with the completion bit set to CPU 901-4. After receiving the response with the completion bit set, the CPU does not wait for additional responses from the local caches.
  • CPU 901-4 forwards a source done message to cache coherence controller 903-5 to home cluster cache coherence controller 921-5, which can then perform updates of its coherence directory. The source done is then forwarded to memory controller 923-1.
  • Figure 9 shows one example of a sequence where only the home cluster needs to be probed.
  • Figure 10 shows one example of a sequence where only a single remote cluster needs to be probed.
  • Figure 10 is a diagrammatic representation depicting an example of transactions for probing a remote cluster as indicated in entries 805, 807, and 815 in Figure 8.
  • processor 1001-1 in a local cluster 1000 sends a data access request such as a read request to a cache coherence controller 1003-1.
  • the cache coherence controller 1003-1 forwards the request to a cache coherence controller 1021-1 in a home cluster 1020.
  • the cache coherence controller 1021-1 at the home cluster 1020 receives the access request and forwards the access request to a memory controller 1023-1, which then probes local nodes 1025, 1027, and cache coherence controller 1021-2.
  • the cache coherence controller 1021-2 determines characteristics associated with the probe from the memory controller 1023-1 to determine whether remote probes are needed and whether a completion bit can be used. Here, the cache coherence controller 1021-2 determines that only a remote cluster needs to be probed and does not forward a probe to request cluster 1000.
  • cache coherence controller 1021-4 After cache coherence controller 1021-4 receives the probes from local nodes as well as the read response from the memory controller 1023-1, a response message is not transmitted to the request cluster because the remote cluster is sending a response message with a completion indicator is transmitted to the request cluster. With the completion indicator, the request cluster does not wait for additional responses from other clusters.
  • the response is forwarded to CPU 1001-4 and a source done message is sent from cache coherence controller 1003-5 to home cluster cache coherence controller 1021-5. With the completion bit set in the response to CPU 1001-4, it does not wait for any other local responses. After all responses from local nodes are received, the source done is then forwarded to memory controller 1023-1, which can then perform updates of its coherence directory.
  • FIG 11 is a process flow diagram showing one example of a technique for handling probe requests at a home cache coherence controller.
  • a probe request associated with a memory line is received.
  • the cache coherence controller forwards the request to the memory controller.
  • the cache coherence controller receives a probe from the memory controller and accesses a coherence directory and probe filter information at 1113 to determine whether the number of probes to various clusters in the system can be reduced.
  • filtering and a completion indicator can be used. In one example, it is determined the filtering a completion indicator can be used by identifying the criteria specified in Figure 8 and by accessing a coherence directory and shown in Figure 7.
  • probes are broadcast to the various nodes with no filtering and no completion bit 1145. If filtering and a completion indicator can be used, it is determined at 1131 if a remote cluster should be probed. If a single remote cluster is the cluster that should be probed, the probe is forwarded with the completion indicator to the remote cluster at 1135. At 1139, home cluster probe responses are received but are not forwarded to the request cluster. The response is not sent to the request cluster from home cluster because a remote cluster is sending a response with a completion indicator to the request cluster.
  • source done information is received from the request cluster and forwarded to the memory controller. If it is determined at 1131 that only the home cluster needs to be probed, then the cache coherence controller at 1141 does not send probes to any request or remote clusters and instead sends a response to the request cluster with a completion indicator. The cache coherence controller sends the response with the completion indicator after receiving home cluster probe responses. At 1149, the cache coherence controller at the home cluster receives source done information from the request cluster and forwards the source done information to the memory controller.
  • the techniques of the present invention provide that when only a remote or request cluster needs to be probed, the memory controller can sometimes be bypassed to allow probing of only the remote or request cluster.
  • a probe is not forwarded to the home cluster memory controller and a probe is forwarded directly to the remote cluster from the home cluster cache coherence controller.
  • Figure 12 is a diagrammatic representation showing exemplary memory controller filter information. Any criterion used to reduce the number of requests forwarded to a memory controller is referred to herein as memory controller filter information. Characteristics of a probe can again be analyzed when a cache coherence controller receives the probe from a request cluster. Transactions such as probe requests can have a variety of characteristics. Some characteristics include whether the probe is a read block (read) 1223 or a read block modify (read/write) 1225. Other characteristics of the probe include the state of the memory line associated with the probe. In some examples, states include invalid 1231, shared 1233, owned 1235, and modified 1237.
  • FIG. 13 shows one example of a sequence where a probe does not need to be forwarded to the home cluster memory controller.
  • processor 1301-1 in a local cluster 1300 sends a data access request such as a read request to a cache coherence controller 1303-1.
  • the cache coherence controller 1303-1 forwards the request to a cache coherence controller 1321-1 in a home cluster 1320.
  • the cache coherence controller 1321-1 at the home cluster 1320 receives the access request and determines whether the memory controller can be bypassed. Forwarding a probe to a remote or request cluster without forwarding the probe to a memory controller is referred to herein as bypassing the memory controller.
  • the determination can be made by using memory controller filter information.
  • the memory controller is bypassed and the probe is forwarded to cache coherence controller 1341-1 in the remote cluster 1340.
  • the probe is forwarded with an indication that a completion bit should be used.
  • the cache coherence controller 1321-1 in the home cluster 1320 is acting as a serialization point in place of the memory controller to maintain cache coherency. Once it is determined that the memory controller can be bypassed, the cache coherence controller 1321-1 blocks all other incoming requests and outgoing probes until a final source done is received from the request cluster.
  • the remote cluster cache coherence controller 1341-1 probes remote cluster nodes and sends a response with a completion indicator to the request cluster 1300. The response is forwarded to CPU 1301-4 and a source done message is sent from cache coherence controller 1303-5 to home cluster cache coherence controller 1321-5. The source done is not forwarded to the memory controller, because the memory controller never processed the transaction.
  • Figure 14 is a flow process diagram showing probe request handling at a home cache coherence controller using memory controller filter information.
  • a probe request associated with a particular memory line is received.
  • characteristics associated with the probe request are identified.
  • memory controller filter information shown in Figure 12 is used to determine whether a memory controller can be bypassed. If it is determined that a memory controller can be bypassed, probe requests associated with the same memory line are blocked at 1415 and a probe is sent to a remote or a request cluster.
  • the memory line is unblocked after receiving a source done from the request cluster. If it is determined at 1411 that a memory controller can not be bypassed, the probe request is forwarded to a serialization point 1405. The transaction sequence can then proceed with or without probe filtering and a completion indicator as shown in 1109 of Figure 11.
  • embodiments of the present invention may be employed with multiple processor clusters connected through a point-to-point, switch, or bus architecture.
  • multiple clusters of processors may share a single cache coherence controller, or multiple cache coherence controllers can be used in " a single cluster. Therefore, the scope of the invention should be determined with reference to the appended claims.

Abstract

According to the present invention, methods and apparatus are provided for increasing the efficiency of data access in a multiple processor, multiple cluster system. Mechanisms for reducing the number of transactions in a multiple cluster system are provided. In one example, probe filter information is used to limit the number of probe requests transmitted to request and remote clusters.

Description

METHODS AND APPARATUS FOR MANAGING PROBE REQUESTS
CROSS-REFERENCE TO RELATED APPLICATIONS The present application is related to filed U.S. Application No. 10/106,426 titled Methods And Apparatus For Speculative Probing At A Request Cluster, U.S. Application No. 10/106,430 titled Methods And Apparatus For Speculative Probing With Early Completion And Delayed Request, and U.S. Application No. 10/106,299 titled Methods And Apparatus For Speculative Probing With Early Completion And Early Request, the entireties of which are incorporated by reference herein for all purposes. The present application is also related to filed U.S. Application Nos. 10/157,340, 10/145,439, 10/145,438, and 10/157,388 titled Methods And Apparatus For Responding To A Request Cluster by David B. Glasco, the entireties of which are incorporated by reference for all purposes. The present application is also related to concurrently filed U.S. Application No. 10/288,399 (Attorney Docket No. NWISP025) with the same title and inventor, the entirety of which is incorporated by reference herein for all purposes.
BACKGROUND OF THE INVENTION
1. Field of the Invention.
The present invention generally relates to accessing data in a multiple processor system. More specifically, the present invention provides techniques for improving data access efficiency while maintaining cache coherency in a multiple processor system having a multiple cluster architecture.
2. Description of Related Art
Data access in multiple processor systems can raise issues relating to cache coherency. Conventional multiple processor computer systems have processors coupled to a system memory through a shared bus. In order to optimize access to data in the system memory, individual processors are typically designed to work with cache memory. In one example, each processor has a cache that is loaded with data that the processor frequently accesses. The cache is read or written by a processor. However, cache coherency problems arise because multiple copies of the same data can co-exist in systems having multiple processors and multiple cache memories. For example, a frequently accessed data block corresponding to a memory line may be loaded into the cache of two different processors. In one example, if both processors attempt to write new values into the data block at the same time, different data values may result. One value may be written into the first cache while a different value is written into the second cache. A system might then be unable to determine what value to write through to system memory.
A variety of cache coherency mechanisms have been developed to address such problems in multiprocessor systems. One solution is to simply force all processor writes to go through to memory immediately and bypass the associated cache. The write requests can then be serialized before overwriting a system memory line. However, bypassing the cache significantly decreases efficiency gained by using a cache. Other cache coherency mechanisms have been developed for specific architectures. In a shared bus architecture, each processor checks or snoops on the bus to determine whether it can read or write a shared cache block. In one example, a processor only writes an object when it owns or has exclusive access to the object. Each corresponding cache object is then updated to allow processors access to the most recent version of the object. Bus arbitration is used when both processors attempt to write the same shared data block in the same clock cycle. Bus arbitration logic decides which processor gets the bus first. Although, cache coherency mechanisms such as bus arbitration are effective, using a shared bus limits the number of processors that can be implemented in a single system with a single memory space. Other multiprocessor schemes involve individual processor, cache, and memory systems connected to other processors, cache, and memory systems using a network backbone such as Ethernet or Token Ring. Multiprocessor schemes involving separate computer systems each with its own address space can avoid many cache coherency problems because each processor has its own associated memory and cache. When one processor wishes to access data on a remote computing system, communication is explicit. Messages are sent to move data to another processor and messages are received to accept data from another processor using standard network protocols such as TCP/IP. Multiprocessor systems using explicit communication including transactions such as sends and receives are referred to as systems using multiple private memories. By contrast, multiprocessor system using implicit communication including transactions such as loads and stores are referred to herein as using a single address space.
Multiprocessor schemes using separate computer systems allow more processors to be interconnected while minimizing cache coherency problems. However, it would take substantially more time to access data held by a remote processor using a network infrastructure than it would take to access data held by a processor coupled to a system bus. Furthermore, valuable network bandwidth would be consumed moving data to the proper processors. This can negatively impact both processor and network performance.
Performance limitations have led to the development of a point-to-point architecture for connecting processors in a system with a single memory space. In one example, individual processors can be directly connected to each other through a plurality of point-to-point links to form a cluster of processors. Separate clusters of processors can also be connected. The point-to-point links significantly increase the bandwidth for coprocessing and multiprocessing functions. However, using a point- to-point architecture to connect multiple processors in a multiple cluster system sharing a single memory space presents its own problems.
Consequently, it is desirable to provide techniques for improving data access and cache coherency in systems having multiple clusters of multiple processors connected using point-to-point links.
SUMMARY OF THE INVENTION According to the present invention, methods and apparatus are provided for increasing the efficiency of data access in a multiple processor, multiple cluster system. Mechanisms for reducing the number of transactions in a multiple cluster system are provided. In one example, probe filter information is used to limit the number of probe requests transmitted to request and remote clusters.
In one embodiment, a computer system is provided. The computer system includes a home cluster having a first plurality of processors and a home cache coherence controller. The first plurality of processors and the home cache coherence controller are interconnected in a point-to-point architecture. The home cache coherence controller is configured to receive a probe request and probe one or more selected clusters. The one or more clusters are selected based on the characteristics associated with the probe request. In another embodiment, a method for managing probes is provided. A probe request is received at a home cache coherence controller in a home cluster. The home cluster includes a first plurality of processors and the home cache coherence controller. The first plurality of processors and the home cache coherence controller are interconnected in a point-to-point architecture. One or more clusters are selected for probing based on the characteristics associated with the probe request. The one or more clusters are probed.
A further understanding of the nature and advantages of the present invention may be realized by reference to the remaining portions of the specification and the drawings.
BRIEF DESCRIPTION OF THE DRAWINGS The invention may best be understood by reference to the following description taken in conjunction with the accompanying drawings, which are illustrative of specific embodiments of the present invention. Figure 1A and IB are diagrammatic representation depicting a system having multiple clusters.
Figure 2 is a diagrammatic representation of a cluster having a plurality of processors.
Figure 3 is a diagrammatic representation of a cache coherence controller. Figure 4 is a diagrammatic representation showing a transaction flow for a data access request from a processor in a single cluster.
Figure 5A-5D are diagrammatic representations showing cache coherence controller functionality.
Figure 6 is a diagrammatic representation depicting a transaction flow for a probe request with multiple probe responses.
Figure 7 is a diagrammatic representation showing a cache coherence directory.
Figure 8 is a diagrammatic representation showing probe filter information that can be used to reduce the number of probes transmitted to various clusters. Figure 9 is a diagrammatic representation showing a transaction flow for probing of a home cluster without probing of other clusters.
Figure 10 is a diagrammatic representation showing a transaction flow for probing of a single remote cluster.
Figure 11 is a flow process diagram showing the handling of a probe request with probe filter information.
Figure 12 is a diagrammatic representation showing memory controller filter information.
Figure 13 is a diagrammatic representation showing a transaction flow for probing a single remote cluster without probing a home cluster. Figure 14 is a flow process diagram showing the handling of a probe request at a home cluster cache coherence controller using memory controller filter information. DETAILED DESCRIPTION OF SPECIFIC EMBODIMENTS Reference will now be made in detail to some specific embodiments of the invention including the best modes contemplated by the inventors for carrying out the invention. Examples of these specific embodiments are illustrated in the accompanying drawings. While the invention is described in conjunction with these specific embodiments, it will be understood that it is not intended to limit the invention to the described embodiments. On the contrary, it is intended to cover alternatives, modifications, and equivalents as may be included within the spirit and scope of the invention as defined by the appended claims. Multi-processor architectures having point-to-point communication among their processors are suitable for implementing specific embodiments of the present invention. In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present invention. The present invention may be practiced without some or all of these specific details. Well-known process operations have not been described in detail in order not to unnecessarily obscure the present invention. Furthermore, the present application's reference to a particular singular entity includes that possibility that the methods and apparatus of the present invention can be implemented using more than one entity, unless the context clearly dictates otherwise.
Techniques are provided for increasing data access efficiency in a multiple processor, multiple cluster system. In a point-to-point architecture, a cluster of processors includes multiple processors directly connected to each other through point-to-point links. By using point-to-point links instead of a conventional shared bus or external network, multiple processors are used efficiently in a system sharing the same memory space. Processing and network efficiency are also improved by avoiding many of the bandwidth and latency limitations of conventional bus and external network based multiprocessor architectures. According to various embodiments, however, linearly increasing the number of processors in a point-to- point architecture leads to an exponential increase in the number of links used to connect the multiple processors. In order to reduce the number of links used and to further modularize a multiprocessor system using a point-to-point architecture, multiple clusters are used. According to various embodiments, the multiple processor clusters are interconnected using a point-to-point architecture. Each cluster of processors includes a cache coherence controller used to handle communications between clusters. In one embodiment, the point-to-point architecture used to connect processors are used to connect clusters as well.
By using a cache coherence controller, multiple cluster systems can be built using processors that may not necessarily support multiple clusters. Such a multiple cluster system can be built by using a cache coherence controller to represent nonlocal nodes in local transactions so that local nodes do not need to be aware of the existence of nodes outside of the local cluster. More detail on the cache coherence controller will be provided below.
In a single cluster system, cache coherency can be maintained by sending all data access requests through a serialization point. Any mechanism for ordering data access requests is referred to herein as a serialization point. One example of a serialization point is a memory controller. Various processors in the single cluster system send data access requests to the memory controller. In one example, the memory controller is configured to serialize or lock the data access requests so that only one data access request for a given memory line is allowed at any particular time. If another processor attempts to access the same memory line, the data access attempt is blocked until the memory line is unlocked. The memory controller allows cache coherency to be maintained in a multiple processor, single cluster system.
A serialization point can also be used in a multiple processor, multiple cluster system where the processors in the various clusters share a single address space. By using a single address space, internal point-to-point links can be used to significantly improve intercluster communication over traditional external network based multiple cluster systems. Various processors in various clusters send data access requests to a memory controller associated with a particular cluster such as a home cluster. The memory controller can similarly serialize all data requests from the different clusters. However, a serialization point in a multiple processor, multiple cluster system may not be as efficient as a serialization point in a multiple processor, single cluster system. That is, delay resulting from factors such as latency from transmitting between clusters can adversely affect the response times for various data access requests. It should be noted that delay also results from the use of probes in a multiple processor environment. Although delay in intercluster transactions in an architecture using a shared memory space is significantly less than the delay in conventional message passing environments using external networks such as Ethernet or Token Ring, even minimal delay is a significant factor. In some applications, there may be millions of data access requests from a processor in a fraction of a second. Any delay can adversely impact processor performance.
According to various embodiments, probe management is used to increase the efficiency of accessing data in a multiple processor, multiple cluster system. A mechanism for eliciting a response from a node to maintain cache coherency in a system is referred to herein as a probe. In one example, a mechanism for snooping a cache is referred to as a probe. A response to a probe can be directed to the source or target of the initiating request. Any mechanism for filtering or reducing the number of probes and probe requests transmitted to various nodes is referred to herein as managing probes. In one example, managing probe entails characterizing a probe request to determine if a probe can be transmitted to a reduced number of entities.
In typical implementations, probe requests are sent to a memory controller that broadcasts probes to various nodes in a system. In such a system, no knowledge of the cache line state is known. All nodes in the system are probed and the request cluster receives a response from each node. In a system with a coherence directory, state information associated with various memory lines can be used to reduce the number of transactions. Any mechanism for maintaining state information associated with various memory lines is referred to herein as a coherence directory. A coherence directory typically includes information for memory lines in a local cluster that are cached in a remote cluster. According to various embodiments, a coherence directory is used to reduce the number of probes to remote quads by inferring the state of local caches. In other embodiments, a coherence directory is used to eliminate the transmission of a request to a memory controller in a home cluster.
Figure 1A is a diagrammatic representation of one example of a multiple cluster, multiple processor system that can use the techniques of the present invention. Each processing cluster 101, 103, 105, and 107 can include a plurality of processors. The processing clusters 101, 103, 105, and 107 are connected to each other through point-to-point links l l la-f. In one embodiment, the multiple processors in the multiple cluster architecture shown in Figure 1A share the same memory space. In this example, the point-to-point links l l la-f are internal system connections that are used in place of a traditional front-side bus to connect the multiple processors in the multiple clusters 101, 103, 105, and 107. The point-to-point links may support any point-to-point coherence protocol.
Figure IB is a diagrammatic representation of another example of a multiple cluster, multiple processor system that can use the techniques of the present invention. Each processing cluster 121, 123, 125, and 127 can be coupled to a switch 131 through point-to-point links 141a-d. It should be noted that using a switch and point- to-point links allows implementation with fewer point-to-point links when connecting multiple clusters in the system. A switch 131 can include a processor with a coherence protocol interface. According to various implementations, a multicluster system shown in Figure 1A is expanded using a switch 131 as shown in Figure IB.
Figure 2 is a diagrammatic representation of a multiple processor cluster, such as the cluster 101 shown in Figure 1A. Cluster 200 includes processors 202a-202d, one or more Basic I/O systems (BIOS) 204, a memory subsystem comprising memory banks 206a-206d, point-to-point communication links 208a-208e, and a service processor 212. The point-to-point communication links are configured to allow interconnections between processors 202a-202d, I/O switch 210, and cache coherence controller 230. The service processor 212 is configured to allow communications with processors 202a-202d, I/O switch 210, and cache coherence controller 230 via a JTAG interface represented in Fig. 2 by links 214a-214f. It should be noted that other interfaces are supported. It should also be noted that in some implementations, a service processor is not included in multiple processor clusters. I/O switch 210 connects the rest of the system to I/O adapters 216 and 220.
According to specific embodiments, the service processor of the present invention has the intelligence to partition system resources according to a previously specified partitioning schema. The partitioning can be achieved through direct manipulation of routing tables associated with the system processors by the service processor which is made possible by the point-to-point communication infrastructure. The routing tables are used to control and isolate various system resources, the connections between which are defined therein.
The processors 202a-d are also coupled to a cache coherence controller 230 through point-to-point links 232a-d. Any mechanism or apparatus that can be used to provide communication between multiple processor clusters while maintaining cache coherence is referred to herein as a cache coherence controller. The cache coherence controller 230 can be coupled to cache coherence controllers associated with other multiprocessor clusters. It should be noted that there can be more than one cache coherence controller in one cluster. The cache coherence controller 230 communicates with both processors 202a-d as well as remote clusters using a point- to-point protocol.
More generally, it should be understood that the specific architecture shown in Figure 2 is merely exemplary and that embodiments of the present invention are contemplated having different configurations and resource interconnections, and a variety of alternatives for each of the system resources shown. However, for purpose of illustration, specific details of server 200 will be assumed. For example, most of the resources shown in Fig. 2 are assumed to reside on a single electronic assembly. In addition, memory banks 206a-206d may comprise double data rate (DDR) memory which is physically provided as dual in-line memory modules (DIMMs). I/O adapter 216 may be, for example, an ultra direct memory access (UDMA) controller or a small computer system interface (SCSI) controller which provides access to a permanent storage device. I/O adapter 220 may be an Ethernet card adapted to provide communications with a network such as, for example, a local area network (LAN) or the Internet.
According to a specific embodiment and as shown in Fig. 2, both of I/O adapters 216 and 220 provide symmetric I/O access. That is, each provides access to equivalent sets of I/O. As will be understood, such a configuration would facilitate a partitioning scheme in which multiple partitions have access to the same types of I/O. However, it should also be understood that embodiments are envisioned in which partitions without I/O are created. For example, a partition including one or more processors and associated memory resources, i.e., a memory complex, could be created for the purpose of testing the memory complex.
According to one embodiment, service processor 212 is a Motorola MPC855T microprocessor which includes integrated chipset functions. The cache coherence controller 230 is an Application Specific Integrated Circuit (ASIC) supporting the local point-to-point coherence protocol. The cache coherence controller 230 can also be configured to handle a non-coherent protocol to allow communication with I/O devices. In one embodiment, the cache coherence controller 230 is a specially configured programmable chip such as a programmable logic device or a field programmable gate array. Figure 3 is a diagrammatic representation of one example of a cache coherence controller 230. According to various embodiments, the cache coherence controller includes a protocol engine 305 configured to handle packets such as probes and requests received from processors in various clusters of a multiprocessor system. The functionality of the protocol engine 305 can be partitioned across several engines to improve performance. In one example, partitioning is done based on packet type (request, probe and response), direction (incoming and outgoing), or transaction flow (request flows, probe flows, etc).
The protocol engine 305 has access to a pending buffer 309 that allows the cache coherence controller to track transactions such as recent requests and probes and associate the transactions with specific processors. Transaction information maintained in the pending buffer 309 can include transaction destination nodes, the addresses of requests for subsequent collision detection and protocol optimizations, response information, tags, and state information. The cache coherence controller has an interface such as a coherent protocol interface 307 that allows the cache coherence controller to communicate with other processors in the cluster as well as external processor clusters. According to various embodiments, each interface 307 and 311 is implemented either as a full crossbar or as separate receive and transmit units using components such as multiplexers and buffers. The cache coherence controller can also include other interfaces such as a non-coherent protocol interface 311 for communicating with I/O devices. It should be noted, however, that the cache coherence controller 230 does not necessarily need to provide both coherent and non-coherent interfaces. It should also be noted that a cache coherence controller in one cluster can communicate with a cache coherence controller in another cluster.
Figure 4 is a diagrammatic representation showing the transactions for a cache request from a processor in a system having a single cluster without using a cache coherence controller. A processor 401-1 sends an access request such as a read memory line request to a memory controller 403-1. The memory controller 403-1 may be associated with this processor, another processor in the single cluster or may be a separate component such as an ASIC or specially configured Programmable Logic Device (PLD). To preserve cache coherence, only one processor is typically allowed to access a memory line corresponding to a shared address space at anyone given time. To prevent other processors from attempting to access the same memory line, the memory line can be locked by the memory controller 403-1. All other requests to the same memory line are blocked or queued. Access by another processor is typically only allowed when the memory controller 403-1 unlocks the memory line. The memory controller 403-1 then sends probes to the local cache memories
405, 407, and 409 to determine cache states. The local cache memories 405, 407, and 409 then in turn send probe responses to the same processor 401-2. The memory controller 403-1 also sends an access response such as a read response to the same processor 401-3. The processor 401-3 can then send a done response to the memory controller 403-2 to allow the memory controller 403-2 to unlock the memory line for subsequent requests. It should be noted that CPU 401-1, CPU 401-2, and CPU 401-3 refer to the same processor.
Figures 5A-5D are diagrammatic representations depicting cache coherence controller operation. The use of a cache coherence controller in multiprocessor clusters allows the creation of a multiprocessor, multicluster coherent domain without affecting the functionality of local nodes such as processors and memory controllers in each cluster. In some instances, processors may only support a protocol that allows for a limited number of processors in a single cluster without allowing for multiple clusters. The cache coherence controller can be used to allow multiple clusters by making local processors believe that the non-local nodes are merely a single local node embodied in the cache coherence controller. In one example, the processors in a cluster do not need to be aware of processors in other clusters. Instead, the processors in the cluster communicate with the cache coherence controller as though the cache coherence controller were representing all non-local nodes. It should be noted that nodes in a remote cluster will be referred to herein as non-local nodes or as remotes nodes. However, non-local nodes refer to nodes not in a request cluster generally and includes nodes in both a remote cluster and nodes in a home cluster. A cluster from which a data access or cache access request originates is referred to herein as a request cluster. A cluster containing a serialization point is referred to herein as a home cluster. Other clusters are referred to as remote clusters. The home cluster and the remote cluster are also referred to herein as non-local clusters.
Figure 5A shows the cache coherence controller acting as an aggregate remote cache. When a processor 501-1 generates a data access request to a local memory controller 503-1, the cache coherence controller 509 accepts the probe from the local memory controller 503-1 and forwards it to non-local node portion 511. It should be noted that a coherence protocol can contain several types of messages. In one example, a coherence protocol includes four types of messages; data or cache access requests, probes, responses or probe responses, and data packets. Data or cache access requests usually target the home node memory controller. Probes are used to query each cache in the system. The probe packet can carry information that allows the caches to properly transition the cache state for a specified line. Responses are used to carry probe response information and to allow nodes to inform other nodes of the state of a given transaction. Data packets carry request data for both write requests and read responses.
According to various embodiments, the memory address resides at the local memory controller. As noted above, nodes including processors and cache coherence controllers outside of a local cluster are referred to herein as non-local nodes. The cache coherence controller 509 then accumulates the response from the non-local nodes and sends a single response in the same manner that local nodes associated with cache blocks 505 and 507 send a single response to processor 501-2. Local processors may expect a single probe response for every local node probed. The use of a cache coherence controller allows the local processors to operate without concern as to whether non-local nodes exist.
It should also be noted that components such as processor 501-1 and processor 501-2 refer herein to the same component at different points in time during a transaction sequence. For example, processor 501-1 can initiate a data access request and the same processor 501-2 can later receive probe responses resulting from the request.
Figure 5B shows the cache coherence controller acting as a probing agent pair. When the cache coherence controller 521-1 receives a probe from non-local nodes 531, the cache coherence controller 521-1 accepts the probe and forwards the probe to local nodes associated with cache blocks 523, 525, and 527. The cache coherence controller 521-2 then forwards a final response to the non-local node portion 531. In this example, the cache coherence controller is both the source and the destination of the probes. The local nodes associated with cache blocks 523, 525, and 527 behave as if the cache coherence controller were a local processor with a local memory request. Figure 5C shows the cache coherence controller acting as a remote memory. When a local processor 541-1 generates an access request that targets remote memory, the cache coherence controller 543-1 forwards the request to the non-local nodes 553. When the remote request specifies local probing, the cache coherence controller 543-1 generates probes to local nodes and the probed nodes provide responses to the processor 541-2. Once the cache coherence controller 543-1 has received data from the non-local node portion 553, it forwards a read response to the processor 541-3. The cache coherence controller also forwards the final response to the remote memory controller associated with non-local nodes 553. Figure 5D shows the cache coherence controller acting as a remote processor.
When the cache coherence controller 561-1 at a first cluster receives a request from a processor in a second cluster, the cache coherence controller acts as a first cluster processor on behalf of the second cluster processor. The cache coherence controller 561-1 accepts the request from portion 575 and forwards it to a memory controller 563-1. The cache coherence controller 561-2 then accumulates all probe responses as well as the data fetched and forwards the final response to the memory controller 563- 2 as well as to non-local nodes 575.
By allowing the cache coherence controller to act as an aggregate remote cache, probing agent pair, remote memory, and remote processor, multiple cluster systems can be built using processors that may not necessarily support multiple clusters. The cache coherence controller can be used to represent non-local nodes in local transactions so that local nodes do not need to be aware of the existence of nodes outside of the local cluster.
Figure 6 is a diagrammatic representation depicting the transactions for a data request from a local processor sent to a non-local cluster using a cache coherence controller. The multicluster system includes a request cluster 600, a home cluster 620, and a remote cluster 640. As noted above, the home cluster 620 and the remote cluster 640 as well as any other clusters excluding the request cluster 600 are referred to herein as non-local clusters. Processors and cache coherence controllers associated with local and non-local clusters are similarly referred to herein as local processors, local cache coherence controllers, non-local processors, and non-local cache coherence controllers, respectively.
According to various embodiments, processor 601-1 in a local cluster 600 sends a data access request such as a read request to a cache coherence controller 603- 1. The cache coherence controller 603-1 tracks the transaction in the pending buffer of Figure 3 and forwards the request to a cache coherence controller 621-1 in a home cluster 620. The cache coherence controller 621-1 at the home cluster 620 receives the access request and tracks the request in its pending buffer. In one example, information associated with the requests are stored in the pending buffer. The cache coherence controller 621-1 forwards the access request to a memory controller 623-1 also associated with the home cluster 620. At this point, the memory controller 623-1 locks the memory line associated with the request. In one example, the memory line is a unique address in the memory space shared by the multiple processors in the request cluster 600, home cluster 620, and the remote cluster 640. The memory controller 623-1 generates a probe associated with the data access request and forwards the probe to local nodes associated with cache blocks 625 and 627 as well as to cache coherence controller 621-2.
It should be noted that although messages associated with requests, probes, responses, and data are described as forwarded from one node to another, the messages themselves may contain variations. In one example, alterations are made to the messages to allow the multiple cluster architecture to be transparent to various local nodes. It should be noted that write requests can be handled as well. In write requests, the targeted memory controller gathers responses and sends the responses to the processor when gathering is complete.
The cache coherence controller 641-1 associated with the remote cluster 640 receives a probe from cache coherence controller 621-2 and probes local nodes associated with cache blocks 645, 647, and 649. Similarly, the cache coherence controller 603-2 associated with the request cluster 600 receives a probe and forwards the probe to local nodes associated with cache blocks 605, 607, and 609 to probe the cache blocks in the request cluster 600. Processor 601-2 receives probe responses from the local nodes associated with cache blocks 605, 607, and 609.
According to various embodiments, cache coherence controller 621-3 accumulates probe responses and sends the probe responses to cache coherence controller 603-3, which in turn forwards the probe responses to the processor 601-3. Cache coherence controller 621-4 also sends a read response to cache coherence controller 603-4, which forwards the read response to processor 601-4. While probes and probe responses carry information for maintaining cache coherency in the system, read responses can carry actual fetched data. After receiving the fetched data, processor 601-4 may send a source done response to cache coherence controller 603- 5. According to various embodiments, the transaction is now complete at the requesting cluster 600. Cache coherence controller 603-5 forwards the source done message to cache coherence controller 621-5. Cache coherence controller 621-5 in turn sends a source done message to memory controller 623-2. Upon receiving the source done message, the memory controller 623-2 can unlock the memory line and the transaction at the home cluster 620 is now complete. Another processor can now access the unlocked memory line.
It should be noted that because the cache coherence controller 621-3 waits for remote cluster probe responses before sending a probe response to cache coherence controller 603-3, delay is introduced into the system. According to various embodiments, probe responses are gathered at cache coherence controller 603-3. By having remote clusters send probe responses through a home cluster, both home cluster probe responses and remote cluster probe responses can be delayed at the home cache coherence controller. In one example, remote cluster probe responses have to travel an additional hop in order to reach a request cluster. The latency for transmission of a probe response between a remote cluster and a request cluster may be substantially less than the latency for transmission of a probe response between a remote cluster and a request cluster through a home cluster. Home cluster probe responses are also delayed as a result of this added hop.
As will be appreciated by one of skill in the art, the specific transaction sequences involving requests, probes, and response messages can vary depending on the specific implementation. In one example, a cache coherence controller 621-3 may wait to receive a read response message from a memory controller 623-1 before transmitting both a probe response message and a read response message to a cache coherence controller 603-3. In other examples, a cache coherence controller may be the actual processor generating the request. Some processors may operate as both a processor and as a cache coherence controller. Furthermore, various data access request messages, probes, and responses associated with reads and writes are contemplated. As noted above, any message for snooping a cache can be referred to as a probe. Similarly, any message for indicating to the memory controller that a memory line should be unlocked can be referred to as a source done message.
It should be noted that the transactions shown in Figure 6 show examples of cache coherence controllers performing many different functions, including functions of remote processors, aggregate local caches, probing agent pairs, and remote memory as described with reference to Figures 5A-5D.
The cache coherence controller 621-1 at the home cluster 620 is acting as a remote processor. When the cache coherence controller receives a request from a request cluster processor, the cache coherence controller is directed to act as the requesting processor on behalf of the request cluster processor. In this case, the cache coherence controller 621-1 accepts a forwarded request from processor 601-1 and sends it to the memory controller 623-1, accumulates responses from all local nodes and the memory controller 623-1, and forwards the accumulated responses and data back to the requesting processor 601-3. The cache coherence controller 621-5 also forwards a source done to the local memory controller 623-2.
The cache coherence controller 603-1 at the request cluster 600 is acting as a remote memory. As remote memory, the cache coherence controller is designed to forward a request from a processor to a proper remote cluster and ensure that local nodes are probed. In this case, the cache coherence controller 603-1 forwards a probe to cache coherence controller 621-1 at a home cluster 620. Cache coherence controller 603-2 also probes local nodes 605, 607, and 609.
The cache coherence controller 641-1 at the request cluster 640 is acting as a probing agent pair. As noted above, when a cache coherence controller acting as a probing agent pair receives a probe from a remote cluster, the cache coherence controller accepts the probe and forwards it to all local nodes. The cache coherence controller accumulates the responses and sends a final response back to the request cluster. Here, the cache coherence controller 641-1 sends a probe to local nodes associated with cache blocks 645, 647, and 649, gathers probe responses and sends the probe responses to cache coherence controller 621-3 at home cluster 620. Similarly, cache coherence controller 603-2 also acts as a probing agent pair at a request cluster 600. The cache coherence controller 603-2 forwards probe requests to local nodes including local nodes associated with cache blocks 605, 607, and 609.
The cache coherence controller 621-2 and 621-3 is also acting as an aggregate remote cache. The cache coherence controller 621-2 is responsible for accepting the probe from the memory controller 623-1 and forwarding the probe to the other processor clusters 600 and 640. More specifically, the cache coherence controller 621-2 forwards the probe to cache coherence controller 603-2 corresponding to request cluster 600 and to cache coherence controller 641-1 corresponding to remote cluster 640. As noted above, using a multiple cluster architecture may introduce delay as well as other undesirable elements such as increased traffic and processing overhead.
Probes are transmitted to all clusters in the multiple cluster system even though not all clusters need to be probed. For example, if a memory line associated with a probe request is invalid or absent from cache, it may not be necessary to probe all of the caches associated with the various clusters. In a system without a coherence directory, it is typically necessary to snoop all clusters. However, by using a coherence directory, the number of transactions in the system can be reduced by probing only a subset of the clusters in a system in order to minimize traffic and processing overhead.
By using a coherence directory, global memory line state information (with respect to each cluster) can be maintained and accessed by a memory controller or a cache coherence controller in a particular cluster. According to various embodiments, the coherence directory tracks and manages the distribution of probes as well as the receipt of responses. If coherence directory information indicates that probing of a specific cluster is not required, the probe to the specific cluster can be eliminated. In one example, a coherence directory indicates that probing of requesting and remote clusters is not necessary. A cache coherence controller in a home cluster probes local nodes without forwarding probes to the request and remote clusters. The cache coherence controller in the home cluster then sends a response to the request cluster after probe responses are received. However, in typical multiple cluster systems, a requesting cluster expects a predetermined number of responses from the various probed clusters. In one example, if the multiple cluster system includes four clusters, a request cluster would expect probe responses associated with nodes in all four clusters.
According to various embodiments, the techniques of the present invention provide a completion bit associated with a probe responses. The completion bit indicates to the requesting cluster that no other probe responses from other clusters should be expected. Any mechanisms for notifying a request cluster that no other probe responses should be expected from other clusters is referred to herein as a completion indicator. In one example, a completion indicator is a completion bit included in the response sent to a request cluster after local nodes are probed. In another example, a completion indicator is separate data transmitted to a request cluster. By using a coherence directory and a completion indicator, the number of transactions associated with probing various clusters can be reduced. For example, with reference to Figure 6, probes to cache coherence controller 603-2 and cache coherence controller 641-1 can be eliminated. A single response with a completion indicator can be transmitted by cache coherence controller 621-4 to the request cluster 600.
Figure 7 is one example of a coherence directory that can be used to allow management and filtering of probes. Various coherence directories are available. In one example, a full directory provides an entry for every memory line in a system. In this example, the coherence directory is maintained at the memory controller and is accessible by a cache coherence controller. However, in a system with a large amount of system memory, a full directory may not be efficient or practical. According to various embodiments, a sparse directory is provided with a limited number of entries associated with a selected set of memory lines. In one example, the coherence directory 701 includes state information 713, dirty data owner information 715, and an occupancy vector 717 associated with the memory lines 711. In some embodiments, the memory line states are modified, owned, shared, and invalid.
In the invalid state, a memory line is not currently available in cache associated with any remote cluster. In the shared state, a memory line may be present in more than one cache, but the memory line has not been modified in any of these caches. When a memory line is in the shared state, an occupancy vector 717 can be checked to determine what caches share the relevant data. An occupancy vector 717 may be implemented as an N-bit string, where each bit represents the availability of the data in the cache of N clusters. Any mechanism for tracking what clusters hold a copy of the relevant memory line in cache is referred to herein as an occupancy vector. The memory line with address 741 is in the shared state, and the occupancy vector 717 indicates that clusters 1 and 3 each have a copy of the shared memory line in cache.
In the modified state, a memory line has been modified and the modified copy exists in cache associated with a particular cluster. When a memory line is modified, dirty data owner information field 715 can be checked to determine the owner of the dirty data. Any mechanism for indicating what cluster owns a modified copy of the memory line in cache is referred to herein as a dirty data owner information field. In one example, the memory line associated with address 781 is modified, and the dirty data owner field 715 indicates that cluster 2 owns the memory line.
In the owned state, a dirty memory line is owned by a single cache but may be resident in multiple caches. It has been read by the owning cache, but has not been modified. In this case, the copy held in memory is stale. If the memory line is in the owned state, dirty data owner field 715 can be accessed to determine which cluster owns the dirty data. In one example, the memory line associated with address 761 is in the owned state and is owned by cluster 4. The occupancy vector 717 can also be checked to determine what other caches may have the relevant data. In this example, the occupancy vector 717 indicates that clusters 2, 3, and 4 each have a copy of the data associated with the memory line in cache.
Although the coherence directory 701 includes the four states of modified, owned, shared, and invalid, it should be noted that particular implementations may use a different set of states. In one example, a system may have the five states of modified, exclusive, owned, shared, and invalid. The techniques of the present invention can be used with a variety of different possible memory line states.
The coherence directory tracks the various transactions such as probe requests and responses in a multiple cluster system to determine when memory lines are added to the coherence directory, when memory lines are removed from the directory, and when information associated with each memory line is updated. By using the coherence directory, the techniques of the present invention recognize that the number of transactions such as probe requests can be reduced by managing or filtering probes that do not need to be sent to specific clusters.
Figure 8 is a diagrammatic representation showing probe filter information that can be used to reduce the number of transactions in a multiple cluster system. Any criterion that can be used to reduce the number of clusters probed from a home cluster is referred to herein as probe filter information. Transactions such as probe requests can have a variety of characteristics. Some characteristics include whether the probe is a read block (read) 823 or a read block modify (read/write) 825. Other characteristics of the probe include the state of the memory line associated with the probe. In some examples, states include invalid 831, shared 833, owned 835, and modified 837. According to various embodiments, a coherence directory maintains information for memory lines in the local cluster that are cached in non-local clusters, where non-local clusters can include request and remote clusters. If the state of the memory line associated with a probe is invalid 831 as indicated in the coherence directory, no copies of the memory line reside in other clusters. Consequently, only the home cluster needs to be probed and a completion bit can be used to indicate to a request cluster that the request cluster should expect only a single response from home cluster instead of a response from each of the clusters. If the memory line associated with the probe is in the shared state 833, and the transaction is a read transaction, only the home cluster needs to be probed and a completion bit can again be used to indicate to the request cluster that only a single response from home cluster should be expected as indicated in entry 803. It should be noted that the completed bit can not be used in entries 811 and 813.
For read transactions on owned memory lines, only the remote cluster with the line cached in the owned state needs to be probed. The remote cluster can transmit the response with a completion bit back to a request cluster. For transactions on modified memory lines, the probe can be sent to the remote cluster with the line cached in the modified state. Although transactions such as read block (read) and read block modify (read/write) are described, it should be noted that other transactions such as test and test and set are contemplated.
Figure 9 is a diagrammatic representation depicting one example of transactions for probing only a home cluster as indicated in entries 801, 809, and 803 in Figure 8. According to various embodiments, processor 901-1 in a local cluster 900 sends a data access request such as a read request to a cache coherence controller 903-1. The cache coherence controller 903-1 forwards the request to a cache coherence controller 921-1 in a home cluster 920. The cache coherence controller 921-1 at the home cluster 920 receives the access request and forwards the access request to a memory controller 923-1, which then probes local nodes 925, 927, and cache coherence controller 921-2. It should be noted that a cache coherency controller 921-1 is typically responsible for updating the coherence directory during various transactions. The cache coherence controller 921-2 determines characteristics associated with the probe from the memory controller 923-1 to determine whether remote probes are needed and whether a completion bit can be used. Here, the cache coherence controller 921-2 determines that no remote probes are needed and does not forward probes to the remote cluster 940 or to request cluster 900.
After cache coherence controller 921-4 receives the probe responses from local nodes as well as the read response from the memory controller 923-1, the response message with a completion indicator is transmitted to the request cluster. With the completion indicator, the request cluster does not wait for additional responses from other clusters. The coherence controller 903-4 forwards the response with the completion bit set to CPU 901-4. After receiving the response with the completion bit set, the CPU does not wait for additional responses from the local caches. CPU 901-4 forwards a source done message to cache coherence controller 903-5 to home cluster cache coherence controller 921-5, which can then perform updates of its coherence directory. The source done is then forwarded to memory controller 923-1. Figure 9 shows one example of a sequence where only the home cluster needs to be probed. Figure 10 shows one example of a sequence where only a single remote cluster needs to be probed. Figure 10 is a diagrammatic representation depicting an example of transactions for probing a remote cluster as indicated in entries 805, 807, and 815 in Figure 8. According to various embodiments, processor 1001-1 in a local cluster 1000 sends a data access request such as a read request to a cache coherence controller 1003-1. The cache coherence controller 1003-1 forwards the request to a cache coherence controller 1021-1 in a home cluster 1020. The cache coherence controller 1021-1 at the home cluster 1020 receives the access request and forwards the access request to a memory controller 1023-1, which then probes local nodes 1025, 1027, and cache coherence controller 1021-2. The cache coherence controller 1021-2 determines characteristics associated with the probe from the memory controller 1023-1 to determine whether remote probes are needed and whether a completion bit can be used. Here, the cache coherence controller 1021-2 determines that only a remote cluster needs to be probed and does not forward a probe to request cluster 1000.
After cache coherence controller 1021-4 receives the probes from local nodes as well as the read response from the memory controller 1023-1, a response message is not transmitted to the request cluster because the remote cluster is sending a response message with a completion indicator is transmitted to the request cluster. With the completion indicator, the request cluster does not wait for additional responses from other clusters. The response is forwarded to CPU 1001-4 and a source done message is sent from cache coherence controller 1003-5 to home cluster cache coherence controller 1021-5. With the completion bit set in the response to CPU 1001-4, it does not wait for any other local responses. After all responses from local nodes are received, the source done is then forwarded to memory controller 1023-1, which can then perform updates of its coherence directory.
Figure 11 is a process flow diagram showing one example of a technique for handling probe requests at a home cache coherence controller. At 1101, a probe request associated with a memory line is received. At 1105, the cache coherence controller forwards the request to the memory controller. At 1109, the cache coherence controller receives a probe from the memory controller and accesses a coherence directory and probe filter information at 1113 to determine whether the number of probes to various clusters in the system can be reduced. At 1121, it is determined whether filtering and a completion indicator can be used. In one example, it is determined the filtering a completion indicator can be used by identifying the criteria specified in Figure 8 and by accessing a coherence directory and shown in Figure 7.
If a completion indicator can not be used, probes are broadcast to the various nodes with no filtering and no completion bit 1145. If filtering and a completion indicator can be used, it is determined at 1131 if a remote cluster should be probed. If a single remote cluster is the cluster that should be probed, the probe is forwarded with the completion indicator to the remote cluster at 1135. At 1139, home cluster probe responses are received but are not forwarded to the request cluster. The response is not sent to the request cluster from home cluster because a remote cluster is sending a response with a completion indicator to the request cluster.
At 1149, source done information is received from the request cluster and forwarded to the memory controller. If it is determined at 1131 that only the home cluster needs to be probed, then the cache coherence controller at 1141 does not send probes to any request or remote clusters and instead sends a response to the request cluster with a completion indicator. The cache coherence controller sends the response with the completion indicator after receiving home cluster probe responses. At 1149, the cache coherence controller at the home cluster receives source done information from the request cluster and forwards the source done information to the memory controller.
According to various embodiments, when the only cluster that needs to be probed is the home cluster, only the nodes in the home cluster are probed. No probes are transmitted to any request or remote cluster. However, when the only cluster that needs to be probed is a remote or request cluster, not only are the nodes in the remote cluster probed, but the nodes in the home cluster are probed as well. According to various embodiments, the techniques of the present invention provide that when only a remote or request cluster needs to be probed, the memory controller can sometimes be bypassed to allow probing of only the remote or request cluster. In one example, a probe is not forwarded to the home cluster memory controller and a probe is forwarded directly to the remote cluster from the home cluster cache coherence controller.
Figure 12 is a diagrammatic representation showing exemplary memory controller filter information. Any criterion used to reduce the number of requests forwarded to a memory controller is referred to herein as memory controller filter information. Characteristics of a probe can again be analyzed when a cache coherence controller receives the probe from a request cluster. Transactions such as probe requests can have a variety of characteristics. Some characteristics include whether the probe is a read block (read) 1223 or a read block modify (read/write) 1225. Other characteristics of the probe include the state of the memory line associated with the probe. In some examples, states include invalid 1231, shared 1233, owned 1235, and modified 1237. When the state of the memory line associated with the probe is invalid 1231, no remote probes are required because not remote clusters have a copy of the memory line in cache. The cache coherence controller also has no knowledge of the home cluster cache state. The probe is consequently forwarded to the memory controller.
For read block transactions on a shared memory line 1203, there is no need to probe the remote clusters as the home cluster contains a valid copy of the memory line in either cache or the memory controller. Consequently the probe is forwarded to the memory controller. For read block modify transactions on shared memory lines 1211, the local node state is unknown and the probe is sent to the memory controller.
For read block transactions on an owned memory line 1205, there is no need to send a probe request to the target or probe local nodes as the owned state implies that the home cluster caches are invalid or shared. A probe is forwarded directly to the owning cluster to acquire the cached data. For read block write transactions on an owned memory line 1213, the local state is unknown and consequently the request is forwarded to the memory controller. When the state of the memory line associated with the request is modified 1237, there is no need to probe local nodes, as a modified state implies the home cluster state is invalid. A probe is forwarded to the cluster owning the memory line.
Figure 13 shows one example of a sequence where a probe does not need to be forwarded to the home cluster memory controller. According to various embodiments, processor 1301-1 in a local cluster 1300 sends a data access request such as a read request to a cache coherence controller 1303-1. The cache coherence controller 1303-1 forwards the request to a cache coherence controller 1321-1 in a home cluster 1320. The cache coherence controller 1321-1 at the home cluster 1320 receives the access request and determines whether the memory controller can be bypassed. Forwarding a probe to a remote or request cluster without forwarding the probe to a memory controller is referred to herein as bypassing the memory controller. In one embodiment, the determination can be made by using memory controller filter information. If the probe characteristics fall within entries 1205, 1207, or 1215, the memory controller is bypassed and the probe is forwarded to cache coherence controller 1341-1 in the remote cluster 1340. In one example, the probe is forwarded with an indication that a completion bit should be used.
The cache coherence controller 1321-1 in the home cluster 1320 is acting as a serialization point in place of the memory controller to maintain cache coherency. Once it is determined that the memory controller can be bypassed, the cache coherence controller 1321-1 blocks all other incoming requests and outgoing probes until a final source done is received from the request cluster. The remote cluster cache coherence controller 1341-1 probes remote cluster nodes and sends a response with a completion indicator to the request cluster 1300. The response is forwarded to CPU 1301-4 and a source done message is sent from cache coherence controller 1303-5 to home cluster cache coherence controller 1321-5. The source done is not forwarded to the memory controller, because the memory controller never processed the transaction.
Figure 14 is a flow process diagram showing probe request handling at a home cache coherence controller using memory controller filter information. At 1401, a probe request associated with a particular memory line is received. At 1403, characteristics associated with the probe request are identified. At 1411, it is determined if the memory controller can be bypassed. According to various embodiments, memory controller filter information shown in Figure 12 is used to determine whether a memory controller can be bypassed. If it is determined that a memory controller can be bypassed, probe requests associated with the same memory line are blocked at 1415 and a probe is sent to a remote or a request cluster. At 1417, the memory line is unblocked after receiving a source done from the request cluster. If it is determined at 1411 that a memory controller can not be bypassed, the probe request is forwarded to a serialization point 1405. The transaction sequence can then proceed with or without probe filtering and a completion indicator as shown in 1109 of Figure 11.
While the invention has been particularly shown and described with reference to specific embodiments thereof, it will be understood by those skilled in the art that changes in the form and details of the disclosed embodiments may be made without departing from the spirit or scope of the invention. For example, embodiments of the present invention may be employed with multiple processor clusters connected through a point-to-point, switch, or bus architecture. In another example, multiple clusters of processors may share a single cache coherence controller, or multiple cache coherence controllers can be used in" a single cluster. Therefore, the scope of the invention should be determined with reference to the appended claims.

Claims

WHAT IS CLAIMED IS:
1. A computer system, comprising: a home cluster including a first plurality of processors and a home cache coherence controller, the first plurality of processors and the home cache coherence controller interconnected in a point-to-point architecture; wherein the home cache coherence controller is configured to receive a probe request and probe one or more selected clusters, wherein the one or more clusters are selected based on the characteristics associated with the probe request.
2. The computer system of claim 1, wherein the one or more clusters are selected using probe filter information associated with a cache coherence directory.
3. The computer system of any of claims 1 -2, wherein the cache coherence directory identifies the state of a plurality of memory lines.
4. The computer system of any of claims 1 -3, wherein the characteristics associated with the probe request include whether the probe request is a read or a read/write request.
5. The computer system of any of claim 1 -4, wherein the characteristics associated with the probe request include the state of the memory line associated with the probe request.
6. The computer system of any of claim 1 -5, wherein probing a selected cluster comprises probing nodes in the selected cluster and sending a response with a completion indicator back to a request cluster.
7. The computer system of claim 6, wherein the completion indicator notifies the request cluster that the probe request is completed.
8. The computer system of claim 6, wherein the completion indicator notifies the request cluster that no other responses associated with the probe request will be transmitted to the request cluster.
9. The computer system of claim 6, wherein the completion indicator is a completion bit.
10. The computer system of claim 6, wherein the completion indicator can be used if the probe request is a read request.
11. The computer system of claim 6, wherein the completion indicator can be used if the probe request is a read/modify request and the state associated with the memory line is invalid or modified.
12. The computer system of claim 6, wherein the completion indicator can not be used if the probe request is a read/modify request and a single remote cluster has the memory line associated with the probe request cached in a shared state.
13. The computer system of claim 6, wherein the completion indicator can not be used if the probe request is a read/modify request and no other clusters other than a single remote cluster has the memory line associated with the probe request cached in an owned state.
14. The computer system of any of claim 1-13, wherein a home cluster is selected for probing.
15. The computer system of claim 14, wherein the home cluster is selected for probing if the memory line associated with the probe request is in an invalid state.
16. The computer system of claim 14j wherein the home cluster is selected for probing if the probe request is a read request and the memory line associated with the probe request is in a shared state.
17. The computer system of claim 16, wherein probing a home cluster comprises probing nodes in the home cluster, receiving probe responses, and sending a response with a completion indicator to a request cluster.
18. The computer system of claim 17, wherein probing nodes in the home cluster includes forwarding the probe request to a home cluster memory controller.
19. The computer system of claim 1 , wherein a remote cluster is selected for probing.
20. The computer system of claim 19, wherein a remote cluster is selected for probing if the memory line associated with the probe request is in a modified state.
21. The computer system of claim 19, wherein a remote cluster is selected for probing if the probe request is a read request and the memory line associated with the probe request is in an owned state.
22. The computer system of claim 19, wherein a remote cluster is selected for probing if the probe request is a read/write request and the single remote cluster has the memory lines associated with the probe request cached in a shared state.
23. The computer system of claim 19, wherein a remote cluster is selected for probing if the probe request is a read/write request and no other clusters other than the single remote cluster has the memory line associated with the probe request cached in an owned state.
24. The computer system of claim 19, wherein probing a remote cluster comprises probing nodes in the home cluster, forwarding a probe request with a completion indicator to the remote cluster, and receiving probe responses from the home cluster nodes.
25. The computer system of claim 24, wherein the remote cluster receives the probe request with the completion indicator, probe remote cluster nodes, and sends a response with a completion indicator to the request cluster.
26. A method for managing probes, the method comprising: receiving a probe request at a home cache coherence controller in a home cluster, the home cluster including a first plurality of processors and the home cache coherence controller, the first plurality of processors and the home cache coherence controller interconnected in a point-to-point architecture; determining if the probe request should bypass a memory controller associated with the home cluster.
27. The method of claim 26, wherein bypassing the memory controller associated with the home cluster comprises forwarding the probe request to a remote cluster without forwarding the probe request to the memory controller.
28. The method of claim 26, wherein bypassing the memory controller associated with the home cluster comprises forwarding the probe request to a request cluster without forwarding the probe request to the memory controller.
29. The method of claim 26, wherein the probe request is associated with a particular memory line.
30. The method of claim 29, wherein cache coherence controller is further configured to block subsequent probe requests associated with the same memory line until a transaction complete indicator is received from a request cluster.
31. The method of claim 30, wherein the transaction complete indicator is a source done message from the home cache coherence controller.
32. An apparatus for managing probes, the apparatus comprising: means for receiving a probe request at a home cache coherence controller in a home cluster, the home cluster including a first plurality of processors and the home cache coherence controller, the first plurality of processors and the home cache coherence controller interconnected in a point-to-point architecture; means for determining if the probe request should bypass a memory controller associated with the home cluster.
33. The apparatus of claim 32, wherein bypassing the memory controller associated with the home cluster comprises forwarding the probe request to a remote cluster without forwarding the probe request to the memory controller.
34. The apparatus of claim 32, wherein bypassing the memory controller associated with the home cluster comprises forwarding the probe request to a request cluster without forwarding the probe request to the memory controller.
35. The apparatus of claim 32, wherein the probe request is associated with a particular memory line.
36. The apparatus of claim 35, wherein cache coherence controller is further configured to block subsequent probe requests associated with the same memory line until a transaction complete indicator is received from a request cluster.
37. The apparatus of claim 36, wherein the transaction complete indicator is a source done message from the home cache coherence controller.
PCT/US2003/034756 2002-11-04 2003-10-30 Methods and apparatus for managing probe requests WO2004042585A2 (en)

Priority Applications (4)

Application Number Priority Date Filing Date Title
AU2003287415A AU2003287415A1 (en) 2002-11-04 2003-10-30 Methods and apparatus for managing probe requests
EP03781650A EP1563387A2 (en) 2002-11-04 2003-10-30 Methods and apparatus for managing probe requests
JP2004550375A JP2006505855A (en) 2002-11-04 2003-10-30 Method and apparatus for managing probe requests
CA002505016A CA2505016A1 (en) 2002-11-04 2003-10-30 Methods and apparatus for managing probe requests

Applications Claiming Priority (4)

Application Number Priority Date Filing Date Title
US10/288,347 2002-11-04
US10/288,399 US7103726B2 (en) 2002-11-04 2002-11-04 Methods and apparatus for managing probe requests
US10/288,399 2002-11-04
US10/288,347 US7003633B2 (en) 2002-11-04 2002-11-04 Methods and apparatus for managing probe requests

Publications (2)

Publication Number Publication Date
WO2004042585A2 true WO2004042585A2 (en) 2004-05-21
WO2004042585A3 WO2004042585A3 (en) 2005-03-03

Family

ID=32314361

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/US2003/034756 WO2004042585A2 (en) 2002-11-04 2003-10-30 Methods and apparatus for managing probe requests

Country Status (6)

Country Link
US (1) US7346744B1 (en)
EP (1) EP1563387A2 (en)
JP (1) JP2006505855A (en)
AU (1) AU2003287415A1 (en)
CA (1) CA2505016A1 (en)
WO (1) WO2004042585A2 (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2007183702A (en) * 2006-01-04 2007-07-19 Hitachi Ltd Cache consistency control method, chip set and multiprocessor system

Families Citing this family (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US7664847B2 (en) * 2003-08-14 2010-02-16 Oracle International Corporation Managing workload by service
US7747717B2 (en) * 2003-08-14 2010-06-29 Oracle International Corporation Fast application notification in a clustered computing system
US20060064400A1 (en) * 2004-09-21 2006-03-23 Oracle International Corporation, A California Corporation Methods, systems and software for identifying and managing database work
US7681104B1 (en) * 2004-08-09 2010-03-16 Bakbone Software, Inc. Method for erasure coding data across a plurality of data stores in a network
US7681105B1 (en) * 2004-08-09 2010-03-16 Bakbone Software, Inc. Method for lock-free clustered erasure coding and recovery of data across a plurality of data stores in a network
US7555605B2 (en) * 2006-09-28 2009-06-30 Freescale Semiconductor, Inc. Data processing system having cache memory debugging support and method therefor
US8713186B2 (en) 2007-03-13 2014-04-29 Oracle International Corporation Server-side connection resource pooling
US20080288725A1 (en) 2007-05-14 2008-11-20 Moyer William C Method and apparatus for cache transactions in a data processing system
US8447934B2 (en) * 2010-06-30 2013-05-21 Advanced Micro Devices, Inc. Reducing cache probe traffic resulting from false data sharing
US10474653B2 (en) 2016-09-30 2019-11-12 Oracle International Corporation Flexible in-memory column store placement

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6085295A (en) * 1997-10-20 2000-07-04 International Business Machines Corporation Method of maintaining data coherency in a computer system having a plurality of interconnected nodes
US6108737A (en) * 1997-10-24 2000-08-22 Compaq Computer Corporation Method and apparatus for reducing latency of inter-reference ordering in a multiprocessor system
US20020046327A1 (en) * 2000-06-10 2002-04-18 Kourosh Gharachorloo Cache coherence protocol engine and method for processing memory transaction in distinct address subsets during interleaved time periods in a multiprocessor system

Family Cites Families (57)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6463529B1 (en) 1989-11-03 2002-10-08 Compaq Computer Corporation, Inc. Processor based system with system wide reset and partial system reset capabilities
US5396635A (en) 1990-06-01 1995-03-07 Vadem Corporation Power conservation apparatus having multiple power reduction levels dependent upon the activity of the computer system
US5195089A (en) 1990-12-31 1993-03-16 Sun Microsystems, Inc. Apparatus and method for a synchronous, high speed, packet-switched bus
US5394555A (en) 1992-12-23 1995-02-28 Bull Hn Information Systems Inc. Multi-node cluster computer system incorporating an external coherency unit at each node to insure integrity of information stored in a shared, distributed memory
JP2819982B2 (en) 1993-03-18 1998-11-05 株式会社日立製作所 Multiprocessor system with cache match guarantee function that can specify range
JPH08185359A (en) 1994-10-31 1996-07-16 Toshiba Corp Memory subsystem
US5692123A (en) 1994-12-07 1997-11-25 Cray Research, Inc. Maintenance channel for modulator, highly interconnected computer systems
US5950226A (en) * 1996-07-01 1999-09-07 Sun Microsystems, Inc. Multiprocessing system employing a three-hop communication protocol
US6658526B2 (en) 1997-03-12 2003-12-02 Storage Technology Corporation Network attached virtual data storage subsystem
US6018791A (en) 1997-04-14 2000-01-25 International Business Machines Corporation Apparatus and method of maintaining cache coherency in a multi-processor computer system with global and local recently read states
FR2763714B1 (en) 1997-05-26 1999-07-02 Bull Sa REPLACEMENT COUNTERS FOR MACHINE WITH NON-UNIFORM ACCESS MEMORY
US6209065B1 (en) 1997-10-24 2001-03-27 Compaq Computer Corporation Mechanism for optimizing generation of commit-signals in a distributed shared-memory system
US6542926B2 (en) 1998-06-10 2003-04-01 Compaq Information Technologies Group, L.P. Software partitioned multi-processor system with flexible resource sharing levels
US6292705B1 (en) 1998-09-29 2001-09-18 Conexant Systems, Inc. Method and apparatus for address transfers, system serialization, and centralized cache and transaction control, in a symetric multiprocessor system
US6633945B1 (en) 1997-12-07 2003-10-14 Conexant Systems, Inc. Fully connected cache coherent multiprocessing systems
US6292906B1 (en) 1997-12-17 2001-09-18 Intel Corporation Method and apparatus for detecting and compensating for certain snoop errors in a system with multiple agents having cache memories
US6334172B1 (en) 1998-02-17 2001-12-25 International Business Machines Corporation Cache coherency protocol with tagged state for modified values
US6192451B1 (en) 1998-02-17 2001-02-20 International Business Machines Corporation Cache coherency protocol for a data processing system including a multi-level memory hierarchy
US6330643B1 (en) 1998-02-17 2001-12-11 International Business Machines Corporation Cache coherency protocols with global and local posted operations
US6631448B2 (en) 1998-03-12 2003-10-07 Fujitsu Limited Cache coherence unit for interconnecting multiprocessor nodes having pipelined snoopy protocol
US6038652A (en) 1998-09-30 2000-03-14 Intel Corporation Exception reporting on function generation in an SIMD processor
US6067603A (en) 1998-10-01 2000-05-23 International Business Machines Corporation Non-uniform memory access (NUMA) data processing system that speculatively issues requests on a node interconnect
US6338122B1 (en) 1998-12-15 2002-01-08 International Business Machines Corporation Non-uniform memory access (NUMA) data processing system that speculatively forwards a read request to a remote processing node
US6490661B1 (en) * 1998-12-21 2002-12-03 Advanced Micro Devices, Inc. Maintaining cache coherency during a memory read operation in a multiprocessing computer system
US6189078B1 (en) 1998-12-22 2001-02-13 Unisys Corporation System and method for increasing data transfer throughput for cache purge transactions using multiple data response indicators to maintain processor consistency
US6167492A (en) 1998-12-23 2000-12-26 Advanced Micro Devices, Inc. Circuit and method for maintaining order of memory access requests initiated by devices coupled to a multiprocessor system
US6279085B1 (en) * 1999-02-26 2001-08-21 International Business Machines Corporation Method and system for avoiding livelocks due to colliding writebacks within a non-uniform memory access system
US6467007B1 (en) 1999-05-19 2002-10-15 International Business Machines Corporation Processor reset generated via memory access interrupt
US6751698B1 (en) 1999-09-29 2004-06-15 Silicon Graphics, Inc. Multiprocessor node controller circuit and method
US6405289B1 (en) 1999-11-09 2002-06-11 International Business Machines Corporation Multiprocessor system in which a cache serving as a highest point of coherency is indicated by a snoop response
US6704842B1 (en) 2000-04-12 2004-03-09 Hewlett-Packard Development Company, L.P. Multi-processor system with proactive speculative data transfer
US20010037435A1 (en) 2000-05-31 2001-11-01 Van Doren Stephen R. Distributed address mapping and routing table mechanism that supports flexible configuration and partitioning in a modular switch-based, shared-memory multiprocessor computer system
US6640287B2 (en) 2000-06-10 2003-10-28 Hewlett-Packard Development Company, L.P. Scalable multiprocessor system and cache coherence method incorporating invalid-to-dirty requests
US6751721B1 (en) 2000-08-31 2004-06-15 Hewlett-Packard Development Company, L.P. Broadcast invalidate scheme
US6633960B1 (en) 2000-08-31 2003-10-14 Hewlett-Packard Development Company, L.P. Scalable directory based cache coherence protocol
JP3860116B2 (en) 2000-10-31 2006-12-20 ミレニアル・ネット・インコーポレーテッド Network processing system with optimized power efficiency
US6738871B2 (en) 2000-12-22 2004-05-18 International Business Machines Corporation Method for deadlock avoidance in a cluster environment
US6738872B2 (en) 2000-12-22 2004-05-18 International Business Machines Corporation Clustered computer system with deadlock avoidance
US6738870B2 (en) 2000-12-22 2004-05-18 International Business Machines Corporation High speed remote storage controller
US7234029B2 (en) 2000-12-28 2007-06-19 Intel Corporation Method and apparatus for reducing memory latency in a cache coherent multi-node architecture
US6615319B2 (en) 2000-12-29 2003-09-02 Intel Corporation Distributed mechanism for resolving cache coherence conflicts in a multi-node computer architecture
US6754782B2 (en) 2001-06-21 2004-06-22 International Business Machines Corporation Decentralized global coherency management in a multi-node computer system
US6760809B2 (en) 2001-06-21 2004-07-06 International Business Machines Corporation Non-uniform memory access (NUMA) data processing system having remote memory cache incorporated within system memory
US6760819B2 (en) 2001-06-29 2004-07-06 International Business Machines Corporation Symmetric multiprocessor coherence mechanism
US6839808B2 (en) 2001-07-06 2005-01-04 Juniper Networks, Inc. Processing cluster having multiple compute engines and shared tier one caches
US7107409B2 (en) 2002-03-22 2006-09-12 Newisys, Inc. Methods and apparatus for speculative probing at a request cluster
US7107408B2 (en) 2002-03-22 2006-09-12 Newisys, Inc. Methods and apparatus for speculative probing with early completion and early request
US7103725B2 (en) 2002-03-22 2006-09-05 Newisys, Inc. Methods and apparatus for speculative probing with early completion and delayed request
US20030195939A1 (en) 2002-04-16 2003-10-16 Edirisooriya Samatha J. Conditional read and invalidate for use in coherent multiprocessor systems
US7395379B2 (en) 2002-05-13 2008-07-01 Newisys, Inc. Methods and apparatus for responding to a request cluster
US7653790B2 (en) 2002-05-13 2010-01-26 Glasco David B Methods and apparatus for responding to a request cluster
US7155525B2 (en) 2002-05-28 2006-12-26 Newisys, Inc. Transaction management in systems having multiple multi-processor clusters
US6865595B2 (en) 2002-05-28 2005-03-08 Newisys, Inc. Methods and apparatus for speculative probing of a remote cluster
US7003633B2 (en) * 2002-11-04 2006-02-21 Newisys, Inc. Methods and apparatus for managing probe requests
US7103726B2 (en) 2002-11-04 2006-09-05 Newisys, Inc. Methods and apparatus for managing probe requests
US6920532B2 (en) * 2002-11-05 2005-07-19 Newisys, Inc. Cache coherence directory eviction mechanisms for modified copies of memory lines in multiprocessor systems
US7069392B2 (en) 2003-06-12 2006-06-27 Newisys, Inc. Methods and apparatus for extended packet communications between multiprocessor clusters

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6085295A (en) * 1997-10-20 2000-07-04 International Business Machines Corporation Method of maintaining data coherency in a computer system having a plurality of interconnected nodes
US6108737A (en) * 1997-10-24 2000-08-22 Compaq Computer Corporation Method and apparatus for reducing latency of inter-reference ordering in a multiprocessor system
US20020046327A1 (en) * 2000-06-10 2002-04-18 Kourosh Gharachorloo Cache coherence protocol engine and method for processing memory transaction in distinct address subsets during interleaved time periods in a multiprocessor system

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2007183702A (en) * 2006-01-04 2007-07-19 Hitachi Ltd Cache consistency control method, chip set and multiprocessor system

Also Published As

Publication number Publication date
WO2004042585A3 (en) 2005-03-03
JP2006505855A (en) 2006-02-16
CA2505016A1 (en) 2004-05-21
EP1563387A2 (en) 2005-08-17
AU2003287415A1 (en) 2004-06-07
US7346744B1 (en) 2008-03-18

Similar Documents

Publication Publication Date Title
US7024521B2 (en) Managing sparse directory evictions in multiprocessor systems via memory locking
US7296121B2 (en) Reducing probe traffic in multiprocessor systems
US7334089B2 (en) Methods and apparatus for providing cache state information
US7003633B2 (en) Methods and apparatus for managing probe requests
US6920532B2 (en) Cache coherence directory eviction mechanisms for modified copies of memory lines in multiprocessor systems
US6934814B2 (en) Cache coherence directory eviction mechanisms in multiprocessor systems which maintain transaction ordering
EP1561162B1 (en) Methods and apparatus for multiple cluster locking
US6925536B2 (en) Cache coherence directory eviction mechanisms for unmodified copies of memory lines in multiprocessor systems
US7107408B2 (en) Methods and apparatus for speculative probing with early completion and early request
US20030225978A1 (en) Methods and apparatus for speculative probing of a remote cluster
US7103725B2 (en) Methods and apparatus for speculative probing with early completion and delayed request
US7272688B1 (en) Methods and apparatus for providing cache state information
US7103636B2 (en) Methods and apparatus for speculative probing of a remote cluster
EP1652091B1 (en) Methods and apparatus for providing early responses from a remote data cache
US7103726B2 (en) Methods and apparatus for managing probe requests
EP1563387A2 (en) Methods and apparatus for managing probe requests
US7337279B2 (en) Methods and apparatus for sending targeted probes
US7162589B2 (en) Methods and apparatus for canceling a memory data fetch

Legal Events

Date Code Title Description
AK Designated states

Kind code of ref document: A2

Designated state(s): AE AG AL AM AT AU AZ BA BB BG BR BY BZ CA CH CN CO CR CU CZ DE DK DM DZ EC EE EG ES FI GB GD GE GH GM HR HU ID IL IN IS JP KE KG KP KR KZ LC LK LR LS LT LU LV MA MD MG MK MN MW MX MZ NI NO NZ OM PG PH PL PT RO RU SC SD SE SG SK SL SY TJ TM TN TR TT TZ UA UG UZ VC VN YU ZA ZM ZW

AL Designated countries for regional patents

Kind code of ref document: A2

Designated state(s): BW GH GM KE LS MW MZ SD SL SZ TZ UG ZM ZW AM AZ BY KG KZ MD RU TJ TM AT BE BG CH CY CZ DE DK EE ES FI FR GB GR HU IE IT LU MC NL PT RO SE SI SK TR BF BJ CF CG CI CM GA GN GQ GW ML MR NE SN TD TG

121 Ep: the epo has been informed by wipo that ep was designated in this application
WWE Wipo information: entry into national phase

Ref document number: 2004550375

Country of ref document: JP

WWE Wipo information: entry into national phase

Ref document number: 2505016

Country of ref document: CA

WWE Wipo information: entry into national phase

Ref document number: 2003287415

Country of ref document: AU

WWE Wipo information: entry into national phase

Ref document number: 2003781650

Country of ref document: EP

WWP Wipo information: published in national office

Ref document number: 2003781650

Country of ref document: EP