Patent Application
20200409709
The disclosure relates generally to electronics, and, more specifically, an embodiment of the disclosure relates to time-multiplexing of a network or processing elements of a configurable spatial accelerator.
A processor, or set of processors, executes instructions from an instruction set, e.g., the instruction set architecture (ISA). The instruction set is the part of the computer architecture related to programming, and generally includes the native data types, instructions, register architecture, addressing modes, memory architecture, interrupt and exception handling, and external input and output (I/O). It should be noted that the term instruction herein may refer to a macro-instruction, e.g., an instruction that is provided to the processor for execution, or to a micro-instruction, e.g., an instruction that results from a processor's decoder decoding macro-instructions.
The present disclosure is illustrated by way of example and not limitation in the figures of the accompanying drawings, in which like references indicate similar elements and in which:
In the following description, numerous specific details are set forth. However, it is understood that embodiments of the disclosure may be practiced without these specific details. In other instances, well-known circuits, structures and techniques have not been shown in detail in order not to obscure the understanding of this description.
References in the specification to “one embodiment,” “an embodiment,” “an example embodiment,” etc., indicate that the embodiment described may include a particular feature, structure, or characteristic, but every embodiment may not necessarily include the particular feature, structure, or characteristic. Moreover, such phrases are not necessarily referring to the same embodiment. Further, when a particular feature, structure, or characteristic is described in connection with an embodiment, it is submitted that it is within the knowledge of one skilled in the art to affect such feature, structure, or characteristic in connection with other embodiments whether or not explicitly described.
A processor (e.g., having one or more cores) may execute instructions (e.g., a thread of instructions) to operate on data, for example, to perform arithmetic, logic, or other functions. For example, software may request an operation and a hardware processor (e.g., a core or cores thereof) may perform the operation in response to the request. One non-limiting example of an operation is a blend operation to input a plurality of vectors elements and output a vector with a blended plurality of elements. In certain embodiments, multiple operations are accomplished with the execution of a single instruction.
Exascale performance, e.g., as defined by the Department of Energy, may require system-level floating point performance to exceed 10{circumflex over ( )}18 floating point operations per second (exaFLOPs) or more within a given (e.g., 20 MW) power budget. Certain embodiments herein are directed to a spatial array of processing elements (e.g., a configurable spatial accelerator (CSA)) that targets high performance computing (HPC), for example, of a processor. Certain embodiments herein of a spatial array of processing elements (e.g., a CSA) target the direct execution of a dataflow graph to yield a computationally dense yet energy-efficient spatial microarchitecture which far exceeds conventional roadmap architectures. Certain embodiments herein overlay (e.g., high-radix) dataflow operations on a communications network, e.g., in addition to the communications network's routing of data between the processing elements, memory, etc. and/or the communications network performing other communications (e.g., not data processing) operations. Certain embodiments herein are directed to a communications network (e.g., a packet switched network) of a (e.g., coupled to) spatial array of processing elements (e.g., a CSA) to perform certain dataflow operations, e.g., in addition to the communications network routing data between the processing elements, memory, etc. or the communications network performing other communications operations. Certain embodiments herein are directed to network dataflow endpoint circuits that (e.g., each) perform (e.g., a portion or all) a dataflow operation or operations, for example, a pick or switch dataflow operation, e.g., of a dataflow graph. Certain embodiments herein include augmented network endpoints (e.g., network dataflow endpoint circuits) to support the control for (e.g., a plurality of or a subset of) dataflow operation(s), e.g., utilizing the network endpoints to perform a (e.g., dataflow) operation instead of a processing element (e.g., core) or arithmetic-logic unit (e.g. to perform arithmetic and logic operations) performing that (e.g., dataflow) operation. In one embodiment, a network dataflow endpoint circuit is separate from a spatial array (e.g. an interconnect or fabric thereof) and/or processing elements.
Below also includes a description of the architectural philosophy of embodiments of a spatial array of processing elements (e.g., a CSA) and certain features thereof. As with any revolutionary architecture, programmability may be a risk. To mitigate this issue, embodiments of the CSA architecture have been co-designed with a compilation tool chain, which is also discussed below.
Exascale computing goals may require enormous system-level floating point performance (e.g., 1 ExaFLOPs) within an aggressive power budget (e.g., 20 MW). However, simultaneously improving the performance and energy efficiency of program execution with classical von Neumann architectures has become difficult: out-of-order scheduling, simultaneous multi-threading, complex register files, and other structures provide performance, but at high energy cost. Certain embodiments herein achieve performance and energy requirements simultaneously. Exascale computing power-performance targets may demand both high throughput and low energy consumption per operation. Certain embodiments herein provide this by providing for large numbers of low-complexity, energy-efficient processing (e.g., computational) elements which largely eliminate the control overheads of previous processor designs. Guided by this observation, certain embodiments herein include a spatial array of processing elements, for example, a configurable spatial accelerator (CSA), e.g., comprising an array of processing elements (PEs) connected by a set of light-weight, back-pressured (e.g., communication) networks. One example of a CSA tile is depicted in
The derivation of a dataflow graph from a sequential compilation flow allows embodiments of a CSA to support familiar programming models and to directly (e.g., without using a table of work) execute existing high performance computing (HPC) code. CSA processing elements (PEs) may be energy efficient. In
Certain embodiments herein provide for performance increases from parallel execution within a (e.g., dense) spatial array of processing elements (e.g., CSA) where each PE and/or network dataflow endpoint circuit utilized may perform its operations simultaneously, e.g., if input data is available. Efficiency increases may result from the efficiency of each PE and/or network dataflow endpoint circuit, e.g., where each PE's operation (e.g., behavior) is fixed once per configuration (e.g., mapping) step and execution occurs on local data arrival at the PE, e.g., without considering other fabric activity, and/or where each network dataflow endpoint circuit's operation (e.g., behavior) is variable (e.g., not fixed) when configured (e.g., mapped). In certain embodiments, a PE and/or network dataflow endpoint circuit is (e.g., each a single) dataflow operator, for example, a dataflow operator that only operates on input data when both (i) the input data has arrived at the dataflow operator and (ii) there is space available for storing the output data, e.g., otherwise no operation is occurring.
Certain embodiments herein include a spatial array of processing elements as an energy-efficient and high-performance way of accelerating user applications. In one embodiment, applications are mapped in an extremely parallel manner. For example, inner loops may be unrolled multiple times to improve parallelism. This approach may provide high performance, e.g., when the occupancy (e.g., use) of the unrolled code is high. However, if there are less used code paths in the loop body unrolled (for example, an exceptional code path like floating point de-normalized mode) then (e.g., fabric area of) the spatial array of processing elements may be wasted and throughput consequently lost.
One embodiment herein to reduce pressure on (e.g., fabric area of) the spatial array of processing elements (e.g., in the case of underutilized code segments) is time multiplexing. In this mode, a single instance of the less used (e.g., colder) code may be shared among several loop bodies, for example, analogous to a function call in a shared library. In one embodiment, spatial arrays (e.g., of processing elements) support the direct implementation of multiplexed codes. However, e.g., when multiplexing or demultiplexing in a spatial array involves choosing among many and distant targets (e.g., sharers), a direct implementation using dataflow operators (e.g., using the processing elements) may be inefficient in terms of latency, throughput, implementation area, and/or energy. Certain embodiments herein describe hardware mechanisms (e.g., network circuitry) supporting (e.g., high-radix) multiplexing or demultiplexing. Certain embodiments herein (e.g., of network dataflow endpoint circuits) permit the aggregation of many targets (e.g., sharers) with little hardware overhead or performance impact. Certain embodiments herein allow for compiling of (e.g., legacy) sequential codes to parallel architectures in a spatial array.
In one embodiment, a plurality of network dataflow endpoint circuits combine as a single dataflow operator, for example, as discussed in reference to
An embodiment of a “Pick” dataflow operator is to select data (e.g., a token) from a plurality of input channels and provide that data as its (e.g., single) output according to control data. Control data for a Pick may include an input selector value. In one embodiment, the selected input channel is to have its data (e.g., token) removed (e.g., discarded), for example, to complete the performance of that dataflow operation (or its portion of a dataflow operation). In one embodiment, additionally, those non-selected input channels are also to have their data (e.g., token) removed (e.g., discarded), for example, to complete the performance of that dataflow operation (or its portion of a dataflow operation).
An embodiment of a “PickSingleLeg” dataflow operator is to select data (e.g., a token) from a plurality of input channels and provide that data as its (e.g., single) output according to control data, but in certain embodiments, the non-selected input channels are ignored, e.g., those non-selected input channels are not to have their data (e.g., token) removed (e.g., discarded), for example, to complete the performance of that dataflow operation (or its portion of a dataflow operation). Control data for a PickSingleLeg may include an input selector value. In one embodiment, the selected input channel is also to have its data (e.g., token) removed (e.g., discarded), for example, to complete the performance of that dataflow operation (or its portion of a dataflow operation).
An embodiment of a “PickAny” dataflow operator is to select the first available (e.g., to the circuit performing the operation) data (e.g., a token) from a plurality of input channels and provide that data as its (e.g., single) output. In one embodiment, PickSingleLeg is also to output the index (e.g., indicating which of the plurality of input channels) had its data selected. In one embodiment, the selected input channel is to have its data (e.g., token) removed (e.g., discarded), for example, to complete the performance of that dataflow operation (or its portion of a dataflow operation). In certain embodiments, the non-selected input channels (e.g., with or without input data) are ignored, e.g., those non-selected input channels are not to have their data (e.g., token) removed (e.g., discarded), for example, to complete the performance of that dataflow operation (or its portion of a dataflow operation). Control data for a PickAny may include a value corresponding to the PickAny, e.g., without an input selector value.
An embodiment of a “Switch” dataflow operator is to steer (e.g., single) input data (e.g., a token) so as to provide that input data to one or a plurality of (e.g., less than all) outputs according to control data. Control data for a Switch may include an output(s) selector value or values. In one embodiment, the input data (e.g., from an input channel) is to have its data (e.g., token) removed (e.g., discarded), for example, to complete the performance of that dataflow operation (or its portion of a dataflow operation).
An embodiment of a “SwitchAny” dataflow operator is to steer (e.g., single) input data (e.g., a token) so as to provide that input data to one or a plurality of (e.g., less than all) outputs that may receive that data, e.g., according to control data. In one embodiment, SwitchAny may provide the input data to any coupled output channel that has availability (e.g., available storage space) in its ingress buffer, e.g., network ingress buffer in
Certain embodiments herein thus provide paradigm-shifting levels of performance and tremendous improvements in energy efficiency across a broad class of existing single-stream and parallel programs, e.g., all while preserving familiar HPC programming models. Certain embodiments herein may target HPC such that floating point energy efficiency is extremely important. Certain embodiments herein not only deliver compelling improvements in performance and reductions in energy, they also deliver these gains to existing HPC programs written in mainstream HPC languages and for mainstream HPC frameworks. Certain embodiments of the architecture herein (e.g., with compilation in mind) provide several extensions in direct support of the control-dataflow internal representations generated by modern compilers. Certain embodiments herein are direct to a CSA dataflow compiler, e.g., which can accept C, C++, and Fortran programming languages, to target a CSA architecture.
Section 1 below discloses embodiments of CSA architecture. In particular, novel embodiments of integrating memory within the dataflow execution model are disclosed. Section 2 delves into the microarchitectural details of embodiments of a CSA. In one embodiment, the main goal of a CSA is to support compiler produced programs. Section 3 below examines embodiments of a CSA compilation tool chain. The advantages of embodiments of a CSA are compared to other architectures in the execution of compiled codes in Section 4. Finally the performance of embodiments of a CSA microarchitecture is discussed in Section 5, further CSA details are discussed in Section 6, and a summary is provided in Section 7.
The goal of certain embodiments of a CSA is to rapidly and efficiently execute programs, e.g., programs produced by compilers. Certain embodiments of the CSA architecture provide programming abstractions that support the needs of compiler technologies and programming paradigms. Embodiments of the CSA execute dataflow graphs, e.g., a program manifestation that closely resembles the compiler's own internal representation (IR) of compiled programs. In this model, a program is represented as a dataflow graph comprised of nodes (e.g., vertices) drawn from a set of architecturally-defined dataflow operators (e.g., that encompass both computation and control operations) and edges which represent the transfer of data between dataflow operators. Execution may proceed by injecting dataflow tokens (e.g., that are or represent data values) into the dataflow graph. Tokens may flow between and be transformed at each node (e.g., vertex), for example, forming a complete computation. A sample dataflow graph and its derivation from high-level source code is shown in
Embodiments of the CSA are configured for dataflow graph execution by providing exactly those dataflow-graph-execution supports required by compilers. In one embodiment, the CSA is an accelerator (e.g., an accelerator in
Turning to embodiments of the CSA, dataflow operators are discussed next.
The key architectural interface of embodiments of the accelerator (e.g., CSA) is the dataflow operator, e.g., as a direct representation of a node in a dataflow graph. From an operational perspective, dataflow operators behave in a streaming or data-driven fashion. Dataflow operators may execute as soon as their incoming operands become available. CSA dataflow execution may depend (e.g., only) on highly localized status, for example, resulting in a highly scalable architecture with a distributed, asynchronous execution model. Dataflow operators may include arithmetic dataflow operators, for example, one or more of floating point addition and multiplication, integer addition, subtraction, and multiplication, various forms of comparison, logical operators, and shift. However, embodiments of the CSA may also include a rich set of control operators which assist in the management of dataflow tokens in the program graph. Examples of these include a “pick” operator, e.g., which multiplexes two or more logical input channels into a single output channel, and a “switch” operator, e.g., which operates as a channel demultiplexor (e.g., outputting a single channel from two or more logical input channels). These operators may enable a compiler to implement control paradigms such as conditional expressions. Certain embodiments of a CSA may include a limited dataflow operator set (e.g., to relatively small number of operations) to yield dense and energy efficient PE microarchitectures. Certain embodiments may include dataflow operators for complex operations that are common in HPC code. The CSA dataflow operator architecture is highly amenable to deployment-specific extensions. For example, more complex mathematical dataflow operators, e.g., trigonometry functions, may be included in certain embodiments to accelerate certain mathematics-intensive HPC workloads. Similarly, a neural-network tuned extension may include dataflow operators for vectorized, low precision arithmetic.
In one embodiment, one or more of the processing elements in the array of processing elements 301 is to access memory through memory interface 302. In one embodiment, pick node 304 of dataflow graph 300 thus corresponds (e.g., is represented by) to pick operator 304A, switch node 306 of dataflow graph 300 thus corresponds (e.g., is represented by) to switch operator 306A, and multiplier node 308 of dataflow graph 300 thus corresponds (e.g., is represented by) to multiplier operator 308A. Another processing element and/or a flow control path network may provide the control signals (e.g., control tokens) to the pick operator 304A and switch operator 306A to perform the operation in
Communications arcs are the second major component of the dataflow graph. Certain embodiments of a CSA describes these arcs as latency insensitive channels, for example, in-order, back-pressured (e.g., not producing or sending output until there is a place to store the output), point-to-point communications channels. As with dataflow operators, latency insensitive channels are fundamentally asynchronous, giving the freedom to compose many types of networks to implement the channels of a particular graph. Latency insensitive channels may have arbitrarily long latencies and still faithfully implement the CSA architecture. However, in certain embodiments there is strong incentive in terms of performance and energy to make latencies as small as possible. Section 2.2 herein discloses a network microarchitecture in which dataflow graph channels are implemented in a pipelined fashion with no more than one cycle of latency. Embodiments of latency-insensitive channels provide a critical abstraction layer which may be leveraged with the CSA architecture to provide a number of runtime services to the applications programmer. For example, a CSA may leverage latency-insensitive channels in the implementation of the CSA configuration (the loading of a program onto the CSA array).
Dataflow architectures generally focus on communication and data manipulation with less attention paid to state. However, enabling real software, especially programs written in legacy sequential languages, requires significant attention to interfacing with memory. Certain embodiments of a CSA use architectural memory operations as their primary interface to (e.g., large) stateful storage. From the perspective of the dataflow graph, memory operations are similar to other dataflow operations, except that they have the side effect of updating a shared store. In particular, memory operations of certain embodiments herein have the same semantics as every other dataflow operator, for example, they “execute” when their operands, e.g., an address, are available and, after some latency, a response is produced. Certain embodiments herein explicitly decouple the operand input and result output such that memory operators are naturally pipelined and have the potential to produce many simultaneous outstanding requests, e.g., making them exceptionally well suited to the latency and bandwidth characteristics of a memory subsystem. Embodiments of a CSA provide basic memory operations such as load, which takes an address channel and populates a response channel with the values corresponding to the addresses, and a store. Embodiments of a CSA may also provide more advanced operations such as in-memory atomics and consistency operators. These operations may have similar semantics to their von Neumann counterparts. Embodiments of a CSA may accelerate existing programs described using sequential languages such as C and Fortran. A consequence of supporting these language models is addressing program memory order, e.g., the serial ordering of memory operations typically prescribed by these languages.
A primary architectural considerations of embodiments of the CSA involve the actual execution of user-level programs, but it may also be desirable to provide several support mechanisms which underpin this execution. Chief among these are configuration (in which a dataflow graph is loaded into the CSA), extraction (in which the state of an executing graph is moved to memory), and exceptions (in which mathematical, soft, and other types of errors in the fabric are detected and handled, possibly by an external entity). Section 2.9 below discusses the properties of a latency-insensitive dataflow architecture of an embodiment of a CSA to yield efficient, largely pipelined implementations of these functions. Conceptually, configuration may load the state of a dataflow graph into the interconnect (and/or communications network (e.g., a network dataflow endpoint circuit thereof)) and processing elements (e.g., fabric), e.g., generally from memory. During this step, all structures in the CSA may be loaded with a new dataflow graph and any dataflow tokens live in that graph, for example, as a consequence of a context switch. The latency-insensitive semantics of a CSA may permit a distributed, asynchronous initialization of the fabric, e.g., as soon as PEs are configured, they may begin execution immediately. Unconfigured PEs may backpressure their channels until they are configured, e.g., preventing communications between configured and unconfigured elements. The CSA configuration may be partitioned into privileged and user-level state. Such a two-level partitioning may enable primary configuration of the fabric to occur without invoking the operating system. During one embodiment of extraction, a logical view of the dataflow graph is captured and committed into memory, e.g., including all live control and dataflow tokens and state in the graph.
Extraction may also play a role in providing reliability guarantees through the creation of fabric checkpoints. Exceptions in a CSA may generally be caused by the same events that cause exceptions in processors, such as illegal operator arguments or reliability, availability, and serviceability (RAS) events. In certain embodiments, exceptions are detected at the level of dataflow operators, for example, checking argument values or through modular arithmetic schemes. Upon detecting an exception, a dataflow operator (e.g., circuit) may halt and emit an exception message, e.g., which contains both an operation identifier and some details of the nature of the problem that has occurred. In one embodiment, the dataflow operator will remain halted until it has been reconfigured. The exception message may then be communicated to an associated processor (e.g., core) for service, e.g., which may include extracting the graph for software analysis.
Embodiments of the CSA computer architectures (e.g., targeting HPC and datacenter uses) are tiled.
In one embodiment, the goal of the CSA microarchitecture is to provide a high quality implementation of each dataflow operator specified by the CSA architecture. Embodiments of the CSA microarchitecture provide that each processing element (and/or communications network (e.g., a network dataflow endpoint circuit thereof)) of the microarchitecture corresponds to approximately one node (e.g., entity) in the architectural dataflow graph. In one embodiment, a node in the dataflow graph is distributed in multiple network dataflow endpoint circuits. In certain embodiments, this results in microarchitectural elements that are not only compact, resulting in a dense computation array, but also energy efficient, for example, where processing elements (PEs) are both simple and largely unmultiplexed, e.g., executing a single dataflow operator for a configuration (e.g., programming) of the CSA. To further reduce energy and implementation area, a CSA may include a configurable, heterogeneous fabric style in which each PE thereof implements only a subset of dataflow operators (e.g., with a separate subset of dataflow operators implemented with network dataflow endpoint circuit(s)). Peripheral and support subsystems, such as the CSA cache, may be provisioned to support the distributed parallelism incumbent in the main CSA processing fabric itself. Implementation of CSA microarchitectures may utilize dataflow and latency-insensitive communications abstractions present in the architecture. In certain embodiments, there is (e.g., substantially) a one-to-one correspondence between nodes in the compiler generated graph and the dataflow operators (e.g., dataflow operator compute elements) in a CSA.
Below is a discussion of an example CSA, followed by a more detailed discussion of the microarchitecture. Certain embodiments herein provide a CSA that allows for easy compilation, e.g., in contrast to an existing FPGA compilers that handle a small subset of a programming language (e.g., C or C++) and require many hours to compile even small programs.
Certain embodiments of a CSA architecture admits of heterogeneous coarse-grained operations, like double precision floating point. Programs may be expressed in fewer coarse grained operations, e.g., such that the disclosed compiler runs faster than traditional spatial compilers. Certain embodiments include a fabric with new processing elements to support sequential concepts like program ordered memory accesses. Certain embodiments implement hardware to support coarse-grained dataflow-style communication channels. This communication model is abstract, and very close to the control-dataflow representation used by the compiler. Certain embodiments herein include a network implementation that supports single-cycle latency communications, e.g., utilizing (e.g., small) PEs which support single control-dataflow operations. In certain embodiments, not only does this improve energy efficiency and performance, it simplifies compilation because the compiler makes a one-to-one mapping between high-level dataflow constructs and the fabric. Certain embodiments herein thus simplify the task of compiling existing (e.g., C, C++, or Fortran) programs to a CSA (e.g., fabric).
Energy efficiency may be a first order concern in modern computer systems. Certain embodiments herein provide a new schema of energy-efficient spatial architectures. In certain embodiments, these architectures form a fabric with a unique composition of a heterogeneous mix of small, energy-efficient, data-flow oriented processing elements (PEs) (and/or a packet switched communications network (e.g., a network dataflow endpoint circuit thereof)) with a lightweight circuit switched communications network (e.g., interconnect), e.g., with hardened support for flow control. Due to the energy advantages of each, the combination of these components may form a spatial accelerator (e.g., as part of a computer) suitable for executing compiler-generated parallel programs in an extremely energy efficient manner. Since this fabric is heterogeneous, certain embodiments may be customized for different application domains by introducing new domain-specific PEs. For example, a fabric for high-performance computing might include some customization for double-precision, fused multiply-add, while a fabric targeting deep neural networks might include low-precision floating point operations.
An embodiment of a spatial architecture schema, e.g., as exemplified in
Programs may be converted to dataflow graphs that are mapped onto the architecture by configuring PEs and the network to express the control-dataflow graph of the program. Communication channels may be flow-controlled and fully back-pressured, e.g., such that PEs will stall if either source communication channels have no data or destination communication channels are full. In one embodiment, at runtime, data flow through the PEs and channels that have been configured to implement the operation (e.g., an accelerated algorithm). For example, data may be streamed in from memory, through the fabric, and then back out to memory.
Embodiments of such an architecture may achieve remarkable performance efficiency relative to traditional multicore processors: compute (e.g., in the form of PEs) may be simpler, more energy efficient, and more plentiful than in larger cores, and communications may be direct and mostly short-haul, e.g., as opposed to occurring over a wide, full-chip network as in typical multicore processors. Moreover, because embodiments of the architecture are extremely parallel, a number of powerful circuit and device level optimizations are possible without seriously impacting throughput, e.g., low leakage devices and low operating voltage. These lower-level optimizations may enable even greater performance advantages relative to traditional cores. The combination of efficiency at the architectural, circuit, and device levels yields of these embodiments are compelling. Embodiments of this architecture may enable larger active areas as transistor density continues to increase.
Embodiments herein offer a unique combination of dataflow support and circuit switching to enable the fabric to be smaller, more energy-efficient, and provide higher aggregate performance as compared to previous architectures. FPGAs are generally tuned towards fine-grained bit manipulation, whereas embodiments herein are tuned toward the double-precision floating point operations found in HPC applications. Certain embodiments herein may include a FPGA in addition to a CSA according to this disclosure.
Certain embodiments herein combine a light-weight network with energy efficient dataflow processing elements (and/or communications network (e.g., a network dataflow endpoint circuit thereof)) to form a high-throughput, low-latency, energy-efficient HPC fabric. This low-latency network may enable the building of processing elements (and/or communications network (e.g., a network dataflow endpoint circuit thereof)) with fewer functionalities, for example, only one or two instructions and perhaps one architecturally visible register, since it is efficient to gang multiple PEs together to form a complete program.
Relative to a processor core, CSA embodiments herein may provide for more computational density and energy efficiency. For example, when PEs are very small (e.g., compared to a core), the CSA may perform many more operations and have much more computational parallelism than a core, e.g., perhaps as many as 16 times the number of FMAs as a vector processing unit (VPU). To utilize all of these computational elements, the energy per operation is very low in certain embodiments.
The energy advantages our embodiments of this dataflow architecture are many. Parallelism is explicit in dataflow graphs and embodiments of the CSA architecture spend no or minimal energy to extract it, e.g., unlike out-of-order processors which must re-discover parallelism each time an instruction is executed. Since each PE is responsible for a single operation in one embodiment, the register files and ports counts may be small, e.g., often only one, and therefore use less energy than their counterparts in core. Certain CSAs include many PEs, each of which holds live program values, giving the aggregate effect of a huge register file in a traditional architecture, which dramatically reduces memory accesses. In embodiments where the memory is multi-ported and distributed, a CSA may sustain many more outstanding memory requests and utilize more bandwidth than a core. These advantages may combine to yield an energy level per watt that is only a small percentage over the cost of the bare arithmetic circuitry. For example, in the case of an integer multiply, a CSA may consume no more than 25% more energy than the underlying multiplication circuit. Relative to one embodiment of a core, an integer operation in that CSA fabric consumes less than 1/30th of the energy per integer operation.
From a programming perspective, the application-specific malleability of embodiments of the CSA architecture yields significant advantages over a vector processing unit (VPU). In traditional, inflexible architectures, the number of functional units, like floating divide or the various transcendental mathematical functions, must be chosen at design time based on some expected use case. In embodiments of the CSA architecture, such functions may be configured (e.g., by a user and not a manufacturer) into the fabric based on the requirement of each application. Application throughput may thereby be further increased. Simultaneously, the compute density of embodiments of the CSA improves by avoiding hardening such functions, and instead provision more instances of primitive functions like floating multiplication. These advantages may be significant in HPC workloads, some of which spend 75% of floating execution time in transcendental functions.
Certain embodiments of the CSA represents a significant advance as a dataflow-oriented spatial architectures, e.g., the PEs of this disclosure may be smaller, but also more energy-efficient. These improvements may directly result from the combination of dataflow-oriented PEs with a lightweight, circuit switched interconnect, for example, which has single-cycle latency, e.g., in contrast to a packet switched network (e.g., with, at a minimum, a 300% higher latency). Certain embodiments of PEs support 32-bit or 64-bit operation. Certain embodiments herein permit the introduction of new application-specific PEs, for example, for machine learning or security, and not merely a homogeneous combination. Certain embodiments herein combine lightweight dataflow-oriented processing elements with a lightweight, low-latency network to form an energy efficient computational fabric.
In order for certain spatial architectures to be successful, programmers are to configure them with relatively little effort, e.g., while obtaining significant power and performance superiority over sequential cores. Certain embodiments herein provide for a CSA (e.g., spatial fabric) that is easily programmed (e.g., by a compiler), power efficient, and highly parallel. Certain embodiments herein provide for a (e.g., interconnect) network that achieves these three goals. From a programmability perspective, certain embodiments of the network provide flow controlled channels, e.g., which correspond to the control-dataflow graph (CDFG) model of execution used in compilers. Certain network embodiments utilize dedicated, circuit switched links, such that program performance is easier to reason about, both by a human and a compiler, because performance is predictable. Certain network embodiments offer both high bandwidth and low latency. Certain network embodiments (e.g., static, circuit switching) provides a latency of 0 to 1 cycle (e.g., depending on the transmission distance.) Certain network embodiments provide for a high bandwidth by laying out several networks in parallel, e.g., and in low-level metals. Certain network embodiments communicate in low-level metals and over short distances, and thus are very power efficient.
Certain embodiments of networks include architectural support for flow control. For example, in spatial accelerators composed of small processing elements (PEs), communications latency and bandwidth may be critical to overall program performance. Certain embodiments herein provide for a light-weight, circuit switched network which facilitates communication between PEs in spatial processing arrays, such as the spatial array shown in
Spatial architectures, such as the one shown in
Operations may be executed based on the availability of their inputs and the status of the PE. A PE may obtain operands from input channels and write results to output channels, although internal register state may also be used. Certain embodiments herein include a configurable dataflow-friendly PE.
Instruction registers may be set during a special configuration step. During this step, auxiliary control wires and state, in addition to the inter-PE network, may be used to stream in configuration across the several PEs comprising the fabric. As result of parallelism, certain embodiments of such a network may provide for rapid reconfiguration, e.g., a tile sized fabric may be configured in less than about 10 microseconds.
Implementing distributed data channels may include two paths, illustrated in
The network may be statically configured, e.g., in addition to PEs being statically configured. During the configuration step, configuration bits may be set at each network component. These bits control, for example, the multiplexer selections and flow control functions. A network may comprise a plurality of networks, e.g., a data path network and a flow control path network. A network or plurality of networks may utilize paths of different widths (e.g., a first width, and a narrower or wider width). In one embodiment, a data path network has a wider (e.g., bit transport) width than the width of a flow control path network. In one embodiment, each of a first network and a second network includes their own data path network and flow control path network, e.g., data path network A and flow control path network A and wider data path network B and flow control path network B.
Certain embodiments of a network are bufferless, and data is to move between producer and consumer in a single cycle. Certain embodiments of a network are also boundless, that is, the network spans the entire fabric. In one embodiment, one PE is to communicate with any other PE in a single cycle. In one embodiment, to improve routing bandwidth, several networks may be laid out in parallel between rows of PEs.
Relative to FPGAs, certain embodiments of networks herein have three advantages: area, frequency, and program expression. Certain embodiments of networks herein operate at a coarse grain, e.g., which reduces the number configuration bits, and thereby the area of the network. Certain embodiments of networks also obtain area reduction by implementing flow control logic directly in circuitry (e.g., silicon). Certain embodiments of hardened network implementations also enjoys a frequency advantage over FPGA. Because of an area and frequency advantage, a power advantage may exist where a lower voltage is used at throughput parity. Finally, certain embodiments of networks provide better high-level semantics than FPGA wires, especially with respect to variable timing, and thus those certain embodiments are more easily targeted by compilers. Certain embodiments of networks herein may be thought of as a set of composable primitives for the construction of distributed, point-to-point data channels.
In certain embodiments, a multicast source may not assert its data valid unless it receives a ready signal from each sink. Therefore, an extra conjunction and control bit may be utilized in the multicast case.
Like certain PEs, the network may be statically configured. During this step, configuration bits are set at each network component. These bits control, for example, the multiplexer selection and flow control function. The forward path of our network requires some bits to swing its muxes. In the example shown in
For the third flow control box from the left in
The network(s) may be statically configured, e.g., in addition to PEs being statically configured during configuration for a dataflow graph. During the configuration step, configuration bits may be set at each network component. These bits may control, for example, the multiplexer selections to control the flow of a dataflow token (e.g., on a data path network) and its corresponding backpressure token (e.g., on a flow control path network). A network may comprise a plurality of networks, e.g., a data path network and a flow control path network. A network or plurality of networks may utilize paths of different widths (e.g., a first width, and a narrower or wider second width). In one embodiment, a data path network has a wider (e.g., bit transport) width than the width of a flow control path network. In one embodiment, each of a first network and a second network includes their own data paths and flow control paths, e.g., data path A and flow control path A and wider data path B and flow control path B. For example, a data path and flow control path for a single output buffer of a producer PE that couples to a plurality of input buffers of consumer PEs. In one embodiment, to improve routing bandwidth, several networks are laid out in parallel between rows of PEs Like certain PEs, the network may be statically configured. During this step, configuration bits may be set at each network component. These bits control, for example, the data path (e.g., multiplexer created data path) and/or flow control path (e.g., multiplexer created flow control path). The forward (e.g., data) path may utilize control bits to swing its switches and/or logic gates.
In certain embodiments, a CSA includes an array of heterogeneous PEs, in which the fabric is composed of several types of PEs each of which implement only a subset of the dataflow operators. By way of example,
PE execution may proceed in a dataflow style. Based on the configuration microcode, the scheduler may examine the status of the PE ingress and egress buffers, and, when all the inputs for the configured operation have arrived and the egress buffer of the operation is available, orchestrates the actual execution of the operation by a dataflow operator (e.g., on the ALU). The resulting value may be placed in the configured egress buffer. Transfers between the egress buffer of one PE and the ingress buffer of another PE may occur asynchronously as buffering becomes available. In certain embodiments, PEs are provisioned such that at least one dataflow operation completes per cycle. Section 2 discussed dataflow operator encompassing primitive operations, such as add, xor, or pick. Certain embodiments may provide advantages in energy, area, performance, and latency. In one embodiment, with an extension to a PE control path, more fused combinations may be enabled. In one embodiment, the width of the processing elements is 64 bits, e.g., for the heavy utilization of double-precision floating point computation in HPC and to support 64-bit memory addressing.
Embodiments of the CSA microarchitecture provide a hierarchy of networks which together provide an implementation of the architectural abstraction of latency-insensitive channels across multiple communications scales. The lowest level of CSA communications hierarchy may be the local network. The local network may be statically circuit switched, e.g., using configuration registers to swing multiplexor(s) in the local network data-path to form fixed electrical paths between communicating PEs. In one embodiment, the configuration of the local network is set once per dataflow graph, e.g., at the same time as the PE configuration. In one embodiment, static, circuit switching optimizes for energy, e.g., where a large majority (perhaps greater than 95%) of CSA communications traffic will cross the local network. A program may include terms which are used in multiple expressions. To optimize for this case, embodiments herein provide for hardware support for multicast within the local network. Several local networks may be ganged together to form routing channels, e.g., which are interspersed (as a grid) between rows and columns of PEs. As an optimization, several local networks may be included to carry control tokens. In comparison to a FPGA interconnect, a CSA local network may be routed at the granularity of the data-path, and another difference may be a CSA's treatment of control. One embodiment of a CSA local network is explicitly flow controlled (e.g., back-pressured). For example, for each forward data-path and multiplexor set, a CSA is to provide a backward-flowing flow control path that is physically paired with the forward data-path. The combination of the two microarchitectural paths may provide a low-latency, low-energy, low-area, point-to-point implementation of the latency-insensitive channel abstraction. In one embodiment, a CSA's flow control lines are not visible to the user program, but they may be manipulated by the architecture in service of the user program. For example, the exception handling mechanisms described in Section 1.2 may be achieved by pulling flow control lines to a “not present” state upon the detection of an exceptional condition. This action may not only gracefully stalls those parts of the pipeline which are involved in the offending computation, but may also preserve the machine state leading up the exception, e.g., for diagnostic analysis. A second network layer, e.g., the mezzanine network, may be a shared, packet switched network. Mezzanine network may include a plurality of distributed network controllers, network dataflow endpoint circuits. The mezzanine network (e.g., the network schematically indicated by the dotted box in
The composability of channels across network layers may be extended to higher level network layers at the inter-tile, inter-die, and fabric granularities.
For example, suppose the operation of this processing (e.g., compute) element is (or includes) what is called call a pick in
For example, suppose the operation of this processing (e.g., compute) element is (or includes) what is called call a switch in
Multiple networks (e.g., interconnects) may be connected to a processing element, e.g., (input) networks 1002, 1004, 1006 and (output) networks 1008, 1010, 1012. The connections may be switches, e.g., as discussed in reference to
Data input buffer 1024 and data input buffer 1026 may perform similarly, e.g., local network 1004 (e.g., set up as a data (as opposed to control) interconnect) is depicted as being switched (e.g., connected) to data input buffer 1024. In this embodiment, a data path (e.g., network as in
A processing element 1000 may be stalled from execution until its operands (e.g., a control input value and its corresponding data input value or values) are received and/or until there is room in the output buffer(s) of the processing element 1000 for the data that is to be produced by the execution of the operation on those operands.
In certain embodiments, a significant source of area and energy reduction is the customization of the dataflow operations supported by each type of processing element. In one embodiment, a proper subset (e.g., most) processing elements support only a few operations (e.g., one, two, three, or four operation types), for example, an implementation choice where a floating point PE only supports one of floating point multiply or floating point add, but not both.
In certain embodiments, data requests (e.g., a load request or a store request) are sent and received by memory interface circuits (e.g., RAF circuits) of a configurable spatial accelerator. In one embodiment, data corresponding to a request (e.g., a load request or a store request) is returned to the same memory interface circuit (e.g., RAF circuit) that issued the request. A request address file (RAF) circuit, a version of which is shown in
Stores may be similar, for example, except both address and data have to arrive (e.g., from one or more PEs) before any operation is sent off to the memory system in certain embodiments.
Local network 1102, 1104, 1106, or 1108 may be a circuit switched network, e.g., as discussed in reference to
Optionally, a translation lookaside buffer (TLB) 1146 may be included to convert a logical address received from an input queue 1122, 1124, or 1126 into a physical address of the memory (e.g., cache). In one embodiment, the memory accessed is one or more of the cache banks discussed herein.
Dataflow graphs may be capable of generating a profusion of (e.g., word granularity) requests in parallel. Thus, certain embodiments of the CSA provide a cache subsystem with sufficient bandwidth to service the CSA. A heavily banked cache microarchitecture, e.g., as shown in
In certain embodiments, accelerator-cache network is further coupled to circuitry 1220 that includes a cache home agent and/or next level cache. In certain embodiments, accelerator-cache network (e.g., interconnect) is separate from any (for example, circuit switched or packet switched) network of an accelerator (e.g., accelerator tile), e.g., RAF is the interface between the processing elements and the cache home agent and/or next level cache. In one embodiment, a cache home agent is to connect to a memory (e.g., separate from the cache banks) to access data from that memory (e.g., memory 202 in
In certain embodiments of a CSA, optimal use is executing the operations associated with a (e.g., main) inner loop in every cycle (for example, of a clock that is driving the execution, e.g., on a clock edge). However, there may be support operations (e.g., outer loops) which do not execute every cycle and which could share same CSA resources without harming overall program performance. Reducing the resources required for these cooler code segments allows more resources to be dedicated to the hot inner loops, benefitting performance.
Certain embodiments herein provide time multiplexing in a spatial array (e.g., CSA) which allow for less frequently execute operations to share spatial array (e.g., fabric) resources. This reduces the overall number of resources needed to execute some codes and improves performance per unit area and performance per unit of power (e.g., Watt). Certain embodiments herein allow the sharing of transmitting and/or receiving processing elements (e.g., that are physically adjacent) to reduce the distance between those processing elements.
Certain embodiments herein allows less utilized portions of a dataflow graph to use fewer resources. For example, on a diverse set of dataflow graphs, a large fraction of the PE network (e.g., latency-insensitive channels (LICs)) can be multiplexed without performance loss, and thus utilize hardware with fewer resources and thus the performance per area is increased.
Certain embodiments herein provide for network time-multiplexing and/or processing element time-multiplexing, e.g., to save implementation area. Certain embodiments herein include bidding to avoid spurious switching (e.g., every cycle) between the multiplexed phases. and/or to permit baseline full-throughput compatibility with no energy overhead. Certain embodiments herein allow for optional time-multiplexing while still permitting baseline full-throughput compatibility with no energy overhead. Certain embodiments herein allow for finer division of switching for quality of service (QoS) instead of using multiplexing-for-timing closure that requires fixed-sized multiplexing slots.
Certain embodiments herein allow for multiplexing inter-PE communications networks, for example, including data path network and the flow control (e.g., backpressure) path network. Certain embodiments herein introduce new configuration within the multiplexor control of the CSA (e.g., circuit-switched network thereof). Certain embodiments herein introduce low-overhead QoS mechanisms to expand the applicability of the PE network. Certain embodiments herein introduce scheduling mechanisms to improve the performance of the PE network.
Certain embodiments herein introduce new configuration state into a multiplexed network.
Certain embodiments herein of time-multiplexing are fully performance-compatible with a not time-multiplexed network, e.g., where baseline performance is achieved simply by setting the configuration registers to statically use the same path (e.g., same phase) which has the effect of providing a dedicated link. In one embodiment, there are no changes to a processing element. In one embodiment (e.g., as discussed further below), a processing element includes time-multiplexing functionality, e.g., where each input and output of the processing element is to comprehend the cycles on which it can communicate, according to the particular time-multiplexing scheme. Certain embodiments herein allow any paths of a CSA to be dynamically switched using time-multiplexing. In one embodiment, by switching all links at once using a shared (e.g., global) control indication (e.g., clock), a coherent link is formed between the communicating endpoints (e.g., PEs, RAFs, etc.). In certain embodiments, time-multiplexing is only used on a proper subset of the networks of a CSA, e.g., where the networks are essentially separate. This allows better design tuning through the mixing of time-multiplexed and not time-multiplexed networks to meet the needs of the certain design embodiments.
In certain embodiments, circuit switched network 1410 includes storage for configuration zero 1450 and configuration one 1452. In one embodiment, the control indication (e.g., clock) value 1456 is provided as control for multiplexer 1454, e.g., to cause the output of configuration zero 1450 in a first time period of control indication (e.g., clock) 1456 and output (e.g., different) configuration one 1452 in a second time period of control indication (e.g., clock) 1456 to the multiplexers of network 1410. In the depicted embodiment, a configuration zero 1450 (e.g., phase) is to cause data output buffer 1 1436A of PE 1400A to be coupled to data input buffer 0 1424C of PE 1400C, e.g., and also flow control (e.g., backpressure) path output 1408C of PE 1400C to couple to flow control (e.g., backpressure) path input 1408A of PE 1400A.
In one embodiment, circuit switched network 1410 includes (i) a data path to send data from first PE 1400A to third PE 1400C and a data path from second PE 1400B to third PE 1400C, and (ii) a flow control path to send control values that controls (or is used to control) the sending of that data from first PE 1400A and second PE 1400B to third PE 1400C. Data path may send a data (e.g., valid) value when data is in an output queue (e.g., buffer) (e.g., when data is in control output buffer 1432A, first data output buffer 1434A, or second data output queue (e.g., buffer) 1436A of first PE 1400A and when data is in control output buffer 1432B, first data output buffer 1434B, or second data output queue (e.g., buffer) 1436B of second PE 1400B). In one embodiment, each output buffer includes its own data path, e.g., for its own data value from producer PE to consumer PE and this data path may be time-multiplexed. Components in PE are examples, for example, a PE may include only a single (e.g., data) input buffer and/or a single (e.g., data) output buffer. Flow control path may send control data that controls (or is used to control) the sending of corresponding data from first PE 1400A and second PE 1400B to third PE 1400C. Flow control data may include a backpressure value from each consumer PE (or aggregated from all consumer PEs, e.g., with an AND logic gate). Flow control data may include a backpressure value, for example, indicating a buffer of the third PE 1400C that is to receive an input value is full.
Turning to the depicted PEs, processing elements 1400A-C include operation configuration registers 1419A-C that may be loaded during configuration (e.g., mapping) and specify the particular operation or operations (for example, to indicate whether to enable in-network pick mode or not). In one embodiment, only the operation configuration register 1419C of the receiving PE 1400C is loaded with the operation configuration value for in-network pick.
Multiple networks (e.g., interconnects) may be connected to a processing element, e.g., networks 1402, 1404, 1406, and 1410. The connections may be switches (e.g., multiplexers). In one embodiment, PEs and a circuit switched network 1410 are configured (e.g., control settings are selected) such that circuit switched network 1410 provides the paths for a desired operation.
A processing element (e.g., or in the network itself) may include a conditional queue (e.g., having only a single slot, of having multiple slots in each conditional queue) as discussed herein. In one embodiment, a single buffer (e.g., or queue) includes its own, respective conditional queue. In the depicted embodiment, conditional queue 1413 is included for control input buffer 1422C, conditional queue 1415 is included for first data input buffer 1424C, and conditional queue 1417 is included for second data input buffer 1426C of PE 1400C. In some embodiments, any conditional queue of a receiver PE (e.g. 1400C) can be used to as a part of the operations described herein. The coupling of conditional queues may also be time-multiplexed.
In certain embodiments, the key performance and energy efficiency enablers of a spatial architecture is dedicated point to point communication. In one embodiment, such communication is local and occurs in a single cycle. However, more distant communication can take multiple cycles to occur. To accommodate this multi-cycle communication and to simplify compilation, certain embodiments of spatial communications networks include in-network storage resources (e.g., elements). In network storage may require that multiplexed channels are separated from one another in order to prevent deadlocks. To solve this issue, certain embodiments herein utilize time-multiplexing within in-network storage resources (e.g., elements). For example, to achieve full throughput in one embodiment, in-fabric storage elements are provisioned with two slots that can be partitioned between the two multiplexed phases (e.g., flows) to form virtual channels.
Depicted in-network storage element 1501 includes a buffer 1542. Buffer 1542 has multiple slots therein. In one embodiment to support time-multiplexing, instead of sharing multiple slots (e.g., as a first-in-first-out buffer), buffer 1542 includes a respective slot or slots for use with each particular phase (e.g., A or B, etc.) of a multi-phased execution in time-multiplexing. In certain embodiments, configuration storage 1556 (e.g., register) has a configuration value stored therein (e.g., during configuration time), and when the configuration value includes a first value, time-multiplexing (e.g., a slot is reserved only for data for a particular phase) is enabled for the in-network storage element 1501 and when the configuration value includes a second value, time-multiplexing (e.g., a slot is reserved only for data for a particular phase) is disabled for the in-network storage element 1501. Additionally or alternatively, when the configuration value includes a third value, a data value from upstream path 1550 is stored into a slot of buffer 1542 (and/or sent from a slot of buffer 1542 to output path 1552, e.g., in a following cycle) and, when the configuration value includes a fourth second value, a data value from upstream path 1550 is steered downstream (e.g., with no delay greater than the delay caused by the physical path) via bypass path 1544 to downstream path 1552 (e.g., to a downstream PE (e.g., receiver PE)), for example, without being stored within (and/or modified by) in-network storage element 1501. In certain embodiments, controller 1540 controls the selection of (i) bypass mode that utilizes bypass path 1544 or (ii) buffer mode that utilizes buffer 1542 storage based on the configuration value stored in configuration storage 1556. In certain embodiments, a data value may include the data itself (e.g., payload) along with an “valid out” value that indicates the data itself is valid for a buffer to store. In certain embodiments, a “valid out” value is sent with the flow control values, e.g., from port 1508A of PE 1500A to port 1508B of PE 1500B or port 1508C of PE 1500C. As discussed further herein, ports (e.g., port 1508A of upstream (TX) PE 1500A to port 1508B of downstream (RX) PE 1500B or port 1508C of downstream (RX) PE 1500C.) may be used to send ready, valid in, and/or valid out values to PEs and/or in-network elements (e.g., in-network element 1501). In one embodiment, the path 1550 extending from buffer 1532A carries a “valid out” value and the data value itself (e.g., the path 1550 may include two, respective wires). In certain embodiments, port 1508A of PE 1500A, port 1508B of PE 1500B, and/or port 1508C of PE 1500C can send and receive data. For example, port 1508A of PE 1500A can (i) receive a flow control value(s) (e.g., a back-pressure value) from a downstream component (e.g., in-network storage element 1501, PE 1500B, or PE 1500C) and/or (ii) send a “valid out” value to a downstream component.
In certain embodiments, (i) when in bypass mode, a flow control value(s) (e.g., a back-pressure value) received on downstream path 1548 is steered upstream by controller 1540 to upstream path 1546 (e.g., to an upstream PE (e.g., transmitter PE)), for example, without being modified by in-network storage element 1501, when in bypass mode or (ii) when in buffer mode, controller 1540 is to send a flow control value to upstream path 1546 (e.g., to an upstream PE (e.g., transmitter PE)) based on the status of buffer 1542 (e.g., a “queue ready” value when storage space is available in buffer 1542).
As one example when in a particular phase, first (e.g., as producer) PE 1500A includes a (e.g., input) port 1508A(1-6) coupled to network 1510, e.g., to receive a backpressure value from second (e.g., as consumer) PE 1500B or third (e.g., as consumer) PE 1500C. In one circuit switched configuration, (e.g., input) port 1508A(1-6) (e.g., having a plurality of parallel inputs (1), (2), (3), (4), (5), and (6)) is to receive a respective backpressure value from each one of control input buffer 1522B, first data input buffer 1524B, and second data input buffer 1526B and/or control input buffer 1522C, first data input buffer 1524C, and second data input buffer 1526C.
In one embodiment (e.g., when in a particular phase), a circuit switched backpressure path (e.g., channel) is formed by setting switches coupled to wires between an input (e.g., input 1, 2, or 3) of port 1508A and an output (e.g., output 1, 2, or 3) of port 1508B to send a backpressure token (e.g., a value indicating no storage is available in an input buffer/queue) for one of control input buffer 1522B, first data input buffer 1524B, or second data input buffer 1526B of second PE 1500B. Additionally or alternatively, a (e.g., different) circuit switched backpressure path (e.g., channel) is formed by setting switches coupled to wires between an input (e.g., a different input of input 1, 2, or 3 (or one of more than 3 inputs in another embodiment) of port 1508A and an output (e.g., output 1, 2, or 3) of port 1508C to send a backpressure token (e.g., a value indicating no storage is available in an input buffer/queue) for one of control input buffer 1522C, first data input buffer 1524C, or second data input buffer 1526C of third PE 1500C.
In certain embodiments (e.g., when in a particular phase), output buffer 1532A of PE 1500A is coupled to in-network storage element, and, (i) when the configuration value in configuration storage 1556 is a certain value, a data value stored in output buffer 1532A of PE 1500A is sent through upstream path 1550 (e.g., upstream from in-network storage element 1501) and is steered downstream (e.g., with no delay greater than the delay caused by the physical path) via bypass path 1544 to downstream path 1552 (e.g., downstream from in-network storage element 1501) to an input buffer of a downstream PE (e.g., PE 1500B or PE 1500C), for example, without being stored within (and/or modified by) in-network storage element 1501, and (ii) when the configuration value in configuration storage 1556 is a different value, a (e.g., different) data value stored in buffer 1532A is sent through upstream path 1550 into a slot of buffer 1542 of in-network storage element 1501. In one embodiment, that (e.g., different) value stored in buffer 1542 is sent to downstream PE (e.g., PE 1500B or PE 1500C), e.g., when the input buffer of the downstream PE has an available storage space (e.g., slot).
Depicted multiplexer 1614 includes multiple inputs (e.g., where multiple buffers of a PE or buffers of other PEs can be selected as an input) and a single output that sources a desired input based on output from control multiplexer 1610. In one embodiment, control multiplexer 1610 sources one of configuration zero 1606 or (e.g., different) configuration one 1606 values and the values are passed as control values (e.g., 00, 01, 10, or 11 to uniquely identify each of four inputs). In certain embodiments, configuration zero 1606 and configuration one 1606 are stored therein (e.g., as a register) via a configuration of a CSA, e.g., via a network as discussed herein. In the depicted embodiment, the control value 1612 is provided from shift register 1620. Additional granularity can be obtained by providing a state machine that chooses amongst the configured LICs according to a defined pattern. In one embodiment with a shift register and two potential configurations, only a single bit is used per multiplexing element (e.g., slot) in the shift register. The use of bandwidth apportionment (e.g., instead of just switching back and forth each cycle between two phases) may reduce spurious dynamic switching. For example, in the case that two low-utilization communications paths are sharing a link, one phase can be assigned one of the slots and the other the remaining slots such that any switching penalties are incurred in fewer cycles, e.g., even if no bidding occurs.
In certain embodiments, a main energy cost of time-multiplexing is data toggling due to switching the network multiplexors. One approach that may reduce the amount of switching is to introduce a communications setup step (e.g., bidding). In this case, multiple flow control networks are constructed such that each point can see the flow control status of both communications endpoints (e.g., PEs). Prior to the cycle in which communications is to occur, the communication end points flow control is examined to determine if communications will actually occur (e.g., output data is available and input storage space for the data is available). In one embodiment, if communications will occur, then this is recorded in a “bid” register which permits the configuration multiplexor to switch, and if no communication can occur, then the multiplexor is not switched. Examples of data and flow control are discussed in reference to
In one embodiment of time-multiplexing of a CSA network, a particular network to be shared equally at a 50% duty cycle among two communication paths. However, this may limits the LICs that can be multiplexed to those that have a throughput of less than 0.5 tokens per cycle, and also remains wasteful of bandwidth in the case that the multiplexed LIC has a duty cycle below 0.5 tokens per cycle. It is possible that more LICs can be accommodated by improving the granularity of multiplexing. One approach is to simply increase the configurations used by the particular network, for example having K configurations rather than just two. In the case that fewer than K paths cross a switch point, bandwidth can be apportioned by assigning one communication path more slots.
Depicted multiplexer 1714 includes multiple inputs (e.g., where multiple buffers of PE or buffers of other PEs can be selected as an input) and a single output that sources a desired input based on output from control multiplexer 1710. In one embodiment, control circuit 1710 sources one of configuration zero 1706 or (e.g., different) configuration one 1708 values and the values are passed as control values (e.g., 00, 01, 10, or 11 to uniquely identify each of four inputs) to generate. In certain embodiments, configuration zero 1706 and configuration one 1708 are stored therein (e.g., as a register) via a configuration of a CSA, e.g., via a network as discussed herein. In one embodiment, the phase 1728 (e.g., clock value) is provided as control for control circuit 1710, e.g., to cause the output of configuration zero 1306 in a first phase (e.g., time period) when bid 1720 indicates corresponding data and storage are available and output (e.g., different) configuration one 1308 in a second phase (e.g., time period) when bid 1720 indicates corresponding data and storage are available. In one embodiment, the phase 1728 is a clock that cycles to cause, when bid is true, the alternating output of configuration zero 1306 and (e.g., different) configuration one 1308 to create a first phase and second phase, respectively. In certain embodiments, bid receiver (RX) multiplexer 1726 and bid transmitter (TX) multiplexer 1724 are included to check for a “not full” value (e.g., from input controller 2500) of a receiver PE and a “valid” (e.g., “not empty”) value (e.g., from output controller 3500) of a transmitter PE. Thus, in one embodiment, in a given phase (e.g., in a first clock cycle), the multiplexer 1712 causes a configuration from storage 1706 (or 1708 in a different clock cycle) to be output as control to bid receiver (RX) multiplexer 1726 and bid transmitter (TX) multiplexer 1724 to source values that indicate if the RX PE has storage space available and the TX PE has data to send to that storage space, and, when the values both indicate yes, the AND gate 1722 outputs a value to bid register 1720. Bid register 1720 then causes a respective output from control circuit 1710 to source one of configuration zero 1706 or (e.g., different) configuration one 1708 values and the values are passed as control values (e.g., 00, 01, 10, or 11 to uniquely identify each of four inputs, to output 1718 the input 1716 that won the current bid.
Additionally or alternatively to time-multiplexing of a CSA network, processing elements may be time-multiplexed. In one embodiment, when a PE is time-multiplexed, its input and output buffers (e.g., queues) are hard partitioned among the time-multiplexed phases (e.g., as contexts). This partitioning creates a pair of virtual channels, one for each phase, which ensures deadlock-freedom between the contexts. Execution occurs for a context when the operands associated with the context become available. True time multiplexing is also acceptable. For stateful dataflow operators, such as repeat (repeat, repeato) or stream compare (scmp) the state of the operation is replicated for each context in certain embodiments.
In one embodiment, data transmission over the local network occurs in a time-multiplexed fashion. In one embodiment, PEs have a known phase which is shared across the PEs (e.g., in a tile). In one embodiment, only the context associated with the phase can transmit, thus the processing elements assert the appropriate flow control status for the appropriate phase. In one embodiment, phases are switched cycle-by-cycle, allowing the multiplexed phases to share network bandwidth. However, network paths need not be switched in certain embodiments. Time division or spatial division network multiplexing is compatible with this approach, but not required.
It is possible to introduce additional configuration in time-multiplexing approach to permit each operation to have its own ALU or logical operation. This is a more general approach, but uses additional configuration state to be provisioned at each processing element in one embodiment. In one embodiment, the choice of configuration is made based on operand availability, and execution for a context can potentially occur whenever the operands of a context are available.
The below discusses two multiplexed phases (e.g., as contexts), but any plurality of time-multiplexing of PEs is possible, e.g., provided that there is enough buffering to establish virtual channels for each context. Certain embodiments herein allow for the multiplexing of stateful dataflow graphs, e.g., with the introduction of virtual channels within the microarchitecture and new dataflow operations to inject tokens into the multiplexed subgraph.
In certain embodiments, an issue in providing virtual channels is that the amount of buffering per channel is reduced. Thus, if the buffer is reduced to one, synchronized time multiplexing may result in throughput loss due to the need to land new data from an upstream PE without having consumed previous data from a downstream PE. To fix this issue, certain embodiments herein use offset phased execution, in which PE execution and network transfer of a phase occur in back-to-back cycles as discussed below in reference to
Note that in one embodiment, the PEs switch phases (e.g., A to B) without the network switching phases. In one embodiment, PEs in phase A (e.g., as programmed by a configuration value and implemented by a scheduler) implement a first operation and, when switched by a control indication (e.g., clock) to phase B (e.g., in consecutive cycles), those same PEs in phase B (e.g., as programmed by a configuration value and implemented by a scheduler) implement a second, different operation. In one embodiment, a separate slot of input buffers and output buffers is reserved for a particular phase, for example, with the upper slot as shown for each of input buffers and output buffers being reserved for PE outputs for phase A and the lower slot for each of input buffers and output buffers being reserved for PE outputs for phase B. In one embodiment, in a same cycle, PEs operate on phase A data (e.g., and not phase B data), and network transmits phase B data (and not phase A data), e.g., and in a next cycle, PEs operate on phase B data (e.g., and not phase A data), and network transmits phase A data (and not phase B data), and this may be repeated.
In certain embodiments, circuit switched network 1810 includes storage for configuration zero 1850 and configuration one 1852. In one embodiment, the control indication (e.g., clock) value 1856 is provided as control for multiplexer 1854, e.g., to cause the output of configuration zero 1850 in a first time period of control indication (e.g., clock) 1856 and output (e.g., different) configuration one 1852 in a second time period of control indication (e.g., clock) 1856 to the multiplexers of network 1810. In the depicted embodiment, a configuration zero 1850 (e.g., phase) is to cause data output buffer 1 1836A of PE 1800A to be coupled to data input buffer 0 1824C of PE 1800C, e.g., and also flow control (e.g., backpressure) path output 1808C of PE 1800C to couple to flow control (e.g., backpressure) path input 1808A of PE 1800A.
In one embodiment, circuit switched network 1810 includes (i) a data path to send data from first PE 1800A to third PE 1800C and a data path from second PE 1800B to third PE 1800C, and (ii) a flow control path to send control values that controls (or is used to control) the sending of that data from first PE 1800A and second PE 1800B to third PE 1800C. Data path may send a data (e.g., valid) value when data is in an output queue (e.g., buffer) (e.g., when data is in control output buffer 1832A, first data output buffer 1834A, or second data output queue (e.g., buffer) 1836A of first PE 1800A and when data is in control output buffer 1832B, first data output buffer 1834B, or second data output queue (e.g., buffer) 1836B of second PE 1800B). In one embodiment, each output buffer includes its own data path, e.g., for its own data value from producer PE to consumer PE and this data path may be time-multiplexed. Components in PE are examples, for example, a PE may include only a single (e.g., data) input buffer and/or a single (e.g., data) output buffer. Flow control path may send control data that controls (or is used to control) the sending of corresponding data from first PE 1800A and second PE 1800B to third PE 1800C. Flow control data may include a backpressure value from each consumer PE (or aggregated from all consumer PEs, e.g., with an AND logic gate). Flow control data may include a backpressure value, for example, indicating a buffer of the third PE 1800C that is to receive an input value is full.
Turning to the depicted PEs, processing elements 1800A-C include operation configuration registers 1819A-C that may be loaded during configuration (e.g., mapping) and specify the particular operation or operations (for example, to indicate whether to enable in-network pick mode or not). In one embodiment, only the operation configuration register 1819C of the receiving PE 1800C is loaded with the operation configuration value for in-network pick.
Multiple networks (e.g., interconnects) may be connected to a processing element, e.g., networks 1802, 1804, 1806, and 1810. The connections may be switches (e.g., multiplexers). In one embodiment, PEs and a circuit switched network 1810 are configured (e.g., control settings are selected) such that circuit switched network 1810 provides the paths for a desired operation.
A processing element (e.g., or in the network itself) may include a conditional queue (e.g., having only a single slot, of having multiple slots in each conditional queue) as discussed herein. In one embodiment, a single buffer (e.g., or queue) includes its own, respective conditional queue. In the depicted embodiment, conditional queue 1813 is included for control input buffer 1822C, conditional queue 1815 is included for first data input buffer 1824C, and conditional queue 1817 is included for second data input buffer 1826C of PE 1800C. In some embodiments, any conditional queue of a receiver PE (e.g. 1800C) can be used to as a part of the operations described herein. The coupling of conditional queues may also be time-multiplexed.
In certain embodiments, there are 3 basic types of entities that may be (e.g., input and/or output) operands to a CSA operation: (i) latency insensitive channels (LICs), (ii) registers, and (iii) literal values. In one embodiment, the size of literals is the size of the operand supported on PEs or other dataflow units, e.g. a 64 bit (64b) operand having a full 64b literal.
The format (e.g., signatures) of operations in the descriptions that follow use the following form: [{name}.]{operand type}{uld}.{data type}[={default value}]. The first part is an optional operand name (e.g., “res.” for a resultant or “ctlseq.” for a control sequence). Next is the operand type, where characters C (Channel), R (Register) or L (Literal) specify what operand types are valid. If there is a d suffix, the operand is an output that is defined, while a u suffix means it is an input that is used. Next is a data type, which reflects the usage in the operation.
For example, res.CRd.s32 means that the operand is called res, it can either a channel (C) or register (R), it is defined (d) by the operation (e.g., it is an output), and uses 32 bits of input, which it treats inside the operation as being signed. Note that this does not mean that input channels smaller than 32 bits are sign extended, although sign extension may be optionally included.
Operands may have default values, denoted by ={default value}, allowing various trailing operands to be omitted in assembly code. This is shown for a given operand description by an = with a default value. Value can be: (i) a numeric value, which is that value (e.g. op2.CRLu.i1=1 means a default value of 1), (ii) the letter I means % ign−ignored/reads as 0, (iii) the letter N means % na−never available, either as input or output (e.g., % na in a field means that field is not utilized for that operation), (iv) the letter R means rounding mode literal ROUND_NEAREST, and (v) the letter T means memory level literal MEMLEVEL_T0 (e.g., closest cache).
In the opcode description semantics, semicolons imply sequencing. If an operand appears by itself, the operation waits for the value to be available. e.g. for memrefs: op2; write(op0,op1); op3=0 means that the operation waits for op2 to be available, performs its access, and then defines op3. The following modifiers can appear for operands: non-consuming use (specified via a “*” prefix in the assembly code). This applies to any storage with empty/full semantics (e.g., LICs, and/or registers), and specifies that the operand is to be reused in the future.
In order to use these new multiplexed processing elements, the operating state architecture is extended to reflect the use of these facilities. In one embodiment, there are two key extensions. First, in this embodiment, each operation is extended with a new field that notes whether the operation is multiplexed or not. Additionally, in this embodiment, two new operations which permit transition between multiplexed and non-multiplexed graph portions are included: multiplex and demultiplex.
In one embodiment, a CSA architecture includes a configuration value that, when stored into the configuration storage (e.g., register), causes the CSA (e.g., a PE thereof) to perform a Multiplex operation according to the following (e.g., semantics and/or description).
In one embodiment, a CSA architecture includes a configuration value that, when stored into the configuration storage (e.g., register), causes the CSA (e.g., a PE thereof) to perform a Demultiplex operation according to the following (e.g., semantics and/or description).
In certain embodiments, an issue in providing virtual channels is that the amount of buffering per channel is reduced. Thus, if the buffer is reduced to one, synchronized time multiplexing will result in throughput loss due to the need to land new data from an upstream PE without having consumed previous data from a downstream PE. To fix this issue, certain embodiments herein use offset phased execution, in which PE execution and network transfer of a phase occur in back-to-back cycles as shown in
Depicted in-network storage element 1901 includes a buffer 1942. Buffer 1942 has multiple slots therein. In one embodiment to support time-multiplexing, instead of sharing multiple slots (e.g., as a first-in-first-out buffer), buffer 1942 includes a respective slot or slots for use with each particular phase (e.g., A or B, etc.) of a multi-phased execution in time-multiplexing. In certain embodiments, configuration storage 1956 (e.g., register) has a configuration value stored therein (e.g., during configuration time), and when the configuration value includes a first value, time-multiplexing (e.g., a slot is reserved only for data for a particular phase) is enabled for the in-network storage element 1901 and when the configuration value includes a second value, time-multiplexing (e.g., a slot is reserved only for data for a particular phase) is disable for the in-network storage element 1901. Additionally or alternatively, when the configuration value includes a third value, a data value from upstream path 1950 is stored into a slot of buffer 1942 (and/or sent from a slot of buffer 1942 to output path 1952, e.g., in a following cycle) and, when the configuration value includes a fourth second value, a data value from upstream path 1950 is steered downstream (e.g., with no delay greater than the delay caused by the physical path) via bypass path 1944 to downstream path 1952 (e.g., to a downstream PE (e.g., receiver PE)), for example, without being stored within (and/or modified by) in-network storage element 1901. In certain embodiments, controller 1940 controls the selection of (i) bypass mode that utilizes bypass path 1944 or (ii) buffer mode that utilizes buffer 1942 storage based on the configuration value stored in configuration storage 1956. In certain embodiments, a data value may include the data itself (e.g., payload) along with an “valid out” value that indicates the data itself is valid for a buffer to store. In certain embodiments, a “valid out” value is sent with the flow control values, e.g., from port 1908A of PE 1900A to port 1908B of PE 1900B or port 1908C of PE 1900C. As discussed further herein, ports (e.g., port 1908A of upstream (TX) PE 1900A to port 1908B of downstream (RX) PE 1900B or port 1908C of downstream (RX) PE 1900C.) may be used to send ready, valid in, and/or valid out values to PEs and/or in-network elements (e.g., in-network element 1901). In one embodiment, the path 1950 extending from buffer 1932A carries a “valid out” value and the data value itself (e.g., the path 1950 may include two, respective wires). In certain embodiments, port 1908A of PE 1900A, port 1908B of PE 1900B, and/or port 1908C of PE 1900C can send and receive data. For example, port 1908A of PE 1900A can (i) receive a flow control value(s) (e.g., a back-pressure value) from a downstream component (e.g., in-network storage element 1901, PE 1900B, or PE 1900C) and/or (ii) send a “valid out” value to a downstream component.
In certain embodiments, (i) when in bypass mode, a flow control value(s) (e.g., a back-pressure value) received on downstream path 1948 is steered upstream by controller 1940 to upstream path 1946 (e.g., to an upstream PE (e.g., transmitter PE)), for example, without being modified by in-network storage element 1901, when in bypass mode or (ii) when in buffer mode, controller 1940 is to send a flow control value to upstream path 1946 (e.g., to an upstream PE (e.g., transmitter PE)) based on the status of buffer 1942 (e.g., a “queue ready” value when storage space is available in buffer 1942).
As one example when in a particular phase, first (e.g., as producer) PE 1900A includes a (e.g., input) port 1908A(1-6) coupled to network 1910, e.g., to receive a backpressure value from second (e.g., as consumer) PE 1900B or third (e.g., as consumer) PE 1900C. In one circuit switched configuration, (e.g., input) port 1908A(1-6) (e.g., having a plurality of parallel inputs (1), (2), (3), (4), (5), and (6)) is to receive a respective backpressure value from each one of control input buffer 1922B, first data input buffer 1924B, and second data input buffer 1926B and/or control input buffer 1922C, first data input buffer 1924C, and second data input buffer 1926C.
In one embodiment (e.g., when in a particular phase), a circuit switched backpressure path (e.g., channel) is formed by setting switches coupled to wires between an input (e.g., input 1, 2, or 3) of port 1908A and an output (e.g., output 1, 2, or 3) of port 1908B to send a backpressure token (e.g., a value indicating no storage is available in an input buffer/queue) for one of control input buffer 1922B, first data input buffer 1924B, or second data input buffer 1926B of second PE 1900B. Additionally or alternatively, a (e.g., different) circuit switched backpressure path (e.g., channel) is formed by setting switches coupled to wires between an input (e.g., a different input of input 1, 2, or 3 (or one of more than 3 inputs in another embodiment) of port 1908A and an output (e.g., output 1, 2, or 3) of port 1908C to send a backpressure token (e.g., a value indicating no storage is available in an input buffer/queue) for one of control input buffer 1922C, first data input buffer 1924C, or second data input buffer 1926C of third PE 1900C.
In certain embodiments (e.g., when in a particular phase), output buffer 1932A of PE 1900A is coupled to in-network storage element, and, (i) when the configuration value in configuration storage 1956 is a certain value, a data value stored in output buffer 1932A of PE 1900A is sent through upstream path 1950 (e.g., upstream from in-network storage element 1901) and is steered downstream (e.g., with no delay greater than the delay caused by the physical path) via bypass path 1944 to downstream path 1952 (e.g., downstream from in-network storage element 1901) to an input buffer of a downstream PE (e.g., PE 1900B or PE 1900C), for example, without being stored within (and/or modified by) in-network storage element 1901, and (ii) when the configuration value in configuration storage 1956 is a different value, a (e.g., different) data value stored in buffer 1932A is sent through upstream path 1950 into a slot of buffer 1942 of in-network storage element 1901. In one embodiment, that (e.g., different) value stored in buffer 1942 is sent to downstream PE (e.g., PE 1900B or PE 1900C), e.g., when the input buffer of the downstream PE has an available storage space (e.g., slot).
Exemplary architectures, systems, etc. that the above may be used in are detailed herein. For example, an instruction, that when decoded and executed, may cause the performance of any of the methods disclosed herein.
At least some embodiments of the disclosed technologies can be described in view of the following examples:
In yet another embodiment, an apparatus comprises a data storage device that stores code that when executed by a hardware processor causes the hardware processor to perform any method disclosed herein. An apparatus may be as described in the detailed description. A method may be as described in the detailed description.
In certain embodiments, a significant source of area and energy reduction is the customization of the dataflow operations supported by each type of processing element. In one embodiment, a proper subset (e.g., most) processing elements support only a few operations (e.g., one, two, three, or four operation types), for example, an implementation choice where a floating point PE only supports one of floating point multiply or floating point add, but not both.
Operation configuration value may be a (e.g., unique) value, for example, according to the format discussed in section 3.5 below, e.g., for the operations discussed in section 3.6 below. In certain embodiments, operation configuration value includes a plurality of bits that cause processing element 2400 to perform a desired (e.g., preselected) operation, for example, performing the desired (e.g., preselected) operation when the incoming operands become available (e.g., in input storage 2424 and/or input storage 2426) and when there is space available to store the output (resultant) operand or operands (e.g., in output storage 2434 and/or output storage 2436). The depicted processing element 2400 includes two sets of operation circuitry 2425 and 2427, for example, to each perform a different operation. In certain embodiments, a PE includes status (e.g., state) storage, for example, within operation circuitry or a status register. Status storage may be modified during the operation in the the course of execution. Status storage may be shared among several operations. See, for example, the status register 1038 in
Depicted processing element 2400 includes an operation configuration storage 2419 (e.g., register(s)) to store an operation configuration value. In one embodiment, all of or a proper subset of a (e.g., single) operation configuration value is sent from the operation configuration storage 2419 (e.g., register(s)) to the multiplexers (e.g., multiplexer 2421 and multiplexer 2423) and/or demultiplexers (e.g., demultiplexer 2441 and demultiplexer 2443) of the processing element 2400 to steer the data according to the configuration.
Processing element 2400 includes a first input storage 2424 (e.g., input queue or buffer) coupled to (e.g., circuit switched) network 2402 and a second input storage 2426 (e.g., input queue or buffer) coupled to (e.g., circuit switched) network 2404. Network 2402 and network 2404 may be the same network (e.g., different circuit switched paths of the same network). Although two input storages are depicted, a single input storage or more than two input storages (e.g., any integer or proper subset of integers) may be utilized (e.g., with their own respective input controllers). Operation configuration value may be sent via the same network that the input storage 2424 and/or input storage 2426 are coupled to.
Depicted processing element 2400 includes input controller 2401, input controller 2403, output controller 2405, and output controller 2407 (e.g., together forming a scheduler for processing element 2400). Embodiments of input controllers are discussed in reference to
In
In certain embodiments, the input data (e.g., dataflow token or tokens) is sent to input storage 2424 and/or input storage 2426 by networks 2402 or networks 2402. In one embodiment, input data is stalled until there is available storage (e.g., in the targeted storage input storage 2424 or input storage 2426) in the storage that is to be utilized for that input data. In the depicted embodiment, operation configuration value (or a portion thereof) is sent to the multiplexers (e.g., multiplexer 2421 and multiplexer 2423) and/or demultiplexers (e.g., demultiplexer 2441 and demultiplexer 2443) of the processing element 2400 as control value(s) to steer the data according to the configuration. In certain embodiments, input operand selection switches 2421 and 2423 (e.g., multiplexers) allow data (e.g., dataflow tokens) from input storage 2424 and input storage 2426 as inputs to either of operation circuitry 2425 or operation circuitry 2427. In certain embodiments, result (e.g., output operand) selection switches 2437 and 2439 (e.g., multiplexers) allow data from either of operation circuitry 2425 or operation circuitry 2427 into output storage 2434 and/or output storage 2436. Storage may be a queue (e.g., first-in-first-out (FIFO) queue). In certain embodiments, an operation takes one input operand (e.g., from either of input storage 2424 and input storage 2426) and produce two resultants (e.g., stored in output storage 2434 and output storage 2436). In certain embodiments, an operation takes two or more input operands (for example, one from each input storage queue, e.g., one from each of input storage 2424 and input storage 2426) and produces a single (or plurality of) resultant (for example, stored in output storage, e.g., output storage 2434 and/or output storage 2436).
In certain embodiments, processing element 2400 is stalled from execution until there is input data (e.g., dataflow token or tokens) in input storage and there is storage space for the resultant data available in the output storage (e.g., as indicated by a backpressure value sent that indicates the output storage is not full). In the depicted embodiment, the input storage (queue) status value from path 2409 indicates (e.g., by asserting a “not empty” indication value or an “empty” indication value) when input storage 2424 contains (e.g., new) input data (e.g., dataflow token or tokens) and the input storage (queue) status value from path 2411 indicates (e.g., by asserting a “not empty” indication value or an “empty” indication value) when input storage 2426 contains (e.g., new) input data (e.g., dataflow token or tokens). In one embodiment, the input storage (queue) status value from path 2409 for input storage 2424 and the input storage (queue) status value from path 2411 for input storage 2426 is steered to the operation circuitry 2425 and/or operation circuitry 2427 (e.g., along with the input data from the input storage(s) that is to be operated on) by multiplexer 2421 and multiplexer 2423.
In the depicted embodiment, the output storage (queue) status value from path 2413 indicates (e.g., by asserting a “not full” indication value or a “full” indication value) when output storage 2434 has available storage for (e.g., new) output data (e.g., as indicated by a backpressure token or tokens) and the output storage (queue) status value from path 2415 indicates (e.g., by asserting a “not full” indication value or a “full” indication value) when output storage 2436 has available storage for (e.g., new) output data (e.g., as indicated by a backpressure token or tokens). In the depicted embodiment, operation configuration value (or a portion thereof) is sent to both multiplexer 2441 and multiplexer 2443 to source the output storage (queue) status value(s) from the output controllers 2405 and/or 2407. In certain embodiments, operation configuration value includes a bit or bits to cause a first output storage status value to be asserted, where the first output storage status value indicates the output storage (queue) is not full or a second, different output storage status value to be asserted, where the second output storage status value indicates the output storage (queue) is full. The first output storage status value (e.g., “not full”) or second output storage status value (e.g., “full”) may be output from output controller 2405 and/or output controller 2407, e.g., as discussed below. In one embodiment, a first output storage status value (e.g., “not full”) is sent to the operation circuitry 2425 and/or operation circuitry 2427 to cause the operation circuitry 2425 and/or operation circuitry 2427, respectively, to perform the programmed operation when an input value is available in input storage (queue) and a second output storage status value (e.g., “full”) is sent to the operation circuitry 2425 and/or operation circuitry 2427 to cause the operation circuitry 2425 and/or operation circuitry 2427, respectively, to not perform the programmed operation even when an input value is available in input storage (queue).
In the depicted embodiment, dequeue (e.g., conditional dequeue) multiplexers 2429 and 2431 are included to cause a dequeue (e.g., removal) of a value (e.g., token) from a respective input storage (queue), e.g., based on operation completion by operation circuitry 2425 and/or operation circuitry 2427. The operation configuration value includes a bit or bits to cause the dequeue (e.g., conditional dequeue) multiplexers 2429 and 2431 to dequeue (e.g., remove) a value (e.g., token) from a respective input storage (queue). In the depicted embodiment, enqueue (e.g., conditional enqueue) multiplexers 2433 and 2435 are included to cause an enqueue (e.g., insertion) of a value (e.g., token) into a respective output storage (queue), e.g., based on operation completion by operation circuitry 2425 and/or operation circuitry 2427. The operation configuration value includes a bit or bits to cause the enqueue (e.g., conditional enqueue) multiplexers 2433 and 2435 to enqueue (e.g., insert) a value (e.g., token) into a respective output storage (queue).
Certain operations herein allow the manipulation of the control values sent to these queues, e.g., based on local values computed and/or stored in the PE.
In one embodiment, the dequeue multiplexers 2429 and 2431 are conditional dequeue multiplexers 2429 and 2431 that, when a programmed operation is performed, the consumption (e.g., dequeuing) of the input value from the input storage (queue) is conditionally performed. In one embodiment, the enqueue multiplexers 2433 and 2435 are conditional enqueue multiplexers 2433 and 2435 that, when a programmed operation is performed, the storing (e.g., enqueuing) of the output value for the programmed operation into the output storage (queue) is conditionally performed.
For example, as discussed herein, certain operations may make dequeuing (e.g., consumption) decisions for an input storage (queue) conditionally (e.g., based on token values) and/or enqueuing (e.g., output) decisions for an output storage (queue) conditionally (e.g., based on token values). An example of a conditional enqueue operation is a PredMerge operation that conditionally writes its outputs, so conditional enqueue multiplexer(s) will be swung, e.g., to store or not store the predmerge result into the appropriate output queue. An example of a conditional dequeue operation is a PredProp operation that conditionally reads its inputs, so conditional dequeue multiplexer(s) will be swung, e.g., to store or not store the predprop result into the appropriate input queue.
In certain embodiments, control input value (e.g., bit or bits) (e.g., a control token) is input into a respective, input storage (e.g., queue), for example, into a control input buffer as discussed herein (e.g., control input buffer 1022 in
Depicted input controller circuitry 2500 includes a Status determiner 2504, a Not Full determiner 2506, and a Not Empty determiner 2508. A determiner may be implemented in software or hardware. A hardware determiner may be a circuit implementation, for example, a logic circuit programmed to produce an output based on the inputs into the state machine(s) discussed below. Depicted (e.g., new) Status determiner 2504 includes a port coupled to queue status register 2502 to read and/or write to input queue status register 2502.
Depicted Status determiner 2504 includes a first input to receive a Valid value (e.g., a value indicating valid) from a transmitting component (e.g., an upstream PE) that indicates if (e.g., when) there is data (valid data) to be sent to the PE that includes input controller circuitry 2500. The Valid value may be referred to as a dataflow token. Depicted Status determiner 2504 includes a second input to receive a value or values from queue status register 2502 that represents that current status of the input queue that input controller circuitry 2500 is controlling. Optionally, Status determiner 2504 includes a third input to receive a value (from within the PE that includes input controller circuitry 2500) that indicates if (when) there is a conditional dequeue, e.g., from operation circuitry 2425 and/or operation circuitry 2427 in
As discussed further below, the depicted Status determiner 2504 includes a first output to send a value on path 2510 that will cause input data (transmitted to the input queue that input controller circuitry 2500 is controlling) to be enqueued into the input queue or not enqueued into the input queue. Depicted Status determiner 2504 includes a second output to send an updated value to be stored in queue status register 2502, e.g., where the updated value represents the updated status (e.g., head value, tail value, count value, or any combination thereof) of the input queue that input controller circuitry 2500 is controlling.
Input controller circuitry 2500 includes a Not Full determiner 2506 that determines a Not Full (e.g., Ready) value and outputs the Not Full value to a transmitting component (e.g., an upstream PE) to indicate if (e.g., when) there is storage space available for input data in the input queue being controlled by input controller circuitry 2500. The Not Full (e.g., Ready) value may be referred to as a backpressure token, e.g., a backpressure token from a receiving PE sent to a transmitting PE.
Input controller circuitry 2500 includes a Not Empty determiner 2508 that determines an input storage (queue) status value and outputs (e.g., on path 2409 or path 2411 in
For example, assume that the operation that is to be performed is to source data from both input storage 2424 and input storage 2426 in
Queue status register 2602 may store any combination of a head value (e.g., pointer) that represents the head (beginning) of the data stored in the queue, a tail value (e.g., pointer) that represents the tail (ending) of the data stored in the queue, and a count value that represents the number of (e.g., valid) values stored in the queue). For example, a count value may be an integer (e.g., two) where the queue is storing the number of values indicated by the integer (e.g., storing two values in the queue). The capacity of data (e.g., storage slots for data, e.g., for data elements) in a queue may be preselected (e.g., during programming), for example, depending on the total bit capacity of the queue and the number of bits in each element. Queue status register 2602 may be updated with the initial values, e.g., during configuration time. Queue status register 2602 may be updated as discussed in reference to
Depicted output controller circuitry 3500 includes a Status determiner 3504, a Not Full determiner 3506, and a Not Empty determiner 3508. A determiner may be implemented in software or hardware. A hardware determiner may be a circuit implementation, for example, a logic circuit programmed to produce an output based on the inputs into the state machine(s) discussed below. Depicted (e.g., new) Status determiner 3504 includes a port coupled to queue status register 3502 to read and/or write to output queue status register 3502.
Depicted Status determiner 3504 includes a first input to receive a Ready value from a receiving component (e.g., a downstream PE) that indicates if (e.g., when) there is space (e.g., in an input queue thereof) for new data to be sent to the PE. In certain embodiments, the Ready value from the receiving component is sent by an input controller that includes input controller circuitry 2500 in
As discussed further below, the depicted Status determiner 3504 includes a first output to send a value on path 3510 that will cause output data (sent to the output queue that output controller circuitry 3500 is controlling) to be enqueued into the output queue or not enqueued into the output queue. Depicted Status determiner 3504 includes a second output to send an updated value to be stored in queue status register 3502, e.g., where the updated value represents the updated status (e.g., head value, tail value, count value, or any combination thereof) of the output queue that output controller circuitry 3500 is controlling.
Output controller circuitry 3500 includes a Not Full determiner 3506 that determines a Not Full (e.g., Ready) value and outputs the Not Full value, e.g., within the PE that includes output controller circuitry 3500, to indicate if (e.g., when) there is storage space available for output data in the output queue being controlled by output controller circuitry 3500. In one embodiment, for an output queue of a PE, a Not Full value that indicates there is no storage space available in that output queue is to cause a stall of execution of the PE (e.g., stall execution that is to cause a resultant to be stored into the storage space) until storage space is available (e.g., and when there is available data in the input queue(s) being sourced from in that PE).
Output controller circuitry 3500 includes a Not Empty determiner 3508 that determines an output storage (queue) status value and outputs (e.g., on path 2445 or path 2447 in
For example, assume that the operation that is to be performed is to send (e.g., sink) data into both output storage 2434 and output storage 2436 in
Queue status register 3602 may store any combination of a head value (e.g., pointer) that represents the head (beginning) of the data stored in the queue, a tail value (e.g., pointer) that represents the tail (ending) of the data stored in the queue, and a count value that represents the number of (e.g., valid) values stored in the queue). For example, a count value may be an integer (e.g., two) where the queue is storing the number of values indicated by the integer (e.g., storing two values in the queue). The capacity of data (e.g., storage slots for data, e.g., for data elements) in a queue may be preselected (e.g., during programming), for example, depending on the total bit capacity of the queue and the number of bits in each element. Queue status register 3602 may be updated with the initial values, e.g., during configuration time. Queue status register 3602 may be updated as discussed in reference to
In certain embodiments, a state machine includes a plurality of single bit width input values (e.g., 0s or 1s), and produces a single output value that has a single bit width (e.g., a 0 or a 1).
In certain embodiments, a first LIC channel may be formed between an output of a first PE to an input of a second PE, and a second LIC channel may be formed between an output of the second PE and an input of a third PE. As an example, a ready value may be sent on a first path of a LIC channel by a receiving PE to a transmitting PE and a valid value may be sent on a second path of the LIC channel by the transmitting PE to the receiving PE. As an example, see
In certain embodiments, processing elements (PEs) communicate using dedicated virtual circuits which are formed by statically configuring a (e.g., circuit switched) communications network. These virtual circuits may be flow controlled and fully back-pressured, e.g., such that a PE will stall if either the source has no data or its destination is full. At runtime, data may flow through the PEs implementing the mapped dataflow graph (e.g., mapped algorithm). For example, data may be streamed in from memory, through the (e.g., fabric area of a) spatial array of processing elements, and then back out to memory.
Such an architecture may achieve remarkable performance efficiency relative to traditional multicore processors: compute, e.g., in the form of PEs, may be simpler and more numerous than cores and communications may be direct, e.g., as opposed to an extension of the memory system. However, the (e.g., fabric area of) spatial array of processing elements may be tuned for the implementation of compiler-generated expression trees, which may feature little multiplexing or demultiplexing. Certain embodiments herein extend (for example, via network resources, such as, but not limited to, network dataflow endpoint circuits) the architecture to support (e.g., high-radix) multiplexing and/or demultiplexing, for example, especially in the context of function calls.
Spatial arrays, such as the spatial array of processing elements 101 in
In one embodiment, a circuit switched network between two points (e.g., between a producer and consumer of data) includes a dedicated communication line between those two points, for example, with (e.g., physical) switches between the two points set to create a (e.g., exclusive) physical circuit between the two points. In one embodiment, a circuit switched network between two points is set up at the beginning of use of the connection between the two points and maintained throughout the use of the connection. In another embodiment, a packet switched network includes a shared communication line (e.g., channel) between two (e.g., or more) points, for example, where packets from different connections share that communication line (for example, routed according to data of each packet, e.g., in the header of a packet including a header and a payload). An example of a packet switched network is discussed below, e.g., in reference to a mezzanine network.
Operations may be executed based on the availability of their inputs and the status of the PE. A PE may obtain operands from input channels and write results to output channels, although internal register state may also be used. Certain embodiments herein include a configurable dataflow-friendly PE.
Instruction registers may be set during a special configuration step. During this step, auxiliary control wires and state, in addition to the inter-PE network, may be used to stream in configuration across the several PEs comprising the fabric. As result of parallelism, certain embodiments of such a network may provide for rapid reconfiguration, e.g., a tile sized fabric may be configured in less than about 10 microseconds.
Further, depicted accelerator tile 4600 includes packet switched communications network 4614, for example, as part of a mezzanine network, e.g., as described below. Certain embodiments herein allow for (e.g., a distributed) dataflow operations (e.g., operations that only route data) to be performed on (e.g., within) the communications network (e.g., and not in the processing element(s)). As an example, a distributed Pick dataflow operation of a dataflow graph is depicted in
As one example, a pick dataflow operation may have a plurality of inputs and steer (e.g., route) one of them as an output, e.g., as in
In the depicted embodiment, packet switched communications network 4614 may handle certain (e.g., configuration) communications, for example, to program the processing elements and/or circuit switched network (e.g., network 4613, which may include switches). In one embodiment, a circuit switched network is configured (e.g., programmed) to perform one or more operations (e.g., dataflow operations of a dataflow graph).
Packet switched communications network 4614 includes a plurality of endpoints (e.g., network dataflow endpoint circuits (4602, 4604, 4606). In one embodiment, each endpoint includes an address or other indicator value to allow data to be routed to and/or from that endpoint, e.g., according to (e.g., a header of) a data packet.
Additionally or alternatively to performing one or more of the above, packet switched communications network 4614 may perform dataflow operations. Network dataflow endpoint circuits (4602, 4604, 4606) may be configured (e.g., programmed) to perform a (e.g., distributed pick) operation of a dataflow graph. Programming of components (e.g., a circuit) are described herein. An embodiment of configuring a network dataflow endpoint circuit (e.g., an operation configuration register thereof) is discussed in reference to
As an example of a distributed pick dataflow operation, network dataflow endpoint circuits (4602, 4604, 4606) in
Network dataflow endpoint circuit 4602 may be configured to receive input data from a plurality of sources (e.g., network dataflow endpoint circuit 4604 and network dataflow endpoint circuit 4606), and to output resultant data, e.g., as in
When network dataflow endpoint circuit 4604 is to transmit input data to network dataflow endpoint circuit 4602 (e.g., when network dataflow endpoint circuit 4602 has available storage room for the data and/or network dataflow endpoint circuit 4604 has its input data), network dataflow endpoint circuit 4604 may generate a packet (e.g., including the input data and a header to steer that data to network dataflow endpoint circuit 4602 on the packet switched communications network 4614 (e.g., as a stop on that (e.g., ring) network 4614). This is illustrated schematically with dashed line 4626 in
When network dataflow endpoint circuit 4606 is to transmit input data to network dataflow endpoint circuit 4602 (e.g., when network dataflow endpoint circuit 4602 has available storage room for the data and/or network dataflow endpoint circuit 4606 has its input data), network dataflow endpoint circuit 4604 may generate a packet (e.g., including the input data and a header to steer that data to network dataflow endpoint circuit 4602 on the packet switched communications network 4614 (e.g., as a stop on that (e.g., ring) network 4614). This is illustrated schematically with dashed line 4618 in
Network dataflow endpoint circuit 4602 (e.g., on receipt of the Input 0 from network dataflow endpoint circuit 4604, Input 1 from network dataflow endpoint circuit 4606, and/or control data) may then perform the programmed dataflow operation (e.g., a Pick operation in this example). The network dataflow endpoint circuit 4602 may then output the according resultant data from the operation, e.g., to processing element 4608 in
In one embodiment, the control data to perform an operation (e.g., pick operation) comes from other components of the spatial array, e.g., a processing element or through network. An example of this is discussed below in reference to
In certain embodiments, a dataflow graph may have certain operations performed by a processing element and certain operations performed by a communication network (e.g., network dataflow endpoint circuit or circuits).
As one description of an embodiment of the microarchitecture, a pick dataflow operator may function to pick one output of resultant data from a plurality of inputs of input data, e.g., based on control data. A network dataflow endpoint circuit 4700 may be configured to consider one of the spatial array ingress buffer(s) 4702 of the circuit 4700 (e.g., data from the fabric being control data) as selecting among multiple input data elements stored in network ingress buffer(s) 4724 of the circuit 4700 to steer the resultant data to the spatial array egress buffer 4708 of the circuit 4700. Thus, the network ingress buffer(s) 4724 may be thought of as inputs to a virtual mux, the spatial array ingress buffer 4702 as the multiplexer select, and the spatial array egress buffer 4708 as the multiplexer output. In one embodiment, when a (e.g., control data) value is detected and/or arrives in the spatial array ingress buffer 4702, the scheduler 4728 (e.g., as programmed by an operation configuration in storage 4726) is sensitized to examine the corresponding network ingress channel. When data is available in that channel, it is removed from the network ingress buffer 4724 and moved to the spatial array egress buffer 4708. The control bits of both ingresses and egress may then be updated to reflect the transfer of data. This may result in control flow tokens or credits being propagated in the associated network. In certain embodiment, all inputs (e.g., control or data) may arise locally or over the network.
Initially, it may seem that the use of packet switched networks to implement the (e.g., high-radix staging) operators of multiplexed and/or demultiplexed codes hampers performance. For example, in one embodiment, a packet-switched network is generally shared and the caller and callee dataflow graphs may be distant from one another. Recall, however, that in certain embodiments, the intention of supporting multiplexing and/or demultiplexing is to reduce the area consumed by infrequent code paths within a dataflow operator (e.g., by the spatial array). Thus, certain embodiments herein reduce area and avoid the consumption of more expensive fabric resources, for example, like PEs, e.g., without (substantially) affecting the area and efficiency of individual PEs to supporting those (e.g., infrequent) operations.
Turning now to further detail of
Depicted network dataflow endpoint circuit 4700 includes a spatial array (e.g., fabric) egress buffer 4708, for example, to output data (e.g., control data) to a (e.g., circuit switched) network. As noted above, although a single spatial array (e.g., fabric) egress buffer 4708 is depicted, a plurality of spatial array (e.g., fabric) egress buffers may be in a network dataflow endpoint circuit. In one embodiment, spatial array (e.g., fabric) egress buffer 4708 is to send (e.g., transmit) data (e.g., control data) onto a communications network of a spatial array (e.g., a spatial array of processing elements), for example, onto one or more of network 4710 and network 4712. In one embodiment, network 4710 is part of network 4613 in
Additionally or alternatively, network dataflow endpoint circuit 4700 may be coupled to another network 4714, e.g., a packet switched network. Another network 4714, e.g., a packet switched network, may be used to transmit (e.g., send or receive) (e.g., input and/or resultant) data to processing elements or other components of a spatial array and/or to transmit one or more of input data or resultant data. In one embodiment, network 4714 is part of the packet switched communications network 4614 in
Network buffer 4718 (e.g., register(s)) may be a stop on (e.g., ring) network 4714, for example, to receive data from network 4714.
Depicted network dataflow endpoint circuit 4700 includes a network egress buffer 4722, for example, to output data (e.g., resultant data) to a (e.g., packet switched) network. As noted above, although a single network egress buffer 4722 is depicted, a plurality of network egress buffers may be in a network dataflow endpoint circuit. In one embodiment, network egress buffer 4722 is to send (e.g., transmit) data (e.g., resultant data) onto a communications network of a spatial array (e.g., a spatial array of processing elements), for example, onto network 4714. In one embodiment, network 4714 is part of packet switched network 4614 in
Depicted network dataflow endpoint circuit 4700 includes a network ingress buffer 4722, for example, to input data (e.g., inputted data) from a (e.g., packet switched) network. As noted above, although a single network ingress buffer 4724 is depicted, a plurality of network ingress buffers may be in a network dataflow endpoint circuit. In one embodiment, network ingress buffer 4724 is to receive (e.g., transmit) data (e.g., input data) from a communications network of a spatial array (e.g., a spatial array of processing elements), for example, from network 4714. In one embodiment, network 4714 is part of packet switched network 4614 in
In one embodiment, the data format (e.g., of the data on network 4714) includes a packet having data and a header (e.g., with the destination of that data). In one embodiment, the data format (e.g., of the data on network 4704 and/or 4706) includes only the data (e.g., not a packet having data and a header (e.g., with the destination of that data)). Network dataflow endpoint circuit 4700 may add (e.g., data output from circuit 4700) or remove (e.g., data input into circuit 4700) a header (or other data) to or from a packet. Coupling 4720 (e.g., wire) may send data received from network 4714 (e.g., from network buffer 4718) to network ingress buffer 4724 and/or multiplexer 4716. Multiplexer 4716 may (e.g., via a control signal from the scheduler 4728) output data from network buffer 4718 or from network egress buffer 4722. In one embodiment, one or more of multiplexer 4716 or network buffer 4718 are separate components from network dataflow endpoint circuit 4700. A buffer may include a plurality of (e.g., discrete) entries, for example, a plurality of registers.
In one embodiment, operation configuration storage 4726 (e.g., register or registers) is loaded during configuration (e.g., mapping) and specifies the particular operation (or operations) this network dataflow endpoint circuit 4700 (e.g., not a processing element of a spatial array) is to perform (e.g., data steering operations in contrast to logic and/or arithmetic operations). Buffer(s) (e.g., 4702, 4708, 4722, and/or 4724) activity may be controlled by that operation (e.g., controlled by the scheduler 4728). Scheduler 4728 may schedule an operation or operations of network dataflow endpoint circuit 4700, for example, when (e.g., all) input (e.g., payload) data and/or control data arrives. Dotted lines to and from scheduler 4728 indicate paths that may be utilized for control data, e.g., to and/or from scheduler 4728. Scheduler may also control multiplexer 4716, e.g., to steer data to and/or from network dataflow endpoint circuit 4700 and network 4714.
In reference to the distributed pick operation in
When network dataflow endpoint circuit 4604 is to transmit input data to network dataflow endpoint circuit 4602 (e.g., when network dataflow endpoint circuit 4602 has available storage room for the data and/or network dataflow endpoint circuit 4604 has its input data), network dataflow endpoint circuit 4604 may generate a packet (e.g., including the input data and a header to steer that data to network dataflow endpoint circuit 4602 on the packet switched communications network 4614 (e.g., as a stop on that (e.g., ring) network). This is illustrated schematically with dashed line 4626 in
When network dataflow endpoint circuit 4606 is to transmit input data to network dataflow endpoint circuit 4602 (e.g., when network dataflow endpoint circuit 4602 has available storage room for the data and/or network dataflow endpoint circuit 4606 has its input data), network dataflow endpoint circuit 4604 may generate a packet (e.g., including the input data and a header to steer that data to network dataflow endpoint circuit 4602 on the packet switched communications network 4614 (e.g., as a stop on that (e.g., ring) network). This is illustrated schematically with dashed line 4618 in
Network dataflow endpoint circuit 4602 (e.g., on receipt of the Input 0 from network dataflow endpoint circuit 4604 in circuit 4602's network ingress buffer(s), Input 1 from network dataflow endpoint circuit 4606 in circuit 4602's network ingress buffer(s), and/or control data from processing element 4608 in circuit 4602's spatial array ingress buffer) may then perform the programmed dataflow operation (e.g., a Pick operation in this example). The network dataflow endpoint circuit 4602 may then output the according resultant data from the operation, e.g., to processing element 4608 in
Depicted receive operation configuration data format 5004 includes an output field 5004A (e.g., indicating which component(s) in a network the (resultant) data is to be sent to), an input field 5004B (e.g., an identifier of the component(s) that is to send the input data), and an operation field 5004C (e.g., indicating which of a plurality of operations are to be performed). In one embodiment, the (e.g., inbound) operation is one of a Pick, PickSingleLeg, PickAny, or Merge dataflow operation, e.g., corresponding to a (e.g., same) dataflow operator of a dataflow graph. In one embodiment, a merge dataflow operation is a pick that requires and dequeues all operands (e.g., with the egress endpoint receiving control).
A configuration data format utilized herein may include one or more of the fields described herein, e.g., in any order.
In one embodiment, circuit 5200 (e.g., network dataflow endpoint circuit) is to receive packet of data in the data format of (e.g., send) operation configuration data format 5202, for example, with the input being the source(s) of the payload (e.g., input data) and the operation field indicating which operation is to be performed (e.g., shown schematically as Switch or SwitchAny). Depicted multiplexer 5204 may select the operation to be performed from a plurality of available operations, e.g., based on the value in operation field 5202D. In one embodiment, circuit 5200 is to perform that operation when both the input data is available and the credit status is a yes (for example, the dependency token indicates) indicating there is room for the output data to be stored, e.g., in a buffer of the destination.
In one embodiment, the send operation does not utilize control beyond checking its input(s) are available for sending. This may enable switch to perform the operation without credit on all legs. In one embodiment, the Switch and/or SwitchAny operation includes a multiplexer controlled by the value stored in the operation field 5202D to select the correct queue management circuitry.
Value stored in operation field 5202D may select among control options, e.g., with different control (e.g., logic) circuitry for each operation, for example, as in
In one embodiment, PickAny executes on the presence of any data and/or selection decoder creates multiplexer selection bits.
In one embodiment, (e.g., as with scheduling) the choice of dequeue is determined by the operation and its dynamic behavior, e.g., to dequeue the operation after performance. In one embodiment, a circuit is to use the operand selection bits to dequeue data (e.g., input, output and/or control data).
Network 5814 may be a circuit switched network, e.g., as discussed herein. Additionally or alternatively, a packet switched network (e.g., as discussed herein) may also be utilized, for example, coupled to network egress buffer 5822, network ingress buffer 5824, or other components herein. Argument queue 5802 may include a control buffer 5802A, for example, to indicate when a respective input queue (e.g., buffer) includes a (new) item of data, e.g., as a single bit. Turning now to
Referring again to
In certain embodiments, an accelerator (e.g., a PE thereof) couples to a RAF circuit or a plurality of RAF circuits through (i) a circuit switched network (for example, as discussed herein, e.g., in reference to
In certain embodiments, a circuit (e.g., a request address file (RAF) circuit) (e.g., each of a plurality of RAF circuits) includes a translation lookaside buffer (TLB) (e.g., TLB circuit). TLB may receive an input of a virtual address and output a physical address corresponding to the mapping (e.g., address mapping) of the virtual address to the physical address (e.g., different than any mapping of a dataflow graph to hardware). A virtual address may be an address as seen by a program running on circuitry (e.g., on an accelerator and/or processor). A physical address may be an (e.g., different than the virtual) address in memory hardware. A TLB may include a data structure (e.g., table) to store (e.g., recently used) virtual-to-physical memory address translations, e.g., such that the translation does not have to be performed on each virtual address present to obtain the physical memory address corresponding to that virtual address. If the virtual address entry is not in the TLB, a circuit (e.g., a TLB manager circuit) may perform a page walk to determine the virtual-to-physical memory address translation. In one embodiment, a circuit (e.g., a RAF circuit) is to receive an input of a virtual address for translation in a TLB (e.g., TLB in RAF circuit) from a requesting entity (e.g., a PE or other hardware component) via a circuit switched network, e.g., as in
Certain HPC applications are characterized by their need for significant floating point bandwidth. To meet this need, embodiments of a CSA may be provisioned with multiple (e.g., between 128 and 256 each) of floating add and multiplication PEs, e.g., depending on tile configuration. A CSA may provide a few other extended precision modes, e.g., to simplify math library implementation. CSA floating point PEs may support both single and double precision, but lower precision PEs may support machine learning workloads. A CSA may provide an order of magnitude more floating point performance than a processor core. In one embodiment, in addition to increasing floating point bandwidth, in order to power all of the floating point units, the energy consumed in floating point operations is reduced. For example, to reduce energy, a CSA may selectively gate the low-order bits of the floating point multiplier array. In examining the behavior of floating point arithmetic, the low order bits of the multiplication array may often not influence the final, rounded product.
Given this maximum carry, if the result of the carry region is less than 2c-g, where the carry region is c bits wide, then the gated region may be ignored since it does not influence the result region. Increasing g means that it is more likely the gated region will be needed, while increasing c means that, under random assumption, the gated region will be unused and may be disabled to avoid energy consumption. In embodiments of a CSA floating multiplication PE, a two stage pipelined approach is utilized in which first the carry region is determined and then the gated region is determined if it is found to influence the result. If more information about the context of the multiplication is known, a CSA more aggressively tune the size of the gated region. In FMA, the multiplication result may be added to an accumulator, which is often much larger than either of the multiplicands. In this case, the addend exponent may be observed in advance of multiplication and the CSDA may adjust the gated region accordingly. One embodiment of the CSA includes a scheme in which a context value, which bounds the minimum result of a computation, is provided to related multipliers, in order to select minimum energy gating configurations.
In certain embodiment, a CSA includes a heterogeneous and distributed fabric, and consequently, runtime service implementations are to accommodate several kinds of PEs in a parallel and distributed fashion. Although runtime services in a CSA may be critical, they may be infrequent relative to user-level computation. Certain implementations, therefore, focus on overlaying services on hardware resources. To meet these goals, CSA runtime services may be cast as a hierarchy, e.g., with each layer corresponding to a CSA network. At the tile level, a single external-facing controller may accepts or sends service commands to an associated core with the CSA tile. A tile-level controller may serve to coordinate regional controllers at the RAFs, e.g., using the ACI network. In turn, regional controllers may coordinate local controllers at certain mezzanine network stops (e.g., network dataflow endpoint circuits). At the lowest level, service specific micro-protocols may execute over the local network, e.g., during a special mode controlled through the mezzanine controllers. The micro-protocols may permit each PE (e.g., PE class by type) to interact with the runtime service according to its own needs. Parallelism is thus implicit in this hierarchical organization, and operations at the lowest levels may occur simultaneously. This parallelism may enables the configuration of a CSA tile in between hundreds of nanoseconds to a few microseconds, e.g., depending on the configuration size and its location in the memory hierarchy. Embodiments of the CSA thus leverage properties of dataflow graphs to improve implementation of each runtime service. One key observation is that runtime services may need only to preserve a legal logical view of the dataflow graph, e.g., a state that can be produced through some ordering of dataflow operator executions. Services may generally not need to guarantee a temporal view of the dataflow graph, e.g., the state of a dataflow graph in a CSA at a specific point in time. This may permit the CSA to conduct most runtime services in a distributed, pipelined, and parallel fashion, e.g., provided that the service is orchestrated to preserve the logical view of the dataflow graph. The local configuration micro-protocol may be a packet-based protocol overlaid on the local network. Configuration targets may be organized into a configuration chain, e.g., which is fixed in the microarchitecture. Fabric (e.g., PE) targets may be configured one at a time, e.g., using a single extra register per target to achieve distributed coordination. To start configuration, a controller may drive an out-of-band signal which places all fabric targets in its neighborhood into an unconfigured, paused state and swings multiplexors in the local network to a pre-defined conformation. As the fabric (e.g., PE) targets are configured, that is they completely receive their configuration packet, they may set their configuration microprotocol registers, notifying the immediately succeeding target (e.g., PE) that it may proceed to configure using the subsequent packet. There is no limitation to the size of a configuration packet, and packets may have dynamically variable length. For example, PEs configuring constant operands may have a configuration packet that is lengthened to include the constant field (e.g., X and Y in
The ability to compile programs written in high-level languages onto a CSA may be essential for industry adoption. This section gives a high-level overview of compilation strategies for embodiments of a CSA. First is a proposal for a CSA software framework that illustrates the desired properties of an ideal production-quality toolchain. Next, a prototype compiler framework is discussed. A “control-to-dataflow conversion” is then discussed, e.g., to converts ordinary sequential control-flow code into CSA dataflow assembly code.
A key portion of the compiler may be implemented in the control-to-dataflow conversion pass, or dataflow conversion pass for short. This pass takes in a function represented in control flow form, e.g., a control-flow graph (CFG) with sequential machine instructions operating on virtual registers, and converts it into a dataflow function that is conceptually a graph of dataflow operations (instructions) connected by latency-insensitive channels (LICs). This section gives a high-level description of this pass, describing how it conceptually deals with memory operations, branches, and loops in certain embodiments.
First, consider the simple case of converting straight-line sequential code to dataflow. The dataflow conversion pass may convert a basic block of sequential code, such as the code shown in
To convert programs with multiple basic blocks and conditionals to dataflow, the compiler generates special dataflow operators to replace the branches. More specifically, the compiler uses switch operators to steer outgoing data at the end of a basic block in the original CFG, and pick operators to select values from the appropriate incoming channel at the beginning of a basic block. As a concrete example, consider the code and corresponding dataflow graph in
Control Equivalence:
Consider a single-entry-single-exit control flow graph G with two basic blocks A and B. A and B are control-equivalent if all complete control flow paths through G visit A and B the same number of times.
LIC Replacement:
In a control flow graph G, suppose an operation in basic block A defines a virtual register x, and an operation in basic block B that uses x. Then a correct control-to-dataflow transformation can replace x with a latency-insensitive channel only if A and B are control equivalent. The control-equivalence relation partitions the basic blocks of a CFG into strong control-dependence regions.
Another important class of CFGs in dataflow conversion are CFGs for single-entry-single-exit loops, a common form of loop generated in (LLVM) IR. These loops may be almost acyclic, except for a single back edge from the end of the loop back to a loop header block. The dataflow conversion pass may use same high-level strategy to convert loops as for branches, e.g., it inserts switches at the end of the loop to direct values out of the loop (either out the loop exit or around the back-edge to the beginning of the loop), and inserts picks at the beginning of the loop to choose between initial values entering the loop and values coming through the back edge.
In one embodiment, the core writes a command into a memory queue and a CSA (e.g., the plurality of processing elements) monitors the memory queue and begins executing when the command is read. In one embodiment, the core executes a first part of a program and a CSA (e.g., the plurality of processing elements) executes a second part of the program. In one embodiment, the core does other work while the CSA is executing its operations.
In certain embodiments, the CSA architecture and microarchitecture provides profound energy, performance, and usability advantages over roadmap processor architectures and FPGAs. In this section, these architectures are compared to embodiments of the CSA and highlights the superiority of CSA in accelerating parallel dataflow graphs relative to each.
The choice of dataflow operators as the fundamental architecture of embodiments of a CSA differentiates those CSAs from a FGPA, and particularly the CSA is as superior accelerator for HPC dataflow graphs arising from traditional programming languages. Dataflow operators are fundamentally asynchronous. This enables embodiments of a CSA not only to have great freedom of implementation in the microarchitecture, but it also enables them to simply and succinctly accommodate abstract architectural concepts. For example, embodiments of a CSA naturally accommodate many memory microarchitectures, which are essentially asynchronous, with a simple load-store interface. One need only examine an FPGA DRAM controller to appreciate the difference in complexity. Embodiments of a CSA also leverage asynchrony to provide faster and more-fully-featured runtime services like configuration and extraction, which are believed to be four to six orders of magnitude faster than an FPGA. By narrowing the architectural interface, embodiments of a CSA provide control over most timing paths at the microarchitectural level. This allows embodiments of a CSA to operate at a much higher frequency than the more general control mechanism offered in a FPGA. Similarly, clock and reset, which may be architecturally fundamental to FPGAs, are microarchitectural in the CSA, e.g., obviating the need to support them as programmable entities. Dataflow operators may be, for the most part, coarse-grained. By only dealing in coarse operators, embodiments of a CSA improve both the density of the fabric and its energy consumption: CSA executes operations directly rather than emulating them with look-up tables. A second consequence of coarseness is a simplification of the place and route problem. CSA dataflow graphs are many orders of magnitude smaller than FPGA net-lists and place and route time are commensurately reduced in embodiments of a CSA. The significant differences between embodiments of a CSA and a FPGA make the CSA superior as an accelerator, e.g., for dataflow graphs arising from traditional programming languages.
The CSA is a novel computer architecture with the potential to provide enormous performance and energy advantages relative to roadmap processors. Consider the case of computing a single strided address for walking across an array. This case may be important in HPC applications, e.g., which spend significant integer effort in computing address offsets. In address computation, and especially strided address computation, one argument is constant and the other varies only slightly per computation. Thus, only a handful of bits per cycle toggle in the majority of cases. Indeed, it may be shown, using a derivation similar to the bound on floating point carry bits described in Section 2.6, that less than two bits of input toggle per computation in average for a stride calculation, reducing energy by 50% over a random toggle distribution. Were a time-multiplexed approach used, much of this energy savings may be lost. In one embodiment, the CSA achieves approximately 3× energy efficiency over a core while delivering an 8× performance gain. The parallelism gains achieved by embodiments of a CSA may result in reduced program run times, yielding a proportionate, substantial reduction in leakage energy. At the PE level, embodiments of a CSA are extremely energy efficient. A second important question for the CSA is whether the CSA consumes a reasonable amount of energy at the tile level. Since embodiments of a CSA are capable of exercising every floating point PE in the fabric at every cycle, it serves as a reasonable upper bound for energy and power consumption, e.g., such that most of the energy goes into floating point multiply and add.
This section discusses further details for configuration and exception handling.
This section discloses examples of how to configure a CSA (e.g., fabric), how to achieve this configuration quickly, and how to minimize the resource overhead of configuration. Configuring the fabric quickly may be of preeminent importance in accelerating small portions of a larger algorithm, and consequently in broadening the applicability of a CSA. The section further discloses features that allow embodiments of a CSA to be programmed with configurations of different length.
Embodiments of a CSA (e.g., fabric) may differ from traditional cores in that they make use of a configuration step in which (e.g., large) parts of the fabric are loaded with program configuration in advance of program execution. An advantage of static configuration may be that very little energy is spent at runtime on the configuration, e.g., as opposed to sequential cores which spend energy fetching configuration information (an instruction) nearly every cycle. The previous disadvantage of configuration is that it was a coarse-grained step with a potentially large latency, which places an under-bound on the size of program that can be accelerated in the fabric due to the cost of context switching. This disclosure describes a scalable microarchitecture for rapidly configuring a spatial array in a distributed fashion, e.g., that avoids the previous disadvantages.
As discussed above, a CSA may include light-weight processing elements connected by an inter-PE network. Programs, viewed as control-dataflow graphs, are then mapped onto the architecture by configuring the configurable fabric elements (CFEs), for example PEs and the interconnect (fabric) networks. Generally, PEs may be configured as dataflow operators and once all input operands arrive at the PE, some operation occurs, and the results are forwarded to another PE or PEs for consumption or output. PEs may communicate over dedicated virtual circuits which are formed by statically configuring the circuit switched communications network. These virtual circuits may be flow controlled and fully back-pressured, e.g., such that PEs will stall if either the source has no data or destination is full. At runtime, data may flow through the PEs implementing the mapped algorithm. For example, data may be streamed in from memory, through the fabric, and then back out to memory. Such a spatial architecture may achieve remarkable performance efficiency relative to traditional multicore processors: compute, in the form of PEs, may be simpler and more numerous than larger cores and communications may be direct, as opposed to an extension of the memory system.
Embodiments of a CSA may not utilize (e.g., software controlled) packet switching, e.g., packet switching that requires significant software assistance to realize, which slows configuration. Embodiments of a CSA include out-of-band signaling in the network (e.g., of only 2-3 bits, depending on the feature set supported) and a fixed configuration topology to avoid the need for significant software support.
One key difference between embodiments of a CSA and the approach used in FPGAs is that a CSA approach may use a wide data word, is distributed, and includes mechanisms to fetch program data directly from memory. Embodiments of a CSA may not utilize JTAG-style single bit communications in the interest of area efficiency, e.g., as that may require milliseconds to completely configure a large FPGA fabric.
Embodiments of a CSA include a distributed configuration protocol and microarchitecture to support this protocol. Initially, configuration state may reside in memory. Multiple (e.g., distributed) local configuration controllers (boxes) (LCCs) may stream portions of the overall program into their local region of the spatial fabric, e.g., using a combination of a small set of control signals and the fabric-provided network. State elements may be used at each CFE to form configuration chains, e.g., allowing individual CFEs to self-program without global addressing.
Embodiments of a CSA include specific hardware support for the formation of configuration chains, e.g., not software establishing these chains dynamically at the cost of increasing configuration time. Embodiments of a CSA are not purely packet switched and do include extra out-of-band control wires (e.g., control is not sent through the data path requiring extra cycles to strobe this information and reserialize this information). Embodiments of a CSA decreases configuration latency by fixing the configuration ordering and by providing explicit out-of-band control (e.g., by at least a factor of two), while not significantly increasing network complexity.
Embodiments of a CSA do not use a serial mechanism for configuration in which data is streamed bit by bit into the fabric using a JTAG-like protocol. Embodiments of a CSA utilize a coarse-grained fabric approach. In certain embodiments, adding a few control wires or state elements to a 64 or 32-bit-oriented CSA fabric has a lower cost relative to adding those same control mechanisms to a 4 or 6 bit fabric.
Embodiments of a CSA include hardware that provides for efficient, distributed, low-latency configuration of a heterogeneous spatial fabric. This may be achieved according to four techniques. First, a hardware entity, the local configuration controller (LCC) is utilized, for example, as in
LCC operation may begin when it receives a pointer to a code segment. Depending on the LCB microarchitecture, this pointer (e.g., stored in pointer register 7506) may come either over a network (e.g., from within the CSA (fabric) itself) or through a memory system access to the LCC. When it receives such a pointer, the LCC optionally drains relevant state from its portion of the fabric for context storage, and then proceeds to immediately reconfigure the portion of the fabric for which it is responsible. The program loaded by the LCC may be a combination of configuration data for the fabric and control commands for the LCC, e.g., which are lightly encoded. As the LCC streams in the program portion, it may interprets the program as a command stream and perform the appropriate encoded action to configure (e.g., load) the fabric.
Two different microarchitectures for the LCC are shown in
In certain embodiments, configuration relies on 2-8 extra, out-of-band control channels to improve configuration speed, as defined below. For example, configuration controller 7502 may include the following control channels, e.g., CFG_START control channel 7508, CFG_VALID control channel 7510, and CFG_DONE control channel 7512, with examples of each discussed in Table 2 below.
Generally, the handling of configuration information may be left to the implementer of a particular CFE. For example, a selectable function CFE may have a provision for setting registers using an existing data path, while a fixed function CFE might simply set a configuration register.
Due to long wire delays when programming a large set of CFEs, the CFG_VALID signal may be treated as a clock/latch enable for CFE components. Since this signal is used as a clock, in one embodiment the duty cycle of the line is at most 50%. As a result, configuration throughput is approximately halved. Optionally, a second CFG_VALID signal may be added to enable continuous programming.
In one embodiment, only CFG_START is strictly communicated on an independent coupling (e.g., wire), for example, CFG_VALID and CFG_DONE may be overlaid on top of other network couplings.
To reduce the overhead of configuration, certain embodiments of a CSA make use of existing network infrastructure to communicate configuration data. A LCC may make use of both a chip-level memory hierarchy and a fabric-level communications networks to move data from storage into the fabric. As a result, in certain embodiments of a CSA, the configuration infrastructure adds no more than 2% to the overall fabric area and power.
Reuse of network resources in certain embodiments of a CSA may cause a network to have some hardware support for a configuration mechanism. Circuit switched networks of embodiments of a CSA cause an LCC to set their multiplexors in a specific way for configuration when the ‘CFG_START’ signal is asserted. Packet switched networks do not require extension, although LCC endpoints (e.g., configuration terminators) use a specific address in the packet switched network. Network reuse is optional, and some embodiments may find dedicated configuration buses to be more convenient.
Each CFE may maintain a bit denoting whether or not it has been configured (see, e.g.,
Internal to the CFE, this bit may be used to drive flow control ready signals. For example, when the configuration bit is de-asserted, network control signals may automatically be clamped to a values that prevent data from flowing, while, within PEs, no operations or other actions will be scheduled.
Dealing with High-Delay Configuration Paths
One embodiment of an LCC may drive a signal over a long distance, e.g., through many multiplexors and with many loads. Thus, it may be difficult for a signal to arrive at a distant CFE within a short clock cycle. In certain embodiments, configuration signals are at some division (e.g., fraction of) of the main (e.g., CSA) clock frequency to ensure digital timing discipline at configuration. Clock division may be utilized in an out-of-band signaling protocol, and does not require any modification of the main clock tree.
Since certain configuration schemes are distributed and have non-deterministic timing due to program and memory effects, different portions of the fabric may be configured at different times. As a result, certain embodiments of a CSA provide mechanisms to prevent inconsistent operation among configured and unconfigured CFEs. Generally, consistency is viewed as a property required of and maintained by CFEs themselves, e.g., using the internal CFE state. For example, when a CFE is in an unconfigured state, it may claim that its input buffers are full, and that its output is invalid. When configured, these values will be set to the true state of the buffers. As enough of the fabric comes out of configuration, these techniques may permit it to begin operation. This has the effect of further reducing context switching latency, e.g., if long-latency memory requests are issued early.
Different CFEs may have different configuration word widths. For smaller CFE configuration words, implementers may balance delay by equitably assigning CFE configuration loads across the network wires. To balance loading on network wires, one option is to assign configuration bits to different portions of network wires to limit the net delay on any one wire. Wide data words may be handled by using serialization/deserialization techniques. These decisions may be taken on a per-fabric basis to optimize the behavior of a specific CSA (e.g., fabric). Network controller (e.g., one or more of network controller 7310 and network controller 7312 may communicate with each domain (e.g., subset) of the CSA (e.g., fabric), for example, to send configuration information to one or more LCCs. Network controller may be part of a communications network (e.g., separate from circuit switched network). Network controller may include a network dataflow endpoint circuit.
Embodiments of a CSA may be an energy-efficient and high-performance means of accelerating user applications. When considering whether a program (e.g., a dataflow graph thereof) may be successfully accelerated by an accelerator, both the time to configure the accelerator and the time to run the program may be considered. If the run time is short, then the configuration time may play a large role in determining successful acceleration. Therefore, to maximize the domain of accelerable programs, in some embodiments the configuration time is made as short as possible. One or more configuration caches may be includes in a CSA, e.g., such that the high bandwidth, low-latency store enables rapid reconfiguration. Next is a description of several embodiments of a configuration cache.
In one embodiment, during configuration, the configuration hardware (e.g., LCC) optionally accesses the configuration cache to obtain new configuration information. The configuration cache may operate either as a traditional address based cache, or in an OS managed mode, in which configurations are stored in the local address space and addressed by reference to that address space. If configuration state is located in the cache, then no requests to the backing store are to be made in certain embodiments. In certain embodiments, this configuration cache is separate from any (e.g., lower level) shared cache in the memory hierarchy.
Demand Caching—In this mode, the configuration cache operates as a true cache. The configuration controller issues address-based requests, which are checked against tags in the cache. Misses are loaded into the cache and then may be re-referenced during future reprogramming.
In-Fabric Storage (Scratchpad) Caching—In this mode the configuration cache receives a reference to a configuration sequence in its own, small address space, rather than the larger address space of the host. This may improve memory density since the portion of cache used to store tags may instead be used to store configuration.
In certain embodiments, a configuration cache may have the configuration data pre-loaded into it, e.g., either by external direction or internal direction. This may allow reduction in the latency to load programs. Certain embodiments herein provide for an interface to a configuration cache which permits the loading of new configuration state into the cache, e.g., even if a configuration is running in the fabric already. The initiation of this load may occur from either an internal or external source. Embodiments of a pre-loading mechanism further reduce latency by removing the latency of cache loading from the configuration path.
Explicit Prefetching—A configuration path is augmented with a new command, ConfigurationCachePrefetch. Instead of programming the fabric, this command simply cause a load of the relevant program configuration into a configuration cache, without programming the fabric. Since this mechanism piggybacks on the existing configuration infrastructure, it is exposed both within the fabric and externally, e.g., to cores and other entities accessing the memory space.
Implicit prefetching—A global configuration controller may maintain a prefetch predictor, and use this to initiate the explicit prefetching to a configuration cache, e.g., in an automated fashion.
Certain embodiments of a CSA (e.g., a spatial fabric) include large amounts of instruction and configuration state, e.g., which is largely static during the operation of the CSA. Thus, the configuration state may be vulnerable to soft errors. Rapid and error-free recovery of these soft errors may be critical to the long-term reliability and performance of spatial systems.
Certain embodiments herein provide for a rapid configuration recovery loop, e.g., in which configuration errors are detected and portions of the fabric immediately reconfigured. Certain embodiments herein include a configuration controller, e.g., with reliability, availability, and serviceability (RAS) reprogramming features. Certain embodiments of CSA include circuitry for high-speed configuration, error reporting, and parity checking within the spatial fabric. Using a combination of these three features, and optionally, a configuration cache, a configuration/exception handling circuit may recover from soft errors in configuration. When detected, soft errors may be conveyed to a configuration cache which initiates an immediate reconfiguration of (e.g., that portion of) the fabric. Certain embodiments provide for a dedicated reconfiguration circuit, e.g., which is faster than any solution that would be indirectly implemented in the fabric. In certain embodiments, co-located exception and configuration circuit cooperates to reload the fabric on configuration error detection.
Some portions of an application targeting a CSA (e.g., spatial array) may be run infrequently or may be mutually exclusive with other parts of the program. To save area, to improve performance, and/or reduce power, it may be useful to time multiplex portions of the spatial fabric among several different parts of the program dataflow graph. Certain embodiments herein include an interface by which a CSA (e.g., via the spatial program) may request that part of the fabric be reprogrammed. This may enable the CSA to dynamically change itself according to dynamic control flow. Certain embodiments herein allow for fabric initiated reconfiguration (e.g., reprogramming). Certain embodiments herein provide for a set of interfaces for triggering configuration from within the fabric. In some embodiments, a PE issues a reconfiguration request based on some decision in the program dataflow graph. This request may travel a network to our new configuration interface, where it triggers reconfiguration. Once reconfiguration is completed, a message may optionally be returned notifying of the completion. Certain embodiments of a CSA thus provide for a program (e.g., dataflow graph) directed reconfiguration capability.
Configure-by-address—In this mode, the fabric makes a direct request to load configuration data from a particular address.
Configure-by-reference—In this mode the fabric makes a request to load a new configuration, e.g., by a pre-determined reference ID. This may simplify the determination of the code to load, since the location of the code has been abstracted.
A CSA may include a higher level configuration controller to support a multicast mechanism to cast (e.g., via network indicated by the dotted box) configuration requests to multiple (e.g., distributed or local) configuration controllers. This may enable a single configuration request to be replicated across larger portions of the fabric, e.g., triggering a broad reconfiguration.
Certain embodiments of a CSA may also experience an exception (e.g., exceptional condition), for example, floating point underflow. When these conditions occur, a special handlers may be invoked to either correct the program or to terminate it. Certain embodiments herein provide for a system-level architecture for handling exceptions in spatial fabrics. Since certain spatial fabrics emphasize area efficiency, embodiments herein minimize total area while providing a general exception mechanism. Certain embodiments herein provides a low area means of signaling exceptional conditions occurring in within a CSA (e.g., a spatial array). Certain embodiments herein provide an interface and signaling protocol for conveying such exceptions, as well as a PE-level exception semantics. Certain embodiments herein are dedicated exception handling capabilities, e.g., and do not require explicit handling by the programmer.
One embodiments of a CSA exception architecture consists of four portions, e.g., shown in
1. PE Exception Generator
2. Local Exception Network
3. Mezzanine Exception Aggregator
4. Tile-Level Exception Aggregator
Processing element 8100 may include processing element 1000 from
The initiation of the exception may either occur explicitly, by the execution of a programmer supplied instruction, or implicitly when a hardened error condition (e.g., a floating point underflow) is detected. Upon an exception, the PE 8100 may enter a waiting state, in which it waits to be serviced by the eventual exception handler, e.g., external to the PE 8100. The contents of the exception packet depend on the implementation of the particular PE, as described below.
A (e.g., local) exception network steers exception packets from PE 8100 to the mezzanine exception network. Exception network (e.g., 8113) may be a serial, packet switched network consisting of a (e.g., single) control wire and one or more data wires, e.g., organized in a ring or tree topology, e.g., for a subset of PEs. Each PE may have a (e.g., ring) stop in the (e.g., local) exception network, e.g., where it can arbitrate to inject messages into the exception network.
PE endpoints needing to inject an exception packet may observe their local exception network egress point. If the control signal indicates busy, the PE is to wait to commence inject its packet. If the network is not busy, that is, the downstream stop has no packet to forward, then the PE will proceed commence injection.
Network packets may be of variable or fixed length. Each packet may begin with a fixed length header field identifying the source PE of the packet. This may be followed by a variable number of PE-specific field containing information, for example, including error codes, data values, or other useful status information.
The mezzanine exception aggregator 8004 is responsible for assembling local exception network into larger packets and sending them to the tile-level exception aggregator 8002. The mezzanine exception aggregator 8004 may pre-pend the local exception packet with its own unique ID, e.g., ensuring that exception messages are unambiguous. The mezzanine exception aggregator 8004 may interface to a special exception-only virtual channel in the mezzanine network, e.g., ensuring the deadlock-freedom of exceptions.
The mezzanine exception aggregator 8004 may also be able to directly service certain classes of exception. For example, a configuration request from the fabric may be served out of the mezzanine network using caches local to the mezzanine network stop.
The final stage of the exception system is the tile-level exception aggregator 8002. The tile-level exception aggregator 8002 is responsible for collecting exceptions from the various mezzanine-level exception aggregators (e.g., 8004) and forwarding them to the appropriate servicing hardware (e.g., core). As such, the tile-level exception aggregator 8002 may include some internal tables and controller to associate particular messages with handler routines. These tables may be indexed either directly or with a small state machine in order to steer particular exceptions.
Like the mezzanine exception aggregator, the tile-level exception aggregator may service some exception requests. For example, it may initiate the reprogramming of a large portion of the PE fabric in response to a specific exception.
Certain embodiments of a CSA include an extraction controller(s) to extract data from the fabric. The below discusses embodiments of how to achieve this extraction quickly and how to minimize the resource overhead of data extraction. Data extraction may be utilized for such critical tasks as exception handling and context switching. Certain embodiments herein extract data from a heterogeneous spatial fabric by introducing features that allow extractable fabric elements (EFEs) (for example, PEs, network controllers, and/or switches) with variable and dynamically variable amounts of state to be extracted.
Embodiments of a CSA include a distributed data extraction protocol and microarchitecture to support this protocol. Certain embodiments of a CSA include multiple local extraction controllers (LECs) which stream program data out of their local region of the spatial fabric using a combination of a (e.g., small) set of control signals and the fabric-provided network. State elements may be used at each extractable fabric element (EFE) to form extraction chains, e.g., allowing individual EFEs to self-extract without global addressing.
Embodiments of a CSA do not use a local network to extract program data. Embodiments of a CSA include specific hardware support (e.g., an extraction controller) for the formation of extraction chains, for example, and do not rely on software to establish these chains dynamically, e.g., at the cost of increasing extraction time. Embodiments of a CSA are not purely packet switched and do include extra out-of-band control wires (e.g., control is not sent through the data path requiring extra cycles to strobe and reserialize this information). Embodiments of a CSA decrease extraction latency by fixing the extraction ordering and by providing explicit out-of-band control (e.g., by at least a factor of two), while not significantly increasing network complexity.
Embodiments of a CSA do not use a serial mechanism for data extraction, in which data is streamed bit by bit from the fabric using a JTAG-like protocol. Embodiments of a CSA utilize a coarse-grained fabric approach. In certain embodiments, adding a few control wires or state elements to a 64 or 32-bit-oriented CSA fabric has a lower cost relative to adding those same control mechanisms to a 4 or 6 bit fabric.
Embodiments of a CSA include hardware that provides for efficient, distributed, low-latency extraction from a heterogeneous spatial fabric. This may be achieved according to four techniques. First, a hardware entity, the local extraction controller (LEC) is utilized, for example, as in
The following sections describe the operation of the various components of embodiments of an extraction network.
LEC operation may begin when it receives a pointer to a buffer (e.g., in virtual memory) where fabric state will be written, and, optionally, a command controlling how much of the fabric will be extracted. Depending on the LEC microarchitecture, this pointer (e.g., stored in pointer register 8404) may come either over a network or through a memory system access to the LEC. When it receives such a pointer (e.g., command), the LEC proceeds to extract state from the portion of the fabric for which it is responsible. The LEC may stream this extracted data out of the fabric into the buffer provided by the external caller.
Two different microarchitectures for the LEC are shown in
Extra Out-of-Band Control Channels (e.g., Wires)
In certain embodiments, extraction relies on 2-8 extra, out-of-band signals to improve configuration speed, as defined below. Signals driven by the LEC may be labelled LEC. Signals driven by the EFE (e.g., PE) may be labelled EFE. Configuration controller 8402 may include the following control channels, e.g., LEC_EXTRACT control channel 8506, LEC_START control channel 8408, LEC_STROBE control channel 8410, and EFE_COMPLETE control channel 8412, with examples of each discussed in Table 3 below.
Generally, the handling of extraction may be left to the implementer of a particular EFE. For example, selectable function EFE may have a provision for dumping registers using an existing data path, while a fixed function EFE might simply have a multiplexor.
Due to long wire delays when programming a large set of EFEs, the LEC_STROBE signal may be treated as a clock/latch enable for EFE components. Since this signal is used as a clock, in one embodiment the duty cycle of the line is at most 50%. As a result, extraction throughput is approximately halved. Optionally, a second LEC_STROBE signal may be added to enable continuous extraction.
In one embodiment, only LEC_START is strictly communicated on an independent coupling (e.g., wire), for example, other control channels may be overlayed on existing network (e.g., wires).
To reduce the overhead of data extraction, certain embodiments of a CSA make use of existing network infrastructure to communicate extraction data. A LEC may make use of both a chip-level memory hierarchy and a fabric-level communications networks to move data from the fabric into storage. As a result, in certain embodiments of a CSA, the extraction infrastructure adds no more than 2% to the overall fabric area and power.
Reuse of network resources in certain embodiments of a CSA may cause a network to have some hardware support for an extraction protocol. Circuit switched networks require of certain embodiments of a CSA cause a LEC to set their multiplexors in a specific way for configuration when the ‘LEC_START’ signal is asserted. Packet switched networks may not require extension, although LEC endpoints (e.g., extraction terminators) use a specific address in the packet switched network. Network reuse is optional, and some embodiments may find dedicated configuration buses to be more convenient.
Each EFE may maintain a bit denoting whether or not it has exported its state. This bit may de-asserted when the extraction start signal is driven, and then asserted once the particular EFE finished extraction. In one extraction protocol, EFEs are arranged to form chains with the EFE extraction state bit determining the topology of the chain. A EFE may read the extraction state bit of the immediately adjacent EFE. If this adjacent EFE has its extraction bit set and the current EFE does not, the EFE may determine that it owns the extraction bus. When an EFE dumps its last data value, it may drives the ‘EFE_DONE’ signal and sets its extraction bit, e.g., enabling upstream EFEs to configure for extraction. The network adjacent to the EFE may observe this signal and also adjust its state to handle the transition. As a base case to the extraction process, an extraction terminator (e.g., extraction terminator 8204 for LEC 8202 or extraction terminator 8208 for LEC 8206 in
Internal to the EFE, this bit may be used to drive flow control ready signals. For example, when the extraction bit is de-asserted, network control signals may automatically be clamped to a values that prevent data from flowing, while, within PEs, no operations or actions will be scheduled.
Dealing with High-Delay Paths
One embodiment of a LEC may drive a signal over a long distance, e.g., through many multiplexors and with many loads. Thus, it may be difficult for a signal to arrive at a distant EFE within a short clock cycle. In certain embodiments, extraction signals are at some division (e.g., fraction of) of the main (e.g., CSA) clock frequency to ensure digital timing discipline at extraction. Clock division may be utilized in an out-of-band signaling protocol, and does not require any modification of the main clock tree.
Ensuring Consistent Fabric Behavior During Extraction
Since certain extraction scheme are distributed and have non-deterministic timing due to program and memory effects, different members of the fabric may be under extraction at different times. While LEC_EXTRACT is driven, all network flow control signals may be driven logically low, e.g., thus freezing the operation of a particular segment of the fabric.
An extraction process may be non-destructive. Therefore a set of PEs may be considered operational once extraction has completed. An extension to an extraction protocol may allow PEs to optionally be disabled post extraction. Alternatively, beginning configuration during the extraction process will have similar effect in embodiments.
In some cases, it may be expedient to extract a single PE. In this case, an optional address signal may be driven as part of the commencement of the extraction process. This may enable the PE targeted for extraction to be directly enabled. Once this PE has been extracted, the extraction process may cease with the lowering of the LEC_EXTRACT signal. In this way, a single PE may be selectively extracted, e.g., by the local extraction controller.
In an embodiment where the LEC writes extracted data to memory (for example, for post-processing, e.g., in software), it may be subject to limited memory bandwidth. In the case that the LEC exhausts its buffering capacity, or expects that it will exhaust its buffering capacity, it may stops strobing the LEC_STROBE signal until the buffering issue has resolved.
Note that in certain figures (e.g.,
In one embodiment, programs, viewed as control data flow graphs, are mapped onto the spatial architecture by configuring PEs and a communications network. Generally, PEs are configured as dataflow operators, similar to functional units in a processor: once the input operands arrive at the PE, some operation occurs, and results are forwarded to downstream PEs in a pipelined fashion. Dataflow operators (or other types of operators) may choose to consume incoming data on a per-operator basis. Simple operators, like those handling the unconditional evaluation of arithmetic expressions often consume all incoming data. It is sometimes useful, however, for operators to maintain state, for example, in accumulation.
The PEs communicate using dedicated virtual circuits, which are formed by statically configuring a circuit-switched communications network. These virtual circuits are flow controlled and fully back pressured, such that PEs will stall if either the source has no data or the destination is full. At runtime, data flows through the PEs implementing a mapped algorithm according to a dataflow graph, also referred to as a subprogram herein. For example, data may be streamed in from memory, through the acceleration hardware 8702, and then back out to memory. Such an architecture can achieve remarkable performance efficiency relative to traditional multicore processors: compute, in the form of PEs, is simpler and more numerous than larger cores and communication is direct, as opposed to an extension of the memory subsystem 8710. Memory system parallelism, however, helps to support parallel PE computation. If memory accesses are serialized, high parallelism is likely unachievable. To facilitate parallelism of memory accesses, the disclosed memory ordering circuit 8705 includes memory ordering architecture and microarchitecture, as will be explained in detail. In one embodiment, the memory ordering circuit 8705 is a request address file circuit (or “RAF”) or other memory request circuitry.
Each memory ordering circuit 8705 may accept read and write requests to the memory subsystem 8710. The requests from the acceleration hardware 8702 arrive at the memory ordering circuit 8705 in a separate channel for each node of the dataflow graph that initiates read or write accesses, also referred to as load or store accesses herein. Buffering is provided so that the processing of loads will return the requested data to the acceleration hardware 8702 in the order it was requested. In other words, iteration six data is returned before iteration seven data, and so forth. Furthermore, note that the request channel from a memory ordering circuit 8705 to a particular cache bank may be implemented as an ordered channel and any first request that leaves before a second request will arrive at the cache bank before the second request.
By considering this sequence of operations, it may be evident that spatial arrays more naturally map to channels. Furthermore, the acceleration hardware 8702 is latency-insensitive in terms of the request and response channels, and inherent parallel processing that may occur. The acceleration hardware may also decouple execution of a program from implementation of the memory subsystem 8710 (
The memory ordering circuit 8705 may further include, but not be limited to, a memory interface 9010, an operations queue 9012, input queue(s) 9016, a completion queue 9020, an operation configuration data structure 9024, and an operations manager circuit 9030 that may further include a scheduler circuit 9032 and an execution circuit 9034. In one embodiment, the memory interface 9010 may be circuit-switched, and in another embodiment, the memory interface 9010 may be packet-switched, or both may exist simultaneously. The operations queue 9012 may buffer memory operations (with corresponding arguments) that are being processed for request, and may, therefore, correspond to addresses and data coming into the input queues 9016.
More specifically, the input queues 9016 may be an aggregation of at least the following: a load address queue, a store address queue, a store data queue, and a dependency queue. When implementing the input queue 9016 as aggregated, the memory ordering circuit 8705 may provide for sharing of logical queues, with additional control logic to logically separate the queues, which are individual channels with the memory ordering circuit. This may maximize input queue usage, but may also require additional complexity and space for the logic circuitry to manage the logical separation of the aggregated queue. Alternatively, as will be discussed with reference to
When shared, the input queues 9016 and the completion queue 9020 may be implemented as ring buffers of a fixed size. A ring buffer is an efficient implementation of a circular queue that has a first-in-first-out (FIFO) data characteristic. These queues may, therefore, enforce a semantical order of a program for which the memory operations are being requested. In one embodiment, a ring buffer (such as for the store address queue) may have entries corresponding to entries flowing through an associated queue (such as the store data queue or the dependency queue) at the same rate. In this way, a store address may remain associated with corresponding store data.
More specifically, the load address queue may buffer an incoming address of the memory 18 from which to retrieve data. The store address queue may buffer an incoming address of the memory 18 to which to write data, which is buffered in the store data queue. The dependency queue may buffer dependency tokens in association with the addresses of the load address queue and the store address queue. Each queue, representing a separate channel, may be implemented with a fixed or dynamic number of entries. When fixed, the more entries that are available, the more efficient complicated loop processing may be made. But, having too many entries costs more area and energy to implement. In some cases, e.g., with the aggregated architecture, the disclosed input queue 9016 may share queue slots. Use of the slots in a queue may be statically allocated.
The completion queue 9020 may be a separate set of queues to buffer data received from memory in response to memory commands issued by load operations. The completion queue 9020 may be used to hold a load operation that has been scheduled but for which data has not yet been received (and thus has not yet completed). The completion queue 9020, may therefore, be used to reorder data and operation flow.
The operations manager circuit 9030, which will be explained in more detail with reference to
From an architectural perspective, there are at least two goals: first, to run general sequential codes correctly, and second, to obtain high performance in the memory operations performed by the microarchitecture 9100. To ensure program correctness, the compiler expresses the dependency between the store operation and the load operation to an array, p, in some fashion, which are expressed via dependency tokens as will be explained. To improve performance, the microarchitecture 9100 finds and issues as many load commands of an array in parallel as is legal with respect to program order.
In one embodiment, the microarchitecture 9100 may include the operations queue 9012, the input queues 9016, the completion queues 9020, and the operations manager circuit 9030 discussed with reference to
The input queues 9016, as mentioned, may include a load address queue 9122, a store address queue 9124, and a store data queue 9126. (The small numbers 0, 1, 2 are channel labels and will be referred to later in
In one embodiment, the completion queues 9020 may include a set of output buffers 9144 and 9146 for receipt of load data from the memory subsystem 8710 and a completion queue 9142 to buffer addresses and data for load operations according to an index maintained by the operations manager circuit 9030. The operations manager circuit 9030 can manage the index to ensure in-order execution of the load operations, and to identify data received into the output buffers 9144 and 9146 that may be moved to scheduled load operations in the completion queue 9142.
More specifically, because the memory subsystem 8710 is out of order, but the acceleration hardware 8702 completes operations in order, the microarchitecture 9100 may re-order memory operations with use of the completion queue 9142. Three different sub-operations may be performed in relation to the completion queue 9142, namely to allocate, enqueue, and dequeue. For allocation, the operations manager circuit 9030 may allocate an index into the completion queue 9142 in an in-order next slot of the completion queue. The operations manager circuit may provide this index to the memory subsystem 8710, which may then know the slot to which to write data for a load operation. To enqueue, the memory subsystem 8710 may write data as an entry to the indexed, in-order next slot in the completion queue 9142 like random access memory (RAM), setting a status bit of the entry to valid. To dequeue, the operations manager circuit 9030 may present the data stored in this in-order next slot to complete the load operation, setting the status bit of the entry to invalid. Invalid entries may then be available for a new allocation.
In one embodiment, the status signals 9048 may refer to statuses of the input queues 9016, the completion queues 9020, the dependency queues 9118, and the dependency token counters 9114. These statuses, for example, may include an input status, an output status, and a control status, which may refer to the presence or absence of a dependency token in association with an input or an output. The input status may include the presence or absence of addresses and the output status may include the presence or absence of store values and available completion buffer slots. The dependency token counters 9114 may be a compact representation of a queue and track a number of dependency tokens used for any given input queue. If the dependency token counters 9114 saturate, no additional dependency tokens may be generated for new memory operations. Accordingly, the memory ordering circuit 8705 may stall scheduling new memory operations until the dependency token counters 9114 becomes unsaturated.
With additional reference to
ldNo[d,x] result.outN, addr.in64, order.in0, order.out0
stNo[d,x] addr.in64, data.inN, order.in0, order.out0
The executable determiner circuit 9200 may be integrated as a part of the scheduler circuit 9032 and which may perform a logical operation to determine whether a given memory operation is executable, and thus ready to be issued to memory. A memory operation may be executed when the queues corresponding to its memory arguments have data and an associated dependency token is present. These memory arguments may include, for example, an input queue identifier 9210 (indicative of a channel of the input queue 9016), an output queue identifier 9220 (indicative of a channel of the completion queues 9020), a dependency queue identifier 9230 (e.g., what dependency queue or counter should be referenced), and an operation type indicator 9240 (e.g., load operation or store operation). A field (e.g., of a memory request) may be included, e.g., in the above format, that stores a bit or bits to indicate to use the hazard checking hardware.
These memory arguments may be queued within the operations queue 9012, and used to schedule issuance of memory operations in association with incoming addresses and data from memory and the acceleration hardware 8702. (See
For a load operation, and by way of example, the memory ordering circuit 8705 may issue a load command when the load operation has an address (input status) and room to buffer the load result in the completion queue 9142 (output status). Similarly, the memory ordering circuit 8705 may issue a store command for a store operation when the store operation has both an address and data value (input status). Accordingly, the status signals 9048 may communicate a level of emptiness (or fullness) of the queues to which the status signals pertain. The operation type may then dictate whether the logic results in an executable signal depending on what address and data should be available.
To implement dependency ordering, the scheduler circuit 9032 may extend memory operations to include dependency tokens as underlined above in the example load and store operations. The control status 9232 may indicate whether a dependency token is available within the dependency queue identified by the dependency queue identifier 9230, which could be one of the dependency queues 9118 (for an incoming memory operation) or a dependency token counter 9114 (for a completed memory operation). Under this formulation, a dependent memory operation requires an additional ordering token to execute and generates an additional ordering token upon completion of the memory operation, where completion means that data from the result of the memory operation has become available to program-subsequent memory operations.
In one embodiment, with further reference to
The priority encoder 9306, for example, may be a circuit (such as a state machine or a simpler converter) that compresses multiple binary inputs into a smaller number of outputs, including possibly just one output. The output of a priority encoder is the binary representation of the original number starting from zero of the most significant input bit. So, in one example, when memory operation 0 (“zero”), memory operation one (“1”), and memory operation two (“2”) are executable and scheduled, corresponding to 9304A, 9304B, and 9304C, respectively. The priority encoder 9306 may be configured to output the selector signal 9307 to the selection circuitry 9308 indicating the memory operation zero as the memory operation that has highest priority. The selection circuitry 9308 may be a multiplexer in one embodiment, and be configured to output its selection (e.g., of memory operation zero) onto the control lines 9310, as a control signal, in response to the selector signal from the priority encoder 9306 (and indicative of selection of memory operation of highest priority). This control signal may go to the multiplexers 9132, 9134, 9136, and/or 9138, as discussed with reference to
An example of memory ordering by the memory ordering circuit 8705 will be illustrated with a simplified example for purposes of explanation with relation to
Assume, for this example, that array p contains 0, 1, 2, 3, 4, 5, 6, and at the end of loop execution, array p will contain 0, 1, 0, 1, 0, 1, 0. This code may be transformed by unrolling the loop, as illustrated in
The way the microarchitecture may perform this reordering is discussed with reference to
In
In
In
Note that the address p[2] for the newest load operation is dependent on the value that first needs to be stored by the store operation for address p[2], which is at the top of the store address queue. Later, the indexed entry in the completion queue for the load operation from address p[2] may remain buffered until the data from the store operation to the address p[2] is completed (see
In
In
In
In
In
In the present embodiment, the process of executing the code of
More specifically, referring to
The method 9800 may continue with the memory ordering circuit scheduling issuance of the second memory operation to the memory in response to receiving the dependency token and the address associated with the dependency token (9840). For example, when the load address queue receives the address for an address argument of a load operation and the dependency queue receives the dependency token for a control argument of the load operation, the memory ordering circuit may schedule issuance of the second memory operation as a load operation. The method 9800 may continue with the memory ordering circuit issuing the second memory operation (e.g., in a command) to the memory in response to completion of the first memory operation (9850). For example, if the first memory operation is a store, completion may be verified by acknowledgement that the data in a store data queue of the set of input queues has been written to the address in the memory. Similarly, if the first memory operation is a load operation, completion may be verified by receipt of data from the memory for the load operation.
Supercomputing at the ExaFLOP scale may be a challenge in high-performance computing, a challenge which is not likely to be met by conventional von Neumann architectures. To achieve ExaFLOPs, embodiments of a CSA provide a heterogeneous spatial array that targets direct execution of (e.g., compiler-produced) dataflow graphs. In addition to laying out the architectural principles of embodiments of a CSA, the above also describes and evaluates embodiments of a CSA which showed performance and energy of larger than 10× over existing products. Compiler-generated code may have significant performance and energy gains over roadmap architectures. As a heterogeneous, parametric architecture, embodiments of a CSA may be readily adapted to all computing uses. For example, a mobile version of CSA might be tuned to 32-bits, while a machine-learning focused array might feature significant numbers of vectorized 8-bit multiplication units. The main advantages of embodiments of a CSA are high performance and extreme energy efficiency, characteristics relevant to all forms of computing ranging from supercomputing and datacenter to the internet-of-things.
In one embodiment, a processor includes a spatial array of processing elements; and a packet switched communications network to route data within the spatial array between processing elements according to a dataflow graph to perform a first dataflow operation of the dataflow graph, wherein the packet switched communications network further comprises a plurality of network dataflow endpoint circuits to perform a second dataflow operation of the dataflow graph. A network dataflow endpoint circuit of the plurality of network dataflow endpoint circuits may include a network ingress buffer to receive input data from the packet switched communications network; and a spatial array egress buffer to output resultant data to the spatial array of processing elements according to the second dataflow operation on the input data. The spatial array egress buffer may output the resultant data based on a scheduler within the network dataflow endpoint circuit monitoring the packet switched communications network. The spatial array egress buffer may output the resultant data based on the scheduler within the network dataflow endpoint circuit monitoring a selected channel of multiple network virtual channels of the packet switched communications network. A network dataflow endpoint circuit of the plurality of network dataflow endpoint circuits may include a spatial array ingress buffer to receive control data from the spatial array that causes a network ingress buffer of the network dataflow endpoint circuit that received input data from the packet switched communications network to output resultant data to the spatial array of processing elements according to the second dataflow operation on the input data and the control data. A network dataflow endpoint circuit of the plurality of network dataflow endpoint circuits may stall an output of resultant data of the second dataflow operation from a spatial array egress buffer of the network dataflow endpoint circuit when a backpressure signal from a downstream processing element of the spatial array of processing elements indicates that storage in the downstream processing element is not available for the output of the network dataflow endpoint circuit. A network dataflow endpoint circuit of the plurality of network dataflow endpoint circuits may send a backpressure signal to stall a source from sending input data on the packet switched communications network into a network ingress buffer of the network dataflow endpoint circuit when the network ingress buffer is not available. The spatial array of processing elements may include a plurality of processing elements; and an interconnect network between the plurality of processing elements to receive an input of the dataflow graph comprising a plurality of nodes, wherein the dataflow graph is to be overlaid into the interconnect network, the plurality of processing elements, and the plurality of network dataflow endpoint circuits with each node represented as a dataflow operator in either of the plurality of processing elements and the plurality of network dataflow endpoint circuits, and the plurality of processing elements and the plurality of network dataflow endpoint circuits are to perform an operation by an incoming operand set arriving at each of the dataflow operators of the plurality of processing elements and the plurality of network dataflow endpoint circuits. The spatial array of processing elements may include a circuit switched network to transport the data within the spatial array between processing elements according to the dataflow graph.
In another embodiment, a method includes providing a spatial array of processing elements; routing, with a packet switched communications network, data within the spatial array between processing elements according to a dataflow graph; performing a first dataflow operation of the dataflow graph with the processing elements; and performing a second dataflow operation of the dataflow graph with a plurality of network dataflow endpoint circuits of the packet switched communications network. The performing the second dataflow operation may include receiving input data from the packet switched communications network with a network ingress buffer of a network dataflow endpoint circuit of the plurality of network dataflow endpoint circuits; and outputting resultant data from a spatial array egress buffer of the network dataflow endpoint circuit to the spatial array of processing elements according to the second dataflow operation on the input data. The outputting may include outputting the resultant data based on a scheduler within the network dataflow endpoint circuit monitoring the packet switched communications network. The outputting may include outputting the resultant data based on the scheduler within the network dataflow endpoint circuit monitoring a selected channel of multiple network virtual channels of the packet switched communications network. The performing the second dataflow operation may include receiving control data, with a spatial array ingress buffer of a network dataflow endpoint circuit of the plurality of network dataflow endpoint circuits, from the spatial array; and configuring the network dataflow endpoint circuit to cause a network ingress buffer of the network dataflow endpoint circuit that received input data from the packet switched communications network to output resultant data to the spatial array of processing elements according to the second dataflow operation on the input data and the control data. The performing the second dataflow operation may include stalling an output of the second dataflow operation from a spatial array egress buffer of a network dataflow endpoint circuit of the plurality of network dataflow endpoint circuits when a backpressure signal from a downstream processing element of the spatial array of processing elements indicates that storage in the downstream processing element is not available for the output of the network dataflow endpoint circuit. The performing the second dataflow operation may include sending a backpressure signal from a network dataflow endpoint circuit of the plurality of network dataflow endpoint circuits to stall a source from sending input data on the packet switched communications network into a network ingress buffer of the network dataflow endpoint circuit when the network ingress buffer is not available. The routing, performing the first dataflow operation, and performing the second dataflow operation may include receiving an input of a dataflow graph comprising a plurality of nodes; overlaying the dataflow graph into the spatial array of processing elements and the plurality of network dataflow endpoint circuits with each node represented as a dataflow operator in either of the processing elements and the plurality of network dataflow endpoint circuits; and performing the first dataflow operation with the processing elements and performing the second dataflow operation with the plurality of network dataflow endpoint circuits when an incoming operand set arrives at each of the dataflow operators of the processing elements and the plurality of network dataflow endpoint circuits. The method may include transporting the data within the spatial array between processing elements according to the dataflow graph with a circuit switched network of the spatial array.
In yet another embodiment, a non-transitory machine readable medium that stores code that when executed by a machine causes the machine to perform a method including providing a spatial array of processing elements; routing, with a packet switched communications network, data within the spatial array between processing elements according to a dataflow graph; performing a first dataflow operation of the dataflow graph with the processing elements; and performing a second dataflow operation of the dataflow graph with a plurality of network dataflow endpoint circuits of the packet switched communications network. The performing the second dataflow operation may include receiving input data from the packet switched communications network with a network ingress buffer of a network dataflow endpoint circuit of the plurality of network dataflow endpoint circuits; and outputting resultant data from a spatial array egress buffer of the network dataflow endpoint circuit to the spatial array of processing elements according to the second dataflow operation on the input data. The outputting may include outputting the resultant data based on a scheduler within the network dataflow endpoint circuit monitoring the packet switched communications network. The outputting may include outputting the resultant data based on the scheduler within the network dataflow endpoint circuit monitoring a selected channel of multiple network virtual channels of the packet switched communications network. The performing the second dataflow operation may include receiving control data, with a spatial array ingress buffer of a network dataflow endpoint circuit of the plurality of network dataflow endpoint circuits, from the spatial array; and configuring the network dataflow endpoint circuit to cause a network ingress buffer of the network dataflow endpoint circuit that received input data from the packet switched communications network to output resultant data to the spatial array of processing elements according to the second dataflow operation on the input data and the control data. The performing the second dataflow operation may include stalling an output of the second dataflow operation from a spatial array egress buffer of a network dataflow endpoint circuit of the plurality of network dataflow endpoint circuits when a backpressure signal from a downstream processing element of the spatial array of processing elements indicates that storage in the downstream processing element is not available for the output of the network dataflow endpoint circuit. The performing the second dataflow operation may include sending a backpressure signal from a network dataflow endpoint circuit of the plurality of network dataflow endpoint circuits to stall a source from sending input data on the packet switched communications network into a network ingress buffer of the network dataflow endpoint circuit when the network ingress buffer is not available. The routing, performing the first dataflow operation, and performing the second dataflow operation may include receiving an input of a dataflow graph comprising a plurality of nodes; overlaying the dataflow graph into the spatial array of processing elements and the plurality of network dataflow endpoint circuits with each node represented as a dataflow operator in either of the processing elements and the plurality of network dataflow endpoint circuits; and performing the first dataflow operation with the processing elements and performing the second dataflow operation with the plurality of network dataflow endpoint circuits when an incoming operand set arrives at each of the dataflow operators of the processing elements and the plurality of network dataflow endpoint circuits. The method may include transporting the data within the spatial array between processing elements according to the dataflow graph with a circuit switched network of the spatial array.
In another embodiment, a processor includes a spatial array of processing elements; and a packet switched communications network to route data within the spatial array between processing elements according to a dataflow graph to perform a first dataflow operation of the dataflow graph, wherein the packet switched communications network further comprises means to perform a second dataflow operation of the dataflow graph.
In one embodiment, a processor includes a core with a decoder to decode an instruction into a decoded instruction and an execution unit to execute the decoded instruction to perform a first operation; a plurality of processing elements; and an interconnect network between the plurality of processing elements to receive an input of a dataflow graph comprising a plurality of nodes, wherein the dataflow graph is to be overlaid into the interconnect network and the plurality of processing elements with each node represented as a dataflow operator in the plurality of processing elements, and the plurality of processing elements are to perform a second operation by a respective, incoming operand set arriving at each of the dataflow operators of the plurality of processing elements. A processing element of the plurality of processing elements may stall execution when a backpressure signal from a downstream processing element indicates that storage in the downstream processing element is not available for an output of the processing element. The processor may include a flow control path network to carry the backpressure signal according to the dataflow graph. A dataflow token may cause an output from a dataflow operator receiving the dataflow token to be sent to an input buffer of a particular processing element of the plurality of processing elements. The second operation may include a memory access and the plurality of processing elements comprises a memory-accessing dataflow operator that is not to perform the memory access until receiving a memory dependency token from a logically previous dataflow operator. The plurality of processing elements may include a first type of processing element and a second, different type of processing element.
In another embodiment, a method includes decoding an instruction with a decoder of a core of a processor into a decoded instruction; executing the decoded instruction with an execution unit of the core of the processor to perform a first operation; receiving an input of a dataflow graph comprising a plurality of nodes; overlaying the dataflow graph into a plurality of processing elements of the processor and an interconnect network between the plurality of processing elements of the processor with each node represented as a dataflow operator in the plurality of processing elements; and performing a second operation of the dataflow graph with the interconnect network and the plurality of processing elements by a respective, incoming operand set arriving at each of the dataflow operators of the plurality of processing elements. The method may include stalling execution by a processing element of the plurality of processing elements when a backpressure signal from a downstream processing element indicates that storage in the downstream processing element is not available for an output of the processing element. The method may include sending the backpressure signal on a flow control path network according to the dataflow graph. A dataflow token may cause an output from a dataflow operator receiving the dataflow token to be sent to an input buffer of a particular processing element of the plurality of processing elements. The method may include not performing a memory access until receiving a memory dependency token from a logically previous dataflow operator, wherein the second operation comprises the memory access and the plurality of processing elements comprises a memory-accessing dataflow operator. The method may include providing a first type of processing element and a second, different type of processing element of the plurality of processing elements.
In yet another embodiment, an apparatus includes a data path network between a plurality of processing elements; and a flow control path network between the plurality of processing elements, wherein the data path network and the flow control path network are to receive an input of a dataflow graph comprising a plurality of nodes, the dataflow graph is to be overlaid into the data path network, the flow control path network, and the plurality of processing elements with each node represented as a dataflow operator in the plurality of processing elements, and the plurality of processing elements are to perform a second operation by a respective, incoming operand set arriving at each of the dataflow operators of the plurality of processing elements. The flow control path network may carry backpressure signals to a plurality of dataflow operators according to the dataflow graph. A dataflow token sent on the data path network to a dataflow operator may cause an output from the dataflow operator to be sent to an input buffer of a particular processing element of the plurality of processing elements on the data path network. The data path network may be a static, circuit switched network to carry the respective, input operand set to each of the dataflow operators according to the dataflow graph. The flow control path network may transmit a backpressure signal according to the dataflow graph from a downstream processing element to indicate that storage in the downstream processing element is not available for an output of the processing element. At least one data path of the data path network and at least one flow control path of the flow control path network may form a channelized circuit with backpressure control. The flow control path network may pipeline at least two of the plurality of processing elements in series.
In another embodiment, a method includes receiving an input of a dataflow graph comprising a plurality of nodes; and overlaying the dataflow graph into a plurality of processing elements of a processor, a data path network between the plurality of processing elements, and a flow control path network between the plurality of processing elements with each node represented as a dataflow operator in the plurality of processing elements. The method may include carrying backpressure signals with the flow control path network to a plurality of dataflow operators according to the dataflow graph. The method may include sending a dataflow token on the data path network to a dataflow operator to cause an output from the dataflow operator to be sent to an input buffer of a particular processing element of the plurality of processing elements on the data path network. The method may include setting a plurality of switches of the data path network and/or a plurality of switches of the flow control path network to carry the respective, input operand set to each of the dataflow operators according to the dataflow graph, wherein the data path network is a static, circuit switched network. The method may include transmitting a backpressure signal with the flow control path network according to the dataflow graph from a downstream processing element to indicate that storage in the downstream processing element is not available for an output of the processing element. The method may include forming a channelized circuit with backpressure control with at least one data path of the data path network and at least one flow control path of the flow control path network.
In yet another embodiment, a processor includes a core with a decoder to decode an instruction into a decoded instruction and an execution unit to execute the decoded instruction to perform a first operation; a plurality of processing elements; and a network means between the plurality of processing elements to receive an input of a dataflow graph comprising a plurality of nodes, wherein the dataflow graph is to be overlaid into the network means and the plurality of processing elements with each node represented as a dataflow operator in the plurality of processing elements, and the plurality of processing elements are to perform a second operation by a respective, incoming operand set arriving at each of the dataflow operators of the plurality of processing elements.
In another embodiment, an apparatus includes a data path means between a plurality of processing elements; and a flow control path means between the plurality of processing elements, wherein the data path means and the flow control path means are to receive an input of a dataflow graph comprising a plurality of nodes, the dataflow graph is to be overlaid into the data path means, the flow control path means, and the plurality of processing elements with each node represented as a dataflow operator in the plurality of processing elements, and the plurality of processing elements are to perform a second operation by a respective, incoming operand set arriving at each of the dataflow operators of the plurality of processing elements.
In one embodiment, a processor includes a core with a decoder to decode an instruction into a decoded instruction and an execution unit to execute the decoded instruction to perform a first operation; and an array of processing elements to receive an input of a dataflow graph comprising a plurality of nodes, wherein the dataflow graph is to be overlaid into the array of processing elements with each node represented as a dataflow operator in the array of processing elements, and the array of processing elements is to perform a second operation when an incoming operand set arrives at the array of processing elements. The array of processing element may not perform the second operation until the incoming operand set arrives at the array of processing elements and storage in the array of processing elements is available for output of the second operation. The array of processing elements may include a network (or channel(s)) to carry dataflow tokens and control tokens to a plurality of dataflow operators. The second operation may include a memory access and the array of processing elements may include a memory-accessing dataflow operator that is not to perform the memory access until receiving a memory dependency token from a logically previous dataflow operator. Each processing element may perform only one or two operations of the dataflow graph.
In another embodiment, a method includes decoding an instruction with a decoder of a core of a processor into a decoded instruction; executing the decoded instruction with an execution unit of the core of the processor to perform a first operation; receiving an input of a dataflow graph comprising a plurality of nodes; overlaying the dataflow graph into an array of processing elements of the processor with each node represented as a dataflow operator in the array of processing elements; and performing a second operation of the dataflow graph with the array of processing elements when an incoming operand set arrives at the array of processing elements. The array of processing elements may not perform the second operation until the incoming operand set arrives at the array of processing elements and storage in the array of processing elements is available for output of the second operation. The array of processing elements may include a network carrying dataflow tokens and control tokens to a plurality of dataflow operators. The second operation may include a memory access and the array of processing elements comprises a memory-accessing dataflow operator that is not to perform the memory access until receiving a memory dependency token from a logically previous dataflow operator. Each processing element may performs only one or two operations of the dataflow graph.
In yet another embodiment, a non-transitory machine readable medium that stores code that when executed by a machine causes the machine to perform a method including decoding an instruction with a decoder of a core of a processor into a decoded instruction; executing the decoded instruction with an execution unit of the core of the processor to perform a first operation; receiving an input of a dataflow graph comprising a plurality of nodes; overlaying the dataflow graph into an array of processing elements of the processor with each node represented as a dataflow operator in the array of processing elements; and performing a second operation of the dataflow graph with the array of processing elements when an incoming operand set arrives at the array of processing elements. The array of processing element may not perform the second operation until the incoming operand set arrives at the array of processing elements and storage in the array of processing elements is available for output of the second operation. The array of processing elements may include a network carrying dataflow tokens and control tokens to a plurality of dataflow operators. The second operation may include a memory access and the array of processing elements comprises a memory-accessing dataflow operator that is not to perform the memory access until receiving a memory dependency token from a logically previous dataflow operator. Each processing element may performs only one or two operations of the dataflow graph.
In another embodiment, a processor includes a core with a decoder to decode an instruction into a decoded instruction and an execution unit to execute the decoded instruction to perform a first operation; and means to receive an input of a dataflow graph comprising a plurality of nodes, wherein the dataflow graph is to be overlaid into the means with each node represented as a dataflow operator in the means, and the means is to perform a second operation when an incoming operand set arrives at the means.
In one embodiment, a processor includes a core with a decoder to decode an instruction into a decoded instruction and an execution unit to execute the decoded instruction to perform a first operation; a plurality of processing elements; and an interconnect network between the plurality of processing elements to receive an input of a dataflow graph comprising a plurality of nodes, wherein the dataflow graph is to be overlaid into the interconnect network and the plurality of processing elements with each node represented as a dataflow operator in the plurality of processing elements, and the plurality of processing elements is to perform a second operation when an incoming operand set arrives at the plurality of processing elements. The processor may further comprise a plurality of configuration controllers, each configuration controller is coupled to a respective subset of the plurality of processing elements, and each configuration controller is to load configuration information from storage and cause coupling of the respective subset of the plurality of processing elements according to the configuration information. The processor may include a plurality of configuration caches, and each configuration controller is coupled to a respective configuration cache to fetch the configuration information for the respective subset of the plurality of processing elements. The first operation performed by the execution unit may prefetch configuration information into each of the plurality of configuration caches. Each of the plurality of configuration controllers may include a reconfiguration circuit to cause a reconfiguration for at least one processing element of the respective subset of the plurality of processing elements on receipt of a configuration error message from the at least one processing element. Each of the plurality of configuration controllers may a reconfiguration circuit to cause a reconfiguration for the respective subset of the plurality of processing elements on receipt of a reconfiguration request message, and disable communication with the respective subset of the plurality of processing elements until the reconfiguration is complete. The processor may include a plurality of exception aggregators, and each exception aggregator is coupled to a respective subset of the plurality of processing elements to collect exceptions from the respective subset of the plurality of processing elements and forward the exceptions to the core for servicing. The processor may include a plurality of extraction controllers, each extraction controller is coupled to a respective subset of the plurality of processing elements, and each extraction controller is to cause state data from the respective subset of the plurality of processing elements to be saved to memory.
In another embodiment, a method includes decoding an instruction with a decoder of a core of a processor into a decoded instruction; executing the decoded instruction with an execution unit of the core of the processor to perform a first operation; receiving an input of a dataflow graph comprising a plurality of nodes; overlaying the dataflow graph into a plurality of processing elements of the processor and an interconnect network between the plurality of processing elements of the processor with each node represented as a dataflow operator in the plurality of processing elements; and performing a second operation of the dataflow graph with the interconnect network and the plurality of processing elements when an incoming operand set arrives at the plurality of processing elements. The method may include loading configuration information from storage for respective subsets of the plurality of processing elements and causing coupling for each respective subset of the plurality of processing elements according to the configuration information. The method may include fetching the configuration information for the respective subset of the plurality of processing elements from a respective configuration cache of a plurality of configuration caches. The first operation performed by the execution unit may be prefetching configuration information into each of the plurality of configuration caches. The method may include causing a reconfiguration for at least one processing element of the respective subset of the plurality of processing elements on receipt of a configuration error message from the at least one processing element. The method may include causing a reconfiguration for the respective subset of the plurality of processing elements on receipt of a reconfiguration request message; and disabling communication with the respective subset of the plurality of processing elements until the reconfiguration is complete. The method may include collecting exceptions from a respective subset of the plurality of processing elements; and forwarding the exceptions to the core for servicing. The method may include causing state data from a respective subset of the plurality of processing elements to be saved to memory.
In yet another embodiment, a non-transitory machine readable medium that stores code that when executed by a machine causes the machine to perform a method including decoding an instruction with a decoder of a core of a processor into a decoded instruction; executing the decoded instruction with an execution unit of the core of the processor to perform a first operation; receiving an input of a dataflow graph comprising a plurality of nodes; overlaying the dataflow graph into a plurality of processing elements of the processor and an interconnect network between the plurality of processing elements of the processor with each node represented as a dataflow operator in the plurality of processing elements; and performing a second operation of the dataflow graph with the interconnect network and the plurality of processing elements when an incoming operand set arrives at the plurality of processing elements. The method may include loading configuration information from storage for respective subsets of the plurality of processing elements and causing coupling for each respective subset of the plurality of processing elements according to the configuration information. The method may include fetching the configuration information for the respective subset of the plurality of processing elements from a respective configuration cache of a plurality of configuration caches. The first operation performed by the execution unit may be prefetching configuration information into each of the plurality of configuration caches. The method may include causing a reconfiguration for at least one processing element of the respective subset of the plurality of processing elements on receipt of a configuration error message from the at least one processing element. The method may include causing a reconfiguration for the respective subset of the plurality of processing elements on receipt of a reconfiguration request message; and disabling communication with the respective subset of the plurality of processing elements until the reconfiguration is complete. The method may include collecting exceptions from a respective subset of the plurality of processing elements; and forwarding the exceptions to the core for servicing. The method may include causing state data from a respective subset of the plurality of processing elements to be saved to memory.
In another embodiment, a processor includes a core with a decoder to decode an instruction into a decoded instruction and an execution unit to execute the decoded instruction to perform a first operation; a plurality of processing elements; and means between the plurality of processing elements to receive an input of a dataflow graph comprising a plurality of nodes, wherein the dataflow graph is to be overlaid into the m and the plurality of processing elements with each node represented as a dataflow operator in the plurality of processing elements, and the plurality of processing elements is to perform a second operation when an incoming operand set arrives at the plurality of processing elements.
In one embodiment, an apparatus (e.g., a processor) includes: a spatial array of processing elements comprising a communications network to receive an input of a dataflow graph comprising a plurality of nodes, wherein the dataflow graph is to be overlaid into the spatial array of processing elements with each node represented as a dataflow operator in the spatial array of processing elements, and the spatial array of processing elements is to perform an operation by a respective, incoming operand set arriving at each of the dataflow operators; a plurality of request address file circuits coupled to the spatial array of processing elements and a cache memory, each request address file circuit of the plurality of request address file circuits to access data in the cache memory in response to a request for data access from the spatial array of processing elements; a plurality of translation lookaside buffers comprising a translation lookaside buffer in each of the plurality of request address file circuits to provide an output of a physical address for an input of a virtual address; and a translation lookaside buffer manager circuit comprising a higher level translation lookaside buffer than the plurality of translation lookaside buffers, the translation lookaside buffer manager circuit to perform a first page walk in the cache memory for a miss of an input of a virtual address into a first translation lookaside buffer and into the higher level translation lookaside buffer to determine a physical address mapped to the virtual address, store a mapping of the virtual address to the physical address from the first page walk in the higher level translation lookaside buffer to cause the higher level translation lookaside buffer to send the physical address to the first translation lookaside buffer in a first request address file circuit. The translation lookaside buffer manager circuit may simultaneously, with the first page walk, perform a second page walk in the cache memory, wherein the second page walk is for a miss of an input of a virtual address into a second translation lookaside buffer and into the higher level translation lookaside buffer to determine a physical address mapped to the virtual address, store a mapping of the virtual address to the physical address from the second page walk in the higher level translation lookaside buffer to cause the higher level translation lookaside buffer to send the physical address to the second translation lookaside buffer in a second request address file circuit. The receipt of the physical address in the first translation lookaside buffer may cause the first request address file circuit to perform a data access for the request for data access from the spatial array of processing elements on the physical address in the cache memory. The translation lookaside buffer manager circuit may insert an indicator in the higher level translation lookaside buffer for the miss of the input of the virtual address in the first translation lookaside buffer and the higher level translation lookaside buffer to prevent an additional page walk for the input of the virtual address during the first page walk. The translation lookaside buffer manager circuit may receive a shootdown message from a requesting entity for a mapping of a physical address to a virtual address, invalidate the mapping in the higher level translation lookaside buffer, and send shootdown messages to only those of the plurality of request address file circuits that include a copy of the mapping in a respective translation lookaside buffer, wherein each of those of the plurality of request address file circuits are to send an acknowledgement message to the translation lookaside buffer manager circuit, and the translation lookaside buffer manager circuit is to send a shootdown completion acknowledgment message to the requesting entity when all acknowledgement messages are received. The translation lookaside buffer manager circuit may receive a shootdown message from a requesting entity for a mapping of a physical address to a virtual address, invalidate the mapping in the higher level translation lookaside buffer, and send shootdown messages to all of the plurality of request address file circuits, wherein each of the plurality of request address file circuits are to send an acknowledgement message to the translation lookaside buffer manager circuit, and the translation lookaside buffer manager circuit is to send a shootdown completion acknowledgment message to the requesting entity when all acknowledgement messages are received.
In another embodiment, a method includes overlaying an input of a dataflow graph comprising a plurality of nodes into a spatial array of processing elements comprising a communications network with each node represented as a dataflow operator in the spatial array of processing elements; coupling a plurality of request address file circuits to the spatial array of processing elements and a cache memory with each request address file circuit of the plurality of request address file circuits accessing data in the cache memory in response to a request for data access from the spatial array of processing elements; providing an output of a physical address for an input of a virtual address into a translation lookaside buffer of a plurality of translation lookaside buffers comprising a translation lookaside buffer in each of the plurality of request address file circuits; coupling a translation lookaside buffer manager circuit comprising a higher level translation lookaside buffer than the plurality of translation lookaside buffers to the plurality of request address file circuits and the cache memory; and performing a first page walk in the cache memory for a miss of an input of a virtual address into a first translation lookaside buffer and into the higher level translation lookaside buffer with the translation lookaside buffer manager circuit to determine a physical address mapped to the virtual address, store a mapping of the virtual address to the physical address from the first page walk in the higher level translation lookaside buffer to cause the higher level translation lookaside buffer to send the physical address to the first translation lookaside buffer in a first request address file circuit. The method may include simultaneously, with the first page walk, performing a second page walk in the cache memory with the translation lookaside buffer manager circuit, wherein the second page walk is for a miss of an input of a virtual address into a second translation lookaside buffer and into the higher level translation lookaside buffer to determine a physical address mapped to the virtual address, and storing a mapping of the virtual address to the physical address from the second page walk in the higher level translation lookaside buffer to cause the higher level translation lookaside buffer to send the physical address to the second translation lookaside buffer in a second request address file circuit. The method may include causing the first request address file circuit to perform a data access for the request for data access from the spatial array of processing elements on the physical address in the cache memory in response to receipt of the physical address in the first translation lookaside buffer. The method may include inserting, with the translation lookaside buffer manager circuit, an indicator in the higher level translation lookaside buffer for the miss of the input of the virtual address in the first translation lookaside buffer and the higher level translation lookaside buffer to prevent an additional page walk for the input of the virtual address during the first page walk. The method may include receiving, with the translation lookaside buffer manager circuit, a shootdown message from a requesting entity for a mapping of a physical address to a virtual address, invalidating the mapping in the higher level translation lookaside buffer, and sending shootdown messages to only those of the plurality of request address file circuits that include a copy of the mapping in a respective translation lookaside buffer, wherein each of those of the plurality of request address file circuits are to send an acknowledgement message to the translation lookaside buffer manager circuit, and the translation lookaside buffer manager circuit is to send a shootdown completion acknowledgment message to the requesting entity when all acknowledgement messages are received. The method may include receiving, with the translation lookaside buffer manager circuit, a shootdown message from a requesting entity for a mapping of a physical address to a virtual address, invalidate the mapping in the higher level translation lookaside buffer, and sending shootdown messages to all of the plurality of request address file circuits, wherein each of the plurality of request address file circuits are to send an acknowledgement message to the translation lookaside buffer manager circuit, and the translation lookaside buffer manager circuit is to send a shootdown completion acknowledgment message to the requesting entity when all acknowledgement messages are received.
In another embodiment, an apparatus includes a spatial array of processing elements comprising a communications network to receive an input of a dataflow graph comprising a plurality of nodes, wherein the dataflow graph is to be overlaid into the spatial array of processing elements with each node represented as a dataflow operator in the spatial array of processing elements, and the spatial array of processing elements is to perform an operation by a respective, incoming operand set arriving at each of the dataflow operators; a plurality of request address file circuits coupled to the spatial array of processing elements and a plurality of cache memory banks, each request address file circuit of the plurality of request address file circuits to access data in (e.g., each of) the plurality of cache memory banks in response to a request for data access from the spatial array of processing elements; a plurality of translation lookaside buffers comprising a translation lookaside buffer in each of the plurality of request address file circuits to provide an output of a physical address for an input of a virtual address; a plurality of higher level, than the plurality of translation lookaside buffers, translation lookaside buffers comprising a higher level translation lookaside buffer in each of the plurality of cache memory banks to provide an output of a physical address for an input of a virtual address; and a translation lookaside buffer manager circuit to perform a first page walk in the plurality of cache memory banks for a miss of an input of a virtual address into a first translation lookaside buffer and into a first higher level translation lookaside buffer to determine a physical address mapped to the virtual address, store a mapping of the virtual address to the physical address from the first page walk in the first higher level translation lookaside buffer to cause the first higher level translation lookaside buffer to send the physical address to the first translation lookaside buffer in a first request address file circuit. The translation lookaside buffer manager circuit may simultaneously, with the first page walk, perform a second page walk in the plurality of cache memory banks, wherein the second page walk is for a miss of an input of a virtual address into a second translation lookaside buffer and into a second higher level translation lookaside buffer to determine a physical address mapped to the virtual address, store a mapping of the virtual address to the physical address from the second page walk in the second higher level translation lookaside buffer to cause the second higher level translation lookaside buffer to send the physical address to the second translation lookaside buffer in a second request address file circuit. The receipt of the physical address in the first translation lookaside buffer may cause the first request address file circuit to perform a data access for the request for data access from the spatial array of processing elements on the physical address in the plurality of cache memory banks. The translation lookaside buffer manager circuit may insert an indicator in the first higher level translation lookaside buffer for the miss of the input of the virtual address in the first translation lookaside buffer and the first higher level translation lookaside buffer to prevent an additional page walk for the input of the virtual address during the first page walk. The translation lookaside buffer manager circuit may receive a shootdown message from a requesting entity for a mapping of a physical address to a virtual address, invalidate the mapping in a higher level translation lookaside buffer storing the mapping, and send shootdown messages to only those of the plurality of request address file circuits that include a copy of the mapping in a respective translation lookaside buffer, wherein each of those of the plurality of request address file circuits are to send an acknowledgement message to the translation lookaside buffer manager circuit, and the translation lookaside buffer manager circuit is to send a shootdown completion acknowledgment message to the requesting entity when all acknowledgement messages are received. The translation lookaside buffer manager circuit may receive a shootdown message from a requesting entity for a mapping of a physical address to a virtual address, invalidate the mapping in a higher level translation lookaside buffer storing the mapping, and send shootdown messages to all of the plurality of request address file circuits, wherein each of the plurality of request address file circuits are to send an acknowledgement message to the translation lookaside buffer manager circuit, and the translation lookaside buffer manager circuit is to send a shootdown completion acknowledgment message to the requesting entity when all acknowledgement messages are received.
In yet another embodiment, a method includes: overlaying an input of a dataflow graph comprising a plurality of nodes into a spatial array of processing elements comprising a communications network with each node represented as a dataflow operator in the spatial array of processing elements; coupling a plurality of request address file circuits to the spatial array of processing elements and a plurality of cache memory banks with each request address file circuit of the plurality of request address file circuits accessing data in the plurality of cache memory banks in response to a request for data access from the spatial array of processing elements; providing an output of a physical address for an input of a virtual address into a translation lookaside buffer of a plurality of translation lookaside buffers comprising a translation lookaside buffer in each of the plurality of request address file circuits; providing an output of a physical address for an input of a virtual address into a higher level, than the plurality of translation lookaside buffers, translation lookaside buffer of a plurality of higher level translation lookaside buffers comprising a higher level translation lookaside buffer in each of the plurality of cache memory banks; coupling a translation lookaside buffer manager circuit to the plurality of request address file circuits and the plurality of cache memory banks; and performing a first page walk in the plurality of cache memory banks for a miss of an input of a virtual address into a first translation lookaside buffer and into a first higher level translation lookaside buffer with the translation lookaside buffer manager circuit to determine a physical address mapped to the virtual address, store a mapping of the virtual address to the physical address from the first page walk in the first higher level translation lookaside buffer to cause the first higher level translation lookaside buffer to send the physical address to the first translation lookaside buffer in a first request address file circuit. The method may include simultaneously, with the first page walk, performing a second page walk in the plurality of cache memory banks with the translation lookaside buffer manager circuit, wherein the second page walk is for a miss of an input of a virtual address into a second translation lookaside buffer and into a second higher level translation lookaside buffer to determine a physical address mapped to the virtual address, and storing a mapping of the virtual address to the physical address from the second page walk in the second higher level translation lookaside buffer to cause the second higher level translation lookaside buffer to send the physical address to the second translation lookaside buffer in a second request address file circuit. The method may include causing the first request address file circuit to perform a data access for the request for data access from the spatial array of processing elements on the physical address in the plurality of cache memory banks in response to receipt of the physical address in the first translation lookaside buffer. The method may include inserting, with the translation lookaside buffer manager circuit, an indicator in the first higher level translation lookaside buffer for the miss of the input of the virtual address in the first translation lookaside buffer and the first higher level translation lookaside buffer to prevent an additional page walk for the input of the virtual address during the first page walk. The method may include receiving, with the translation lookaside buffer manager circuit, a shootdown message from a requesting entity for a mapping of a physical address to a virtual address, invalidating the mapping in a higher level translation lookaside buffer storing the mapping, and sending shootdown messages to only those of the plurality of request address file circuits that include a copy of the mapping in a respective translation lookaside buffer, wherein each of those of the plurality of request address file circuits are to send an acknowledgement message to the translation lookaside buffer manager circuit, and the translation lookaside buffer manager circuit is to send a shootdown completion acknowledgment message to the requesting entity when all acknowledgement messages are received. The method may include receiving, with the translation lookaside buffer manager circuit, a shootdown message from a requesting entity for a mapping of a physical address to a virtual address, invalidate the mapping in a higher level translation lookaside buffer storing the mapping, and sending shootdown messages to all of the plurality of request address file circuits, wherein each of the plurality of request address file circuits are to send an acknowledgement message to the translation lookaside buffer manager circuit, and the translation lookaside buffer manager circuit is to send a shootdown completion acknowledgment message to the requesting entity when all acknowledgement messages are received.
In another embodiment, a system includes a core with a decoder to decode an instruction into a decoded instruction and an execution unit to execute the decoded instruction to perform a first operation; a spatial array of processing elements comprising a communications network to receive an input of a dataflow graph comprising a plurality of nodes, wherein the dataflow graph is to be overlaid into the spatial array of processing elements with each node represented as a dataflow operator in the spatial array of processing elements, and the spatial array of processing elements is to perform a second operation by a respective, incoming operand set arriving at each of the dataflow operators; a plurality of request address file circuits coupled to the spatial array of processing elements and a cache memory, each request address file circuit of the plurality of request address file circuits to access data in the cache memory in response to a request for data access from the spatial array of processing elements; a plurality of translation lookaside buffers comprising a translation lookaside buffer in each of the plurality of request address file circuits to provide an output of a physical address for an input of a virtual address; and a translation lookaside buffer manager circuit comprising a higher level translation lookaside buffer than the plurality of translation lookaside buffers, the translation lookaside buffer manager circuit to perform a first page walk in the cache memory for a miss of an input of a virtual address into a first translation lookaside buffer and into the higher level translation lookaside buffer to determine a physical address mapped to the virtual address, store a mapping of the virtual address to the physical address from the first page walk in the higher level translation lookaside buffer to cause the higher level translation lookaside buffer to send the physical address to the first translation lookaside buffer in a first request address file circuit. The translation lookaside buffer manager circuit may simultaneously, with the first page walk, perform a second page walk in the cache memory, wherein the second page walk is for a miss of an input of a virtual address into a second translation lookaside buffer and into the higher level translation lookaside buffer to determine a physical address mapped to the virtual address, store a mapping of the virtual address to the physical address from the second page walk in the higher level translation lookaside buffer to cause the higher level translation lookaside buffer to send the physical address to the second translation lookaside buffer in a second request address file circuit. The receipt of the physical address in the first translation lookaside buffer may cause the first request address file circuit to perform a data access for the request for data access from the spatial array of processing elements on the physical address in the cache memory. The translation lookaside buffer manager circuit may insert an indicator in the higher level translation lookaside buffer for the miss of the input of the virtual address in the first translation lookaside buffer and the higher level translation lookaside buffer to prevent an additional page walk for the input of the virtual address during the first page walk. The translation lookaside buffer manager circuit may receive a shootdown message from a requesting entity for a mapping of a physical address to a virtual address, invalidate the mapping in the higher level translation lookaside buffer, and send shootdown messages to only those of the plurality of request address file circuits that include a copy of the mapping in a respective translation lookaside buffer, wherein each of those of the plurality of request address file circuits are to send an acknowledgement message to the translation lookaside buffer manager circuit, and the translation lookaside buffer manager circuit is to send a shootdown completion acknowledgment message to the requesting entity when all acknowledgement messages are received. The translation lookaside buffer manager circuit may receive a shootdown message from a requesting entity for a mapping of a physical address to a virtual address, invalidate the mapping in the higher level translation lookaside buffer, and send shootdown messages to all of the plurality of request address file circuits, wherein each of the plurality of request address file circuits are to send an acknowledgement message to the translation lookaside buffer manager circuit, and the translation lookaside buffer manager circuit is to send a shootdown completion acknowledgment message to the requesting entity when all acknowledgement messages are received.
In yet another embodiment, a system includes a core with a decoder to decode an instruction into a decoded instruction and an execution unit to execute the decoded instruction to perform a first operation; a spatial array of processing elements comprising a communications network to receive an input of a dataflow graph comprising a plurality of nodes, wherein the dataflow graph is to be overlaid into the spatial array of processing elements with each node represented as a dataflow operator in the spatial array of processing elements, and the spatial array of processing elements is to perform a second operation by a respective, incoming operand set arriving at each of the dataflow operators; a plurality of request address file circuits coupled to the spatial array of processing elements and a plurality of cache memory banks, each request address file circuit of the plurality of request address file circuits to access data in (e.g., each of) the plurality of cache memory banks in response to a request for data access from the spatial array of processing elements; a plurality of translation lookaside buffers comprising a translation lookaside buffer in each of the plurality of request address file circuits to provide an output of a physical address for an input of a virtual address; a plurality of higher level, than the plurality of translation lookaside buffers, translation lookaside buffers comprising a higher level translation lookaside buffer in each of the plurality of cache memory banks to provide an output of a physical address for an input of a virtual address; and a translation lookaside buffer manager circuit to perform a first page walk in the plurality of cache memory banks for a miss of an input of a virtual address into a first translation lookaside buffer and into a first higher level translation lookaside buffer to determine a physical address mapped to the virtual address, store a mapping of the virtual address to the physical address from the first page walk in the first higher level translation lookaside buffer to cause the first higher level translation lookaside buffer to send the physical address to the first translation lookaside buffer in a first request address file circuit. The translation lookaside buffer manager circuit may simultaneously, with the first page walk, perform a second page walk in the plurality of cache memory banks, wherein the second page walk is for a miss of an input of a virtual address into a second translation lookaside buffer and into a second higher level translation lookaside buffer to determine a physical address mapped to the virtual address, store a mapping of the virtual address to the physical address from the second page walk in the second higher level translation lookaside buffer to cause the second higher level translation lookaside buffer to send the physical address to the second translation lookaside buffer in a second request address file circuit. The receipt of the physical address in the first translation lookaside buffer may cause the first request address file circuit to perform a data access for the request for data access from the spatial array of processing elements on the physical address in the plurality of cache memory banks. The translation lookaside buffer manager circuit may insert an indicator in the first higher level translation lookaside buffer for the miss of the input of the virtual address in the first translation lookaside buffer and the first higher level translation lookaside buffer to prevent an additional page walk for the input of the virtual address during the first page walk. The translation lookaside buffer manager circuit may receive a shootdown message from a requesting entity for a mapping of a physical address to a virtual address, invalidate the mapping in a higher level translation lookaside buffer storing the mapping, and send shootdown messages to only those of the plurality of request address file circuits that include a copy of the mapping in a respective translation lookaside buffer, wherein each of those of the plurality of request address file circuits are to send an acknowledgement message to the translation lookaside buffer manager circuit, and the translation lookaside buffer manager circuit is to send a shootdown completion acknowledgment message to the requesting entity when all acknowledgement messages are received. The translation lookaside buffer manager circuit may receive a shootdown message from a requesting entity for a mapping of a physical address to a virtual address, invalidate the mapping in a higher level translation lookaside buffer storing the mapping, and send shootdown messages to all of the plurality of request address file circuits, wherein each of the plurality of request address file circuits are to send an acknowledgement message to the translation lookaside buffer manager circuit, and the translation lookaside buffer manager circuit is to send a shootdown completion acknowledgment message to the requesting entity when all acknowledgement messages are received.
In another embodiment, an apparatus (e.g., a processor) includes: a spatial array of processing elements comprising a communications network to receive an input of a dataflow graph comprising a plurality of nodes, wherein the dataflow graph is to be overlaid into the spatial array of processing elements with each node represented as a dataflow operator in the spatial array of processing elements, and the spatial array of processing elements is to perform an operation by a respective, incoming operand set arriving at each of the dataflow operators; a plurality of request address file circuits coupled to the spatial array of processing elements and a cache memory, each request address file circuit of the plurality of request address file circuits to access data in the cache memory in response to a request for data access from the spatial array of processing elements; a plurality of translation lookaside buffers comprising a translation lookaside buffer in each of the plurality of request address file circuits to provide an output of a physical address for an input of a virtual address; and a means comprising a higher level translation lookaside buffer than the plurality of translation lookaside buffers, the means to perform a first page walk in the cache memory for a miss of an input of a virtual address into a first translation lookaside buffer and into the higher level translation lookaside buffer to determine a physical address mapped to the virtual address, store a mapping of the virtual address to the physical address from the first page walk in the higher level translation lookaside buffer to cause the higher level translation lookaside buffer to send the physical address to the first translation lookaside buffer in a first request address file circuit.
In yet another embodiment, an apparatus includes a spatial array of processing elements comprising a communications network to receive an input of a dataflow graph comprising a plurality of nodes, wherein the dataflow graph is to be overlaid into the spatial array of processing elements with each node represented as a dataflow operator in the spatial array of processing elements, and the spatial array of processing elements is to perform an operation by a respective, incoming operand set arriving at each of the dataflow operators; a plurality of request address file circuits coupled to the spatial array of processing elements and a plurality of cache memory banks, each request address file circuit of the plurality of request address file circuits to access data in (e.g., each of) the plurality of cache memory banks in response to a request for data access from the spatial array of processing elements; a plurality of translation lookaside buffers comprising a translation lookaside buffer in each of the plurality of request address file circuits to provide an output of a physical address for an input of a virtual address; a plurality of higher level, than the plurality of translation lookaside buffers, translation lookaside buffers comprising a higher level translation lookaside buffer in each of the plurality of cache memory banks to provide an output of a physical address for an input of a virtual address; and a means to perform a first page walk in the plurality of cache memory banks for a miss of an input of a virtual address into a first translation lookaside buffer and into a first higher level translation lookaside buffer to determine a physical address mapped to the virtual address, store a mapping of the virtual address to the physical address from the first page walk in the first higher level translation lookaside buffer to cause the first higher level translation lookaside buffer to send the physical address to the first translation lookaside buffer in a first request address file circuit.
In another embodiment, an apparatus comprises a data storage device that stores code that when executed by a hardware processor causes the hardware processor to perform any method disclosed herein. An apparatus may be as described in the detailed description. A method may be as described in the detailed description.
In yet another embodiment, a non-transitory machine readable medium that stores code that when executed by a machine causes the machine to perform a method comprising any method disclosed herein.
An instruction set (e.g., for execution by a core) may include one or more instruction formats. A given instruction format may define various fields (e.g., number of bits, location of bits) to specify, among other things, the operation to be performed (e.g., opcode) and the operand(s) on which that operation is to be performed and/or other data field(s) (e.g., mask). Some instruction formats are further broken down though the definition of instruction templates (or subformats). For example, the instruction templates of a given instruction format may be defined to have different subsets of the instruction format's fields (the included fields are typically in the same order, but at least some have different bit positions because there are less fields included) and/or defined to have a given field interpreted differently. Thus, each instruction of an ISA is expressed using a given instruction format (and, if defined, in a given one of the instruction templates of that instruction format) and includes fields for specifying the operation and the operands. For example, an exemplary ADD instruction has a specific opcode and an instruction format that includes an opcode field to specify that opcode and operand fields to select operands (source1/destination and source2); and an occurrence of this ADD instruction in an instruction stream will have specific contents in the operand fields that select specific operands. A set of SIMD extensions referred to as the Advanced Vector Extensions (AVX) (AVX1 and AVX2) and using the Vector Extensions (VEX) coding scheme has been released and/or published (e.g., see Intel® 64 and IA-32 Architectures Software Developer's Manual, June 2016; and see Intel® Architecture Instruction Set Extensions Programming Reference, February 2016).
Embodiments of the instruction(s) described herein may be embodied in different formats. Additionally, exemplary systems, architectures, and pipelines are detailed below. Embodiments of the instruction(s) may be executed on such systems, architectures, and pipelines, but are not limited to those detailed.
A vector friendly instruction format is an instruction format that is suited for vector instructions (e.g., there are certain fields specific to vector operations). While embodiments are described in which both vector and scalar operations are supported through the vector friendly instruction format, alternative embodiments use only vector operations the vector friendly instruction format.
While embodiments of the disclosure will be described in which the vector friendly instruction format supports the following: a 64 byte vector operand length (or size) with 32 bit (4 byte) or 64 bit (8 byte) data element widths (or sizes) (and thus, a 64 byte vector consists of either 16 doubleword-size elements or alternatively, 8 quadword-size elements); a 64 byte vector operand length (or size) with 16 bit (2 byte) or 8 bit (1 byte) data element widths (or sizes); a 32 byte vector operand length (or size) with 32 bit (4 byte), 64 bit (8 byte), 16 bit (2 byte), or 8 bit (1 byte) data element widths (or sizes); and a 16 byte vector operand length (or size) with 32 bit (4 byte), 64 bit (8 byte), 16 bit (2 byte), or 8 bit (1 byte) data element widths (or sizes); alternative embodiments may support more, less and/or different vector operand sizes (e.g., 256 byte vector operands) with more, less, or different data element widths (e.g., 128 bit (16 byte) data element widths).
The class A instruction templates in
The generic vector friendly instruction format 9900 includes the following fields listed below in the order illustrated in
Format field 9940—a specific value (an instruction format identifier value) in this field uniquely identifies the vector friendly instruction format, and thus occurrences of instructions in the vector friendly instruction format in instruction streams. As such, this field is optional in the sense that it is not needed for an instruction set that has only the generic vector friendly instruction format.
Base operation field 9942—its content distinguishes different base operations.
Register index field 9944—its content, directly or through address generation, specifies the locations of the source and destination operands, be they in registers or in memory. These include a sufficient number of bits to select N registers from a P×Q (e.g. 32×512, 16×128, 32×1024, 64×1024) register file. While in one embodiment N may be up to three sources and one destination register, alternative embodiments may support more or less sources and destination registers (e.g., may support up to two sources where one of these sources also acts as the destination, may support up to three sources where one of these sources also acts as the destination, may support up to two sources and one destination).
Modifier field 9946—its content distinguishes occurrences of instructions in the generic vector instruction format that specify memory access from those that do not; that is, between no memory access 9905 instruction templates and memory access 9920 instruction templates. Memory access operations read and/or write to the memory hierarchy (in some cases specifying the source and/or destination addresses using values in registers), while non-memory access operations do not (e.g., the source and destinations are registers). While in one embodiment this field also selects between three different ways to perform memory address calculations, alternative embodiments may support more, less, or different ways to perform memory address calculations.
Augmentation operation field 9950—its content distinguishes which one of a variety of different operations to be performed in addition to the base operation. This field is context specific. In one embodiment of the disclosure, this field is divided into a class field 9968, an alpha field 9952, and a beta field 9954. The augmentation operation field 9950 allows common groups of operations to be performed in a single instruction rather than 2, 3, or 4 instructions.
Scale field 9960—its content allows for the scaling of the index field's content for memory address generation (e.g., for address generation that uses 2scale*index+base).
Displacement Field 9962A—its content is used as part of memory address generation (e.g., for address generation that uses 2scale*index+base+displacement).
Displacement Factor Field 9962B (note that the juxtaposition of displacement field 9962A directly over displacement factor field 9962B indicates one or the other is used)—its content is used as part of address generation; it specifies a displacement factor that is to be scaled by the size of a memory access (N)—where N is the number of bytes in the memory access (e.g., for address generation that uses 2scale*index+base+scaled displacement). Redundant low-order bits are ignored and hence, the displacement factor field's content is multiplied by the memory operands total size (N) in order to generate the final displacement to be used in calculating an effective address. The value of N is determined by the processor hardware at runtime based on the full opcode field 9974 (described later herein) and the data manipulation field 9954C. The displacement field 9962A and the displacement factor field 9962B are optional in the sense that they are not used for the no memory access 9905 instruction templates and/or different embodiments may implement only one or none of the two.
Data element width field 9964—its content distinguishes which one of a number of data element widths is to be used (in some embodiments for all instructions; in other embodiments for only some of the instructions). This field is optional in the sense that it is not needed if only one data element width is supported and/or data element widths are supported using some aspect of the opcodes.
Write mask field 9970—its content controls, on a per data element position basis, whether that data element position in the destination vector operand reflects the result of the base operation and augmentation operation. Class A instruction templates support merging-writemasking, while class B instruction templates support both merging- and zeroing-writemasking. When merging, vector masks allow any set of elements in the destination to be protected from updates during the execution of any operation (specified by the base operation and the augmentation operation); in other one embodiment, preserving the old value of each element of the destination where the corresponding mask bit has a 0. In contrast, when zeroing vector masks allow any set of elements in the destination to be zeroed during the execution of any operation (specified by the base operation and the augmentation operation); in one embodiment, an element of the destination is set to 0 when the corresponding mask bit has a 0 value. A subset of this functionality is the ability to control the vector length of the operation being performed (that is, the span of elements being modified, from the first to the last one); however, it is not necessary that the elements that are modified be consecutive. Thus, the write mask field 9970 allows for partial vector operations, including loads, stores, arithmetic, logical, etc. While embodiments of the disclosure are described in which the write mask field's 9970 content selects one of a number of write mask registers that contains the write mask to be used (and thus the write mask field's 9970 content indirectly identifies that masking to be performed), alternative embodiments instead or additional allow the mask write field's 9970 content to directly specify the masking to be performed.
Immediate field 9972—its content allows for the specification of an immediate. This field is optional in the sense that is it not present in an implementation of the generic vector friendly format that does not support immediate and it is not present in instructions that do not use an immediate.
Class field 9968—its content distinguishes between different classes of instructions. With reference to
In the case of the non-memory access 9905 instruction templates of class A, the alpha field 9952 is interpreted as an RS field 9952A, whose content distinguishes which one of the different augmentation operation types are to be performed (e.g., round 9952A.1 and data transform 9952A.2 are respectively specified for the no memory access, round type operation 9910 and the no memory access, data transform type operation 9915 instruction templates), while the beta field 9954 distinguishes which of the operations of the specified type is to be performed. In the no memory access 9905 instruction templates, the scale field 9960, the displacement field 9962A, and the displacement scale filed 9962B are not present.
In the no memory access full round control type operation 9910 instruction template, the beta field 9954 is interpreted as a round control field 9954A, whose content(s) provide static rounding. While in the described embodiments of the disclosure the round control field 9954A includes a suppress all floating point exceptions (SAE) field 9956 and a round operation control field 9958, alternative embodiments may support may encode both these concepts into the same field or only have one or the other of these concepts/fields (e.g., may have only the round operation control field 9958).
SAE field 9956—its content distinguishes whether or not to disable the exception event reporting; when the SAE field's 9956 content indicates suppression is enabled, a given instruction does not report any kind of floating-point exception flag and does not raise any floating point exception handler.
Round operation control field 9958—its content distinguishes which one of a group of rounding operations to perform (e.g., Round-up, Round-down, Round-towards-zero and Round-to-nearest). Thus, the round operation control field 9958 allows for the changing of the rounding mode on a per instruction basis. In one embodiment of the disclosure where a processor includes a control register for specifying rounding modes, the round operation control field's 9950 content overrides that register value.
In the no memory access data transform type operation 9915 instruction template, the beta field 9954 is interpreted as a data transform field 9954B, whose content distinguishes which one of a number of data transforms is to be performed (e.g., no data transform, swizzle, broadcast).
In the case of a memory access 9920 instruction template of class A, the alpha field 9952 is interpreted as an eviction hint field 9952B, whose content distinguishes which one of the eviction hints is to be used (in
Vector memory instructions perform vector loads from and vector stores to memory, with conversion support. As with regular vector instructions, vector memory instructions transfer data from/to memory in a data element-wise fashion, with the elements that are actually transferred is dictated by the contents of the vector mask that is selected as the write mask.
Temporal data is data likely to be reused soon enough to benefit from caching. This is, however, a hint, and different processors may implement it in different ways, including ignoring the hint entirely.
Non-temporal data is data unlikely to be reused soon enough to benefit from caching in the 1st-level cache and should be given priority for eviction. This is, however, a hint, and different processors may implement it in different ways, including ignoring the hint entirely.
In the case of the instruction templates of class B, the alpha field 9952 is interpreted as a write mask control (Z) field 9952C, whose content distinguishes whether the write masking controlled by the write mask field 9970 should be a merging or a zeroing.
In the case of the non-memory access 9905 instruction templates of class B, part of the beta field 9954 is interpreted as an RL field 9957A, whose content distinguishes which one of the different augmentation operation types are to be performed (e.g., round 9957A.1 and vector length (VSIZE) 9957A.2 are respectively specified for the no memory access, write mask control, partial round control type operation 9912 instruction template and the no memory access, write mask control, VSIZE type operation 9917 instruction template), while the rest of the beta field 9954 distinguishes which of the operations of the specified type is to be performed. In the no memory access 9905 instruction templates, the scale field 9960, the displacement field 9962A, and the displacement scale filed 9962B are not present.
In the no memory access, write mask control, partial round control type operation 9910 instruction template, the rest of the beta field 9954 is interpreted as a round operation field 9959A and exception event reporting is disabled (a given instruction does not report any kind of floating-point exception flag and does not raise any floating point exception handler).
Round operation control field 9959A—just as round operation control field 9958, its content distinguishes which one of a group of rounding operations to perform (e.g., Round-up, Round-down, Round-towards-zero and Round-to-nearest). Thus, the round operation control field 9959A allows for the changing of the rounding mode on a per instruction basis. In one embodiment of the disclosure where a processor includes a control register for specifying rounding modes, the round operation control field's 9950 content overrides that register value.
In the no memory access, write mask control, VSIZE type operation 9917 instruction template, the rest of the beta field 9954 is interpreted as a vector length field 9959B, whose content distinguishes which one of a number of data vector lengths is to be performed on (e.g., 128, 256, or 512 byte).
In the case of a memory access 9920 instruction template of class B, part of the beta field 9954 is interpreted as a broadcast field 9957B, whose content distinguishes whether or not the broadcast type data manipulation operation is to be performed, while the rest of the beta field 9954 is interpreted the vector length field 9959B. The memory access 9920 instruction templates include the scale field 9960, and optionally the displacement field 9962A or the displacement scale field 9962B.
With regard to the generic vector friendly instruction format 9900, a full opcode field 9974 is shown including the format field 9940, the base operation field 9942, and the data element width field 9964. While one embodiment is shown where the full opcode field 9974 includes all of these fields, the full opcode field 9974 includes less than all of these fields in embodiments that do not support all of them. The full opcode field 9974 provides the operation code (opcode).
The augmentation operation field 9950, the data element width field 9964, and the write mask field 9970 allow these features to be specified on a per instruction basis in the generic vector friendly instruction format.
The combination of write mask field and data element width field create typed instructions in that they allow the mask to be applied based on different data element widths.
The various instruction templates found within class A and class B are beneficial in different situations. In some embodiments of the disclosure, different processors or different cores within a processor may support only class A, only class B, or both classes. For instance, a high performance general purpose out-of-order core intended for general-purpose computing may support only class B, a core intended primarily for graphics and/or scientific (throughput) computing may support only class A, and a core intended for both may support both (of course, a core that has some mix of templates and instructions from both classes but not all templates and instructions from both classes is within the purview of the disclosure). Also, a single processor may include multiple cores, all of which support the same class or in which different cores support different class. For instance, in a processor with separate graphics and general purpose cores, one of the graphics cores intended primarily for graphics and/or scientific computing may support only class A, while one or more of the general purpose cores may be high performance general purpose cores with out of order execution and register renaming intended for general-purpose computing that support only class B. Another processor that does not have a separate graphics core, may include one more general purpose in-order or out-of-order cores that support both class A and class B. Of course, features from one class may also be implement in the other class in different embodiments of the disclosure. Programs written in a high level language would be put (e.g., just in time compiled or statically compiled) into an variety of different executable forms, including: 1) a form having only instructions of the class(es) supported by the target processor for execution; or 2) a form having alternative routines written using different combinations of the instructions of all classes and having control flow code that selects the routines to execute based on the instructions supported by the processor which is currently executing the code.
It should be understood that, although embodiments of the disclosure are described with reference to the specific vector friendly instruction format 10000 in the context of the generic vector friendly instruction format 9900 for illustrative purposes, the disclosure is not limited to the specific vector friendly instruction format 10000 except where claimed. For example, the generic vector friendly instruction format 9900 contemplates a variety of possible sizes for the various fields, while the specific vector friendly instruction format 10000 is shown as having fields of specific sizes. By way of specific example, while the data element width field 9964 is illustrated as a one bit field in the specific vector friendly instruction format 10000, the disclosure is not so limited (that is, the generic vector friendly instruction format 9900 contemplates other sizes of the data element width field 9964).
The generic vector friendly instruction format 9900 includes the following fields listed below in the order illustrated in
EVEX Prefix (Bytes 0-3) 10002—is encoded in a four-byte form.
Format Field 9940 (EVEX Byte 0, bits [7:0])—the first byte (EVEX Byte 0) is the format field 9940 and it contains 0x62 (the unique value used for distinguishing the vector friendly instruction format in one embodiment of the disclosure).
The second-fourth bytes (EVEX Bytes 1-3) include a number of bit fields providing specific capability.
REX field 10005 (EVEX Byte 1, bits [7-5])—consists of a EVEX.R bit field (EVEX Byte 1, bit [7]—R), EVEX.X bit field (EVEX byte 1, bit [6]—X), and 9957BEX byte 1, bit[5]—B). The EVEX.R, EVEX.X, and EVEX.B bit fields provide the same functionality as the corresponding VEX bit fields, and are encoded using is complement form, i.e. ZMM0 is encoded as 1211B, ZMM15 is encoded as 0000B. Other fields of the instructions encode the lower three bits of the register indexes as is known in the art (rrr, xxx, and bbb), so that Rrrr, Xxxx, and Bbbb may be formed by adding EVEX.R, EVEX.X, and EVEX.B.
REX′ field 9910—this is the first part of the REX′ field 9910 and is the EVEX.R′ bit field (EVEX Byte 1, bit [4]—R′) that is used to encode either the upper 16 or lower 16 of the extended 32 register set. In one embodiment of the disclosure, this bit, along with others as indicated below, is stored in bit inverted format to distinguish (in the well-known x86 32-bit mode) from the BOUND instruction, whose real opcode byte is 62, but does not accept in the MOD R/M field (described below) the value of 11 in the MOD field; alternative embodiments of the disclosure do not store this and the other indicated bits below in the inverted format. A value of 1 is used to encode the lower 16 registers. In other words, R′Rrrr is formed by combining EVEX.R′, EVEX.R, and the other RRR from other fields.
Opcode map field 10015 (EVEX byte 1, bits [3:0]—mmmm)—its content encodes an implied leading opcode byte (0F, 0F 38, or 0F 3).
Data element width field 9964 (EVEX byte 2, bit [7]—W)—is represented by the notation EVEX.W. EVEX.W is used to define the granularity (size) of the datatype (either 32-bit data elements or 64-bit data elements).
EVEX.vvvv 10020 (EVEX Byte 2, bits [6:3]—vvvv)—the role of EVEX.vvvv may include the following: 1) EVEX.vvvv encodes the first source register operand, specified in inverted (1s complement) form and is valid for instructions with 2 or more source operands; 2) EVEX.vvvv encodes the destination register operand, specified in 1s complement form for certain vector shifts; or 3) EVEX.vvvv does not encode any operand, the field is reserved and should contain 1211b. Thus, EVEX.vvvv field 10020 encodes the 4 low-order bits of the first source register specifier stored in inverted (1s complement) form. Depending on the instruction, an extra different EVEX bit field is used to extend the specifier size to 32 registers.
EVEX.U 9968 Class field (EVEX byte 2, bit [2]—U)—If EVEX.0=0, it indicates class A or EVEX.U0; if EVEX.0=1, it indicates class B or EVEX.U1.
Prefix encoding field 10025 (EVEX byte 2, bits [1:0]—pp)—provides additional bits for the base operation field. In addition to providing support for the legacy SSE instructions in the EVEX prefix format, this also has the benefit of compacting the SIMD prefix (rather than requiring a byte to express the SIMD prefix, the EVEX prefix requires only 2 bits). In one embodiment, to support legacy SSE instructions that use a SIMD prefix (66H, F2H, F3H) in both the legacy format and in the EVEX prefix format, these legacy SIMD prefixes are encoded into the SIMD prefix encoding field; and at runtime are expanded into the legacy SIMD prefix prior to being provided to the decoder's PLA (so the PLA can execute both the legacy and EVEX format of these legacy instructions without modification). Although newer instructions could use the EVEX prefix encoding field's content directly as an opcode extension, certain embodiments expand in a similar fashion for consistency but allow for different meanings to be specified by these legacy SIMD prefixes. An alternative embodiment may redesign the PLA to support the 2 bit SIMD prefix encodings, and thus not require the expansion.
Alpha field 9952 (EVEX byte 3, bit [7]—EH; also known as EVEX.EH, EVEX.rs, EVEX.RL, EVEX.write mask control, and EVEX.N; also illustrated with α)—as previously described, this field is context specific.
Beta field 9954 (EVEX byte 3, bits [6:4]—SSS, also known as EVEX.s2_0, EVEX.r2-0, EVEX.rr1, EVEX.LL0, EVEX.LLB; also illustrated with βββ)—as previously described, this field is context specific.
REX′ field 9910—this is the remainder of the REX′ field and is the EVEX.V′ bit field (EVEX Byte 3, bit [3]—V′) that may be used to encode either the upper 16 or lower 16 of the extended 32 register set. This bit is stored in bit inverted format. A value of 1 is used to encode the lower 16 registers. In other words, V′VVVV is formed by combining EVEX.V′, EVEX.vvvv.
Write mask field 9970 (EVEX byte 3, bits [2:0]—kkk)—its content specifies the index of a register in the write mask registers as previously described. In one embodiment of the disclosure, the specific value EVEX kkk=000 has a special behavior implying no write mask is used for the particular instruction (this may be implemented in a variety of ways including the use of a write mask hardwired to all ones or hardware that bypasses the masking hardware).
Real Opcode Field 10030 (Byte 4) is also known as the opcode byte. Part of the opcode is specified in this field.
MOD R/M Field 10040 (Byte 5) includes MOD field 10042, Reg field 10044, and R/M field 10046. As previously described, the MOD field's 10042 content distinguishes between memory access and non-memory access operations. The role of Reg field 10044 can be summarized to two situations: encoding either the destination register operand or a source register operand, or be treated as an opcode extension and not used to encode any instruction operand. The role of R/M field 10046 may include the following: encoding the instruction operand that references a memory address, or encoding either the destination register operand or a source register operand.
Scale, Index, Base (SIB) Byte (Byte 6)—As previously described, the scale field's 5450 content is used for memory address generation. SIB.xxx 10054 and SIB.bbb 10056—the contents of these fields have been previously referred to with regard to the register indexes Xxxx and Bbbb.
Displacement field 9962A (Bytes 7-10)—when MOD field 10042 contains 10, bytes 7-10 are the displacement field 9962A, and it works the same as the legacy 32-bit displacement (disp32) and works at byte granularity.
Displacement factor field 9962B (Byte 7)—when MOD field 10042 contains 01, byte 7 is the displacement factor field 9962B. The location of this field is that same as that of the legacy x86 instruction set 8-bit displacement (disp8), which works at byte granularity. Since disp8 is sign extended, it can only address between −128 and 127 bytes offsets; in terms of 64 byte cache lines, disp8 uses 8 bits that can be set to only four really useful values −128, −64, 0, and 64; since a greater range is often needed, disp32 is used; however, disp32 requires 4 bytes. In contrast to disp8 and disp32, the displacement factor field 9962B is a reinterpretation of disp8; when using displacement factor field 9962B, the actual displacement is determined by the content of the displacement factor field multiplied by the size of the memory operand access (N). This type of displacement is referred to as disp8*N. This reduces the average instruction length (a single byte of used for the displacement but with a much greater range). Such compressed displacement is based on the assumption that the effective displacement is multiple of the granularity of the memory access, and hence, the redundant low-order bits of the address offset do not need to be encoded. In other words, the displacement factor field 9962B substitutes the legacy x86 instruction set 8-bit displacement. Thus, the displacement factor field 9962B is encoded the same way as an x86 instruction set 8-bit displacement (so no changes in the ModRM/SIB encoding rules) with the only exception that disp8 is overloaded to disp8*N. In other words, there are no changes in the encoding rules or encoding lengths but only in the interpretation of the displacement value by hardware (which needs to scale the displacement by the size of the memory operand to obtain a byte-wise address offset). Immediate field 9972 operates as previously described.
When U=1, the alpha field 9952 (EVEX byte 3, bit [7]—EH) is interpreted as the write mask control (Z) field 9952C. When U=1 and the MOD field 10042 contains 11 (signifying a no memory access operation), part of the beta field 9954 (EVEX byte 3, bit [4]—S0) is interpreted as the RL field 9957A; when it contains a 1 (round 9957A.1) the rest of the beta field 9954 (EVEX byte 3, bit [6-5]—S2-1) is interpreted as the round operation field 9959A, while when the RL field 9957A contains a 0 (VSIZE 9957.A2) the rest of the beta field 9954 (EVEX byte 3, bit [6-5]—S2_1) is interpreted as the vector length field 9959B (EVEX byte 3, bit [6-5]—L1-0). When U=1 and the MOD field 10042 contains 00, 01, or 10 (signifying a memory access operation), the beta field 9954 (EVEX byte 3, bits [6:4]—SSS) is interpreted as the vector length field 9959B (EVEX byte 3, bit [6-5]—L1-0) and the broadcast field 9957B (EVEX byte 3, bit [4]—B).
In other words, the vector length field 9959B selects between a maximum length and one or more other shorter lengths, where each such shorter length is half the length of the preceding length; and instructions templates without the vector length field 9959B operate on the maximum vector length. Further, in one embodiment, the class B instruction templates of the specific vector friendly instruction format 10000 operate on packed or scalar single/double-precision floating point data and packed or scalar integer data. Scalar operations are operations performed on the lowest order data element position in an zmm/ymm/xmm register; the higher order data element positions are either left the same as they were prior to the instruction or zeroed depending on the embodiment.
Write mask registers 10115—in the embodiment illustrated, there are 8 write mask registers (k0 through k7), each 64 bits in size. In an alternate embodiment, the write mask registers 10115 are 16 bits in size. As previously described, in one embodiment of the disclosure, the vector mask register k0 cannot be used as a write mask; when the encoding that would normally indicate k0 is used for a write mask, it selects a hardwired write mask of 0xFFFF, effectively disabling write masking for that instruction.
General-purpose registers 10125—in the embodiment illustrated, there are sixteen 64-bit general-purpose registers that are used along with the existing x86 addressing modes to address memory operands. These registers are referenced by the names RAX, RBX, RCX, RDX, RBP, RSI, RDI, RSP, and R8 through R15.
Scalar floating point stack register file (x87 stack) 10145, on which is aliased the MMX packed integer flat register file 10150—in the embodiment illustrated, the x87 stack is an eight-element stack used to perform scalar floating-point operations on 32/64/80-bit floating point data using the x87 instruction set extension; while the MMX registers are used to perform operations on 64-bit packed integer data, as well as to hold operands for some operations performed between the MMX and XMM registers.
Alternative embodiments of the disclosure may use wider or narrower registers. Additionally, alternative embodiments of the disclosure may use more, less, or different register files and registers.
Processor cores may be implemented in different ways, for different purposes, and in different processors. For instance, implementations of such cores may include: 1) a general purpose in-order core intended for general-purpose computing; 2) a high performance general purpose out-of-order core intended for general-purpose computing; 3) a special purpose core intended primarily for graphics and/or scientific (throughput) computing. Implementations of different processors may include: 1) a CPU including one or more general purpose in-order cores intended for general-purpose computing and/or one or more general purpose out-of-order cores intended for general-purpose computing; and 2) a coprocessor including one or more special purpose cores intended primarily for graphics and/or scientific (throughput). Such different processors lead to different computer system architectures, which may include: 1) the coprocessor on a separate chip from the CPU; 2) the coprocessor on a separate die in the same package as a CPU; 3) the coprocessor on the same die as a CPU (in which case, such a coprocessor is sometimes referred to as special purpose logic, such as integrated graphics and/or scientific (throughput) logic, or as special purpose cores); and 4) a system on a chip that may include on the same die the described CPU (sometimes referred to as the application core(s) or application processor(s)), the above described coprocessor, and additional functionality. Exemplary core architectures are described next, followed by descriptions of exemplary processors and computer architectures.
In
The front end unit 10230 includes a branch prediction unit 10232 coupled to an instruction cache unit 10234, which is coupled to an instruction translation lookaside buffer (TLB) 10236, which is coupled to an instruction fetch unit 10238, which is coupled to a decode unit 10240. The decode unit 10240 (or decoder or decoder unit) may decode instructions (e.g., macro-instructions), and generate as an output one or more micro-operations, micro-code entry points, micro-instructions, other instructions, or other control signals, which are decoded from, or which otherwise reflect, or are derived from, the original instructions. The decode unit 10240 may be implemented using various different mechanisms. Examples of suitable mechanisms include, but are not limited to, look-up tables, hardware implementations, programmable logic arrays (PLAs), microcode read only memories (ROMs), etc. In one embodiment, the core 10290 includes a microcode ROM or other medium that stores microcode for certain macro-instructions (e.g., in decode unit 10240 or otherwise within the front end unit 10230). The decode unit 10240 is coupled to a rename/allocator unit 10252 in the execution engine unit 10250.
The execution engine unit 10250 includes the rename/allocator unit 10252 coupled to a retirement unit 10254 and a set of one or more scheduler unit(s) 10256. The scheduler unit(s) 10256 represents any number of different schedulers, including reservations stations, central instruction window, etc. The scheduler unit(s) 10256 is coupled to the physical register file(s) unit(s) 10258. Each of the physical register file(s) units 10258 represents one or more physical register files, different ones of which store one or more different data types, such as scalar integer, scalar floating point, packed integer, packed floating point, vector integer, vector floating point, status (e.g., an instruction pointer that is the address of the next instruction to be executed), etc. In one embodiment, the physical register file(s) unit 10258 comprises a vector registers unit, a write mask registers unit, and a scalar registers unit. These register units may provide architectural vector registers, vector mask registers, and general purpose registers. The physical register file(s) unit(s) 10258 is overlapped by the retirement unit 10254 to illustrate various ways in which register renaming and out-of-order execution may be implemented (e.g., using a reorder buffer(s) and a retirement register file(s); using a future file(s), a history buffer(s), and a retirement register file(s); using a register maps and a pool of registers; etc.). The retirement unit 10254 and the physical register file(s) unit(s) 10258 are coupled to the execution cluster(s) 10260. The execution cluster(s) 10260 includes a set of one or more execution units 10262 and a set of one or more memory access units 10264. The execution units 10262 may perform various operations (e.g., shifts, addition, subtraction, multiplication) and on various types of data (e.g., scalar floating point, packed integer, packed floating point, vector integer, vector floating point). While some embodiments may include a number of execution units dedicated to specific functions or sets of functions, other embodiments may include only one execution unit or multiple execution units that all perform all functions. The scheduler unit(s) 10256, physical register file(s) unit(s) 10258, and execution cluster(s) 10260 are shown as being possibly plural because certain embodiments create separate pipelines for certain types of data/operations (e.g., a scalar integer pipeline, a scalar floating point/packed integer/packed floating point/vector integer/vector floating point pipeline, and/or a memory access pipeline that each have their own scheduler unit, physical register file(s) unit, and/or execution cluster—and in the case of a separate memory access pipeline, certain embodiments are implemented in which only the execution cluster of this pipeline has the memory access unit(s) 10264). It should also be understood that where separate pipelines are used, one or more of these pipelines may be out-of-order issue/execution and the rest in-order.
The set of memory access units 10264 is coupled to the memory unit 10270, which includes a data TLB unit 10272 coupled to a data cache unit 10274 coupled to a level 2 (L2) cache unit 10276. In one exemplary embodiment, the memory access units 10264 may include a load unit, a store address unit, and a store data unit, each of which is coupled to the data TLB unit 10272 in the memory unit 10270. The instruction cache unit 10234 is further coupled to a level 2 (L2) cache unit 10276 in the memory unit 10270. The L2 cache unit 10276 is coupled to one or more other levels of cache and eventually to a main memory.
By way of example, the exemplary register renaming, out-of-order issue/execution core architecture may implement the pipeline 10200 as follows: 1) the instruction fetch 10238 performs the fetch and length decoding stages 10202 and 10204; 2) the decode unit 10240 performs the decode stage 10206; 3) the rename/allocator unit 10252 performs the allocation stage 10208 and renaming stage 10210; 4) the scheduler unit(s) 10256 performs the schedule stage 10212; 5) the physical register file(s) unit(s) 10258 and the memory unit 10270 perform the register read/memory read stage 10214; the execution cluster 10260 perform the execute stage 10216; 6) the memory unit 10270 and the physical register file(s) unit(s) 10258 perform the write back/memory write stage 10218; 7) various units may be involved in the exception handling stage 10222; and 8) the retirement unit 10254 and the physical register file(s) unit(s) 10258 perform the commit stage 10224.
The core 10290 may support one or more instructions sets (e.g., the x86 instruction set (with some extensions that have been added with newer versions); the MIPS instruction set of MIPS Technologies of Sunnyvale, Calif.; the ARM instruction set (with optional additional extensions such as NEON) of ARM Holdings of Sunnyvale, Calif.), including the instruction(s) described herein. In one embodiment, the core 10290 includes logic to support a packed data instruction set extension (e.g., AVX1, AVX2), thereby allowing the operations used by many multimedia applications to be performed using packed data.
It should be understood that the core may support multithreading (executing two or more parallel sets of operations or threads), and may do so in a variety of ways including time sliced multithreading, simultaneous multithreading (where a single physical core provides a logical core for each of the threads that physical core is simultaneously multithreading), or a combination thereof (e.g., time sliced fetching and decoding and simultaneous multithreading thereafter such as in the Intel® Hyperthreading technology).
While register renaming is described in the context of out-of-order execution, it should be understood that register renaming may be used in an in-order architecture. While the illustrated embodiment of the processor also includes separate instruction and data cache units 10234/10274 and a shared L2 cache unit 10276, alternative embodiments may have a single internal cache for both instructions and data, such as, for example, a Level 1 (L1) internal cache, or multiple levels of internal cache. In some embodiments, the system may include a combination of an internal cache and an external cache that is external to the core and/or the processor. Alternatively, all of the cache may be external to the core and/or the processor.
The local subset of the L2 cache 10304 is part of a global L2 cache that is divided into separate local subsets, one per processor core. Each processor core has a direct access path to its own local subset of the L2 cache 10304. Data read by a processor core is stored in its L2 cache subset 10304 and can be accessed quickly, in parallel with other processor cores accessing their own local L2 cache subsets. Data written by a processor core is stored in its own L2 cache subset 10304 and is flushed from other subsets, if necessary. The ring network ensures coherency for shared data. The ring network is bi-directional to allow agents such as processor cores, hf caches and other logic blocks to communicate with each other within the chip. Each ring data-path is 1012-bits wide per direction.
Thus, different implementations of the processor 10400 may include: 1) a CPU with the special purpose logic 10408 being integrated graphics and/or scientific (throughput) logic (which may include one or more cores), and the cores 10402A-N being one or more general purpose cores (e.g., general purpose in-order cores, general purpose out-of-order cores, a combination of the two); 2) a coprocessor with the cores 10402A-N being a large number of special purpose cores intended primarily for graphics and/or scientific (throughput); and 3) a coprocessor with the cores 10402A-N being a large number of general purpose in-order cores. Thus, the processor 10400 may be a general-purpose processor, coprocessor or special-purpose processor, such as, for example, a network or communication processor, compression engine, graphics processor, GPGPU (general purpose graphics processing unit), a high-throughput many integrated core (MIC) coprocessor (including 30 or more cores), embedded processor, or the like. The processor may be implemented on one or more chips. The processor 10400 may be a part of and/or may be implemented on one or more substrates using any of a number of process technologies, such as, for example, BiCMOS, CMOS, or NMOS.
The memory hierarchy includes one or more levels of cache within the cores, a set or one or more shared cache units 10406, and external memory (not shown) coupled to the set of integrated memory controller units 10414. The set of shared cache units 10406 may include one or more mid-level caches, such as level 2 (L2), level 3 (L3), level 4 (L4), or other levels of cache, a last level cache (LLC), and/or combinations thereof. While in one embodiment a ring based interconnect unit 10412 interconnects the integrated graphics logic 10408, the set of shared cache units 10406, and the system agent unit 10410/integrated memory controller unit(s) 10414, alternative embodiments may use any number of well-known techniques for interconnecting such units. In one embodiment, coherency is maintained between one or more cache units 10406 and cores 10402-A-N.
In some embodiments, one or more of the cores 10402A-N are capable of multi-threading. The system agent 10410 includes those components coordinating and operating cores 10402A-N. The system agent unit 10410 may include for example a power control unit (PCU) and a display unit. The PCU may be or include logic and components needed for regulating the power state of the cores 10402A-N and the integrated graphics logic 10408. The display unit is for driving one or more externally connected displays.
The cores 10402A-N may be homogenous or heterogeneous in terms of architecture instruction set; that is, two or more of the cores 10402A-N may be capable of execution the same instruction set, while others may be capable of executing only a subset of that instruction set or a different instruction set.
Referring now to
The optional nature of additional processors 10515 is denoted in
The memory 10540 may be, for example, dynamic random access memory (DRAM), phase change memory (PCM), or a combination of the two. For at least one embodiment, the controller hub 10520 communicates with the processor(s) 10510, 10515 via a multi-drop bus, such as a frontside bus (FSB), point-to-point interface such as QuickPath Interconnect (QPI), or similar connection 10595.
In one embodiment, the coprocessor 10545 is a special-purpose processor, such as, for example, a high-throughput MIC processor, a network or communication processor, compression engine, graphics processor, GPGPU, embedded processor, or the like. In one embodiment, controller hub 10520 may include an integrated graphics accelerator.
There can be a variety of differences between the physical resources 10510, 10515 in terms of a spectrum of metrics of merit including architectural, microarchitectural, thermal, power consumption characteristics, and the like.
In one embodiment, the processor 10510 executes instructions that control data processing operations of a general type. Embedded within the instructions may be coprocessor instructions. The processor 10510 recognizes these coprocessor instructions as being of a type that should be executed by the attached coprocessor 10545. Accordingly, the processor 10510 issues these coprocessor instructions (or control signals representing coprocessor instructions) on a coprocessor bus or other interconnect, to coprocessor 10545. Coprocessor(s) 10545 accept and execute the received coprocessor instructions.
Referring now to
Processors 10670 and 10680 are shown including integrated memory controller (IMC) units 10672 and 10682, respectively. Processor 10670 also includes as part of its bus controller units point-to-point (P-P) interfaces 10676 and 10678; similarly, second processor 10680 includes P-P interfaces 10686 and 10688. Processors 10670, 10680 may exchange information via a point-to-point (P-P) interface 10650 using P-P interface circuits 10678, 10688. As shown in
Processors 10670, 10680 may each exchange information with a chipset 10690 via individual P-P interfaces 10652, 10654 using point to point interface circuits 10676, 10694, 10686, 10698. Chipset 10690 may optionally exchange information with the coprocessor 10638 via a high-performance interface 10639. In one embodiment, the coprocessor 10638 is a special-purpose processor, such as, for example, a high-throughput MIC processor, a network or communication processor, compression engine, graphics processor, GPGPU, embedded processor, or the like.
A shared cache (not shown) may be included in either processor or outside of both processors, yet connected with the processors via P-P interconnect, such that either or both processors' local cache information may be stored in the shared cache if a processor is placed into a low power mode.
Chipset 10690 may be coupled to a first bus 10616 via an interface 10696. In one embodiment, first bus 10616 may be a Peripheral Component Interconnect (PCI) bus, or a bus such as a PCI Express bus or another third generation I/O interconnect bus, although the scope of the present disclosure is not so limited.
As shown in
Referring now to
Referring now to
Embodiments (e.g., of the mechanisms) disclosed herein may be implemented in hardware, software, firmware, or a combination of such implementation approaches. Embodiments of the disclosure may be implemented as computer programs or program code executing on programmable systems comprising at least one processor, a storage system (including volatile and non-volatile memory and/or storage elements), at least one input device, and at least one output device.
Program code, such as code 10630 illustrated in
The program code may be implemented in a high level procedural or object oriented programming language to communicate with a processing system. The program code may also be implemented in assembly or machine language, if desired. In fact, the mechanisms described herein are not limited in scope to any particular programming language. In any case, the language may be a compiled or interpreted language.
One or more aspects of at least one embodiment may be implemented by representative instructions stored on a machine-readable medium which represents various logic within the processor, which when read by a machine causes the machine to fabricate logic to perform the techniques described herein. Such representations, known as “IP cores” may be stored on a tangible, machine readable medium and supplied to various customers or manufacturing facilities to load into the fabrication machines that actually make the logic or processor.
Such machine-readable storage media may include, without limitation, non-transitory, tangible arrangements of articles manufactured or formed by a machine or device, including storage media such as hard disks, any other type of disk including floppy disks, optical disks, compact disk read-only memories (CD-ROMs), compact disk rewritable's (CD-RWs), and magneto-optical disks, semiconductor devices such as read-only memories (ROMs), random access memories (RAMs) such as dynamic random access memories (DRAMs), static random access memories (SRAMs), erasable programmable read-only memories (EPROMs), flash memories, electrically erasable programmable read-only memories (EEPROMs), phase change memory (PCM), magnetic or optical cards, or any other type of media suitable for storing electronic instructions.
Accordingly, embodiments of the disclosure also include non-transitory, tangible machine-readable media containing instructions or containing design data, such as Hardware Description Language (HDL), which defines structures, circuits, apparatuses, processors and/or system features described herein. Such embodiments may also be referred to as program products.
In some cases, an instruction converter may be used to convert an instruction from a source instruction set to a target instruction set. For example, the instruction converter may translate (e.g., using static binary translation, dynamic binary translation including dynamic compilation), morph, emulate, or otherwise convert an instruction to one or more other instructions to be processed by the core. The instruction converter may be implemented in software, hardware, firmware, or a combination thereof. The instruction converter may be on processor, off processor, or part on and part off processor.
