Patent Application
20130275735
In a multi-sequencer computer system or computer platform, technical as well as economic constraints may justify an asymmetric organization of specialized computing resources or processors or processing units. In this application, a processor or processing unit may also be referred to, hereinafter, as a sequencer or a processing core. In general, an operating system may not scale well on a computer platform having implemented a large number of sequencers. In particular, the operating system may not be able to handle an asymmetric organization of multiple sequencers efficiently.
The organization of a computer platform may be asymmetric with regard to the types of sequencer's that the platform has implemented. For example, there may be sequencers that are “visible” to the operating system (OS), referred to herein as OS-visible sequencers, and are managed by OS kernels. The OS kernels may be able to control the privilege states of OS-visible sequencers and to provide exception handlers to process interrupts and/or exceptions during execution of threads by the OS-visible sequencers. On the other hand, there may be sequencers that are “invisible” to the OS, referred to herein as OS-sequestered sequencers. OS-sequestered sequencers may not execute operating system codes and may be managed by a sequencer manager.
Recently, a multi-sequencer multiple-instruction-flow-multiple-data flow (MIMI)) ISA (Instruction Set Architecture) extension has been proposed that defines a set of instructions allowing OS-sequestered sequencers to be treated as ting-3 user-level architectural resources for concurrent execution of multiple user-level threads in the single OS thread context. Therefore, OS-sequestered sequencers become available to user-level applications. By using the user-level multi-sequencer MIMI ISA extension, an application thread running on an OS-visible sequencer may be able to use the additional sequencers which are sequestered from the OS The multi-sequencer MIMD ISA extension allows applications to scale, independent of the OS, for a large number of sequencers by submitting work to sequencers that may or may not be visible to the operating system. In the mean time, the operating system may continue to run on a subset of sequence is that are visible to the OS. The user-level MIMD ISA extension allows execution of user-level threads to run on OS-sequestered sequencers, and the user-level threads may run concurrently with the OS thread that created the user-level threads and tuns on an OS-visible sequencer.
The invention will be understood and appreciated more fully from the following detailed description of embodiments of the invention, taken in conjunction with the accompanying drawings of which:
It will be appreciated that for simplicity and clarity of illustration, elements shown in the figures have not necessarily been drawn to scale. For example, the dimensions of some of the elements may be exaggerated relative to other elements for clarity.
In the following detailed description, numerous specific details are set forth in order to provide a thorough understanding of embodiments of the invention However it will be under stood by those of ordinary skill in the art that the embodiments of the invention may be practiced without these specific details. In other instances, well-known methods and procedures have not been described in detail so as not to obscure the embodiments of the invention.
Some portions of the detailed description in the following are presented in terms of algorithms and symbolic representations of operations on data bits or binary digital signals within a computer memory. These algorithmic descriptions and representations may be the techniques used by those skilled in the data processing arts to convey the substance of the work to others skilled in the art.
An algorithm is here, and generally, considered to be a self-consistent sequence of acts or operations leading to a desired result. These include physical manipulations of physical quantities. Usually, though not necessarily, these quantities take the form of electrical or magnetic signals capable of being stored, transferred, combined, compared, and otherwise manipulated. It has proven convenient at times, principally for reasons of common usage, to refer to these signals as bits, values, elements, symbols, characters, terms, numbers or the like. It should be understood, however, that all of these and similar terms are to be associated with the appropriate physical quantities and are merely convenient labels applied to these quantities.
Unless specifically stated otherwise, as apparent from the following discussions, it is appreciated that throughout the specification discussions utilizing terms such as “processing”, “computing”, “calculating”, “determining”, or the like, refer to the action and/or processes of a computer or computing system, or similar electronic computing device, that manipulate and/or transform data represented as physical, such as electronic, quantities within the computing system's registers and/or memories into other data similarly represented as physical quantities within the computing system's memories, registers or other such information storage, transmission or display devices.
Some embodiments of the invention may be implemented, for example, using a machine-readable medium or article which may store an instruction or a set of instructions that, if executed by a machine, cause the machine to perform a method and/or operations in accordance with embodiments of the invention Such machine may include, for example, any suitable processing platform, computing platform, computing device, processing device, computing system, processing system, computer, processor, or the like, and may be implemented using any suitable combination of hardware and/or software. The machine-readable medium or article may include, for example, any suitable type of memory unit, memory device, memory article, memory medium, storage device, storage article, storage medium and/or storage unit, e.g., memory, removable or non-removable media, erasable or non-erasable media, writeable or re-writeable media, digital or analog media, hard disk, floppy disk, Compact Disk Read Only Memory (CD-ROM), Compact Disk Recordable (CD-R), Compact Disk Rewriteable (CD-RW), optical disk, magnetic media, various types of Digital Versatile Disks (DVDs), a tape, a cassette, or the like The instructions may include any suitable type of code, for, example, source code, target code, compiled code, interpreted code, executable code, static code, dynamic code, or the like, and may be implemented using any suitable high-level, low-level, object-oriented, visual, compiled and/or interpreted programming language, e.g., C, C++, Java, BASIC, Pascal, Fortran, Cobol, assembly language, machine code, or the like.
Embodiments of the invention may include apparatuses for performing the operations herein. These apparatuses may be specially constructed for the desired purposes, or they may include a general-purpose computer selectively activated or reconfigured by a computer program stored in the computer. Such a computer program may be stored in a computer readable storage medium, such as, but is not limited to, any type of disk including floppy disks, optical disks, CD-ROMs, magnetic-optical disks, read-only memories (ROM), random access memories (RAM), electrically programmable read-only memories (EPROM), electrically erasable and programmable read only memories (EEPROM), magnetic or optical cards, or any other type of media suitable for storing electronic instructions, and capable of being coupled to a computer system bus.
The processes and displays presented herein are not inherently related to any particular computer or other apparatus. Various general-purpose systems may be used with programs in accordance with the teachings herein, or it may prove convenient to construct a more specialized apparatus to perform the desired method. The desired structure for a variety of these systems will appear from the description below. In addition, embodiments of the invention are not described with reference to any particular programming language. It will be appreciated that a variety of programming languages may be used to implement the teachings of the invention as described herein.
In the following description, various figures, diagrams, flowcharts, models, and descriptions are presented as different means to effectively convey and illustrate different embodiments of the invention that are proposed in this application. It shall be understood by those skilled in the art that they are provided merely as illustrative samples, and shall not be constructed as limiting.
Multi-sequencer system 100, as shown in
Sequencer manager 114 may be implemented as a driver, an extension of an operating system such as an extension of OS 112, any other hardware or a combination of hardware and software. Multi-sequencer hardware 104 may be symmetric or asymmetric in terms of the types of sequencers 116, 118, 120 and/or 122 one or more of the sequencers 116, 118, 120 and/or 122 may be a physical processor with its own set of execution resources. Alternatively, one or more of the sequencers 116, 118, 120 and/or 122 may be a logical processor (of, for example, a hyper-threaded processor), that may share physical resources with other logical processors. Sequencers 116, 118, 120 and 122 may be implemented in a single hardware core or in multiple separate hardware cores.
A non-exhaustive list of examples for multi-sequencer system 100 may include a desktop personal computer, a work station, a server computer, a laptop computer, a notebook computer, a hand-held computer, a personal digital assistant (PDA), a mobile telephone, a game console, and the like.
A non-exhaustive list of examples for memory 102 may include one or any combination of the following semiconductor devices, such as synchronous dynamic random access memory (SDRAM) devices, RAMBUS dynamic random access memory (RDRAM) devices, double data rate (DDR) memory devices, static random access memory (SRAM) devices, flash memory (FM) devices, electrically erasable programmable read only memory (E-EPROM) devices, non-volatile random access memory (NVRAM) devices, universal serial bus (USB) removable memory devices, and the like; optical devices, such as compact disk read only memory (CD ROM), and the like; and magnetic devices, such as a hard disk, a floppy disk, a magnetic tape, and the like. Memory 102 may be fixed within or removable from system 100.
According to one illustrative embodiment of the invention, user-level program 106 may access a thread library via Application Program Interface (API) 110 API 110 may provide creation, control, and synchronization of threads to user-level program 106. Scheduler 108 may schedule instructions of the threads for execution on multi-sequencer hardware 104.
A non-exhaustive list of examples for multi-sequencer hardware 104 may include a plurality of single-threaded or multi-threaded central processing unit (CPU), digital signal processor (DSP), reduced instruction set computer (RISC), complex instruction set computer (CISC) and the like. Moreover, multi-sequencer 104 may be part of an application specific integrated circuit (ASIC) or may be a part of an application specific standard product (ASSP).
According to one illustrative embodiment of the invention, multi-sequencer hardware 104 may be a single-core processor having implemented sequencers 116, 118, 120 and 122 (SID0-SID3) that may be, for example, logical processors. Single-core processor 104 may be able to support, for example, concurrent, simultaneous, and/or switch-on-event multi-threading. According to one illustrative embodiment, each of the sequencers SID0-SID3 may have its own next-instruction-pointer logic, while the same single-core processor 104 may execute all the thread instructions. Each logical processor SID1-SID3 may maintain its own version of the architecture state, although execution resources of single-core processor 104 may be shared among concurrent, simultaneous, and/or switch-on-event threads.
While the illustrative embodiment of multi-sequencer hardware 104 discussed above may refer to single thread per sequencer, a person skilled in the art shall appreciate that the application herein is not limited to single-threaded processors. The techniques discussed herein may equally be employed in any Chip Multiprocessing (CMP) or Simultaneous Multithreading Processor (SMP) system, including in a hybrid system with CMP processors and SMP processors where each core of a CMP processor is a SMP processor or a Switch-On-Event Multiprocessor (SoeMP). For example, the techniques disclosed herein may be used in a system that includes multi-threaded processing cores in a single chip hardware package 104.
As is discussed above, sequencers SID0-SID3 are not necessarily uniform and may be asymmetric with respect to factors that may affect computation quality such as, for example, processing speed, processing capability, and power consumption For example, sequencer SID0 may be a “heavyweight” sequencer in that it may be designed to process most of the instructions in a given instruction set architecture (e.g. IA-32 Instruction Set Architecture associated with a 32-bit processor). Whereas, sequencer SID1 may be a “lightweight” sequencer in that it may process a selected subset of those instructions. In another embodiment of the invention, a heavyweight processor may be one that processes instructions at a faster late than a lightweight processor. In addition, some sequencers may be visible and other sequencers may be invisible to an operating system. For example, sequencer SID0 may be visible to OS 112, whereas sequencers SID1 to SID3 may be sequestered from OS 112 However, this does not mean that heavyweight sequencers are OS-visible and/or lightweight sequencers are OS-sequestered. An operating system does not schedule instructions on a sequencer that is in a sequestered state (such a sequencer is referred to herein as an OS-sequestered sequencer).
According to some illustrative embodiments of the invention, multi-sequencer system 100 may be able to provide services of persistent user-level threads (PULTs), as detailed below with reference to
Hardware 200 may correspond to multi-sequencer hardware 104 that forms a part of multi-sequencer system 100 in
According to one embodiment of the invention, execution of an application 301, which may be, for example, user-level program 106 of
According to one embodiment of the invention, a sequencer manager 310 may have control access to both sequencers 311 and 312 that are OS-visible, and another set of sequencers, for example, sequencers 313, 314, 315, 316, 317 and 318 that are sequestered from OS 302 and not managed by the kernels of OS 302. Sequencers 313, 314, 315, 316, 317 and 318 are OS-sequestered, or “OS-invisible”, sequencers.
According to one embodiment of the invention, an OS-scheduled thread, (e.g., a native thread that potentially includes a user-level thread) may interact with sequencer manager 310 through an interface that may be provided through hardware or software or a combination of both. In order to use sequencers that are not visible to OS 302, such as sequencers 313, 314, 315, 316, 317 and 318, an OS-scheduled thread, for example the native thread 303, may send a request to sequencer manager 310 to create a user-level thread, for example, thread 304, to turn on one of the OS-sequestered sequencers, for example, sequencer 313. According to one embodiment of the invention, user-level thread 304 may run on sequencer 313 independent of the OS context switch activities of the native thread 303 and therefore may be a persistent user-level thread (PULT).
In the following description, native thread 303 is used as an example to illustrate the creation and execution of a persistent user-level thread. A person skilled in the art will appreciate that following description also applies to other threads such as, for example a user level thread that belongs to a native thread.
According to one illustrative embodiment of the invention, OS-sequestered sequencer 313 may execute PULT 304 in a virtual address space (VAS) 320 that is shared by user-level thread 303 running on OS-visible sequencer 311. OS 302 may create VAS 320 for the execution of application 301. According to one illustrative embodiment of the invention, by sharing a common VAS 320, sequencer manager 310 may be able to capture an execution environment, for example, a virtual address map, of user-level thread 303 that runs on OS-visible sequencer 311, and then apply this execution environment to PULT 304 that runs on OS-sequestered sequencer 313. According to one embodiment of the invention, OS-sequestered sequencer 313 may have one or more snoopy TLB's (Translation Lookaside Buffer) that are able to automatically track future changes in the virtual address space. A TLB may be a small cache present in a processor that keeps track of virtual to physical address translation for quick lookup. A processor may use this TLB to avoid unnecessary page walks to translate virtual address to physical address.
According to one embodiment of the invention, multi-sequencer system 400 may include a sequencer manager 410 that manages a plurality of OS-sequestered sequencers, for example, sequencers 411, 412, 413 and 414, and other OS-visible sequencers, for example, sequencers 405 and 406. An operating system (OS) 402 may manage execution of one or mote user-level applications, for example, applications 401A and 401B. OS 402 manages applications 401A and 401B by creating one or more threads including, for example, user-level threads 403 and 404, which run on OS-visible sequencers 405 and 406, and other native threads.
Sequencer manager 410 may provide execution resources to applications 401A and 401B through access to sequencers 411, 412, 413 and 414, which are sequestered from OS 402 and not managed by OS 402. The access may be provided through requests made by threads 403 and 404 running on OS-visible sequencers 405 and 406 or other native threads scheduled by the OS. As a result, applications 401A and 401B may have access to OS-sequestered sequencers 411, 412, 413 and 414 via user-level threads 403 and/or 404 or other native threads.
According to one embodiment of the invention, sequencer manager 410 may allocate various time slots on sequencers 411, 412, 413 and 414 for execution of one or more persistent user-level threads created by user-level threads 403 and 404 for applications 401A and 401B. For example, symbols t1, t2, t3, t4, t5 and t6 in
According to one embodiment of the invention, sequencer manager 410 may set up one or more interrupt descriptor tables, for example, tables 421, 422, 423 and 424, for one or more OS-sequestered sequencers, for example, sequencers 411, 412, 413 and 414. An interrupt descriptor table, for example, table 421, may include an event handler 431 to handle events occurring on sequencer 411. During execution time, for example, t3 of PULT 13 on an OS-sequestered sequencer, for example, sequencer 411, PULT T3 may encounter an event or situation such as, for example, a page fault or a system call, that requires handling or attention of OS 402. Event handler 431 may then suspend the execution of PULT T3 and save the execution state, at the time the event happens, of PULT T3.
According to another embodiment of the invention, an OS-invisible sequencer, for example, sequencer 411 may directly interrupt the execution of an OS-visible sequencer, for example, sequencer 405 that invoked the PULT T3, when encountering a page fault or a system call. Sequencer 411 may then send a trigger to sequencer manager 410 requesting handling of the interruption. According to yet another embodiment of the invention, OS-sequestered sequencer 411 may program a ring-0 service channel or channel to handle a page fault or system call. A channel may be a register holding architectural state that includes, for example, a trigger-response mapping where the trigger is an architecturally defined set of one or more processor conditions and the response is a service routine. In addition, a channel may be programmed, for example, by a user-level instruction. When the page fault or system call occurs, OS-sequestered sequencer-411 may invoke a handler whose address is specified in the channel.
PULT T3 may remain suspended until a time when a new user-level thread, for example, thread 409, which runs the same application 401A, is scheduled to run on an OS-visible sequencer. Thread 409, also referred to as a service thread, may make a request to sequencer manager 410 to confirm or verify whether there are any PULTs that have been suspended and are waiting for OS services which may be provided, for example, through a proxy execution as described below in detail. A proxy execution may be performed through thread 409 impersonating or imitating PULT T3, by picking up the state of PULT T3 from sequencer manager 410. The execution state of PULT T3 may be resumed by service thread 409 which shares a common address space, for example, a commonly shared virtual address space (VAS) (
The proxy execution discussed above may create an illusion of symmetry to an application programmer and therefore hide the inherent asymmetric nature of how a multi-sequencer system, for example, system 100, is built.
An ISA defines a logical view of a system as seen by an assembly language programmer, binary translator, assembler or the like. ISA 500 may include a logical storage 502 and an instruction set 504. Logical storage 502 may define a visible memory hierarchy, addressing scheme, register set, etc for a multi-sequencer system such as, for example, multi-sequencer system 100 as in
Instruction set 504 may include the instruction set known as IA-32 instruction set and, according to illustrative embodiments of the invention, its extension although the present invention is not limited in this respect and other instruction sets are possible. According to one embodiment of the invention, instruction set 504 may include, for example, an E-SXFR instruction that includes a SXFR control-transfer instruction and its extension, and an E-SEMONITOR instruction that includes a SEMONITOR monitoring instruction and its extension. Compared to SXFR and SEMONITOR, which can be performed between two sequencers belonging to the same OS native thread, E-SXFR and E-SEMONITOR can be performed between two sequencers that belong to two distinct OS native threads respectively, which may run at different privilege levels. For example, one thread may be a user-level application that is subject to OS context switch, and the other thread may be a PULT that runs persistently in the privilege level, like a device driver, and is not subject to OS context switch.
According to one embodiments of the invention, the E-SXFR instruction may be used to send a signal from a first sequencer to a second sequencer, and the E-SEMONITOR instruction may be used to configure the second sequencer to monitor for signals coming from the first sequencer. Furthermore, the E-SXFR control transfer and E-MONITOR monitoring instructions are sequencer aware, and may be used to construct other sequencer aware composite instructions.
According to one embodiment of the invention, an E-SXFR instruction may have a sample instruction format as shown in
According to one embodiment, operand 604 may correspond to a sequencer ID (SID) for a destination/target sequencer to which the E-SXFR instruction signal is sent. Operand 606 may include a scenario or control message, which may be an architecturally defined identifier code representing a condition or anticipated event. A scenario may be used to affect asynchronous control transfer as described in detail with reference to TABLE 1.
According to one embodiment, operand 608 may include a parameter that conditions the execution of instructions on a sequencer that executes the E-SXFR instruction. Examples of parameters, as shown in
According to one embodiment, operand 610 may include a scenario-specific payload or data message. For example, in the case of a FORK scenario, the payload may include an instruction pointer at which execution on the sequencer identified by operand 604 is to commence. According to another embodiment, payload operand 610 may include an instruction pointer, a stack pointer, a set of control registers, etc. Addresses contained in payload operand 610 may be expressed in a variety of addressing modes such as, for example, literal, register indirect, and/or base/offset addressing.
According to one embodiment, operand 612 may specify a Touting method or function on the SID contained in operand 604. The touting function controls whether the signal, which is created as a result of executing the E-SXFR instruction, is sent as a broadcast, a unicast, or a multicast signal. The touting function may also encode information such as, for example, topology-specific hint that may be used to assist an underlying inter-sequencer interconnect in routing to deliver the signal.
According to one embodiment, the E-SEMONITOR instruction may have a sample instruction format as shown in
The E-SEMONIIOR instruction may map a scenario specified in operand 704 to a SIDEIP specified in operand 706. Mapping of a scenario to a SIDEIP may be referred to herein as mapping of a “service channel”. Operand 708 may allow a programmer to enter one or more control parameters to specify or control how a particular service channel is to be serviced. For example, a programmer may use the E-SEMONITOR instruction to program the service channels that a sequencer may monitor.
Based on the above description, it will be appreciated that both the E-SXFR and E-SEMONITOR instructions are “sequencer-aware” instructions in that they include operands that may identify particular sequencers.
According to one embodiment of the invention, an OS-sequestered sequencer 822 may execute a persistent user-level thread as indicated at block 801. The persistent user-level thread may be referred to herein as PULT-1 for convenience of description. PULT-1 may be created, for example, by a first user-level thread that is executed by a first OS-visible sequencer. During execution, PULT-1 may encounter one or more events that require services or attention of an OS. A second user-level thread, which may or may not be the same user-level thread as the first user-level thread that created PULT-1 and execute on a second OS-visible sequencer for example OS-visible sequencer 821, may provide the requested OS services to PULT-1 as described below in detail.
According to one embodiment of the invention, PULT-1 may encounter an event at time t1 requiring OS services, as indicated at block 802 in
According to one illustrative embodiment of the invention, at some point of time, an OS-scheduled thread such as a new user-level thread may start to run on OS-visible sequencer 821 as indicated at block 805. This new user-level thread may be executed for the same application as PULT-1 This thread may be able to provide OS services through, for example, proxy execution, to PULT-I and therefore may be referred to herein as a service thread. Before providing OS services, the service thread may check or verify with sequencer manager 820, as indicated at block 806, whether there are any suspended PULTs that are waiting for OS services.
The service thread may identify PULT-1 as a thread waiting for OS services. According to one embodiment of the invention, the service thread may then proceed to imitate, in other words, impersonate, PULT-1 by picking up the execution state of PULT-1 through, for example, their commonly shared virtual address space (VAS) as indicated at block 807. PULT-1, imitated by the service thread, may then get executed on the OS-visible sequencer executing the service thread to receive OS services. As a result, OS services are provided to PULT-1, as indicated at block 808, through proxy execution of the service thread and the condition under which PULT-1 was suspended is now serviced by the operating system. According to one embodiment of the invention, a user level thread that runs on an OS-visible sequencer, for example, OS-visible sequencer 821, may detect a suspended PULT by monitoring an address location using an scenario. A service scenario is a condition that may be used to monitor, for example, cache misses, and if a cache miss occurs which causes, for example, a threshold being crossed, a handler in the service channel may be invoked.
According to one embodiment of the invention, after the operating system completes servicing the condition of the suspended PULT-1, the post-execution state of PULT-1 may be saved and subsequently returned to OS-sequestered sequencer 822 through sequencer manager 820 as indicated at block 809. After receiving the post execution state from the service thread, OS-sequestered sequencer. 822 may resume its execution of PULT-1 as indicated at block 810.
Table 1 contains a list of sample scenarios according to some illustrative embodiments of the invention.
According to illustrative embodiments of the invention, scenarios in Table 1 may be divided into egress services and ingress services. Among each service, operations and control messages may be characterized by their OS-visibility which may be part of the characteristics of a shred processing unit (SPU). Among egress services, there are scenarios of operations that fall into the category of resource not available (RNA), which is a category for events generated during execution on a sequencer due to access to a resource not being available on an OS-sequestered sequencer. According to one embodiment of the invention, scenarios that fall into the category of RNA may include, for example, page fault handling and/or I/O access on the OS-sequestered sequencer which is incapable of directly activating OS service.
According to illustrative embodiments of the invention, accompanying these scenarios of operations described above is a set of outgoing control messages. In one illustrative embodiment, the messages may include, for example, the messages as listed under egress service scenarios in Table 1. However, the invention is not limited in this respect and other messages may be used. The control messages go to one or more OS-visible sequencers.
In relation to the egress service scenarios discussed above, there are ingress service scenarios that fall into the category of an OS-visible sequencer providing resource to service requests from an OS-sequestered sequencer. As is shown in Table 1 under ingress service scenarios, these services correspond directly to the scenarios under egress service but are available from OS-visible sequencers in one illustrative embodiment, the incoming control messages, accompanying the ingress services, may include the messages as listed under ingress service scenarios in Table 1 however, the invention is not limited in this respect and other messages may be used the incoming messages may come from one or more OS-sequestered sequencers.
According to illustrative embodiments of the invention, instructions that serve the egress services are, for example, relatively simple SSE13 instructions in terms of core complexity, and instructions that serve, for example, the ingress services or requests from an OS-sequestered sequencer are, in general, more complex.
While certain features of the invention have been illustrated and described herein, many modifications, substitutions, changes, and equivalents will now occur to those of ordinary skill in the art. It is, therefore, to be understood that the appended claims are intended to cover all such modifications and changes as fall within the spirit of the invention.
This application is a continuation of U.S. patent application Ser. No. 13/221,119, filed Aug. 30, 2011, which is a continuation of U.S. patent application Ser. No. 11/239,475, filed Sep. 30, 2005, now U.S. Pat. No. 8,028,295, issued Sep. 27, 2011, the content of which is hereby incorporated by reference.
| Number | Date | Country | |
|---|---|---|---|
| Parent | 13221119 | Aug 2011 | US |
| Child | 13914830 | US | |
| Parent | 11239475 | Sep 2005 | US |
| Child | 13221119 | US |
