International Business Machines Corporation

Jerry Don Lewis

Ravi Kumar Arimilli

John Steven Dodson

Jody B Joyner

Guy L Guthrie

Derek E Williams

Leo James Clark

Lakshminarayana Baba Arimilli

John Michael Kaiser

Timothy Michael Skergan

John Steve Dodson

James Stephen Fields, Jr

Lakshminarayanan Baba Arimilli

Warren Edward Maule

Bernard Charles Drerup

Vicente Enrique Chung

Bradley David McCredie

Gary Alan Morrison

Michael Stephen Floyd

According to method of data processing in a multiprocessor data processing system, in response to a processor request to modify a target granule of a target cache line of data containing multiple granules, a processing unit originates on an interconnect of the multiprocessor data processing system a data-claim-partial request that requests permission to promote only the target granule of the target cache line to a unique copy with an intent to modify the target granule. In response to a combined response to the data-claim-partial request indicating success (the combined response representing a system-wide response to the data-claim-partial-request), the processing unit promotes only the target granule of the target cache line to a unique copy by updating a coherency state of the target granule and retaining a coherency state of at least one other granule of the target cache line.

Claiming coherency ownership of a partial cache line of data

According to at least one embodiment, a method of data processing in a multiprocessor data processing system includes a requesting processing unit initiating an interconnect operation including a memory access request that indicates an acceptability of a variable amount of data to service the interconnect request for data. In response to snooping the memory access request on an interconnect, a snooper selects an amount of data to supply to the requesting processing unit and transmits the selected amount of data to the requesting processing unit. The requesting processing unit receives the selected amount of data and utilizes at least some of the selected amount of data to service a processor request.

Interconnect operation indicating acceptability of partial data delivery

A method of data processing in a cache memory includes caching a plurality of cache lines of data in a corresponding plurality of entries in a cache array, where each of the plurality of cache lines includes multiple data granules. For each of the plurality of cache entries, a plurality of line coherency state fields indicates an associated coherency state applicable to two or more data granules. For at least a particular cache line among the plurality of cache lines, a granule coherency state field indicates a coherency state for a particular granule of the multiple data granules in the particular cache line, where the coherency state field indicated by the granule coherency state field differs from that indicated for the particular cache line by its line coherency state field.

Cache management for partial cache line operations

Mechanisms for performing dynamic request routing based on broadcast depth queue information are provided. Each processor chip in the system may use a synchronized heartbeat signal it generates to provide queue depth information to each of the other processor chips in the system. The queue depth information identifies a number of requests or amount of data in each of the queues of a processor chip that originated the heartbeat signal. The queue depth information from each of the processor chips in the system may be used by the processor chips in determining optimal routing paths for data from a source processor chip to a destination processor chip. As a result, the congestion of data for processing at each of the processor chips along each possible routing path may be taken into account when selecting to which processor chip to forward data.

Performing dynamic request routing based on broadcast queue depths

According to a method of data processing in a multiprocessor data processing system, in response to a processor request to read a target granule of a target cache line of data containing multiple granules, a processing unit originates on an interconnect of the multiprocessor data processing system a partial read request that requests permission to read only the target granule of the target cache line. In response to a combined response to the partial read request indicating success, the combined response representing a system-wide response to the partial read request, the processing unit receives the target granule of the target cache line, supplies the target granule to a requesting processor core, and updates a coherency state of the target granule while retaining a coherency state of at least one other granule of the target cache line.

Partial cache line accesses based on memory access patterns

A method of data processing in a processing unit supported by a memory hierarchy includes the processing unit performing a plurality of memory accesses to the memory hierarchy. The plurality of memory accesses includes one or more memory accesses targeting a full cache line of data. The processing unit monitors utilization of data accessed by the plurality of memory accesses, and based upon the utilization of the data, dynamically alters a memory access mode of operation so that a subsequent storage-modifying memory access targets less than a full cache line of data.

Dynamic selection of a memory access size

Mechanisms for providing a cluster-wide system clock in a multi-tiered full graph (MTFG) interconnect architecture are provided. Heartbeat signals transmitted by each of the processor chips in the computing cluster are synchronized. Internal system clock signals are generated in each of the processor chips based on the synchronized heartbeat signals. As a result, the internal system clock signals of each of the processor chips are synchronized since the heartbeat signals, that are the basis for the internal system clock signals, are synchronized. Mechanisms are provided for performing such synchronization using direct couplings of processor chips within the same processor book, different processor books in the same supernode, and different processor books in different supernodes of the MTFG interconnect architecture.

Cluster-wide system clock in a multi-tiered full-graph interconnect architecture

A system for providing a cluster-wide system clock in a multi-tiered full graph (MTFG) interconnect architecture are provided. Heartbeat signals transmitted by each of the processor chips in the computing cluster are synchronized. Internal system clock signals are generated in each of the processor chips based on the synchronized heartbeat signals. As a result, the internal system clock signals of each of the processor chips are synchronized since the heartbeat signals, that are the basis for the internal system clock signals, are synchronized. Mechanisms are provided for performing such synchronization using direct couplings of processor chips within the same processor book, different processor books in the same supernode, and different processor books in different supernodes of the MTFG interconnect architecture.

System for providing a cluster-wide system clock in a multi-tiered full-graph interconnect architecture

A mechanism for performing dynamic request routing based on broadcast source request information is provided. Each processor chip in the system may use a synchronized heartbeat signal it generates to provide source request information to each of the other processor chips in the system. The source request information identifies the number of active source requests sent by the processor chip that originated the heartbeat signal. The source request information from each of the processor chips in the system may be used by the processor chips in determining optimal routing paths for data from a source processor chip to a destination processor chip. As a result, the congestion of data for processing at each of the processor chips along each possible routing path may be taken into account when selecting to which processor chip to forward data.

Dynamic routing based on information of not responded active source requests quantity received in broadcast heartbeat signal and stored in local data structure for other processor chips

A processor book designed to support both commercial workloads and technical workloads based on a dynamic or static mechanism of reconfiguring the external wiring interconnect. The processor book is configured as a building block for commercial workload processing systems with external connector buses (ECBs). The processor book is also provided with routing logic to enable to ECBs to be utilized for either book-to-book routing or routing within the same processor book. A table specific wiring scheme is provided for coupling the ECBs running off the chips of one MCM to the chips of the second MCM on the processor book so that the chips of the first MCM are connected directly to the chips of a second MCM that is logically furthest away and vice versa. Once the wiring of the ECBs are completed according to the wiring scheme, the operational and functional characteristics reflect those of a processor book configured for technical workloads.

Data processing system with backplane and processor books configurable to support both technical and commercial workloads

The data interconnect and routing mechanism reduces data communication latency, supports dynamic route determination based upon processor activity level/traffic, and implements an architecture that supports scalable improvements in communication frequencies. In one application, a data processing system includes first and second processing books, each including at least first and second processing units. Each of the first and second processing units has a respective first output data bus. The first output data bus of the first processing unit is coupled to the second processing unit, and the first output data bus of the second processing unit is coupled to the first processing unit. At least the first processing unit of the first processing book and the second processing unit of the second processing book each have a respective second output data bus. The second output data bus of the first processing unit of the first processing book is coupled to the first processing unit of the second processor book, and the second output data bus of the second processing unit of the second processor book is coupled to the second processing unit of the first processor book.

Multiprocessor data processing system having scalable data interconnect and data routing mechanism

An apparatus for performing bus tracing with scalable bandwidth in a distributed memory symmetric multiprocesssor system is disclosed. The distributed memory symmetric multiprocessor system includes multiple processing units, each coupled to a memory module. Each of the processing units includes a memory controller and a bus trace macro (BTM) module. The memory controller is coupled to an interconnect for the symmetric multiprocessor system, and the BTM module is connected between the interconnect and the memory controller via two multiplexors. A subset of the BTM modules within the symmetric multiprocessor system is enabled for performing tracing operations such that address transactions on the interconnect are divided among the subset of the BTM modules to be selectively and separately intercepted by each BTM module within the subset of the BTM modules.

Method and apparatus for performing bus tracing with scalable bandwidth in a data processing system having a distributed memory

An apparatus for performing imprecise bus tracing in a distributed memory symmetric multiprocessor system is disclosed. The apparatus includes a bus trace macro (BTM) module that can control the snoop traffic seen by one or more of the memory controllers in the data processing system and utilize a local memory attached to the memory controller for storing trace records. After the BTM module is enabled for tracing operations, the BTM module snoops transactions on the interconnect and packs information contained within these transactions into a block of data of a size that matches the write buffers within the memory controller. In addition, the BTM module also includes a dropped record counter for counting the number of address transactions that were not converted to trace records because all the write buffers were completely full. After an occurence of the write buffers full condition, a time stamp trace record is inserted before a new trace record can be written. The time stamp trace record includes a count of the number of address transactions that were not converted to trace records.

Method and apparatus for performing imprecise bus tracing in a data processing system having a distributed memory

A data interconnect and routing mechanism reduces data communication latency, supports dynamic route determination based upon processor activity level/traffic, and implements an architecture that supports scalable improvements in communication frequencies. In one implementation, a data processing system includes at least first through third processing units, data storage coupled to the plurality of processing units, and an interconnect fabric. The interconnect fabric includes at least a first data bus coupling the first processing unit to the second processing unit and a second data bus coupling the third processing unit to the second processing unit so that the first and third processing units can transmit data traffic to the second processing unit. The data processing system further includes a control channel coupling the first and third processing units. The first processing unit requests approval from the third processing unit via the control channel to transmit a data communication to the second processing unit, and the third processing unit approves or delays transmission of the data communication in a response transmitted via the control channel.

Multiprocessor data processing system having a data routing mechanism regulated through control communication

The symmetric multiprocessor system includes multiple processing nodes, with multiple agents at each node, connected to each other via an interconnect. A request transaction is initiated by a master agent in a master node to all receiving nodes. A write counter number is generated for associating with the request transaction. The master agent then waits for a combined response from the receiving nodes. After the receipt of the combined response, a data packet is sent from the master agent to all intended one of the receiving nodes according to the combined response. After the data packet has been sent, the master agent in the master node is ready to send another request transaction along with a new write counter number, without the necessity of waiting for an acknowledgement from the receiving node.

Growing community of inventors

Round Rock, TX, United States of America

Jerry Don Lewis