GB2583574A - Handling exceptions in a multi-tile processing arrangement - Google Patents
Handling exceptions in a multi-tile processing arrangement Download PDFInfo
- Publication number
- GB2583574A GB2583574A GB2002863.5A GB202002863A GB2583574A GB 2583574 A GB2583574 A GB 2583574A GB 202002863 A GB202002863 A GB 202002863A GB 2583574 A GB2583574 A GB 2583574A
- Authority
- GB
- United Kingdom
- Prior art keywords
- exception
- tile
- tiles
- instruction
- debug
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F11/00—Error detection; Error correction; Monitoring
- G06F11/07—Responding to the occurrence of a fault, e.g. fault tolerance
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F11/00—Error detection; Error correction; Monitoring
- G06F11/07—Responding to the occurrence of a fault, e.g. fault tolerance
- G06F11/0703—Error or fault processing not based on redundancy, i.e. by taking additional measures to deal with the error or fault not making use of redundancy in operation, in hardware, or in data representation
- G06F11/0766—Error or fault reporting or storing
- G06F11/0772—Means for error signaling, e.g. using interrupts, exception flags, dedicated error registers
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F11/00—Error detection; Error correction; Monitoring
- G06F11/07—Responding to the occurrence of a fault, e.g. fault tolerance
- G06F11/0703—Error or fault processing not based on redundancy, i.e. by taking additional measures to deal with the error or fault not making use of redundancy in operation, in hardware, or in data representation
- G06F11/0706—Error or fault processing not based on redundancy, i.e. by taking additional measures to deal with the error or fault not making use of redundancy in operation, in hardware, or in data representation the processing taking place on a specific hardware platform or in a specific software environment
- G06F11/0715—Error or fault processing not based on redundancy, i.e. by taking additional measures to deal with the error or fault not making use of redundancy in operation, in hardware, or in data representation the processing taking place on a specific hardware platform or in a specific software environment in a system implementing multitasking
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F11/00—Error detection; Error correction; Monitoring
- G06F11/07—Responding to the occurrence of a fault, e.g. fault tolerance
- G06F11/0703—Error or fault processing not based on redundancy, i.e. by taking additional measures to deal with the error or fault not making use of redundancy in operation, in hardware, or in data representation
- G06F11/0706—Error or fault processing not based on redundancy, i.e. by taking additional measures to deal with the error or fault not making use of redundancy in operation, in hardware, or in data representation the processing taking place on a specific hardware platform or in a specific software environment
- G06F11/0721—Error or fault processing not based on redundancy, i.e. by taking additional measures to deal with the error or fault not making use of redundancy in operation, in hardware, or in data representation the processing taking place on a specific hardware platform or in a specific software environment within a central processing unit [CPU]
- G06F11/0724—Error or fault processing not based on redundancy, i.e. by taking additional measures to deal with the error or fault not making use of redundancy in operation, in hardware, or in data representation the processing taking place on a specific hardware platform or in a specific software environment within a central processing unit [CPU] in a multiprocessor or a multi-core unit
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F11/00—Error detection; Error correction; Monitoring
- G06F11/36—Preventing errors by testing or debugging software
- G06F11/362—Software debugging
- G06F11/3636—Software debugging by tracing the execution of the program
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F11/00—Error detection; Error correction; Monitoring
- G06F11/36—Preventing errors by testing or debugging software
- G06F11/362—Software debugging
- G06F11/3648—Software debugging using additional hardware
- G06F11/3656—Software debugging using additional hardware using a specific debug interface
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F11/00—Error detection; Error correction; Monitoring
- G06F11/36—Preventing errors by testing or debugging software
- G06F11/362—Software debugging
- G06F11/366—Software debugging using diagnostics
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F15/00—Digital computers in general; Data processing equipment in general
- G06F15/16—Combinations of two or more digital computers each having at least an arithmetic unit, a program unit and a register, e.g. for a simultaneous processing of several programs
- G06F15/177—Initialisation or configuration control
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Quality & Reliability (AREA)
- Computer Hardware Design (AREA)
- Software Systems (AREA)
- Debugging And Monitoring (AREA)
Abstract
A multi-tile processing system having an interconnect wherein tiles participate in a bulk synchronous parallel (BSP) scheme wherein a debug tool accesses an exception raised by one of the tiles. In BSP each of the tiles in a group perform an on-tile compute phase followed by an inter-tile exchange phase with the exchange phase being held back until all tiles have completed the compute phase. A debug tool (1120) accesses an exception raised signal (1114) of one of the tiles (4) to ascertain a cause of the exception raised by the tile, performs a debug operation to resolve the exception, and deasserts the exception. The tiles which are not in an excepted state continue to perform their on-tile compute phase, but do not commence their inter-tile exchange phase while one or more other tile is in an excepted state. After the excepted state has been resolved, the tile that issued the exception raised signal continues with its compute phase until it reaches a synchronisation point at which it issues a synchronisation request.
Description
Handling Exceptions in a Multi-Tile Processing Arrangement
Technical Field
The present disclosure relates to handling exceptions while processing workloads of multiple different tiles in a multi-tile processing arrangement, each tile comprising its own processing unit and memory. Particularly, the disclosure relates to bulk synchronous parallel (BSP) communication schemes whereby each of a group of tiles must complete a compute phase before any of the tiles in the group can proceed to an exchange phase. The techniques are also applicable to multithreaded operation in a single tile.
Background
A multi-threaded processor is a processor which is capable of executing multiple program threads alongside one another. The processor may comprise some hardware that is common to the multiple different threads (e.g. a common instruction memory, data memory and/or execution unit); but to support the multi-threading, the processor also comprises some dedicated hardware specific to each thread.
The dedicated hardware comprises at least a respective context register file for each of the number of threads that can be executed at once. A "context", when talking about multithreaded processors, refers to the program state of a respective one of the threads being executed alongside one another (e.g. program counter value, status and current operand values).The context register file refers to the respective collection of registers for representing this program state of the respective thread. Registers in a register file are distinct from general purpose memory in that register addresses are fixed as bits in instruction words, whereas memory addresses can be computed by executing instructions. The registers of a given context typically comprise a respective program counter for the respective thread, and a respective set of operand registers for temporarily holding the data acted upon and output by the respective thread during the computations performed by that thread. Each context may also have a respective status register for storing a status of the respective thread (e.g. whether it is paused or running). Thus each of the currently running threads has its own separate program counter, and optionally operand registers and status register(s).
One possible form of multi-threading is parallelism. That is, as well as multiple contexts, multiple execution pipelines are provided: i.e. a separate execution pipeline for each stream of instructions to be executed in parallel. However, this requires a great deal of duplication in terms of hardware.
Instead, another form of multi-threaded processor employs concurrency rather than parallelism, whereby the threads share a common execution pipeline (or at least a common part of a pipeline) and different threads are interleaved through this same, shared execution pipeline. Performance of a multi-threaded processor may still be improved compared to no concurrency or parallelism, thanks to increased opportunities for hiding pipeline latency. Also, this approach does not require as much extra hardware dedicated to each thread as a fully parallel processor with multiple execution pipelines, and so does not incur so much extra silicon.
One form of parallelism can be achieved by means of a processor comprising an arrangement of multiple tiles on the same chip (i.e. same die), each tile comprising its own separate respective processing unit and memory (including program memory and data memory). Thus, separate portions of program code can be run in parallel on different ones of the tiles. The tiles are connected together via an on-chip interconnect which enables the code run on the different tiles to communicate. In some cases, the processing unit on each tile may itself run multiple concurrent threads on-tile, each tile having its own respective set of contexts and corresponding pipeline as described above in order to support interleaving of multiple threads on the same tile through the same pipeline.
In general, there may exist dependencies between the portions of a program running on different tiles. A technique is therefore required to prevent a piece of code on one tile running ahead of data upon which it is dependent being made available by another piece of code on another tile. There are a number of possible schemes for achieving this, but the scheme of interest herein is known as "bulk synchronous parallel" (BSP). According to BSP, each tile performs a compute phase and an exchange phase in an alternating cycle. During the compute phase each tile performs one or more computation tasks locally on tile, but does not communicate any results of its computations with any others of the tiles. In the exchange phase, each tile is allowed to exchange one or more results of the computations from the preceding compute phase to and/or from one or more others of the tiles in the group, but does not yet proceed to the next compute phase. Further, according to the BSP principle, a barrier synchronization is placed at the juncture transitioning from the compute phase into the exchange phase, or transitioning from the exchange phase into the compute phase, or both. That is to say, either: (a) all tiles are required to complete their respective compute phases before any in the group is allowed to proceed to the next exchange phase, or (b) all tiles in the group are required to complete their respective exchange phases before any tile in the group is allowed to proceed to the next compute phase, or (c) both. In some scenarios a tile performing computation may be allowed to communicate with other system resources such as a network card or storage disk, as long as no communication with other tiles in the group is involved.
An example use of multi-threaded and/or multi-tiled processing is found in machine intelligence. As will be familiar to those skilled in the art of machine intelligence, a machine intelligence algorithm is based around performing iterative updates to a "knowledge model", which can be represented by a graph of multiple interconnected nodes. Each node represents a function of its inputs. Some nodes receive the inputs to the graph and some receive inputs from one or more other nodes, whilst the output of some nodes form the inputs of other nodes, and the output of some nodes provide the output of the graph (and in some cases a given node may even have all of these: inputs to the graph, outputs from the graph and connections to other nodes). Further, the function at each node is parameterized by one or more respective parameters, e.g. weights. During a learning stage the aim is, based on a set of experiential input data, to find values for the various parameters such that the graph as a whole will generate a desired output for a range of possible inputs. Various algorithms for doing this are known in the art, such as a back propagation algorithm based on stochastic gradient descent. Over multiple iterations based on the input data, the parameters are gradually tuned to decrease their errors, and thus the graph converges toward a solution. In a subsequent stage, the learned model can then be used to make predictions of outputs given a specified set of inputs or to make inferences as to inputs (causes) given a specified set of outputs.
The implementation of each node will involve the processing of data, and the interconnections of the graph correspond to data to be exchanged between the nodes. Typically, at least some of the processing of each node can be carried out independently of some or all others of the nodes in the graph, and therefore large graphs expose great opportunities for con currency and/or parallelism.
Particular challenges arise in the debugging of such processors.
Summary
Machine intelligence graphs consist of many nodes connected to one another via edges, where nodes represent a certain processing requirement on their individual data and edges represent data to be transferred between nodes. This structure allows for a high level of parallelism to be performed. One method of parallelism is multithreading, in which multiple processes are executed concurrently.
A first aspect of the present invention is directed towards a processing system comprising an arrangement of tiles and an interconnect for communication between the tiles, wherein: each tile comprises an execution unit for executing machine code instructions, each being an instance of a predefined set of instruction types in an instruction set of the processing system; wherein each tile is configured when in an excepted state to raise an exception signal and to pause instruction issue until the excepted state has been resolved; wherein each tile comprises a local debug interface to enable access by the debug tool to resolve the excepted state, and wherein the processing system further comprises exception handling logic which is connected to receive the exception signal raised by any of the tiles, and to provide an asserted exception indicator when at least one of the inputs to the logic receives an exception signal from a tile.
In some embodiments, the interconnect is operable to conduct communications between the group of some or all of the tiles according to a bulk synchronous parallel scheme, whereby each of the tiles in said group perform an on-tile compute phase followed by an inter-tile exchange phase with the exchange phase being held back until all tiles in said group have completed the compute phase wherein each tile in the group is configured to generate a synchronisation request when it has completed its compute phase, and to pause issue of instructions until it receives a synchronisation acknowledgement signal; and the processing system comprises synchronisation logic which is configured to issue a synchronisation acknowledgement signal to each of the tiles in the group when it has received synchronisation requests from each of the tiles in the group.
In some embodiment, tiles in the processing system which are not in an excepted state continue to perform their on-tile compute phase up to issuance of a synchronisation request while the excepted state is being resolved in the tile with the excepted state.
Each tile in the processing system mentioned above may comprise a set of debug registers which enable the execution unit to recognise an excepted state. These debug registers may comprise an instruction data break register which identifies data-related criteria at which to raise an exception and/or one or more program count at which an exception is to be raised. In embodiments, one of the set of debug registers may comprise an exception enable flag which governs operation of a synchronisation instruction which forms part of the instruction set of each tile, wherein in a first state of the exception enable flag execution of the synchronisation instruction causes the execution unit to send the synchronisation request to the synchronisation logic, and in the second state of the exception enable flag execution of the synchronisation instruction does not send the synchronisation request and sets the exception signal.
In one embodiment, each tile in the group comprises: multiple context register sets, each context register set arranged to store a program state of the respective one of multiple threads; and a scheduler arranged to schedule the execution of the respective one of a plurality of worker threads in each of a plurality of time slots in a repeating sequence of interleaved time slots, with the program state of each of the worker threads being stored in a respective one of the context register sets; wherein according to the bulk synchronous parallel scheme, the exchange phase is held back until all the worker threads on all the tiles in the group have issued a synchronisation request to indicate they have completed the compute phase.
In embodiments, the exchange phase of the processing system mentioned above may be arranged to be performed by a supervisor thread separate to the worker threads, wherein the supervisor thread causes the synchronisation request to be issued from the file. Additionally, the context register sets on each tile may comprise multiple worker context register sets arranged to represent the program stage of respective ones of said plurality of worker threads, and to indicate an exception type of an exception encountered by a worker thread, and an additional supervisor context register comprising an additional set of registers arranged to represent the program state of the supervisor thread, and an exception type of an exception encountered by the supervisor thread.
In embodiments, the set of debug registers of the processing system may comprise an exception context register which indicates for each worker context and the supervisor whether or not it is in an excepted state.
A further aspect of the present invention is directed towards a method of debugging a processing system comprising an arrangement of tiles and an interconnect for communicating between the tiles, wherein each tile comprises an execution unit for executing machine code instructions, each being an instance of a predefined set of instruction types in an instruction set of the processing system, the interconnect being operable to conduct communications between a group of some or all of the tiles according to a bulk synchronous parallel scheme whereby each of the tiles in the group perform an on-tile compute phase followed by an inter-tile exchange phase with the exchange phase being held back until all of the tiles in the group have completed the compute phase, the method comprising reading an exception indicator in a debug tool, and accessing an exception raised signal of one of the tiles, accessing debug hardware on the one of the tiles to ascertain a cause of the exception raised by the tile, performing a debug operation to resolve the exception, and deasserting the exception raised signal by the tile when the exception has been resolved.
In embodiments, the method may comprise dealing with a plurality of exception raised signals in sequence until the exception indicator is deasserted.
In embodiments, the tiles mentioned in the method described above which are not in an excepted state may continue to perform their on-tile compute phase, but may not commence their inter-tile exchange phase while one or more other tile is in an excepted state.
After the excepted state has been resolved, the tile that issued the exception raised signal may continue with its compute phase until it reaches a synchronisation point at which it issues a synchronisation request.
In embodiments, resolving the excepted state in the method described above comprises accessing at least one of a set of debug registers to identify the cause of the excepted state.
The present application is a divisional from GB Application No. 1821067.4. A first aspect of the invention of the parent application is directed towards a processing system comprising an arrangement of tiles and an interconnect for communication between the tiles, wherein: each tile comprises an execution unit for executing machine code instructions, each being an instance of a predefined set of instruction types in an instruction set of the processing system; the interconnect is operable to conduct communications between the group of some or all of the tiles according to a bulk synchronous scheme, whereby each of the tiles in said group perform an on-tile compute phase followed by an inter-tile exchange phase with the exchange phase being held back until all tiles in said group have completed the compute phase wherein each tile of the group is configured to generate a synchronisation request when it has completed its compute phase, and to pause an issue of instructions until it receives a synchronisation acknowledgement signal; and the processing system comprises synchronisation logic whi ch is configured to issue a synchronisation acknowledgement signal to each of the files in the group when it has received synchronisation requests from each of the tiles in the group, wherein each tile is configured when in an excepted state to raise an exception signal and to pause instruction issue until the excepted state has been resolved wherein files in the processing system which are not in an excepted state continue to perform their on-tile compute phase up to issuance of a synchronisation request while the excepted state is being resolved in the tile with the excepted state.
In embodiments of the invention of the parent application, the processing system may comprise a debug interface to enable access to the processing system to the debug tool to resolve the excepted state. This debug interface may be local to enable access by the debug tool.
In embodiments of the invention of the parent, the processing system may also comprise exception handling logic which is connected to receive the exception signal raised by any of the tiles, and to provide an asserted exception indicator when at least one of the inputs to the logic receives an exception signal from a tile.
Brief Description of the Drawings
To aid understanding of the present disclosure and to show how embodiments may be put into effect, reference is made by way of example to the accompanying drawings in which: Figure 1 is a schematic block diagram of a multi-threaded processing unit, Figure 2 is a schematic block diagram of a plurality of thread contexts, Figure 3 schematically illustrates a scheme of interleaved execution time slots, Figure 4 shows more detail of debug circuitry, Figure 5 schematically illustrates synchronization amongst worker threads on the same tile, Figure 6 is a schematic block diagram of a processor chip comprising multiple tiles, Figure 7 is a schematic illustration of a bulk synchronous parallel (BSP) computing model, Figure 8 is another schematic illustration of a BSP model, Figure 9 is a schematic illustration of system of multiple interconnected processor chips, and Figure 10 is a schematic illustration of a graph used in a machine intelligence algorithm.
Detailed Description of Embodiments
The present disclosure addresses challenges of inspecting / debugging tiles in a multitile, multithreaded processor architecture using a hardware configurable mechanism.
The following describes a processor architecture which includes a dedicated instruction in its instruction set for performing a barrier synchronization, where execution depends on the state of a configurable flag. First however an example processor in which this may be incorporated is described with reference to Figures 1 to 4.
Figure 1 illustrates an example of a processor module 4 in accordance with embodiments of the present disclosure. For instance the processor module 4 may be one tile of an array of like processor tiles on a same chip, or may be implemented as a stand-alone processor on its own chip. The processor module 4 comprises a multi-threaded processing unit 10 in the form of a barrel-threaded processing unit, and a local memory 11 (i.e. on the same tile in the case of a multi-tile array, or same chip in the case of a single-processor chip). A barrel-threaded processing unit is a type of multi-threaded processing unit in which the execution time of the pipeline is divided into a repeating sequence of interleaved time slots, each of which can be owned by a given thread. This will be discussed in more detail shortly. The memory 11 comprises an instruction memory 12 and a data memory 22 (which may be implemented in different addressable memory unit or different regions of the same addressable memory unit). The instruction memory 12 stores machine code to be executed by the processing unit 10, whilst the data memory 22 stores both data to be operated on by the executed code and data output by the executed code (e.g as a result of such operations).
The memory 12 stores a variety of different threads of a program, each thread comprising a respective sequence of instructions for performing a certain task or tasks. Note that an instruction as referred to herein means a machine code instruction, i.e. an instance of one of the fundamental instructions of the processor's instruction set, consisting of a single opcode and zero or more operands.
The program described herein comprises a plurality of worker threads, and a supervisor subprogram which may be structured as one or more supervisor threads. These will be discussed in more detail shortly. In embodiments, each of some or all of the worker threads takes the form of a respective "codelet". A codelet is a particular type of thread, sometimes also referred to as an "atomic" thread. It has all the input information it needs to execute from the beginning of the thread (from the time of being launched), i.e. it does not take any input from any other part of the program or from memory after being launched. Further, no other part of the program will use any outputs (results) of the thread until it has terminated (finishes). Unless it encounters an error, it is guaranteed to finish. N.B. some literature also defines a codelet as being stateless, i.e. if run twice it could not inherit any information from its first run, but that additional definition is not adopted here. Note also that not all of the worker threads need be codelets (atomic), and in embodiments some or all of the workers may instead be able to communicate with one another.
Within the processing unit 10, multiple different ones of the threads from the instruction memory 12 can be interleaved through a single execution pipeline 13 (though typically only a subset of the total threads stored in the instruction memory can be interleaved at any given point in the overall program). The multi-threaded processing unit 10 comprises: a plurality of context register files 26 each arranged to represent the state (context) of a different respective one of the threads to be executed concurrently; a shared execution pipeline 13 that is common to the concurrently executed threads; and a scheduler 24 for scheduling the concurrent threads for execution through the shared pipeline in an interleaved manner, preferably in a round robin manner. The processing unit 10 is connected to a shared instruction memory 12 common to the plurality of threads, and a shared data memory 22 that is again common to the plurality of threads.
The execution pipeline 13 comprises a fetch stage 14, a decode stage 16, and an execution stage 18 comprising an execution unit which may perform arithmetic and logical operations, address calculations, load and store operations, and other operations, as defined by the instruction set architecture. Each of the context register files 26 comprises a respective set of registers for representing the program state of a respective thread.
The tile also has a debug interface 120 and a set of debug hardware contexts 122 described later.
An example of the registers making up each of the context register files 26 is illustrated schematically in Figure 2. Each of the context register files 26 comprises a respective one or more control registers 28, comprising at least a program counter (PC) for the respective thread (for keeping track of the instruction address at which the thread is currently executing), and in embodiments also a set of one or more status registers (SR) recording a current status of the respective thread (such as whether it is currently running or paused, e.g. because it has encountered an exception event). Each of the context register files 26 also comprises a respective set of operand registers (OP) 32, for temporarily holding operands of the instructions executed by the respective thread, i.e. values operated upon or resulting from operations defined by the opcodes of the respective thread's instructions when executed. It will be appreciated that each of the context register files 26 may optionally comprise a respective one or more other types of register (not shown). Note also that whilst the term "register file" is sometimes used to refer to a group of registers in a common address space, this does not necessarily have to be the case in the present disclosure and each of the hardware contexts 26 (each of the register sets 26 representing each context) may more generally comprise one or multiple such register files.
As will be discussed in more detail later, the disclosed arrangement has one worker context register file CXO... CX(M-1) for each of the number M of threads that can be executed concurrently (NI=3 in the example illustrated but this is not limiting), and one additional supervisor context register file CXS. The worker context register files are reserved for storing the contexts of worker threads, and the supervisor context register file is reserved for storing the context of a supervisor thread. Note that in embodiments the supervisor context is special, in that it has a different number of registers than each of the workers. Each of the worker contexts preferably have the same number of status registers and operand registers as one another. In embodiments the supervisor context may have fewer operand registers than each of the workers. Examples of operand registers the worker context may have that the supervisor does not include: floating point registers, accumulate registers, and/or dedicated weight registers (for holding weights of a neural network). In embodiments the supervisor may also have a different number of status registers. Further, in embodiments the instruction set architecture of the processor module 4 may be configured such that the worker threads and supervisor thread(s) execute some different types of instruction but also share some instruction types.
The fetch stage 14 is connected so as to fetch instructions to be executed from the instruction memory 12, under control of the scheduler 24. The scheduler 24 is configured to control the fetch stage 14 to fetch an instruction from each of a set of concurrently executing threads in turn in a repeating sequence of time slots, thus dividing the resources of the pipeline 13 into a plurality of temporally interleaved time slots, as will be discussed in more detail shortly. For example the scheduling scheme could be round-robin or weighted round-robin. Another term for a processor operating in such a manner is a barrel threaded processor.
In some embodiments, the scheduler 24 may have access to one of the status registers SR of each thread indicating whether the thread is paused, so that the scheduler 24 in fact controls the fetch stage 14 to fetch the instructions of only those of the threads that are currently active. In embodiments, preferably each time slot (and corresponding context register file) is always owned by one thread or another, i.e. each slot is always occupied by some thread, and each slot is always included in the sequence of the scheduler 24; though the thread occupying any given slot may happen to be paused at the time, in which case when the sequence comes around to that slot, the instruction fetch for the respective thread is passed over. Alternatively it is not excluded for example that in alternative, less preferred implementations, some slots can be temporarily vacant and excluded from the scheduled sequence. Where reference is made to the number of time slots the execution unit is operable to interleave, or such like, this refers to the maximum number of slots the execution is capable of executing concurrently, i.e. the number unit's hardware supports.
The fetch stage 14 has access to the program counter (PC) of each of the contexts. For each respective thread, the fetch stage 14 fetches the next instruction of that thread from the next address in the program memory 12 as indicated by the program counter. The program counter increments each execution cycle unless branched by a branch instruction. The fetch stage 14 then passes the fetched instruction to the decode stage 16 to be decoded, and the decode stage 16 then passes an indication of the decoded instruction to the execution unit 18 along with the decoded addresses of any operand registers 32 specified in the instruction, in order for the instruction to be executed. The execution unit 18 has access to the operand registers 32 and the control registers 28, which it may use in executing the instruction based on the decoded register addresses, such as in the case of an arithmetic instruction (e.g. by adding, multiplying, subtracting or dividing the values in two operand registers and outputting the result to another operand register of the respective thread). Or if the instruction defines a memory access (load or store), the load/store logic of the execution unit 18 loads a value from the data memory into an operand register of the respective thread, or stores a value from an operand register of the respective thread into the data memory 22, in accordance with the instruction. Or if the instruction defines a branch or a status change, the execution unit changes value in the program counter PC or one of the status registers SR accordingly. Note that while one thread's instruction is being executed by the execution unit 18, an instruction from the thread in the next time slot in the interleaved sequence can be being decoded by the decode stage 16; and/or while one instruction is being decoded by the decode stage 16, the instruction from the thread in the next time slot after that can be being fetched by the fetch stage 14 (though in general the scope of the disclosure is not limited to one instruction per time slot, e.g. in alternative scenarios a batch of two or more instructions could be issued from a given thread per time slot). Thus the interleaving advantageously hides latency in the pipeline 13, in accordance with known barrel threaded processing techniques.
An example of the interleaving scheme implemented by the scheduler 24 is illustrated in Figure 3. Here the concurrent threads are interleaved according to a round-robin scheme whereby, within each round of the scheme, the round is divided into a sequence of time slots SO, Si, S2..., each for executing a respective thread. Typically each slot is one processor cycle long and the different slots are evenly sized, though not necessarily so in all possible embodiments, e.g. a weighted round-robin scheme is also possible whereby some threads get more cycles than others per execution round. In general the barrel-threading may employ either an even round-robin or a weighted round-robin schedule, where in the latter case the weighting may be fixed or adaptive.
Whatever the sequence per execution round, this pattern then repeats, each round comprising a respective instance of each of the time slots. Note therefore that a time slot as referred to herein means the repeating allocated place in the sequence, not a particular instance of the time slot in a given repetition of the sequence. Put another way, the scheduler 24 apportions the execution cycles of the pipeline 13 into a plurality of temporally interleaved (time-division multiplexed) execution channels, with each comprising a recurrence of a respective time slot in a repeating sequence of time slots. In the illustrated embodiment, there are four time slots, but this is just for illustrative purposes and other numbers are possible. E.g. in one preferred embodiment there are in fact six time slots.
Whatever the number of time slots the round-robin scheme is divided into, then according to present disclosure, the processing unit 10 comprises one more context register file 26 than there are time slots, i.e. it supports one more context than the number of interleaved timeslots it is capable of barrel-threading.
This is illustrated by way of example in Figure 2: if there are four time slots SO... S3 as shown in Figure 3, then there are five context register files, labelled here CXO, CX1, CX2, CX3 and CXS. That is, even though there are only four execution time slots SO... S3 in the barrel-threaded scheme and so only four threads can be executed concurrently, it is disclosed herein to add a fifth context register file CXS, comprising a fifth program counter (PC), a fifth set of operand registers 32, and in embodiments also a fifth set of one or more status registers (SR). Though note that as mentioned, in embodiments the supervisor context may differ from the others CXO...3, and the supervisor thread may support a different set of instructions for operating the execution pipeline 13.
Each of the first four contexts CXO... CX3 is used to represent the state of a respective one of a plurality of "worker threads" currently assigned to one of the four execution time slots SO.S3, for performing whatever application-specific computation tasks are desired by the programmer (note again this may only be subset of the total number of worker threads of the program as stored in the instruction memory 12). The fifth context CXS however, is reserved for a special function, to represent the state of a "supervisor thread" (SV) whose role it is to coordinate the execution of the worker threads, at least in the sense of assigning which of the worker threads W is to be executed in which of the time slots SO, Sl, S2... at what point in the overall program. According to the present embodiments the supervisor thread has additional coordinating responsibilities. The supervisor thread is responsible for performing barrier synchronisations to ensure a certain order of execution. E.g. in a case where one or more second threads are dependent on data to be output by one or more first threads run on the same processor module 4, the supervisor may perform a barrier synchronization to ensure that none of the second threads begin until the first threads have finished. In addition to, or instead of this first example, the supervisor may perform a barrier synchronization to ensure that one or more threads on the processor module 4 do not begin until a certain external source of data, such as another tile or processor chip, has completed the processing required to make that data available. The supervisor thread may also be used to perform other functionality relating to the multiple worker threads. For example, the supervisor thread may be responsible for communicating data externally to the processor module 4 (to receive external data to be acted on by one or more of the threads, and/or to transmit data output by one or more of the worker threads). In general, the supervisor thread may be used to provide any kind of overseeing or coordinating function desired by the programmer. For instance, as another example, the supervisor may oversee transfer between the tile local memory 12 and one or more resources in the wider system (external to the array 6) such as a storage disk or network card.
The supervisor thread SV does not have its own time slot per se in the scheme of interleaved time slots. Nor do the workers have their own timeslots per se, as allocation of slots to worker threads is flexibly defined. Rather, each time slot has its own dedicated context register file (CX0...CXM-1) for storing worker context, which is used by the worker when the slot is allocated to the worker, but not used when the slot is allocated to the supervisor. When a given slot is allocated to the supervisor, that slot instead uses the context register file CVS of the supervisor. Note that the supervisor always has access to its own context and no workers are able to occupy the supervisor context register file CXS.
The supervisor thread SV has the ability to run in any and all of the time slots SO....S3 (or more generally SO... SM-1). The scheduler 24 is configured io, when the program as a whole starts, begin by allocating the supervisor thread to all of the time slots, i.e. so the supervisor SV starts out running in all of SO... S3. However, the supervisor thread is provided with a mechanism for, at some subsequent point (either straight away or after performing one or more supervisor tasks), temporarily relinquishing each of the slots in which it is running to a respective one of the worker threads. This is achieved by the supervisor thread executing a relinquish instruction, called "RUN" by way of example herein. In embodiments this instruction takes two operands: an address of a worker thread in the instruction memory 12 and an address of some data for that worker thread in the data memory 22: RUN task addr, data addr The worker threads are portions of code that can be run concurrently with one another, each representing one or more respective computation tasks to be performed. The data address may specify some data to be acted upon by the worker thread. Alternatively, the relinquish instruction may take only a single operand specifying the address of the worker thread, and the data address could be included in the code of the worker thread; or in another example the single operand could point to a data structure specifying the addresses of the worker thread and data. As mentioned, in embodiments at least some of the workers may take the form of codelets, i.e. atomic units of concurrently executable code. Alternatively or additionally, some of the workers need not be codelets and may instead be able to communicate with one another.
The relinquish instruction ("RUN") acts on the scheduler 24 so as to relinquish the current time slot, in which this instruction is itself executed, to the worker thread specified by the operand. Note that it is implicit in the relinquish instruction that it is the time slot in which the instruction is executed that is being relinquished (implicit in the context of machine code instructions means it doesn't need an operand to specify this -it is understood implicitly from the opcode itself). Thus the time slot which is given away is the time slot in which the supervisor executes the relinquish instruction. Or put another way, the supervisor is executing in the same space that it gives away. The supervisor says "run this piece of code at this location", and then from that point onwards the recurring slot is owned (temporarily) by the relevant worker thread.
The supervisor thread SV performs a similar operation in each of one or more others of the time slots, to give away some or all of its time slots to different respective ones of the worker threads WO... W3 (selected from a larger set WO.wj in the instruction memory 12). Once it Is has done so for the last slot, the supervisor is suspended (then later will resume where it left off when one of the slots is handed back by a worker W).
The supervisor thread SV is thus able to allocate different worker threads, each performing one or more tasks, to different ones of the interleaved execution time slots SO... S3. When the supervisor thread determines it is time to run a worker thread, it uses the relinquish instruction ("RUN") to allocate this worker to the time slot in which the RUN instruction was executed.
In some embodiments, the instruction set also comprises a variant of the run instruction, RUNALL ("run all"). This instruction is used to launch a set of more than one worker together, all executing the same code. In embodiments this launches a worker in every one of the processing unit's slots SO. S3 (or more generally SO... S(1\4-1)).
Further, in some embodiments the RUN and/or RUNALL instruction, when executed, also automatically copies some status from one or more of the supervisor status registers CXS(SR) to a corresponding one or more status registers of the worker thread(s) launched by the RUN or RUNALL. For instance the copied status may comprise one or more modes, such as a floating point rounding mode (e.g. round to nearest or round to zero) and/or an overflow mode (e.g. saturate or use a separate value representing infinity). The copied status or mode then controls the worker in question to operate in accordance with the copied status or mode. In embodiments, the worker can later overwrite this in its own status register (but cannot change the supervisor's status). In further alternative or additional embodiments, that the workers can choose to read some status from one or more status registers of the supervisor (and again may change their own status later) E g again this could be to adopt a mode from the supervisor status register, such as a floating point mode or a rounding mode. In embodiments however, the supervisor cannot read any of the context registers CXO of the workers.
Once launched, each of the currently allocated worker threads WO W3 proceeds to perform the one or more computation tasks defined in the code specified by the respective relinquish instruction. At the end of this, the respective worker thread then hands the time slot in which it is running back to the supervisor thread. This is achieved by executing an exit instruction ("EXIT').
The EXIT instruction takes at least one operand and preferably only a single operand, exit state (e.g. a binary value), to be used for any purpose desired by the programmer to indicate a state of the respective codelet upon ending (e.g. to indicate whether a certain condition was met): EXIT exit state The EXIT instruction acts on the scheduler 24 so that the time slot in which it is executed is returned back to the supervisor thread. The supervisor thread can then perform one or more subsequent supervisor tasks (e.g. barrier synchronization and/or exchange of data with external resources such as other tiles), and/or continue to execute another relinquish instruction to allocate a new worker thread (W4, etc.) to the slot in question. Note again therefore that the total number of threads in the instruction memory 12 may be greater than the number that barrel-threaded processing unit 10 can interleave at any one time. It is the role of the supervisor thread SV to schedule which of the worker threads WO...Wj from the instruction memory 12, at which stage in the overall program, are to be assigned to which of the interleaved time slots SO. .SN4 in the round robin schedule of the scheduler 24.
Referring to Figure 5, in embodiments a SYNC (synchronization) instruction is provided in the processor's instruction set. The SYNC instruction has the effect of causing the supervisor thread SV to wait until all currently executing workers W have exited by means of an EXIT instruction. In embodiments the SYNC instruction takes a mode as an operand (in embodiments its only operand), the mode specifying whether the SYNC is to act only locally in relation to only those worker threads running locally on the same processor module 4, e.g. same tile, as the supervisor as part of which the SYNC is executed on (i.e. only threads through the same pipeline 13 of the same barrel-threaded processing unit 10); or whether instead it is to apply across multiple tiles or even across multiple chips.
SYNC mode // mode E {tile, chip, zone_1, zone_2{ For the purposes of Figure 5 a local SYNC will be assumed ("SYNC tile", i.e. a synchronization within a single tile).
The workers do not need to be identified as operands of the SYNC instruction, as it is implicit that the supervisor S V is then caused to automatically wait until none of the time slots SO, Si, ... of the barrel-threaded processing unit l0 is occupied by a worker. As shown in Figure 5, once each of a current batch of workers WLn have all been launched by the supervisor, the supervisor then executes a SYNC instruction. If the supervisor SV launches workers Win all the slots S0...3 of the barrel-threaded processing unit 10 (all four in the example illustrated, but that is just one example implementation), then the SYNC will be executed by the supervisor once the first of the current batch of worker threads WLn has exited, thus handing back control of at least one slot to the supervisor SV. Otherwise if the workers do not take up all of the slots, the SYNC will simply be executed immediately after the last thread of the current batch WLn has been launched. Either way, the SYNC causes the supervisor SV to wait for all others of the current batch of workers WLn-1 to execute an EXIT before the supervisor can proceed. This waiting by the supervisor thread is imposed in hardware once the SYNC has been executed. I.e. in response to the opcode of the SYNC instruction, the logic in the execution unit (EXU) of the execution stage 18 causes the fetch stage 14 and scheduler 24 to pause from issuing instructions of the supervisor thread until all outstanding worker threads have executed an EXIT instruction.
The present application discusses exception handling within a computer of the type described herein. There are two situations in which exceptions can arise. One group of exceptions arise during normal operation, and another set of exceptions arise in the context of debugging. Exceptions can halt or suspend processing of a thread, and in the case of worker threads, this means that they do not reach their EXIT instruction until the exception has been resolved. Figure 5 illustrates, by crosshatching, that the worker executing in slot SI has encountered an exception, and will not get to its EXIT instruction until some time later than would be normal after it has been debugged.
One particular exception described herein is a BREAK_on_Sync exception which is raised by the SYNC instruction when a BREAK on Sync exception flag is set. The BREAK on Sync exception is described in more detail later.
Exceptions indicate unusual and potentially problematic execution conditions, a divergence from the normal execution behaviour of a program. When in a debug mode, exceptions are deliberately created in order to allow a debugging process to be carried out. Whether or not the chip is in a debug mode, it is important to be able to ascertain the causes of exceptions, and in some cases to recover from such exceptions. The multithreaded BSP execution model discussed herein permits exceptions to be inspected, or debug to take place in one part of the computer, without interrupting normal operation in another part. For example, the exception handling mechanism discussed herein enables debug to be implemented on one tile or a set of tiles without the other tiles on the same chip being aware that a debug process is in operation. For a chip with such a large number of tiles, this is an extremely valuable feature. Moreover, the exception handling mechanism permits one thread to encounter an exception without interrupting other threads on the same tile.
Characteristics of the exception handling mechanism discussed herein include the following: Exceptions are specific to a particular context.
Exception causes are recorded within architectural state (an exception type register) and can be interrogated by external agents to ascertain the source of the exception. One such external agent may be a debug tool which can interrogate the state via the debug interface 120.
An exception may give rise to an exception event, depending on the nature of the exception. Exception events cause instruction issue of the excepting context to stall immediately following exception detection, and become visible at system and host and/or debug at level.
Moreover, exception events cause the system-wide barrier synchronisation mechanism to stall, either indefinitely, or until the exception has been recovered. In some embodiments, following an exception event anywhere in the system, all tiles will stall at the end of the current superstep. Note, however, that the stall is accommodated by the BSP mechanism -threads and tiles other than the excepted one will progress to their next natural sync point, issue a sync request and then wait for a sync ack. The sync ack may take longer than if the exception had not occurred but there is no defined expected time so the system does not consider this to be outside normal operation. Note also that, other threads/ tiles may take longer to get to their sync point in any event.
Where possible, exception events are made recoverable, and in particular exception events raised by a debugging tool and by BREAK-on Sync are recoverable by the following described mechanism.
The phases of the exception handling model will now be described.
A first phase is exception detection. Exception detection is performed by debug hardware 122 on each tile.
According to the next phase, exceptions are logged in context registers associated with the threads such that the state can later be interrogated to ascertain the cause of the exception. Context exception registers are denoted 38 in Figure 2, and one is shown in each worker context and one for the supervisor context. Note that an exception is logged in two places. As shown in Figure 4, the debug hardware 122 comprises an exception status register 1108 which holds a set of context entries. Each context entry indicates whether or not a particular context (worker or supervisor) has excepted. The context entries 110 may each be only a single bit to denote whether there has been an exception in the corresponding context. However, the context entries may have a greater number of bits, two or more, in which case the additional bit or bits may give some additional information about the nature of the exception. For example, they could indicate whether or not the exception is recoverable. The exception status register 1108 may have additional entries 1112 denoting for example different types of errors. For example, a memory error flag may be asserted if the particular exception relates to a parity or error check in the memory, since such errors are likely to fatally compromise operation. The exception status register 1108 enables a quick look up to be made when dealing with an exception raised by that tile. The debug handler tool only needs to further investigate context registers of contexts which are indicated as having been excepted. Note that the exception register 38 for each context may indicate the type of the exception (e-type), in addition to an indication as to whether or not an exception should be raised.
That is, for certain exceptions, additional architectural state 1112 is modified to indicate the precise cause. These exceptions include instruction breaks, instruction data breaks, and BREAK on Sync exceptions. It will readily be appreciated that other exceptions may be logged in the debug hardware.
For exceptions that do not give rise to an exception event, exception handling effectively terminates after the phases of exception detection and exception logging. For exceptions that do give rise to an exception event, the next phase is the raising of an exception event. Exception events occur whenever it is deemed necessary to halt or suspend the execution of a thread that caused an exception. One particular type of exception event discussed herein are break exceptions. Some of these (instruction breaks, instruction data breaks) Break on Sync exceptions are triggered by information held in the debug hardware 122.
The exception model discussed herein defines that break exceptions are recoverable. The system guarantees the ability to restart execution following a break exception event, without requiring a reset event. When a break occurs, architectural state indicating the cause of the break is accessible via a debug system. This architectural state may be in the exception register 38 of each context and/or in the exception context register 1108 in the debug hardware 122.
In the exception event raise phase, when an exception event occurs during the execution of any thread (supervisor or worker), then the instruction issue for this context is stalled. This means that the thread does not get to its normal endpoint (an EXIT instruction for workers or a SYNC instruction for the supervisor). This causes the system-wide synchronisation to stall. A tile output signal exception_raise 1114 is asserted. At this point, an exception event will be raised to the host and/or debug tool. However, all other threads continue execution as normal until they become quiescent. Note that the architecture described herein is of the type that the worker threads would in any event reach a state where they waited for a next compute phase to start following a synchronisation and exchange phase. Therefore normal operation of the processor (other than the excepted thread(s) is not disturbed just because they may have to wait longer because one or more other threads have raised an exception. Moreover, regardless of the execution state of the other threads, the debug tool is able to interrogate the debug visible architectural state of the tile that it is currently dealing with. In particular, the debugger can ascertain the cause of the exception and can restart execution of the excepting thread. This is described in the exception recovery phase.
In the exception decode phase, architectural state in the exception status register 1108 and the exception registers 38 is interrogated to ascertain the cause of the exception.
In the exception recovery phase, there is a clearing down step. The final step in recovering from an exception event on a break is initiated by a write to a clear register 1102. There is one clear register per worker and one for the supervisor context. This write initiates a sequence of events which is dependent on the source of the exception event.
If the exception event was raised by a worker, the steps which are followed are dependent on the state of the supervisor.
If the supervisor is in a state in which it is still waiting for workers, the exception flag in the exception status register 1108 for the worker which raised the event is de-asserted and the worker will continue to execute. Instruction issue for that worker context is re-enabled. The instruction of the program count prior to the exception will be re-fetched and issued. No modification to the supervisor is made. The supervisor continues to wait for all workers to become inactive (reach their EXIT point).
If the supervisor had also raised an exception the same sequence will be followed, except that subsequent exception events such as that raised by the supervisor context will now become visible outside the tile, due to the exception_rai se signal. This is explained in more detail later.
If the supervisor was not waiting for workers, and had not raised an exception the same steps are followed, and the supervisor thread continues to run (either to wait for workers or execute its SYNC instruction).
If the current exception event was raised by the supervisor, its exception flag in the exception status register 1108 is de-asserted and the supervisor resumes execution. The instruction at the program count prior to the exception is re-fetched and issued.
The mechanism by which exceptions are handled on the chip will now be described with reference to Figures 2, 4 and 6.
Figure 6 illustrates schematically the architecture of a computer comprising a plurality of tiles 4. Aspects of the computer shown in Figure 6 are described in more detail later, but for the present time the debugging aspect will be described. As already mentioned, each tile has a debug interface 120 which allows access to the tile through a computer debug interface 1118 on the computer. A debug tool 1120 can access any tile via this computer debug interface 1118. Different mechanisms can be used to enable access by the tool to the tile, but the main function is a read from architectural state on the tile. As explained, each tile can issue an exception raised signal 1114. This is raised on a wire or set of wires which is connected from each individual tile to the debug interface 1118. There, a logic circuit which functions as an OR gate receives all of the exception raised wires and generates a global exception raised signal 1140 to the debug tool, the state of which depends on the state of the inputs to the logic. The logic is such that while any exception raised signal 1114 from an individual tile is active, the global exception raised signal 1140 will be active. When the global raised signal 1140 is active, the debug tool reads the exception status register in each tile to ascertain which tiles have raised an exception. The debug tool resolves an issue for a context, and clears the exception for that context. If on the same tile another context has excepted, the tile exception raised signal will remain asserted. The tile exception raised signal will only be deasserted when no contexts have an unresolved exception. That is, once all issues have been resolved on that tile, an exception flag 1142 in the debug hardware 122 is cleared and the exception raised signal is deasserted for that tile. In this way, each tile that has raised an exception raised signal is dealt with in turn. The debug tool operates until the global exception raised signal 1140 is deasserted which indicates that each tile had caused an exception has now been dealt with. This indicates the end of that particular debug phase. As mentioned, while any individual file is being subject to debug, the other tiles may continue to operate until they reach their own sync request point and are waiting for a sync acknowledgement. For tiles where a thread has been excepted, the particular thread may be debugged while the other threads continue to run to their own end point.
As mentioned, one particular type of exception event which can be caused is a BREAK on Sync event. Such an event is triggered when the supervisor executes a SYNC instruction and a break on sync flag 1100 is set on the tile. If the flag is not set, the synch instruction executes as normal, as described herein. If the flag is set, then a break exception event is raised whenever the supervisor context executes a SYNC instruction (for either internal or external SYNC zones). The break exception event is only raised once all of the worker contexts are inactive, but before the SYNC instruction initiates its synchronisation handshake. The tile therefore will be in a suspended or stalled state because it has not generated a sync request, and therefore will not receive a SYNC ack. However, other tiles on the chip can continue to process their workloads until such time as they have all reached their SYNC points. While that is happening, state on the stalled tile can be inspected in the exception status register 1108 of the debug hardware 122 and exception register 38 of the context and possibly debugged.
Note also that when a worker context reaches a break point, it will effectively stall until the exception event is cleared, and it will not reach its exit state. However, other worker context on the same tile will be unaffected. The consequence of a worker being stalled (and not getting to its exit instruction) is that when the other workers have issued their exit states, the tile will fail to sync because one worker context will still be 'hanging'. However, individual threads and individual tiles can tolerate delays of this kind, because their operation is not dependent on completing programs within a certain time frame. Once the issue has been resolved, and the exception has been cleared, normal operation is resumed.
Reverting to the BREAK on Sync scenario, clearing down a BREAK on Sync allows a tile to continue execution following a BREAK _ on _Sync exception event. The exception event is cleared in the supervisor event clear register 1102 and the supervisor recommences execution. Supervisor execution will be restarted by resuming execution of the SYNC instruction, which will cause the SYNC request to be generated and initiate normal operation.
As explained above, the debug interface 120 on each tile supports the BREAK_on_Sync state configuration, and implements other debug features. The debug interface 120 provides a mechanism by which architectural state on the tile can be interrogated and modified by the external debug tool 1120. Furthermore, the mechanism supports instruction breakpoint channels and data breakpoint channels. In addition, the debug interface enables the BREAK on Sync flag 1100 to be set.
Instruction break point channels are implemented by one or more registers 1104 on the tile in the debug hardware which can hold a program count number. When enabled, an instruction break channel causes a break exception event to be raised prior to issue of any solo instruction (or group of instructions) when the value of the program count register (in the control register 28 of each context) matches the value in the instruction break register, and the criteria specified by instruction break channel configurations registers are met. If no match is detected, normal instruction execution ensues.
Each tile also supports data breakpoint channels 1 1 06. These are implemented as one or more register 1106 in the debug hardware which enable accesses to the file memory to be detected. Data break channels cause break exception events to be raised if any memory class instruction has a b eh avi our which matches criteria specified by data break channel configuration registers. If no match is detected, normal instruction execution ensues. For example, data break registers may specify the start point of a contiguous address range, or specify a range size or a data value.
There follows a fuller explanation of a processor in which the exception handling mechanism may be implemented.
As mentioned, in embodiments the processor module 4 may be implemented as one of an array of interconnected tiles forming a multi-tile processor, wherein each of the tiles may be configured as described above in relation to Figures t to 5.
This is illustrated further in Figure 6 which shows a single chip processor 2, i.e. a single die, comprising an array 6 of multiple processor tiles 4 and an on-chip interconnect 34 connecting between the tiles 4. The chip 2 may be implemented alone on its own single-chip integrated circuit package, or as one of multiple dies packaged in the same IC package. The on-chip interconnect may also be referred to herein as the "exchange fabric" 34 as it enables the tiles 4 to exchange data with one another. Each tile 4 comprises a respective instance of the barrel-threaded processing unit 10 and memory 11, each arranged as described above in relation to Figures 1 to 6. For instance, by way of illustration the chip 2 may comprise of the order of hundreds of tiles 4, or even over a thousand. For completeness, note also that an "array" as referred to herein does not necessarily imply any particular number of dimensions or physical layout of the files 4.
In embodiments each chip 2 also comprises one or more external links 8, enabling the chip 2 to be connected to one or more, external processors on different chips (e.g. one or more other instances of the same chip 2). These external links 8 may comprise any one or more of: one or more chip-to-host links for connecting the chip 2 to a host processor, and/or one or more chip-to-chip links for connecting together with one or more other instances of the chip 2 on the same IC package or card, or on different cards. In one example arrangement, the chip 2 receives work from a host processor (not shown) which is connected to the chip via one of the chip-to-host links in the form of input data to be processed by the chip 2. Multiple instances of the chip 2 can be connected together into cards by chip-to-chip links. Thus a host may access a computer which is architected as a single chip processor 2or as multiple single chip processors 2 possibly arranged on multiple interconnected cards, depending on the workload required for the host application.
The interconnect 34 is configured to enable the different processor tiles 4 in the array 6 to communicate with one another on-chip 2. However, as well as there potentially being dependencies between threads on the same tile 4, there may also be dependencies between the portions of the program running on different tiles 4 in the array 6. A technique is therefore required to prevent a piece of code on one tile 4 running ahead of data upon which it is dependent being made available by another piece of code on another tile 4.
In embodiments, this is achieved by implementing a bulk synchronous parallel (BSP) exchange scheme, as illustrated schematically in Figures 7 and 8.
According to one version of BSP, each tile 4 performs a compute phase 52 and an exchange phase 50 in an alternating cycle, separated from one to the other by a barrier synchronization 30 between tiles. In the case illustrated a barrier synchronization is placed between each compute phase 52 and the following exchange phase 50. During the compute phase 52 each tile 4 performs one or more computation tasks locally on-tile, but does not communicate any results of these computations with any others of the tiles 4. In the exchange phase 50 each tile 4 is allowed to exchange one or more results of the computations from the preceding compute phase to and/or from one or more others of the tiles in the group, but does not perform any new computations until it has received from other tiles 4 any data on which its task(s) has/have dependency It is not excluded that other operations such as internal control-related operations may be performed in the exchange phase. In embodiments the exchange phase 50 does not include any non-time-deterministic computations, but a small number of time-deterministic computations may optionally be allowed during the exchange phase 50. Note also that a tile 4 performing computation may be allowed during the compute phase 52 to communicate with other external system resources external to the array of tiles 4 being synchronized -e.g. a network card, disk drive, or field programmable gate array (FPGA) -as long as this does not involve communication with other tiles 4 within the group being synchronized. The communication external to the tile group may optionally utilise the BSP mechanism, but alternatively may not utilize BSP and may instead use some other synchronization mechanism of its own.
According to the BSP principle, a barrier synchronization 30 is placed at the juncture transitioning from the compute phases 52 into the exchange phase 50, or the juncture transitioning from the exchange phases 50 into the compute phase 52, or both. That is to say, either: (a) all tiles 4 are required to complete their respective compute phases 52 before any in the group is allowed to proceed to the next exchange phase 50, or (b) all files 4 in the group are required to complete their respective exchange phases 50 before any tile in the group is allowed to proceed to the next compute phase 52, or (c) both of these conditions is enforced. In all three variants it is the individual processors which alternate between phases, and the whole assembly which synchronizes. The sequence of exchange and compute phases may then repeat over multiple repetitions. In BSP terminology, each repetition of exchange phase and compute phase is sometimes referred to as a "superstep" (though note that in the literature the terminology is not always used consistently: sometimes each individual exchange phase and compute phase individually is called a superstep, whereas elsewhere, as in the terminology adopted herein, the exchange and compute phases together are referred to as a superstep).
Figure 8 illustrates the BSP principle as implemented amongst a group 4i, 4ii, 4iii of some or all of the tiles in the array 6, in the case which imposes: (a) a barrier synchronization from compute phase 52 to exchange phase 50 (see above). Note that in this arrangement, some tiles 4 are allowed to begin computing 52 whilst some others are still exchanging.
According to embodiments disclosed herein, this type of BSP may be facilitated by incorporating additional, special, dedicated functionality into a machine code instruction for performing barrier synchronization, i.e. the SYNC instruction.
In embodiments, the SYNC function takes this functionality when qualified by an inter-tile mode as an operand, e.g. the on-chip mode: SYNC chip.
The execution unit (EXU) of the execution stage 18 is configured so as, in response to the opcode of the SYNC instruction, when qualified by the on-chip (inter-tile) operand, to cause the supervisor thread in which the "SYNC chip" was executed to be paused until all the tiles 4 in the array 6 have finished running workers. This can be used to implement a barrier to the next BSP superstep. I.e. after all tiles 4 on the chip 2 have passed the barrier, the cross-tile program as a whole can progress to the next exchange phase 50.
Once the supervisor has launched (RUN) all of the threads which it intends to run in the current compute cycle 52, it then executes a SYNC instruction. This triggers the following functionality to be triggered in dedicated synchronization logic 39 on the tile 4, and in a synchronization controller 36 implemented in the hardware interconnect 34. This functionality of both the on-tile sync logic 39 and the synchronization controller 36 in the interconnect 34 is implemented in dedicated hardware circuitry such that, once the SYNC instruction is executed, the rest of the functionality proceeds without further instructions being executed to do so.
Firstly, the on-tile sync logic 39 causes the instruction issue for the supervisor on the tile 4 in question to automatically pause (causes the fetch stage 14 and scheduler 24 to suspend issuing instructions of the supervisor). Once all the outstanding worker threads on the local tile 4 have performed an EXIT, then depending on the status of the BREAK-on-Sync flag, the sync logic 39 sends a synchronization request "sync_req" to the synchronization controller 36 in the interconnect 34. The local tile 4 then continues to wait with the supervisor instruction issue paused. A similar process is also implemented on each of the other tiles 4 in the array 6 (each comprising its own instance of the sync logic 39). Thus at some point, once all the final workers in the current compute phase 52 have EXITed on all the tiles 4 in the array 6, the synchronization controller 36 will have received a respective synchronization request (sync req) from all the tiles 4 in the array 6 that are not subject to an exception. The chip is 'paused' until exceptions have been cleared, and the sync_req issued from cleared tiles. Only then, in response to receiving the sync req from every tile 4 in the array 6 on the same chip 2, the synchronization controller 36 sends a synchronization acknowledgement signal "sync acl<" back to the sync logic 39 on each of the tiles 4. Up until this point, each of the tiles 4 has had its supervisor instruction issue paused waiting for the synchronization acknowledgment signal (sync_ack). Upon receiving the sync_ack signal, the sync logic 39 in the tile 4 automatically unpauses the supervisor instruction issue for the respective supervisor thread on that tile 4. The supervisor is then free to proceed with exchanging data with other tiles 4 via the interconnect 34 in a subsequent exchange phase 50.
Preferably the sync req and sync ack signals are transmitted and received to and from the synchronization controller, respectively, via one or more dedicated sync wires connecting each tile 4 to the synchronization controller 36 in the interconnect 34.
As mentioned previously, in embodiments multiple instances of the chip 2 can be connected together to form an even larger array of tiles 4 spanning multiple chips 2. This is illustrated in Figure 9. Some or all of the chips 2 may be implemented on the same IC package or some or all of the chips 2 may be implemented on different IC packages. The chips 2 are connected together by an external interconnect 72 (via the external links 8 shown in Fig. 7). As well as providing a conduit for exchange of data between tiles 4 on different chips, the external exchange peripheral 72 also provides hardware support for performing barrier synchronization between the tiles 4 on different chips 2.
In embodiments, the SYNC instruction can take at least one further possible value of its mode operand to specify an external, i.e. inter-chip, synchronization: SYNC zone_n, wherein zone n represents an external sync zone. The external interconnect 72 comprises similar hardware logic to that described in relation to Figure 11, but on an external, inter-chip scale. When the SYNC instruction is executed with an external sync zone of two or more chips 2 specified in its operand, this causes the logic in the external interconnect 72 to operate in a similar manner to that described in relation to the internal interconnect 34, but across the tiles 4 on the multiple different chips 2 in the specified sync zone.
That is, in response to an external SYNC, the supervisor instruction issue is paused until all tiles 4 on all chips 2 in the external sync zone have completed their compute phase 52, are not subject to an exception and submitted a sync request. Further, logic in the external interconnect 72 aggregates the local exit states of all these tiles 4, across the multiple chips 2 in the zone in question. Once all tiles 4 in the external sync zone have made the sync request, the external interconnect 72 signals a sync acknowledgment back to the tiles 4. In response to the sync acknowledgement, the tiles 4 on all the chips 2 in the zone resume instruction issue for the supervisor.
In embodiments the functionality of the interconnect 72 may be implemented in the chips 2, i.e. the logic may be distributed among the chips 2 such that only wired connections between chips are required (Figure 9 is schematic).
All tiles 4 within the mentioned sync zone are programmed to indicate the same sync zone via the mode operand of their respective SYNC instructions. In embodiments the sync logic in the external interconnect 72 peripheral is configured such that, if this is not the case due to a programming error or other error (such as a memory parity error), then some or all tiles 4 will not receive an acknowledgement, and therefore that the system will come to a halt at the next external bather, thus allowing a managing external CPU (e.g. the host) to intervene for debug or system recovery. In other embodiments an error is raised in the case where the sync zones do not match. Preferably however the compiler is configured to ensure the tiles in the same zone all indicate the same, correct sync zone at the relevant time.
Figure 10 illustrates an example application of the processor architecture disclosed herein, namely an application to machine intelligence.
As will be familiar to a person skilled in the art of machine intelligence, machine intelligence begins with a learning stage where the machine intelligence algorithm learns a knowledge model. The model comprises a graph of interconnected nodes (i.e. vertices) 102 and edges (i.e. links) 104. Each node 102 in the graph has one or more input edges and one or more output edges. Some of the input edges of some of the nodes 102 are the output edges of some others of the nodes, thereby connecting together the nodes to form the graph. Further, one or more of the input edges of one or more of the nodes 102 form the inputs to the graph as a whole, and one or more of the output edges of one or more of the nodes 102 form the outputs of the graph as a whole. Sometimes a given node may even have all of these: inputs to the graph, outputs from the graph and connections to other nodes. Each edge 104 communicates a value or more often a tensor (n-dimensional array {a matrix is a 2-dimensional array, tensors are n-dimensional arrays}, these forming the inputs and outputs provided to and from the nodes 102 on their input and output edges, respectively.
Each node 102 represents a function of its one or more inputs as received on its input edge or edges, with the result of this function being the output(s) provided on the output edge or edges. Each function is parameterized by one or more respective parameters (sometimes referred to as weights, though they need not necessarily be multiplicative weights). In general the functions represented by the different nodes 102 may be different forms of function and/or may be parameterized by different parameters.
Further, each of the one or more parameters of each node's function is characterized by a respective error value. Moreover, a respective condition may be associated with the error(s) in the parameter(s) of each node 102. For a node 102 representing a function parameterized by a single parameter, the condition may be a simple threshold, i.e. the condition is satisfied if the error is within the specified threshold but not satisfied if the error is beyond the threshold. For a node 102 parameterized by more than one respective parameter, the condition for that node 102 having reached an acceptable level of error may be more complex. For example, the condition may be satisfied only if each of the parameters of that node 102 falls within respective threshold. As another example, a combined metric may be defined combining the errors in the different parameters for the same node 102, and the condition may be satisfied if the value of the combined metric falls within a specified threshold, but otherwise the condition is not satisfied if the value of the combined metric is beyond the threshold (or vice versa depending on the definition of the metric). Whatever the condition, this gives a measure of whether the error in the parameter(s) of the node falls below a certain level or degree of acceptability. In general, any suitable metric may be used. The condition or metric may be the same for all nodes, or different for different respective ones of the nodes.
In the learning stage the algorithm receives experience data, i.e. multiple data points representing different possible combinations of inputs to the graph. As more and more experience data is received, the algorithm gradually tunes the parameters of the various nodes 102 in the graph based on the experience data so as to try to minimize the errors in the parameters. The goal is to find values of the parameters such that the output of the graph is as close as possible to a desired output for a given input. As the graph as a whole tends toward such a state, the graph is said to converge. After a suitable degree of convergence the graph can then be used to perform predictions or inferences, i.e. to predict an outcome for some given input or infer a cause for some given output.
The learning stage can take a number of different possible forms. For instance, in a supervised approach, the input experience data takes the form of training data, i.e. inputs which correspond to known outputs. With each data point, the algorithm can tune the parameters such that the output more closely matches the known output for the given input. In the subsequent prediction stage, the graph can then be used to map an input query to an approximate predicted output (or vice versa if'making an inference). Other approaches are also possible. For instance, in an unsupervised approach, there is no concept of a reference result per input datum, and instead the machine intelligence algorithm is left to identify its own structure in the output data. Or in a reinforcement approach, the algorithm tries out at least one possible output for each data point in the input experience data, and is told whether this output is positive or negative (and potentially a degree to which it is positive or negative), e.g. win or lose, or reward or punishment, or such like. Over many trials the algorithm can gradually tune the parameters of the graph to be able to predict inputs that will result in a positive outcome. The various approaches and algorithms for learning a graph will be known to a person skilled in the art of machine learning.
According to an exemplary application of the techniques disclosed herein, each worker thread is programmed to perform the computations associated with a respective individual one of the nodes 102 in a machine intelligence graph. In this case at least some of the edges 104 between nodes 102 correspond to the exchanges of data between threads, and some may involve exchanges between tiles. Furthermore, the individual exit states of the worker threads are used by the programmer to represent whether or not the respective node 102 has satisfied its respective condition for convergence of the parameter(s) of that node, i.e. whether or not the error in the parameter or parameters has fallen within the acceptable level or region in error space.
In the case of a multi-tile arrangement 6, each tile runs a subgraph of the graph. Each subgraph comprises a supervisor subprogram comprising one or more supervisor threads, and a set of worker threads in which some or all of the workers may take the form of codelets.
In such applications, or indeed any graph-based application where each worker thread is used to represent a respective node in a graph, the "codelet" comprised by each worker may be defined as a software procedure operating on the persistent state and the inputs and/or outputs of one vertex, wherein the codelet: * is launched on one worker thread register context, to run in one barrel slot, by the supervisor thread executing a "run" instruction; * runs to completion without communication with other codelets or the supervisor (except for the return to the supervisor when the codelet exits); * has access to the persistent state of a vertex via a memory pointer provided by the "run" instruction, and to a non-persistent working area in memory which is private to that barrel slot; and * executes "EXIT" as its last instruction, whereupon the barrel slot which it was using is returned to the supervisor.
To update a graph (or sub-graph) means to update each constituent vertex once, in any order consistent with the causality defined by the edges. To update a vertex means to run a codelet on the vertex state. A codelet is an update procedure for vertices -one codelet is usually associated with many vertices. The supervisor executes one RUN instruction per vertex, each such instruction specifying a vertex state address and a codelet address.
It will be appreciated that the above embodiments have been described by way of example only.
Further, the terms "supervisor" and "worker" do not imply any specific responsibilities except where otherwise explicitly stated, and particularly do not in themselves necessarily limit to the above-described scheme in which a supervisor thread relinquishes its time slot to a worker, and so forth. In general, worker thread may refer to any thread to which some computational task is allocated. The supervisor may represent any kind of overseeing or coordinating thread responsible for actions such as: assigning workers to barrel slots, and/or performing barrier synchronizations between multiple threads, and/or performing any control-flow operation (such as a branch) in dependence on the outcome of more than one thread.
Where reference is made to a sequence of interleaved time slots, or the like, this does not necessarily imply that the sequence referred to makes up all possible or available slots. For instance, the sequence in question could be all possible slots or only those currently active. It is not necessarily precluded that there may be other potential slots that are not currently included in the scheduled sequence.
The term tile as used herein does not necessarily limit to any particular topography or the like, and in general may refer to any modular unit of processing resource, comprising a processing unit 10 and corresponding memory 11, in an array of like modules, typically at least some of which are on the same chip (i.e. same die).
Further, the scope of the present disclosure is not limited to a time-deterministic internal interconnect or a non-time-deterministic external interconnect. The synchronization and debug mechanisms disclosed herein can also be used in a completely time-deterministic arrangement, or a completely non-time-deterministic arrangement.
Other applications and variants of the disclosed techniques may become apparent to a person skilled in the art once given the disclosure herein. The scope of the present disclosure is not limited by the described embodiments but only by the accompanying claims.
Claims (9)
- Claims 1. A method of debugging a processing system comprising an arrangement of tiles and an interconnect for communicating between the tiles, wherein each tile comprises an execution unit for executing machine code instructions, each being an instance of a predefined set of instruction types in an instruction set of the processing system, [the interconnect being operable to conduct communications between a group of some or all of the tiles according to a bulk synchronous parallel scheme whereby each of the tiles in the group perform an on-tile compute phase followed by an inter-tile exchange phase with the exchange phase being held back until all of the tiles in the group have completed the compute phase], the method comprising reading an exception indicator in a debug tool, and accessing an exception raised signal of one of the tiles, accessing debug hardware on the one of the tiles to ascertain a cause of the exception raised by the tile, performing a debug operation to resolve the exception, and deasserting the exception raised signal by the tile when the exception has been resolved.
- 2. A method according to claim 1 comprising dealing with a plurality of exception raised signals in sequence until the exception indicator is deasserted.
- 3. A method according to claim 1 or 2 wherein tiles which are not in an excepted state continue to perform their on-tile compute phase, but do not commence their inter-tile exchange phase while one or more other tile is in an excepted state.
- 4. A method according to any preceding claim wherein, after the excepted state has been resolved, the tile that issued the exception raised signal continues with its compute phase until it reaches a synchronisation point at which it issues a synchronisation request.
- 5. A method according to any preceding claim wherein resolving the excepted state comprises accessing at least one of a set of debug registers to identify the cause of the excepted state.
- 6. A processing system comprising an arrangement of tiles and an interconnect for communication between the tiles, wherein: each tile comprises an execution unit for executing machine code instructions, each being an instance of a predefined set of instruction types in an instruction set of the processing system; wherein each tile is configured when in an excepted state to raise an exception signal and to pause instruction issue until the excepted state has been resolved; wherein each tile comprises a local debug interface to enable access by the debug tool to resolve the excepted state, and wherein the processing system further comprises exception handling logic which is connected to receive the exception signal raised by any of the tiles, and to provide an asserted exception indicator when at least one of the inputs to the logic receives an exception signal from a tile.
- 7. A processing system according to claim 6 wherein each tile comprises a set of debug registers which enable the execution unit to recognise an excepted state.
- 8. A processing system according to claim 7 wherein the set of debug registers comprises an instruction data break register which identifies data related criteria at which to raise an exception.
- 9. A processing system according to claim 7 wherein one of the set of debug registers comprises an instruction break register which identifies one or more program count at which an exception is to be raised.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
GB2002863.5A GB2583574B (en) | 2018-12-21 | 2018-12-21 | Handling Exceptions in a Multi-Tile Processing Arrangement |
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
GB1821067.4A GB2580136B (en) | 2018-12-21 | 2018-12-21 | Handling exceptions in a multi-tile processing arrangement |
GB2002863.5A GB2583574B (en) | 2018-12-21 | 2018-12-21 | Handling Exceptions in a Multi-Tile Processing Arrangement |
Publications (3)
Publication Number | Publication Date |
---|---|
GB202002863D0 GB202002863D0 (en) | 2020-04-15 |
GB2583574A true GB2583574A (en) | 2020-11-04 |
GB2583574B GB2583574B (en) | 2021-05-19 |
Family
ID=70278779
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
GB2002863.5A Active GB2583574B (en) | 2018-12-21 | 2018-12-21 | Handling Exceptions in a Multi-Tile Processing Arrangement |
Country Status (1)
Country | Link |
---|---|
GB (1) | GB2583574B (en) |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20070180322A1 (en) * | 2006-01-11 | 2007-08-02 | Seiko Epson Corporation | Debug support device, and program for directing computer to perform debugging method |
US20160062863A1 (en) * | 2013-04-09 | 2016-03-03 | Siemens Aktiengesellschaft | Multicore processor system having an error analysis function |
-
2018
- 2018-12-21 GB GB2002863.5A patent/GB2583574B/en active Active
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20070180322A1 (en) * | 2006-01-11 | 2007-08-02 | Seiko Epson Corporation | Debug support device, and program for directing computer to perform debugging method |
US20160062863A1 (en) * | 2013-04-09 | 2016-03-03 | Siemens Aktiengesellschaft | Multicore processor system having an error analysis function |
Also Published As
Publication number | Publication date |
---|---|
GB2583574B (en) | 2021-05-19 |
GB202002863D0 (en) | 2020-04-15 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US11893390B2 (en) | Method of debugging a processor that executes vertices of an application, each vertex being assigned to a programming thread of the processor | |
US11593185B2 (en) | Synchronization in a multi-tile processing arrangement | |
CA3021407C (en) | Combining states of multiple threads in a multi-threaded processor | |
US11550591B2 (en) | Scheduling tasks in a multi-threaded processor | |
US11645081B2 (en) | Handling exceptions in a multi-tile processing arrangement | |
US10628377B2 (en) | Synchronization in a multi-tile processing arrangement | |
US11449338B2 (en) | Handling exceptions in a multi-tile processing arrangement | |
US10606641B2 (en) | Scheduling tasks in a multi-threaded processor | |
US11635966B2 (en) | Pausing execution of a first machine code instruction with injection of a second machine code instruction in a processor | |
GB2583574A (en) | Handling exceptions in a multi-tile processing arrangement |