Method, system, and computer program product for allocating an exclusive memory resource
By identifying and ignoring exclusive memory resources in cache coherence protocols, the method addresses the scalability and performance issues in parallel computer systems, improving resource management and reducing overhead in multiprocessor environments.
Patent Information
- Application Number
- PCT/US2024/027631
- Authority / Receiving Office
- WO · WO
- Patent Type
- Applications
- Current Assignee / Owner
- Priority Date
- 2023-05-17
- Filing Date
- 2024-05-03
- Publication Date
- 2024-11-21
AI Technical Summary
In parallel computer systems, the addition of processors can lead to increased overhead and resource intensity in maintaining cache coherence, limiting scalability and potentially negatively impacting performance due to the need for additional hardware checks and complex configurations, especially when not all data resources require tracking by a cache coherence protocol.
A method and system for allocating memory resources based on instructions, where exclusive memory resources are identified and ignored by the cache coherence protocol, reducing the need for coherence operations and conserving computational resources by filtering data resources that do not require coherence tracking.
This approach enhances scalability and performance by reducing unnecessary cache coherence operations and resource usage, allowing for more efficient memory management in multiprocessor systems, especially when dealing with exclusive memory resources that do not need to be tracked.
Smart Images

Figure US2024027631_21112024_PF_FP_ABST
Abstract
Description
METHOD, SYSTEM, AND COMPUTER PROGRAM PRODUCT FOR ALLOCATING AN EXCLUSIVE MEMORY RESOURCEBACKGROUND
[0001] This International Application is related to and claims priority to U.S. Provisional Application No. 63 / 502,748, filed on May 17, 2023, the entire contents of which are incorporated herein by reference.1. Field
[0002] This disclosure relates generally to multiprocessing systems, managing cache memory in a parallel processing environment, and scalable multiprocessor systems with processor cache memory including, in some non-limiting embodiments or aspects, systems, methods, and computer program products for allocating memory based on data and / or software instructions used in a cache coherence protocol. A cache coherence protocol may include a method and / or implementation (e.g., a software implementation) to maintain coherent and / or consistent cache memory with independent processor caches in a scalable multiprocessor system. Some non-limiting embodiments or aspects may generally relate to systems, methods, and computer program products for allocating memory resources based on data instructions with a cache coherence protocol independent of a cache coherence protocol implementation (e.g., independent of how a cache coherence protocol is implemented).2. Technical Considerations
[0003] Some computers (e.g., computing devices) may be configured to perform multiple calculations and / or processes simultaneously using a plurality of processors (e.g., parallel computer systems). In some instances, parallel computer systems may allow for larger (e.g., large amounts of data, more complex) problems to be solved in a shorter amount of time due to calculations and / or processes being performed simultaneously, rather than the calculations and / or processes being performed sequentially by a single processor of a computing device. With a parallel computer system, each processor of the plurality of processors may have access to shared memory within a computing device and each processor may have a cache memory such that data and / or values (e.g., data resources) stored in shared memory may beread by each processor and written into the cache memory of each processor to perform the calculations and / or processes on the data resources. Once each processor has completed performing the calculations and / or processes on the data, the data may be written back to shared memory.
[0004] When performing calculations and / or processes simultaneously with a parallel computer system, there may be a risk that some data resources may be processed out of order and / or improperly overwritten by another read / write operation of one of the processors of the plurality of processors. For example, two or more processors may read the same data resource residing at a shared memory address in shared memory such that each of the two or more processors generates a local copy of the data resource in the cache memory (e.g., each processor writes a copy of the data resource to their own cache memory). The two or more processors may each perform a different operation on their own local copy of the data resource in their own cache memory simultaneously to generate different results (e.g., new and / or updated data resources), and each of the two or more processors may attempt to write the different results back to the shared memory address (e.g., each processor may attempt to write to the same shared memory address) in shared memory. Generating different results from simultaneous calculation and / or processing by the two or more processors may create inconsistencies in the data resources residing in shared memory and / or in each cache memory of each processor. In some instances, the inconsistencies in the data resources may cause problems in calculations and / or incorrect results of a calculation and / or process. Thus, the parallel computer system may need to make sure that all of the data residing in shared memory and in each cache memory of each processor remains current, coherent, and / or consistent.
[0005] Some parallel computer systems may implement a hardware cache coherence protocol to remedy some of the inconsistencies that may occur between the data residing in shared memory and / or in each cache memory of each processor. Such a cache coherence protocol may track and / or monitor data instructions and / or data resources that are passed between shared memory, cache memory, and each processor of the parallel computing system.
[0006] However, a hardware cache coherence protocol may be dependent on an initial configuration of a plurality of processors in a parallel computer system. For example, a parallel computer system that implements a hardware cache coherenceprotocol may only function with a specific configuration of processors (e.g., multiprocessors) and may not be scalable (e.g., expanding a configuration of a system such that the system may be able to handle a larger load of requests, calculations processing, and / or the like) with the addition of more processors to the parallel computer system. The parallel computer system may not be scalable with the addition of more processors because adding each additional processor having a connection to shared memory would introduce additional overhead, requiring a hardware cache coherence protocol to monitor another cache and requiring additional hardware checks. Further, with each addition of another processor to the parallel computer system, the overhead needed to keep all caches coherent and / or consistent increases significantly. For example, in a parallel computer system including a first processor having a first cache and a second processor having a second cache, the first cache must be kept coherent and / or consistent with the second cache. Adding a third processor with a third cache to the parallel computer system would then require the first cache to be coherent and / or consistent with the second cache and the third cache, the second processor to be coherent and / or consistent with the first cache and the third cache, and the third cache to be coherent and / or consistent with the first cache and the second cache. With more processors that are added to the parallel computer system, maintaining coherent and / or consistent caches via a hardware cache coherence protocol becomes increasingly more resource intensive.
[0007] Sometimes, scaling (e.g., expanding a configuration of a system) a parallel computer system may be accomplished by adding memory and / or increasing an amount of memory resources. However, adding memory and / or increasing the amount of memory resources only allows for additional storage and does not increase processing capabilities for each processor in a parallel computer system. In some instances, the performance of a parallel computer system may be negatively impacted by the addition of processors to a parallel computer system implementing a hardware cache coherence protocol. The addition of more processors to a parallel computer system may require additional configuration, such as modifying the hardware cache coherence protocol and providing the additional processors with access to cache memory of existing processors in the parallel computer system. Thus, scaling a parallel computer system by adding memory and / or increasing an amount of memory resources and / or adding processors may not significantly improve systemperformance and may actually negatively impact system performance, requiring complex configuration and / or modifications to the system and / or software.
[0008] In other instances, data may not need to be tracked and / or monitored by a cache coherence protocol (e.g., a hardware cache coherence protocol, or other type of cache coherence protocol implementation). For example, data resources may be only accessible and / or processed by a single processor in a parallel computer system. The parallel computer system may process data resources that are not shared resources, or data resources that are limited to processing by only one processor in a multiprocessor system. Limitations placed on computing resources (e.g., hardware, virtual resources, and / or the like) or on data resources may be developed and / or enforced by many different sources, such as various software applications, hardware components, etc. In instances where a data resource is limited and / or designated as an exclusive resource (e.g., a data resource that is not shared and can only be processed by one processor), the data resource does not need to be tracked and / or monitored by a cache coherence protocol. Tracking and / or monitoring such an exclusive data resource using a cache coherence protocol may burden a parallel computing system implementing a cache coherence protocol, and tracking and / or monitoring of the exclusive data resource may constitute a waste of computing resources.SUMMARY
[0009] Accordingly, provided are methods, systems, and computer program products for allocating a memory resource based on instructions and / or requests for data used in a cache coherence protocol and for allocating exclusive data resources independent of a cache coherence protocol used in a multiprocessor system.
[0010] According to non-limiting embodiments or aspects, provided is a computer- implemented method. The computer-implemented method may include generating a request for allocating a memory resource. The method may further include determining that the memory resource associated with the request is an exclusive memory resource. The method may further include allocating a data resource associated with the memory resource based on determining that the memory resource is an exclusive memory resource, such that the data resource is ignored by a cache coherence protocol.
[0011] According to non-limiting embodiments or aspects, provided is a system for coherent memory, including at least one processor. The system may include one or more cache memories. Each cache memory may include a cache controller programmed or configured to execute a cache coherence protocol. The one or more processors may be programmed or configured to generate a memory allocation instruction allocating a memory resource such that when the memory resource is referenced in an instruction the at least one processor transmits identifiers in simultaneous signals to the cache controller such that the identifiers enable the cache controller to determine the location of values of the memory resource when performing coherence.
[0012] According to non-limiting embodiments or aspects, provided is a computer- implemented method. The method may include generating a request for allocating a memory resource. The method may include determining the possible current locations of the value of the memory resource. The method may further include transmitting a signal to the cache controller. The signal may enable the cache controller to determine the location of the current value of the memory resource via the cache coherence protocol.
[0013] According to non-limiting embodiments or aspects, provided is a system including at least one processor and at least one non-transitory computer-readable medium storing instructions that, when executed by the at least one processor, cause the at least one processor to perform the computer-implemented method disclosed herein.
[0014] According to non-limiting embodiments or aspects, provided is a computer program product including at least one non-transitory computer-readable medium including program instructions that, when executed by at least one processor, cause the at least one processor to perform the computer-implemented method disclosed herein.
[0015] Other non-limiting embodiments or aspects will be set forth in the following numbered clauses:
[0016] Clause 1 : A computer-implemented method, comprising: generating, with at least one processor, a request for allocating a memory resource; determining, with at least one processor, that the memory resource associated with the request is an exclusive memory resource; filtering, with at least one processor, a data resourceassociated with the memory resource based on determining that the memory resource is an exclusive memory resource, such that the data resource is ignored by a cache coherence protocol.
[0017] Clause 2: The computer-implemented method of clause 1 , wherein the request for allocating the memory resource is a cache memory allocation request.
[0018] Clause 3: The computer-implemented method of clause 1 or 2, further comprising transmitting the request for allocating the memory resource to a queue of requests.
[0019] Clause 4: The computer-implemented method of any one of clauses 1-3, further comprising: performing, with at least one processor, an atomic HSP on a shared memory location; and adding, with at least one processor, the request to a queue of cache memory allocation requests.
[0020] Clause 5: The computer-implemented method of any one of clauses 1-4, wherein the exclusive memory resource comprises a cache memory resource of exactly one processor.
[0021] Clause 6: The computer-implemented method of any one of clauses 1-5, wherein the cache coherence protocol is a hardware cache coherence protocol.
[0022] Clause 7: The computer-implemented method of any one of clauses 1-6, wherein filtering comprises: transmitting the data resource to the memory resource; and foregoing detecting the data resource with the cache coherence protocol.
[0023] Clause 8: The computer-implemented method of any one of clauses 1-7, wherein determining that the memory resource associated with the request is an exclusive memory resource comprises: detecting an address included in the request for allocating the memory resource; and determining that the address included in the request is a cache memory address.
[0024] Clause 9: A system for coherent memory, comprising: at least one processor; one or more cache memories, each cache memory comprising a cache controller programmed or configured to execute a cache coherence protocol; wherein the one or more processors are programmed or configured to: generate a memory allocation instruction allocating a memory resource such that when the memory resource is referenced in an instruction the at least one processor transmits identifiers in simultaneous signals to the cache controller such that the identifiers enable thecache controller to determine the location of values of the memory resource when performing coherence.
[0025] Clause 10: A computer-implemented method, comprising: generating, with at least one processor, a request for allocating a memory resource; determining, with at least one processor, the possible current locations of the value of the memory resource; transmitting, with at least one processor, a signal to the cache controller, wherein the signal enables the cache controller to determine the location of the current value of the memory resource via the cache coherence protocol.
[0026] Clause 11 : The computer-implemented method of any of clauses 1-10, further comprising: detecting, with at least one processor, a processor identifier associated with a processor; and transmitting, with at least one processor, the processor identifier to the cache controller as an additional input to the cache coherence protocol.
[0027] Clause 12: A computer-implemented method, including: generating, with at least one processor, a request for allocating a memory resource; detecting, with at least one processor, an indicator included in the request for allocating the memory resource; determining, with at least one processor, that the indicator included in the request is associated with a global memory address; transmitting, with at least one processor, a signal to the cache controller, wherein the signal causes the cache controller to read from shared memory and write to shared memory and forgo coherence on the memory resource via a cache coherence protocol.
[0028] Clause 13: The computer-implemented method of clause 12, including: protecting, with at least one processor, the memory resource via a hardware synchronization primitive (HSP) or an enqueue, wherein the memory resource comprises a global memory resource that is accessible by other tasks executing on one or more processors.
[0029] Clause 14: A system for coherent memory, including: at least one processor; one or more cache memories, each cache memory comprising a cache controller programmed or configured to execute a cache coherence protocol; wherein the one or more processors are programmed or configured to: generate a global memory allocation instruction allocating a memory resource such that when the memory resource is referenced in an instruction the at least one processor transmits a signalto the cache controller to read and write from shared memory and to forgo coherence on the memory resource via the cache coherence protocol.
[0030] Clause 15: A system comprising at least one processor programmed or configured to perform the computer-implemented method of any one of clauses 1 -13.
[0031] Clause 16: A computer program product comprising at least one non- transitory computer-readable medium including program instructions that, when executed by at least one processor, cause the at least one processor to perform the computer-implemented method of any one of clauses 1-13.
[0032] These and other features and characteristics of the present disclosure, as well as the methods of operation and functions of the related elements of structures and the combination of parts and economies of manufacture, will become more apparent upon consideration of the following description and the appended claims with reference to the accompanying drawings, all of which form a part of this specification, wherein like reference numerals designate corresponding parts in the various figures. It is to be expressly understood, however, that the drawings are for the purpose of illustration and description only and are not intended as a definition of the limits of the disclosed subject matter.BRIEF DESCRIPTION OF THE DRAWINGS
[0033] Additional advantages and details are explained in greater detail below with reference to the non-limiting embodiments that are illustrated in the accompanying schematic figures, in which:
[0034] FIG. 1 is a schematic diagram of a system for maintaining coherent and / or consistent memory with independent processor caches in a scalable multiprocessor system with a cache coherence protocol according to some non-limiting embodiments or aspects;
[0035] FIG. 2A is a flow diagram of an exemplary existing process for determining whether a data resource requires coherence by a parallel computing system;
[0036] FIG. 2B is a flow diagram of a process for determining whether a data resource requires coherence by a parallel computing system according to some nonlimiting embodiments or aspects;
[0037] FIG. 3 is a diagram of an exemplary environment in which methods, systems, and / or computer program products, described herein, may be implemented according to some non-limiting embodiments or aspects;
[0038] FIG. 4 is a schematic diagram of example components of one or more devices of FIG. 1 and / or FIG. 3 according to some non-limiting embodiments or aspects;
[0039] FIG. 5 is a schematic diagram of an exemplary existing system for implementing a cache coherence protocol in a multiprocessor system.
[0040] FIG. 6A is a flow diagram of a process for allocating a memory resource independent of a cache coherence protocol;
[0041] FIG. 6B is a diagram of an example system according to non-limiting embodiments or aspects depicting how a processor may pass coherence data to a coherence unit;
[0042] FIG. 7A is a diagram pertaining to a CPU identifier (ID) showing differences between execution of exclusive memory addresses and current memory addresses, and the relationship to cache memories;
[0043] FIG. 7B is a diagram of a system for cache coherence for an instruction executing for an exclusive memory address and a non-exclusive memory address;
[0044] FIG. 7C is a diagram of a system to forgo cache coherence for an exclusive memory address and a global memory address; and
[0045] FIG. 8 is a schematic diagram of a coherent memory system which includes one or more processors with exclusive memory that allocate global memory in shared memory and execute process tasks from assigned task queues residing in shared memory.DETAILED DESCRIPTION
[0046] For purposes of the description hereinafter, the terms “end,” “upper,” “lower,” “right,” “left,” “vertical,” “horizontal,” “top,” “bottom,” “lateral,” “longitudinal,” and derivatives thereof shall relate to the embodiments as they are oriented in the drawing figures. However, it is to be understood that the embodiments may assume various alternative variations and step sequences, except where expressly specified to the contrary. It is also to be understood that the specific devices and processes illustrated in the attached drawings, and described in the following specification, are simplyexemplary embodiments or aspects of the disclosed subject matter. Hence, specific dimensions and other physical characteristics related to the embodiments or aspects disclosed herein are not to be considered as limiting.
[0047] No aspect, component, element, structure, act, step, function, instruction, and / or the like used herein should be construed as critical or essential unless explicitly described as such. Also, as used herein, the articles “a” and “an” are intended to include one or more items and may be used interchangeably with “one or more” and “at least one.” Furthermore, as used herein, the term “set” is intended to include one or more items (e.g., related items, unrelated items, a combination of related and unrelated items, and / or the like) and may be used interchangeably with “one or more” or “at least one.” Where only one item is intended, the term “one” or similar language is used. Also, as used herein, the terms “has,” “have,” “having,” or the like are intended to be open-ended terms. Further, the phrase “based on” is intended to mean “based at least partially on” unless explicitly stated otherwise.
[0048] As used herein, the term “communication” may refer to the reception, receipt, transmission, transfer, provision, and / or the like of data (e.g., information, signals, messages, instructions, commands, and / or the like). For one unit (e.g., a device, a system, a component of a device or system, combinations thereof, and / or the like) to be in communication with another unit means that the one unit is able to directly or indirectly receive information from and / or transmit information to the other unit. This may refer to a direct or indirect connection (e.g., a direct communication connection, an indirect communication connection, and / or the like) that is wired and / or wireless in nature. Additionally, two units may be in communication with each other even though the information transmitted may be modified, processed, relayed, and / or routed between the first and second unit. For example, a first unit may be in communication with a second unit even though the first unit passively receives information and does not actively transmit information to the second unit. As another example, a first unit may be in communication with a second unit if at least one intermediary unit processes information received from the first unit and communicates the processed information to the second unit.
[0049] As used herein, the term “computing device” may refer to one or more electronic devices configured to process data. A computing device may, in some examples, include the necessary components to receive, process, and output data,such as a processor, a display, a memory, an input device, a network interface, and / or the like. A computing device may be a mobile device. As an example, a mobile device may include a cellular phone (e.g., a smartphone or standard cellular phone), a portable computer, a wearable device (e.g., watches, glasses, lenses, clothing, and / or the like), a personal digital assistant (PDA), and / or other like devices. A computing device may also be a desktop computer or other form of non-mobile computer.
[0050] As used herein, the terms “client” and “client device” may refer to one or more client-side devices or systems used to initiate or facilitate a network connection. As an example, a “client device” may refer to one or more computing devices used by a user, one or more personal computers used by a user, one or more mobile devices used by a user, and / or the like. In some non-limiting embodiments or aspects, a client device may be an electronic device configured to communicate with one or more networks. For example, a client device may include one or more computers, portable computers, laptop computers, tablet computers, mobile devices, cellular phones, wearable devices (e.g., watches, glasses, lenses, clothing, and / or the like), PDAs, and / or the like. Moreover, a “client” may also refer to an entity (e.g., a user, a corporation, and / or the like) that owns, utilizes, and / or operates a client device.
[0051] As used herein, the term “server” may refer to or include one or more computing devices that are operated by or faci litate communication and processing for multiple parties (e.g., clients, client devices, users, and / or the like) in a network environment, such as the Internet, although it will be appreciated that communication may be facilitated over one or more public or private network environments and that various other arrangements are possible. Further, multiple computing devices (e.g., servers, mobile devices, etc.) directly or indirectly communicating in the network environment may constitute a “system.” Reference to “a server” or “a processor,” as used herein, may refer to a previously-recited server and / or processor that is recited as performing a previous step or function, a different server and / or processor, and / or a combination of servers and / or processors. For example, as used in the specification and the claims, a first server and / or a first processor that is recited as performing a first step or function may refer to the same or different server and / or a processor recited as performing a second step or function.
[0052] FIG. 1 depicts a system 100 for maintaining coherent and / or consistent memory with independent processor caches in a scalable multiprocessor system witha cache coherence protocol according to some non-limiting embodiments or aspects. System 100 may include coherent memory system 102, processor 104-1 to processor 104-n (e.g., a plurality of processors, referred to individually as processor 104 and collectively as processors 104 where appropriate), cache memory 106-1 to cache memory 106-n (e.g., a plurality of cache memories, referred to individually as cache memory 106 and collectively as cache memories 106 where appropriate), and shared memory 108.
[0053] Coherent memory system 102 may include a computing device, such as a server (e.g., a single server), a group of servers, and / or other like devices. In some non-limiting embodiments or aspects, coherent memory system 102 may include a processor and / or memory as described herein. In some non-limiting embodiments or aspects, coherent memory system 102 may include one or more client devices. In some non-limiting embodiments or aspects, coherent memory system 102 may include one or more servers and / or one or more client devices executing instructions (e.g., software instructions) that cause coherent memory system 102 to perform one or more steps of methods and processes as described herein.
[0054] Processor 104 (e.g., processor 104-1 to processor 104-n) may include at least one processor (e.g., a multi-core processor), such as a central processing unit (CPU), an accelerated processing unit (APU), a graphics processing unit (GPU), a microprocessor, and / or the like. In some non-limiting embodiments or aspects, processor 104 may include at least one processor having a single core (e.g., a singlecore processor) or at least one processor having multiple cores (e.g., a multiprocessor, a multi-core processor, a processor including more than one core, and / or the like). In some non-limiting embodiments or aspects, processor 104 may include at least one core (e.g., a core of a processor associated with a dedicated L1 cache) that is a component of (e.g., part of) a single-core processor or a multiprocessor.
[0055] In some non-limiting embodiments or aspects, processor 104 (e.g., processors 104-1 to 104-n) may be programmed to perform one or more steps of methods described herein. In some non-limiting embodiments or aspects, processors 104 may include one or more processors executing instructions (e.g., software instructions) that cause processors 104 to perform one or more steps of methods as described herein. In some non-limiting embodiments or aspects, processor 104 may be in communication with cache memory 106 (e.g., cache memories 106-1 to 106-n).In some non-limiting embodiments or aspects, processor 104 may be capable of receiving information (e.g., data, data resources, and / or the like) from and / or communicating (e.g., transmitting) information to cache memory 106. In some nonlimiting embodiments or aspects, processor 104 may execute instructions to identify exclusive memory and / or exclusive memory resources (e.g., an instance of an algorithm configured to keep information and / or data resources stored in cache memory 106 coherent and / or consistent with information and / or data resources stored in shared memory 108 for processing by processor 104, while identifying exclusive memory to reduce the need for coherence operations, and / or the like).
[0056] In some non-limiting embodiments or aspects, at least one processor 104 may execute an operating system where the operating system is a multitasking operating system which may require a cache coherence protocol (e.g., a hardware cache coherence protocol), but the operating system may be modified to execute coherent memory system 102. For example, at least one processor 104 may execute the operating system where the operating system is modified, while implementing a hardware cache coherence protocol, to execute coherent memory system 102 (e.g., a modified multitasking operating system). In some non-limiting embodiments or aspects, at least one processor 104 may execute an operating system where the operating system is a non-multitasking atomic operating system which executes tasks from a shared memory queue (e.g., a queue of one or more tasks stored in shared memory 108). In some non-limiting embodiments or aspects, at least one processor 104 may execute either a modified multitasking operating system ora non-multitasking atomic operating system that may execute tasks from a shared memory queue. In some non-limiting embodiments or aspects, at least one processor 104 may add tasks to a shared memory queue, independent of whether processor 104 is executing a modified multitasking operating system or a non-multitasking atomic operating system.
[0057] In some non-limiting embodiments or aspects, at least one processor 104 may continue execution while the at least one processor 104 requires a result (e.g., waits for a result that may be required for further execution) from a task that the at least one processor 104 added to a shared memory queue. To continue execution while the at least one processor 104 requires a result from the task, the at least one processor 104 may wait until the task added to the shared memory queue has a result (e.g., a result associated with the task added to the shared memory queue) stored ina shared memory location of shared memory 108. In this manner, the at least one processor 104 and tasks the at least one processor 104 has initiated may run and / or be executed concurrently. In some non-limiting embodiments or aspects, processor 104 may not execute any operating systems that require multitasking. In this way, multitasking of conventional computing systems may be replaced by concurrent processing.
[0058] In some non-limiting embodiments or aspects, each processor 104 of a plurality of processors 104 may not be configured to interact with other processors 104 of the plurality of processors 104 to perform a hardware cache coherence protocol. For example, each processor 104 of the plurality of processors 104 may execute independently where exclusive memory has been identified to cause each processor 104 to interact with cache memory 106 of each processor respectively (e.g., processor 104-1 may interact with cache memory 106-1 as exclusive memory to execute a first task, processor 104-2 may interact with cache memory 106-2 as exclusive memory to execute a second task, etc.).
[0059] Cache memory 106 (e.g., cache memories 106-1 to 106-n) may include cache memory internal to and / or associated with a processor (e.g., processors 104-1 to 104-n). In some non-limiting embodiments or aspects, cache memory 106 may include a level 1 and / or primary cache, a level 2 and / or secondary cache, and / or a level 3 and / or tertiary cache. In some non-limiting embodiments or aspects, cache memory 106 may include a storage component (e.g., a volatile storage component) that stores information and / or instructions for use by processor 104. In some nonlimiting embodiments or aspects, cache memory 106 may include CPU memory of processor 104. In some non-limiting embodiments or aspects, cache memory 106 may include random access memory (RAM) and / or another type of static storage device. In some non-limiting embodiments or aspects, cache memory 106 may be identified as exclusive memory and / or may be determined to include at least one exclusive memory resource storing an exclusive data resource.
[0060] In some non-limiting embodiments or aspects, cache memory 106 may store information and / or software related to the operation and use of coherent memory system 102 and / or processor 104. For example, cache memory 106 may include a type of computer-readable medium. In some non-limiting embodiments or aspects, each cache memory 106 may transmit information to and / or receive information fromeach processor 104 respectively (e.g., cache memory 106-1 may transmit information to and / or receive information from processor 104-1 , cache memory 106-2 may transmit information to and / or receive information from processor 104-2, etc.). In some non-limiting embodiments or aspects, each cache memory 106 may be implemented by (e.g., part of) each processor 104 respectively (e.g., cache memory 106-1 may be part of processor 104-1 , cache memory 106-2 may be part of processor 104-2, etc.). In some non-limiting embodiments or aspects, cache memory 106 may be referred to as exclusive memory, as each cache memory 106 may only be accessible by processor 104 associated with cache memory 106 (e.g., cache memory 106-1 is exclusive memory to processor 104-1 as it is only accessible by processor 104-1 , cache memory 106-2 is exclusive memory to processor 104-2 as it is only accessible by processor 104-2, etc.).
[0061] In some non-limiting embodiments or aspects, cache memory 106 may be exclusive memory due to hardware constraints (e.g., one hardware-based cache memory associated with each processor, other processors are not physically connected to the hardware-based cache memory and cannot communicate with the hardware-based cache memory). Additionally or alternatively, cache memory 106 may be exclusive memory due to logical constraints (e.g., a hardware-based pool of cache memory may be logically divided into different cache memory resources, each cache memory resource only accessible by one processor and determined to be exclusive memory). In some non-limiting embodiments, cache memories 106 (e.g., a plurality of cache memories 106) may include a number of cache memories 106 including hardware constraints, a number of cache memories 106 including logical constraints, or a combination of both hardware constraints and logical constraints.
[0062] Shared memory 108 may include RAM, read only memory (ROM), and / or another type of dynamic or static storage device (e.g., flash memory, magnetic memory, optical memory, etc.) that stores information and / or instructions for use by coherent memory system 102 and / or processor 104. In some non-limiting embodiments or aspects, shared memory 108 may include a computer readable medium that may store information and / or software related to the operation and use of coherent memory system 102 and / or processor 104. For example, shared memory 108 may include a hard disk (e.g., a magnetic disk, an optical disk, a magneto-optic disk, a solid-state disk, etc.) and / or another type of computer-readable medium. Insome non-limiting embodiments or aspects, shared memory 108 may include memory that is mapped to one or more processes (e.g., processors executing software instructions, software applications, and / or the like) such that the one or more processors may read and / or write data to the memory without interacting with an operating system of a computing device (e.g., without calling functions of the operating system). In some non-limiting embodiments or aspects, shared memory 108 may transmit information to and / or receive information from processor 104.
[0063] In some non-limiting embodiments or aspects, shared memory 108 may include a plurality of shared memory addresses. Each shared memory address may include an identifier of a shared memory location (e.g., a shared memory block) in shared memory 108. In some non-limiting embodiments or aspects, shared memory 108 may include exclusive memory such that shared memory 108 includes at least one exclusive memory resource. In some non-limiting embodiments or aspects, system 100 may include only one instance (e.g., one copy) of shared memory 108. Alternatively, system 100 may include a plurality of instances of shared memory 108, where each instance of shared memory 108 is identical (e.g., multiple copies of shared memory 108 where each instance is a mirrored copy of shared memory 108). Shared memory 108 may not include a cache memory (e.g., a separate and / or integrated cache memory) associated with shared memory 108. In some non-limiting embodiments or aspects, shared memory 108 without a cache memory associated with shared memory 108 may be referred to as cacheless shared memory.
[0064] In some non-limiting embodiments or aspects, shared memory may refer to memory of a computing device that is accessible by a plurality of processors. Shared memory may be accessible by the plurality of processors simultaneously. In some nonlimiting embodiments or aspects, shared memory may include a physical memory device (e.g., RAM) that may be accessible by the plurality of processors. The physical memory device may include a plurality of memory locations and each memory location may include and / or correspond to a physical memory address of a plurality of physical memory addresses. In some non-limiting embodiments or aspects, shared memory may include virtual memory (e.g., virtual memory configured in software instructions, and / or the like). Virtual memory may include a plurality of virtual memory locations. Each virtual memory location may include and / or correspond to a virtual memory address of a plurality of virtual memory addresses. Each virtual memory address maybe associated with (e.g., may be mapped to) a physical memory address of the plurality of physical memory addresses. In this way, shared memory may include virtual memory which corresponds to a plurality of physical memory addresses for a plurality of physical memory locations where the plurality of physical memory locations span across one or more physical memory devices (e.g., RAM, disk storage, and / or the like).
[0065] In non-limiting embodiments or aspects, there are a plurality of locations in system 100 where an updated data resource (e.g., the most current and / or up-to-date value of a data resource) may be located. The updated data resource may be located in (1 ) shared memory (e.g., shared memory 108), in (2) a first processor’s cache memory (e.g., cache memory 106-1 of processor 104-1 ), or in (3) another processor’s cache memory (e.g., cache memory 106-2 through cache memory 106-n of processor 104-2 through 104-n). Shared memory 108 may consist of memory resources (e.g., shared memory resources) that may be modified by an atomic hardware synchronization primitive (HSR) and / or memory resources that may be locked with a semaphore. Cache memory resources may consist of memory resources (e.g., cache memory resources) that may be exclusive memory resources and / or memory resources that may be common static memory resources. In existing systems, a hardware cache coherence protocol may be used to ensure that data resources stored in shared memory and / or stored in each cache memory associated with each processor remain updated.
[0066] As used herein, a protocol may refer to a set of rules that allows a plurality of connected devices (e.g., computing devices, processors 104, cache memories 106, and / or the like) to communicate.
[0067] In some non-limiting embodiments or aspects, processor 104 may execute an instruction that causes processor 104 to generate a pointer (e.g., a shared pointer) associated with a shared memory block and causes processor 104 to store the pointer in cache memory 106. In this way, any operation and / or instruction of coherent memory system 102 that references the pointer may read the shared memory address (e.g., the shared memory block) that is associated with the pointer and results of the operation and / or instruction may be stored in shared memory 108 and / or cache memory 106.
[0068] In some non-limiting embodiments or aspects, processor 104 may execute an instruction that reads (e.g., retrieves) a data resource from shared memory 108 and writes (e.g., stores) the data resource to shared memory 108 and / or cache memory 106. In some non-limiting embodiments, processor 104 may execute an instruction that reads a data resource from cache memory 106 and writes the data resource to shared memory 108 and / or cache memory 106. For example, processor 104 (e.g., processor 104-1 ) may execute an instruction that causes processor 104 to read a data resource from cache memory 106 associated with processor 104 (e.g., cache memory 106-1 ) and causes processor 104 to write the data resource to shared memory 108. Cache memory 106 and / or shared memory 108 may include exclusive memory, indicating that the data resource does not reside in multiple memory resources (e.g., the data resource does not reside in multiple caches or in multiple shared memory 108 locations). In some non-limiting embodiments or aspects, processor 104 may execute an instruction that causes processor 104 to read a plurality of data resources from one or more input sources and causes processor 104 to write the plurality of data resources to one or more output destinations.
[0069] In some non-limiting embodiments or aspects, input source may refer to one input source or a prioritized list of input sources. For example, if a prioritized list of input sources includes cache memory 106 as a first prioritized input source and the prioritized list of input sources includes shared memory 108 as a second prioritized input source, processor 104 may attempt to retrieve an input from cache memory 106 as an input source first, and if no input exists in cache memory 106 or if cache memory 106 is not present, then processor 104 may default to shared memory 108 as the input source to retrieve the input. In some non-limiting embodiments or aspects, output destination may refer to one output destination, a prioritized list of output destinations, and / or more than one output destination.
[0070] In some non-limiting embodiments or aspects, a cache coherence protocol may refer to a method and / or protocol that may be implemented in hardware and / or software components of a computing device such that components of the computing device may maintain a coherent and / or a consistent view of data (e.g., a consistent definition, a consistent value of a data resource) that is transmitted between components and read and / or written by components of the computing device (e.g., shared by components of the computing device). In some non-limiting embodimentsor aspects, maintaining a coherent and / or a consistent view of data that is shared by components of a computing device may include tracking the status of the data that is shared. In some non-limiting embodiments or aspects, a cache coherence protocol may be based on a directory in memory, snooping (e.g., detecting addresses that are passed via a bus) on a bus, and / or the like.
[0071] The number and arrangement of systems and devices shown in FIG. 1 are provided as an example. There may be additional systems and / or devices, fewer systems and / or devices, different systems and / or devices, and / or differently arranged systems and / or devices than those shown in FIG. 1 . Furthermore, two or more systems or devices shown in FIG. 1 may be implemented within a single system or device, or a single system or device shown in FIG. 1 may be implemented as multiple, distributed systems or devices. Additionally or alternatively, a set of systems (e.g., one or more systems) or a set of devices (e.g., one or more devices) of system 100 may perform one or more functions described as being performed by another set of systems or another set of devices of system 100.
[0072] Referring now to FIGS. 2A and 2B, FIG. 2A shows a flow diagram of an existing process 200A for determining whether a data resource requires coherence by a parallel computing system (e.g., by a multiprocessor system). FIG. 2B shows a flow diagram of a process 200B for determining whether a data resource requires coherence by a parallel computing system (e.g., by multiprocessor system augmented with a coherent memory system, such as coherent memory system 102) according to some non-limiting embodiments or aspects.
[0073] As shown in FIG. 2A, at step 202A, a processor may execute an instruction. In step 204A, a processor may determine that the instruction executed by the processor is a read operation or a write operation. In step 206A, a system (e.g., a hardware system and / or component) may monitor a data resource associated with the instruction using a cache coherence protocol.
[0074] As shown in FIG. 2B, at step 202B, a processor (e.g., processor 104 as disclosed herein) may execute an instruction. For example, processor 104 may execute an instruction by first generating an instruction (e.g., a request) to allocate a memory resource (e.g., a cache memory resource or a shared memory resource). Processor 104 may execute a memory allocation instruction. Processor 104 may execute a shared memory allocation instruction to allocate memory in shared memory108. Processor 104 may execute a cache memory allocation instruction to allocate memory in cache memory 106.
[0075] In some non-limiting embodiments or aspects, processor 104 may execute a memory allocation instruction that may be classified as a shared memory allocation instruction or a cache memory allocation instruction based on an indicator (e.g., an indicator of an address of memory) and / or based on an address of memory that is to be allocated included in the memory allocation instruction (e.g., whether the address is an address to cache memory or an address to shared memory). Processor 104 may use the indicator and / or the address of memory that is to be allocated to determine a type of classification (e.g., cache memory or shared memory) for the memory allocation instruction. In this way, such classification may be used for filtering data instructions, data resources, and / or requests for data used with a cache coherence protocol in a multiprocessor system such that the multiprocessor system may perform cache coherence operations on only specific instructions where cache coherence is required, and the multiprocessor system may ignore specific instructions where cache coherence is not required.
[0076] In some non-limiting embodiments or aspects, processor 104 may generate a request for allocating a memory resource. For example, processor 104 may generate a request for allocating a memory resource that includes a read operation instruction or a write operation instruction.
[0077] As shown in FIG. 2B, at step 204B, processor 104 may determine that a memory resource associated with the instruction does not require coherence. For example, processor 104 may determine that the instruction is a write operation to an exclusive memory resource. Processor 104 may determine that the memory resource associated with the instruction is an exclusive memory resource based on the instruction. To make this determination, processor 104 may first determine that the instruction is a request and / or instruction to perform a write operation to a memory resource. Processor 104 may then determine whether the memory resource for the write operation is an exclusive memory resource (e.g., a cache memory resource accessible by only one processor 104) or a shared memory resource in shared memory 108. Processor 104 may determine whether the memory resource for the write operation is an exclusive memory resource or a shared memory resource based on an address included in the request and / or instruction (e.g., the request to allocatea memory resource). For example, processor 104 may determine that the request and / or instruction is associated with an exclusive memory resource by detecting and / or identifying the address included in the request and / or instruction and by determining that the address is an address to the exclusive memory resource.
[0078] In some non-limiting embodiments or aspects, processor 104 may determine that the memory resource associated with the request does not require coherence via a cache coherence protocol executed by a cache controller. In some non-limiting embodiments or aspects, processor 104 may determine that the memory resource associated with the request requires coherence via a cache coherence protocol executed by a cache controller (e.g., coherence unit 102).
[0079] As shown in FIG. 2B, at step 206B, processor 104 may transmit a signal to a cache controller or coherence unit. For example, processor 104 may filter a data resource associated with the memory resource by transmitting a signal to the cache controller to filter a data resource associated with the memory resource that was determined to not require coherence. Processor 104 may transmit a signal to forgo and / or prevent coherence on a memory resource or processor 104 may transmit a signal to cause coherence to be performed on a memory resource in order to filter a data resource associated with the memory resource based on determining that the memory resource is an exclusive memory resource. In this way, processor 104 and / or a coherent memory system (e.g., coherent memory system 102) may flag or ignore a data resource for cache coherence, thus saving computational resources.
[0080] In some non-limiting embodiments or aspects, processor 104 may filter a data resource by tagging, identifying, and / or labeling the data resource as a data resource that does not require coherence by a cache coherence protocol. Processor 104 may tag, identify, and / or label the memory resource (e.g., the exclusive memory resource) such that a data resource that is previously or subsequently stored in the memory resource is also tagged, identified, and / or labeled as an exclusive data resource. In this way, a system (e.g., system 100) may save computing resources and may operate more efficiently by forgoing coherence of a tagged, identified, and / or labeled memory resource and / or a tagged, identified, and / or labeled data resource such that a cache coherence protocol does not track and / or monitor any tagged, identified, and / or labeled resource (e.g., memory resource and / or data resource) that is identified as not requiring coherence. When the tagged, identified, and / or labeleddata resource or tagged, identified, and / or labeled memory resource is no longer in use by the system and / or processor 704 or when a status of requiring coherence changes, then a tag, identifier, and / or label may be removed by processor 104.
[0081] Additionally or alternatively, a status of a memory resource and / or a data resource may be tracked by a processor 104 such that the processor 104 prevents the data resource and / or the memory resource from being tracked and / or monitored by a cache coherence protocol. When the data resource and / or the memory resource is no longer in use by the system and / or processor 104, then processor 104 may forgo tracking the data resource and / or the memory resource such that the data resource and / or the memory resource may be subject to coherence via the cache coherence protocol.
[0082] In some non-limiting embodiments or aspects, processor 104 may transmit a signal to the cache controller. The signal may cause the cache controller to forgo coherence on the memory resource via the cache coherence protocol.
[0083] As shown in FIG. 2B, at step 208B, processor 104 may forgo and / or prevent monitoring of the data resource and / or the memory resource associated with the instruction by a cache coherence protocol. For example, processor 104 may forgo and / or prevent monitoring of the data resource by a cache coherence protocol such that the data resource is ignored by the cache coherence protocol (e.g., by the implementation of the cache coherence protocol, either in hardware or software). Alternatively, processor 104 may prevent monitoring of the data resource by the cache coherence protocol such that the data resource is not monitored by coherence (e.g., by a coherence system, such as coherent memory system 102). In this way, a system (e.g., system 100) may filter data resources and / or memory resources from coherence when coherence is not required for the data resource and / or the memory resource in order to reduce computational resources required to carry out a cache coherence protocol and monitor data resources and / or memory resources in a parallel computing system.
[0084] Steps of process 200B may be performed in various orders and sequences and are not limited to being performed in the order shown in FIG. 2B. Accordingly, steps of process 200B are not limited to any particular order and may be performed by various components, whether implemented on a single computing device ormultiple, distributed computing devices, having a multiprocessor architecture. Steps of process 200B may also be performed by a single processor or by multiple processors.
[0085] Referring now to FIG. 3, shown is a diagram of a non-limiting embodiment or aspect of an exemplary environment 300 in which systems, products, and / or methods, as described herein, may be implemented. As shown in FIG. 3, environment 300 may include coherent memory system 302, computing device 304, client device 306, server 308, and communication network 310. In some non-limiting embodiments or aspects, each of computing device 304, client device 306, server 308, and / or communication network 310 may be implemented by (e.g., part of) coherent memory system 302. In some non-limiting embodiments or aspects, at least one of each of computing device 304, client device 306, server 308, and / or communication network 310 may be implemented by (e.g., part of) another system, another device, another group of systems, or another group of devices, separate from or including coherent memory system 302, such as computing device 304, client device 306, server 308, and / or the like.
[0086] Coherent memory system 302 may include one or more devices capable of receiving information from and / or communicating information to computing device 304, client devices 306, and / or server 308 via communication network 310. For example, coherent memory system 302 may include a computing device, such as a server, a group of servers, and / or other like devices. In some non-limiting embodiments or aspects, coherent memory system 302 may be associated with a server as described herein. In some non-limiting embodiments or aspects, coherent memory system 302 may be in communication with a data storage device (e.g., database, memory, shared memory, cache memory, and / or the like), which may be local or remote to coherent memory system 302. In some non-limiting embodiments or aspects, coherent memory system 302 may be capable of receiving information from, storing information in, communicating information to, or searching information stored in the data storage device. In some non-limiting embodiments or aspects, coherent memory system 302 may be the same as or similar to coherent memory system 102.
[0087] Computing device 304 may include one or more devices capable of receiving information and / or communicating information to coherent memory system 302, client device 306, and / or server 308 via communication network 310. For example, computing device 304 may include a computing device, such as a server, agroup of servers, and / or other like devices. In some non-limiting embodiments or aspects, computing device 304 may be associated with a server, a client device, and / or a user device as described herein.
[0088] Client device 306 may include one or more devices capable of receiving information from and / or communicating information to coherent memory system 302, computing device 304, and / or server 308 via communication network 310. Additionally or alternatively, one or more client devices 306 may include a device capable of receiving information from and / or communicating information to other client devices 306 via communication network 310, another network (e.g., an ad hoc network, a local network, a private network, a virtual private network, and / or the like), and / or any other suitable communication technique. For example, client device 306 may include a user device and / or the like.
[0089] Communication network 310 may include one or more wired and / or wireless networks. For example, communication network 310 may include a cellular network (e.g., a long-term evolution (LTE®) network, a third generation (3G) network, a fourth generation (4G) network, a fifth generation (5G) network, a code division multiple access (CDMA) network, and / or the like), a public land mobile network (PLMN), a local area network (LAN), a wide area network (WAN), a metropolitan area network (MAN), a telephone network (e.g., the public switched telephone network (PSTN)), a private network (e.g., a private network associated with coherent memory system 302), an ad hoc network, an intranet, the Internet, a fiber optic-based network, a cloud computing network, and / or the like, and / or a combination of these or other types of networks.
[0090] The number and arrangement of systems, devices, and / or networks shown in FIG. 3 are provided as an example. There may be additional systems, devices, and / or networks; fewer systems, devices, and / or networks; different systems, devices, and / or networks; and / or differently arranged systems, devices, and / or networks than those shown in FIG. 3. Furthermore, two or more systems or devices shown in FIG. 3 may be implemented within a single system or device, or a single system or device shown in FIG. 3 may be implemented as multiple, distributed systems or devices. Additionally or alternatively, a set of systems (e.g., one or more systems) or a set of devices (e.g., one or more devices) of environment 300 may perform one or more functions described as being performed by another set of systems or another set of devices of environment 300.
[0091] Referring now to FIG. 4, shown is a diagram of example components of a device 400 according to non-limiting embodiments or aspects. Device 400 (and / or at least one component of device 400) may correspond to at least one of coherent memory system 102, processor 104, cache memory 106, and / or shared memory 108 in FIG. 1 and / or at least one of coherent memory system 302, computing device 304, client device 306, and / or server 308 in FIG. 3, as an example. In some non-limiting embodiments or aspects, such systems or devices in FIG. 1 or FIG. 3 may include at least one device 400 and / or at least one component of device 400. In some nonlimiting embodiments or aspects, device 400 may not implement a hardware cache coherence protocol. The number and arrangement of components shown in FIG. 4 are provided as an example. In some non-limiting embodiments or aspects, device 400 may include additional components, fewer components, different components, or differently arranged components than those shown in FIG. 4. Additionally, or alternatively, a set of components (e.g., one or more components) of device 400 may perform one or more functions described as being performed by another set of components of device 400.
[0092] As shown in FIG. 4, device 400 may include bus 402, processor 404, memory 406, storage component 408, input component 410, output component 412, and communication interface 414. Bus 402 may include a component that permits communication among the components of device 400. In some non-limiting embodiments or aspects, bus 402 may or may not implement a hardware cache coherence protocol. In some non-limiting embodiments or aspects, processor 604 may be implemented in hardware, software (e.g., firmware), or a combination of hardware and software. For example, processor 404 may include a processor (e.g., a central processing unit (CPU), a graphics processing unit (GPU), an accelerated processing unit (APU), etc.), a microprocessor, a digital signal processor (DSP), and / or any processing component (e.g., a field-programmable gate array (FPGA), an applicationspecific integrated circuit (ASIC), etc.) that can be programmed to perform a function. In some non-limiting embodiments or aspects, processor 404 may be the same as or similar to processor 104. Memory 406 may include RAM, ROM, and / or another type of dynamic or static storage device (e.g., flash memory, magnetic memory, optical memory, etc.) that stores information and / or instructions for use by processor 404. Insome non-limiting embodiments or aspects, memory 406 may be the same as or similar to cache memory 106 and / or shared memory 108.
[0093] With continued reference to FIG. 4, storage component 408 may store information and / or software related to the operation and use of device 400. For example, storage component 408 may include a hard disk (e.g., a magnetic disk, an optical disk, a magneto-optic disk, a solid state disk, etc.) and / or another type of computer-readable medium. In some non-limiting embodiments or aspects, storage component 408 may be the same as or similar to cache memory 106 and / or shared memory 108. Input component 410 may include a component that permits device 400 to receive information, such as via user input (e.g., a touch screen display, a keyboard, a keypad, a mouse, a button, a switch, a microphone, etc.). Additionally, or alternatively, input component 410 may include a sensor for sensing information (e.g., a global positioning system (GPS) component, an accelerometer, a gyroscope, an actuator, etc.). Output component 412 may include a component that provides output information from device 400 (e.g., a display, a speaker, one or more light-emitting diodes (LEDs), etc.). Communication interface 414 may include a transceiver-like component (e.g., a transceiver, a receiver, a transmitter, a separate receiver and transmitter pair, etc.) that enables device 400 to communicate with other devices, such as via a wired connection, a wireless connection, or a combination of wired and wireless connections. Communication interface 414 may permit device 400 to receive information from another device and / or provide information to another device. For example, communication interface 414 may include an Ethernet interface, an optical interface, a coaxial interface, an infrared interface, a radio frequency (RF) interface, a universal serial bus (USB) interface, a Wi-Fi® interface, a cellular network interface, and / or the like.
[0094] Device 400 may perform one or more processes described herein. Device 400 may perform these processes based on processor 404 executing software instructions stored by a computer-readable medium, such as memory 406 and / or storage component 408. A computer-readable medium may include any non-transitory memory device. A memory device includes memory space located inside of a single physical storage device or memory space spread across multiple physical storage devices. Software instructions may be read into memory 406 and / or storage component 408 from another computer-readable medium, or from another device viacommunication interface 414. When executed, software instructions stored in memory 406 and / or storage component 408 may cause processor 404 to perform one or more processes described herein. Additionally, or alternatively, hardwired circuitry may be used in combination with software instructions to perform one or more processes described herein. Thus, embodiments described herein are not limited to any specific combination of hardware circuitry and software. The term “programmed or configured,” as used herein, refers to an arrangement of software, hardware circuitry, or any combination thereof on one or more devices.
[0095] Referring now to FIG. 5, FIG. 5 shows an exemplary existing system 500 for implementing a cache coherence protocol in a multiprocessor system. System 500 may include coherence unit 502-1 to coherence unit 502-n (e.g., a plurality of coherence units, referred to individually as coherence unit 502 and collectively as coherence units 502 where appropriate), processor 504-1 to processor 504-n (e.g., a plurality of processors, referred to individually as processor 502 and collectively as processor 502 where appropriate), cache memory 506-1 to cache memory 506-n (e.g., a plurality of cache memories, referred to individually as cache memory 506 and collectively as cache memories 506 where appropriate), and shared memory 508.
[0096] Coherence unit 502 may include a computing device, such as a server (e.g. , a single server), a group of servers, and / or other like devices. Processor 504 (e.g., processor 504-1 to processor 504-n) may include at least one processor (e.g., a multicore processor), such as a CPU. In some non-limiting embodiments or aspects, processors 504 may include one or more processors executing instructions (e.g., software instructions) that cause processors 504 to perform one or more steps of methods and / or processes as described herein. Processor 504 may be in communication with cache memory 506 (e.g., cache memories 506-1 to 506-n). Processor 504 may be capable of receiving information (e.g., data, data resources, and / or the like) from and / or communicating (e.g., transmitting) information to cache memory 506.
[0097] At least one processor 504 may execute an operating system where the operating system is a multitasking operating system which may require a cache coherence protocol. For example, at least one processor 504 may execute the operating system where the operating system implements a cache coherence protocol and the operating system executes at least one coherence unit 502. At least oneprocessor 504 may add tasks to a shared memory queue, independent of whether processor 504 is executing a multitasking operating system or a non-multitasking atomic operating system.
[0098] Cache memory 506 (e.g., cache memories 506-1 to 506-n) may include cache memory internal to and / or associated with a processor (e.g., processors 504-1 to 504-n). Cache memory 506 may include a level 1 and / or primary cache, a level 2 and / or secondary cache, and / or a level 3 and / or tertiary cache. Cache memory 506 may include a storage component (e.g., a volatile storage component) that stores information and / or instructions for use by processor 504.
[0099] Shared memory 508 may include RAM, read only memory (ROM), and / or another type of dynamic or static storage device (e.g., flash memory, magnetic memory, optical memory, etc.) that stores information and / or instructions for use by system 500 and / or processor 504.
[0100] Although FIG. 5 shows an exemplary existing system 500 for implementing a cache coherence protocol in a multiprocessor system 500, it should be understood that methods and systems disclosed herein (e.g., coherent memory system 102 and / or coherent memory system 302) may be implemented in conjunction with an exemplary existing system 500 to augment and / or modify functions of coherence units 502.
[0101] Referring to FIG. 6A, FIG. 6A shows a flow diagram of a process 600A for allocating a memory resource independent of a cache coherence protocol. As shown in FIG. 6A, at step 602A, coherent memory system 102 (e.g., processor 104 thereof) may generate and / or execute a memory allocation instruction. Coherent memory system 102 may determine whether the memory allocation instruction is an instruction to allocate exclusive memory. If coherent memory system 102 determines the memory allocation instruction is for exclusive memory (e.g., an exclusive memory resource), then, at step 606A, coherent memory system 102 may execute the memory allocation instruction to allocate exclusive memory (e.g., cache memory). If coherent memory system 102 determines the memory allocation instruction is not a request to allocate exclusive memory, then, at step 604A, coherent memory system 102 may execute the memory allocation instruction to allocate global memory (e.g., shared memory). Once coherent memory system 102 has executed the memory allocation instruction, at step 608A, processor 104 of coherent memory system 102 may return a pointer (e.g., avariable storing a memory address) to the location of the allocated memory (e.g., exclusive memory or global memory).
[0102] Steps of process 600A may be performed in various orders and sequences and are not limited to being performed in the order shown in FIG. 6A. Accordingly, steps of process 600A are not limited to any particular order and may be performed by various components, whether implemented on a single computing device or multiple, distributed computing devices, having a multiprocessor architecture. Steps of process 600A may also be performed by a single processor or by multiple processors.
[0103] Referring to FIG. 6B, FIG. 6B shows a diagram of an example system 600B according to non-limiting embodiments or aspects depicting how a processor may pass coherence data (e.g., data stored in a memory resource transmitted to a coherence unit for coherence processing by a cache coherence protocol) to a coherence unit. System 600B may include coherence unit 602, processor 604, and cache memory or exclusive memory 606. Processor 604 may store a data resource in cache memory or exclusive memory 606. Processor 604 may determine whether the data resource is stored in an exclusive memory resource and processor 604 may tag and / or identify the data resource if processor 604 determines the data resource is stored in an exclusive memory resource. If processor 604 determines that the data resource is stored in an exclusive memory resource, processor 604 then may transmit a signal to coherence unit 602 where the signal causes coherence unit 602 to forgo performing cache coherence on the data resource. The data resource may be transmitted from cache memory or exclusive memory 606 to coherence unit 602 prior to or after processor 604 transmits the signal to forgo coherence to coherence unit 602. In some instances, processor 604 may also transmit a CPU identifier (ID) to coherence unit 602. The CPU ID may be used to determine whether the data resource is stored in an exclusive memory resource or in multiple cache memory resources.
[0104] Referring to FIG. 7 A, FIG. 7A shows a diagram of a system 700A pertaining to a CPU ID showing differences between execution of exclusive memory addresses and current memory addresses, and a relationship to cache memories. System 700A may include coherence unit 702, processor 704, cache memory 706, and other cache memory 710-1 to other cache memory 710-n (referred to collectively as other cache memories 710 and individually as cache memory 710, where appropriate). In some non-limiting embodiments or aspects, cache memory 706 may be the same as orsimilar to other cache memory 710 (e.g., at least one other cache memory 710 may be associated with processor 704). Other cache memories 710 may be present in a multiprocessor system, where each cache memory 710 is associated with a processor separate from processor 704.
[0105] In current systems, as shown in the left side of FIG. 7A, processor 704 may store data resources in cache memory 706. Coherence unit 702, having no knowledge and / or information of the status of the data resource, may perform cache coherence on every cache memory including cache memory 706 and / or other cache memories 710 in a multiprocessor system. Coherence unit 702 performs cache coherence on every cache memory 706 for every data resource stored in cache memories 706. In contrast, according to methods and system as disclosed herein, as shown on the right side of FIG. 7A, processor 704 may provide and / or associate a CPU ID with a data resource when processor 704 stores the data resource in cache memory 706 associated with processor 704. Coherence unit 702 may read the data resource and may detect and / or identify the CPU ID. Detection and / or identification of the CPU ID may cause coherence unit 702 to recognize that the data resource is stored in an exclusive memory resource and coherence unit 702 may be configured to forgo and / or prevent performing cache coherence on the data resource when coherence unit 702 detects and / or identifies the CPU ID associated with the data resource. In this way, the CPU ID may tag a memory resource where the data resource is stored as an exclusive memory resource so the memory resource is not subject to cache coherence, thus reducing computational resources and requirements to propagate data resources to other cache memories in a multiprocessor system.
[0106] Referring to FIG. 7B, FIG. 7B shows a diagram of a system 700B for cache coherence for an instruction executing an exclusive memory address and a nonexclusive memory address. System 700B may include coherence unit 702, processor 704, cache memory 706, shared memory 708, other cache memory 710, and exclusive memory 712. In some non-limiting embodiments or aspects, cache memory 706 may be the same as or similar to other cache memory 710 (e.g., at least one other cache memory 710 may be associated with processor 704) and / or exclusive memory 712 (e.g., cache memory 706 is labeled as exclusive memory 712 if / when cache memory 706 is identified as and / or determined to include a memory resource that does not require cache coherence. Other cache memories 710 may be present in amultiprocessor system, where each cache memory 710 is associated with a processor separate from processor 704. As an example, a shared memory resource in shared memory 708 may include read only memory. Coherence unit 702 may read the shared memory resource in shared memory 708 and, because the shared memory resource is read only, coherence unit 702 may forgo performing cache coherence on the data resource read from the shared memory resource (e.g., coherence is not needed because the shared memory resource is read only and any data resource stored therein cannot be written and / or altered).
[0107] Processor 704 may read the shared memory resource (e.g., the same shared memory resource storing the data resource that did not have cache coherence performed) and processor 704 may store the data resource (e.g., a copy of the data resource) in cache memory 706. In some non-limiting embodiments or aspects, processor 702 may provide and / or associate an exclusive address with the data resource and / or memory resource when storing the data resource in cache memory 706. If processor 704 provides and / or associates an exclusive address with the data resource when storing the data resource in cache memory 706, cache memory 706 may be identified as an exclusive memory resource, thus indicating and / or flagging to coherence unit 702 that cache memory 706 is exclusive memory (e.g., exclusive memory 712) and the data resource stored in cache memory 706 is stored in an exclusive memory resource that does not require cache coherence. Coherence unit 702 may detect the exclusive address associated with the data resource and detection of the exclusive address may cause coherence unit 702 to forgo performing cache coherence on the data resource. In this way, redundant coherence operations can be eliminated in computing systems (e.g., multiprocessing systems), for example, when read only data resources are copied to and stored in various cache memories in a multiprocessor system. Redundant coherence operations may be eliminated in various methods of memory resource and data resource handling, based on an exclusive address.
[0108] Alternatively, if processor 704 did not provide and / or associate an exclusive address with the data resource, then cache memory 706 may not be identified as an exclusive memory resource, thus indicating and / or flagging to coherence unit 702 that cache memory 706 is not exclusive memory and the data resource stored in cache memory 706 is not stored in an exclusive memory resource. Coherence unit 702 maythen perform cache coherence as normal on the data resource. In this way, normal cache coherence operations may continue to operate when they may be required to ensure data resource are current across multiple cache memories (e.g., in other cache memory 710) and / or shared memory (e.g., shared memory 708).
[0109] Referring to FIG. 7C, FIG. 7C shows a diagram of a system 700C to forgo cache coherence for an exclusive memory address and a global memory address. System 700C may include coherence unit 702, processor 704, shared memory 708, global memory 710, and exclusive memory 712 (e.g., the same as or similar to cache memory 706). In some non-limiting embodiments or aspects, shared memory 708 may include global memory 710. Global memory 710 may be protected memory such that global memory 710 is locked and a data resource can only be written to a global memory resource when global memory 710 is unlocked. In this way, global memory 710 that is protected (e.g., locked) may be treated as temporary exclusive memory.
[0110] As shown in FIG. 7C, on the left side of FIG. 7C, the diagram may depict the same or similar system and / or process to the left side of FIG. 7B. Thus, coherence unit 702 may read a shared memory resource in shared memory 708 and, because the shared memory resource is read only, coherence unit 702 may forgo performing cache coherence on the data resource read from the shared memory resource (e.g., coherence is not needed because the shared memory resource is read only and any data resource stored therein cannot be written and / or altered).
[0111] As shown in FIG. 7C, on the right side of FIG. 7C, system 700C differs from system 700B in that coherence unit 702 in system 700C may read a memory resource from shared memory 708 and / or global memory 710. A data resource having no indication of exclusivity (e.g., a CPU ID and / or an exclusive address, among various indicators) that originates and / or is read from global memory 710 may be treated as a data resource from an exclusive memory resource. This is because global memory 710 may be protected from being overwritten by processor 704 and / or threads of other processors while global memory 710 is locked. Thus, there may be no need to undergo cache coherence for a data resource stored in global memory 710 because the data resource in global memory 710 is unique and protected. Processor 704 may control updates to global memory 710, but processor 704 may not have permission (e.g., via software logic) to update global memory 710 when global memory 710 is protected. Global memory 710 may be identified by software logic and may change memorylocations in shared memory 708 as software logic for permissions changes. Coherence may not be required whenever a current copy of a data resource resides in shared memory 708.
[0112] The number and arrangement of systems and devices shown in FIGS. 7A- 7C are provided as an example. There may be additional systems and / or devices, fewer systems and / or devices, different systems and / or devices, and / or differently arranged systems and / or devices than those shown in FIG. 7A-7C. Furthermore, two or more systems or devices shown in FIG. 7A-7C may be implemented within a single system or device, or a single system or device shown in FIG. 7A-7C may be implemented as multiple, distributed systems or devices. Additionally or alternatively, a set of systems (e.g., one or more systems) or a set of devices (e.g., one or more devices) of system 700A-700C may perform one or more functions described as being performed by another set of systems or another set of devices of system 700A-700C.
[0113] Referring to FIG. 8, FIG. 8 shows a schematic diagram of a coherent memory system 800 which includes one or more processors with exclusive memory that allocate global memory in shared memory and process tasks from assigned task queues residing in shared memory, according to non-limiting embodiments and aspects disclosed herein.
[0114] Coherent memory system 800 may include processing unit (Pll) 804-1 to PU 804-n (referred to collectively as PUs 804 and individually as Pll 804, where appropriate), shared memory 808, PU-1 global memory 810-1 to Pll-n global memory 810-n (referred to collectively as PU global memories 810 and individually as PU global memory 810, where appropriate), and exclusive memory 812-1 to exclusive memory 812-n (referred to collectively as exclusive memories 812 and individually as exclusive memory 812, where appropriate, where each exclusive memory 812 is associated with each corresponding PU 804). In some non-limiting embodiments or aspects, exclusive memories 812 may be the same as or similar to cache memories associated with PUs 804, such that exclusive memories 812 are cache memories that have been identified as and / or determined to include exclusive memory resources. Coherent memory system 800 is shown in FIG. 8 as including a queue of global memory allocation requests, a queue of specific processor tasks, and a queue of general processor tasks (e.g., queues in software logic configured to organize data requests based on type of request, time of request, source of request, etc.).
[0115] In some non-limiting embodiments or aspects, coherent memory system 800 does not perform multitask operations and / or multitask processing. A processor (e.g., processor 704 and / or PU 804) may be depicted (e.g., in FIG. 7C) as executing multiprocessing software, thus any operating system software changes may be used for performance but are not required for migration. Coherent memory system 800 may not perform multitasking operations and / or multitask processing because each PU 804 may execute tasks without cache coherence operations being performed on data resources because of the use of exclusive memories 812 associated with each PU 804. In some non-limiting embodiments or aspects, when a number of tasks (e.g., a number of tasks combined between each queue) exceeds a number of PUs 804 available for processing in the system, then some remaining tasks may remain in a queue. Otherwise, each task may have a dedicated PU 804 to process and / or execute the task. Once each PU 804 has a task to execute, tasks may not be swapped out (e.g., shared) between PUs 804 and / or the queues and exclusive memories 812 and any data resources stored therein will not be shared among cache memories associated with PUs 804, thus keeping each exclusive memory 812 identified and / or determined to be an exclusive memory (e.g., a cache memory identified as including an exclusive memory resource), thereby not requiring performance of cache coherence on the cache memory of any PU 804.
[0116] In some non-limiting embodiments or aspects, each PU 804 may be associated with its own portion of shared memory 808 labeled as PU global memory 810. Global memory 808 (and each portion of shared memory 808 designated as PU global memory 810) may be protected by software logic, such as by using a hardware synchronization primitive (HSP). Shared memory 808 may also be protected because shared memory 808 includes one or more PU global memories 810 that are protected.
[0117] FIG. 8 shows system 800 having at least four locations of shared memory 808 that are designated as PU global memory 810 (PU global memory 810-1 , PU global memory 810-2, PU global memory 810-3, up to PU global memory 810-n) when memory locations of shared memory 808 are allocated by a respective processor (e.g., PU 804-1 allocates PU global memory 810-1 , PU 804-2 allocates PU global memory 810-2, etc.). FIG. 8 also shows at least two types of task queues, a specific tasks queue and a general tasks queue, and shows a queue representing global memory allocation requests.
[0118] As an example, at startup of system 800, Pll 804-1 may select and / or receive a first task from the Queue of Specific Processor T asks. Pll 804-1 may receive a first task including instructions to handle global memory allocation requests. Tasks from the Queue of Specific Processor Tasks may all be executed by a single processor (e.g., PU 804-1). In this example, since PU 804-1 is the sole Pll executing the first task to handle global memory allocation requests, PU 804-1 may execute atomically and without locks. PU 804-1 may execute multiple tasks received from the Queue of Specific Processor Tasks, but each task may only be executed on a single processor. Executing tasks without locks may enable tables to reside in cache memory. In this way, management of shared memory for system 800 may be accomplished without the use of any locks.
[0119] In a further example, at startup of system 800, another processor such as PU 804-2 may select and / or receive a second task from the Queue of Specific Processor Tasks. PU 804-2 may receive a second task including instructions to perform all accounting and billing for system 800. System 800 may include all program information in a trusted computing base (TCB) for the second task. In this example, since a current executing program may assume billing may be shared by multiple computing devices, billing updates may be contentious and may have been independently minimized by program software. By transferring billing tasks to another processor, execution may occur simultaneously if the program does not require immediate results.
[0120] In this example, PU 804-3 and / or PU 804-n may execute general tasks. General tasks may not require updating of shared information in shared memory 808 and / or general tasks may not require use of a system routine (e.g., a routine instructed to be performed by the operating system) to perform updating of shared memory 808. An example of a general task may include a user request such as a request to access and / or load a web page. Typically, general tasks may execute in exclusive memory and would thus not require coherence on data resources used for executing general tasks.
[0121] The number and arrangement of systems and devices shown in FIG. 8 are provided as an example. There may be additional systems and / or devices, fewer systems and / or devices, different systems and / or devices, and / or differently arranged systems and / or devices than those shown in FIG. 8. Furthermore, two or more systemsor devices shown in FIG. 8 may be implemented within a single system or device, or a single system or device shown in FIG. 8 may be implemented as multiple, distributed systems or devices. Additionally or alternatively, a set of systems (e.g., one or more systems) or a set of devices (e.g., one or more devices) of system 800 may perform one or more functions described as being performed by another set of systems or another set of devices of system 800.
[0122] Although embodiments have been described in detail for the purpose of illustration, it is to be understood that such detail is solely for that purpose and that the disclosure is not limited to the disclosed embodiments or aspects, but, on the contrary, is intended to cover modifications and equivalent arrangements that are within the spirit and scope of the appended claims. For example, it is to be understood that the present disclosure contemplates that, to the extent possible, one or more features of any embodiment or aspect can be combined with one or more features of any other embodiment or aspect.
Claims
WHAT IS CLAIMED IS:1 . A computer-implemented method, comprising: generating, with at least one processor, a request for allocating a memory resource; determining, with at least one processor, that the memory resource associated with the request does not require coherence via a cache coherence protocol executed by a cache controller; and transmitting, with at least one processor, a signal to the cache controller, wherein the signal causes the cache controller to forgo coherence on the memory resource via the cache coherence protocol.
2. The computer-implemented method of claim 1 , wherein the request for allocating the memory resource is an exclusive memory allocation request.
3. The computer-implemented method of claim 1 , further comprising transmitting the request for allocating the memory resource to a queue of requests.
4. The computer-implemented method of claim 3, further comprising: performing, with at least one processor, an atomic HSP on a shared memory location; and adding, with at least one processor, the request to a queue of memory allocation requests.
5. The computer-implemented method of claim 1 , wherein the memory resource comprises an exclusive memory resource of exactly one processor.
6. The computer-implemented method of claim 1 , wherein the cache coherence protocol is a hardware cache coherence protocol.
7. The computer-implemented method of claim 1 , further comprising:detecting, with at least one processor, a processor identifier associated with a processor; and transmitting, with at least one processor, the processor identifier to the cache controller.
8. The computer-implemented method of claim 1 , wherein determining that the memory resource associated with the request does not require coherence comprises: detecting an indicator included in the request for allocating the memory resource; and determining that the indicator included in the request is for an exclusive memory address.
9. A system for coherent memory, comprising: at least one processor; one or more cache memories, each cache memory comprising a cache controller programmed or configured to execute a cache coherence protocol; wherein the one or more processors are programmed or configured to: generate an exclusive memory allocation instruction allocating a memory resource such that when the memory resource is referenced in an instruction the at least one processor transmits coherence data in simultaneous signals to the cache controller such that the identifiers enable the cache controller to determine the location of values of the memory resource when performing coherence.
10. A computer-implemented method, comprising: generating, with at least one processor, a request for allocating a memory resource; determining, with at least one processor, the possible current locations of the value of the memory resource.; transmitting, with at least one processor, a signal to the cache controller, wherein the signal enables the cache controller to determine the location of values of the memory resource via the cache coherence protocol.
11. The computer-implemented method of claim 10, further comprising: detecting, with at least one processor, a processor identifier associated with a processor; and transmitting, with at least one processor, the processor identifier to the cache controller as an additional input to the cache coherence protocol.
12. A computer-implemented method, comprising: generating, with at least one processor, a request for allocating a memory resource; detecting, with at least one processor, an indicator included in the request for allocating the memory resource; determining, with at least one processor, that the indicator included in the request is associated with a global memory address; transmitting, with at least one processor, a signal to the cache controller, wherein the signal causes the cache controller to read from shared memory and write to shared memory and forgo coherence on the memory resource via a cache coherence protocol.
13. The computer-implemented method of claim 12, further comprising: protecting, with at least one processor, the memory resource via a hardware synchronization primitive (HSP) or an enqueue, wherein the memory resource comprises a global memory resource that is accessible by other tasks executing on one or more processors.
14. A system for coherent memory, comprising: at least one processor; one or more cache memories, each cache memory comprising a cache controller programmed or configured to execute a cache coherence protocol; wherein the one or more processors are programmed or configured to:generate a global memory allocation instruction allocating a memory resource such that when the memory resource is referenced in an instruction the at least one processor transmits a signal to the cache controller to read and write from shared memory and to forgo coherence on the memory resource via the cache coherence protocol.
Citation Information
Patent Citations
Private Memory Management using Utility Thread
US20220050790A1
Merging data for write allocate
US20220164217A1
Cache for Storing Coherent and Non-Coherent Data
US20220382679A1
US194862635027P
