Synchronizing system resources in a multi-socket data processing system - Patents.com

By synchronizing system resources across multiple SOCs in multi-socket data processing systems, the method addresses inconsistent values, ensuring stable system operation and accurate event timelines.

JP2024543389APending Publication Date: 2024-11-21XILINX INC
View PDF 5 Cites 0 Cited by

Patent Information

Application Number
JP2024527426
Authority / Receiving Office
JP · JP
Patent Type
Applications
Current Assignee / Owner
Priority Date
2021-11-16
Filing Date
2022-11-09
Publication Date
2024-11-21

AI Technical Summary

Technical Problem

In multi-socket data processing systems, system resources such as system counters and debug trace timestamps often operate independently, leading to inconsistent values across different sockets, causing unpredictable system behavior and inaccurate event timelines, which hampers proper system operation and debugging.

Method used

A method and system for synchronizing system resources across multiple system-on-chip (SOC) units in a multi-socket data processing system by designating a primary SOC to write updated values to other SOCs and initiate operations, using communication buses and sideband channels for simultaneous synchronization.

Benefits of technology

Ensures consistent and synchronized system resource values across all sockets, improving system stability and accuracy of event timelines, thereby enhancing system operation and debugging capabilities.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 2024543389000001_ABST
    Figure 2024543389000001_ABST
Patent Text Reader

Abstract

Synchronizing system resources of a multi-socket data processing system can include providing a trigger event from a primary system-on-chip (SOC) to a global synchronization circuit. The primary SOC is one of the multiple SOCs, and the trigger event is provided over a first sideband channel. In response to the trigger event, the global synchronization circuit can broadcast a synchronization event to the multiple SOCs over a second sideband channel. In response to the synchronization event, system resources of each SOC of the multiple SOCs are programmed with a common value. The programming synchronizes the system resources of the multiple SOCs.
Need to check novelty before this filing date? Find Prior Art

Description

[Technical field]

[0001] FIELD OF THE DISCLOSURE This disclosure relates to multiprocessing, and more particularly, to synchronizing system resources in a multi-socket data processing system. [Background technology]

[0002] A symmetric multi-processing (SMP) data processing system is a data processing system that has a hardware and software architecture that includes multiple processors (e.g., two or more processors). The processors are identical and are controlled by a single operating system instance that treats each processor equally; that is, no processor is reserved for a special purpose. The processors are connected to a single shared main memory and have full access to all input / output devices.

[0003] A multi-socket (MS) data processing system refers to a data processing system having multiple interconnected processors, each disposed in a different physical socket. The processors are connected by a coherent communication link initiated over a communication bus. One example of such a communication protocol is Cache Coherent Interconnect for Accelerators (CCIX). The coherent communication link supports shared main memory access.

[0004] In some MS and / or SMP data processing systems, each processor may be implemented as a System-on-Chip (SOC). A SOC is an integrated circuit (IC) that includes multiple processor cores. Each processor core may be configured to execute program code. The IC may be implemented as a single die within a package or as a multi-die IC implemented in a single package. In an SMP data processing system, the operating system and applications running therein have a unified view across all SOCs. The operating system further sees each processor core in a unified manner as if each processor core were part of the same SOC.

[0005] One aspect of certain multiprocessor data processing systems is that certain system resources must be synchronized across SOCs. One example of a system resource that must be synchronized across SOCs in various types of systems, including MS-SMP data processing systems, is a system counter. As an example, an operating system of an MS-SMP data processing system may migrate a process from a first processor core in a first SOC disposed in a first socket to another processor core in a second SOC disposed in a second socket. The migrated process should see a consistent state of the particular system resource before, during, and after the migration. For example, in the case of a system counter, the system counter used by the first SOC should be synchronized with the system counter of the second SOC such that the values ​​of each system counter match (e.g., are synchronized) during operation. This allows the migrated process to see a consistent state with respect to the system counter of the first SOC before the migration and the system counter of the second SOC after the migration. Summary of the Invention

[0006] In one or more example implementations, a method of synchronizing system resources of a multi-socket data processing system can include stopping each of a plurality of system resources under control of a primary system-on-chip (SOC). Each system resource is located in a different SOC of a plurality of SOCs of a multi-SOC system. The primary SOC is one of the plurality of SOCs. The method can include using the primary SOC to write updated values ​​to the system resources of each other SOC of the plurality of SOCs while stopped. The method can also include starting operation of the system resources in each SOC of the plurality of SOCs using the primary SOC after the writing.

[0007] In one aspect, the method includes, after stopping and before writing, using the primary SOC to read a value from each of a plurality of system resources for each SOC of the plurality of SOCs other than the primary SOC, and determining, by the primary SOC, a difference between the value read from the system resource of the primary SOC and the values ​​read from the system resources of the other SOCs.

[0008] In another aspect, the writing includes the primary SOC writing an updated value to each of the other SOCs, the updated value written to each of the other SOCs depending on a difference between a value read from the system resource of the primary SOC and a value read from the system resource of the other SOC.

[0009] In another aspect, writing the updated values ​​and initiating the operations to each system resource of the other SOCs takes into account the latency in the primary SOC communicating with each system resource of the multiple system resources in each other SOC.

[0010] In another embodiment, the multiple SOCs are identical.

[0011] In another aspect, the stopping, writing and starting are communicated as part of the boot process.

[0012] In another embodiment, communications between the primary SOC and other SOCs are conveyed via a communications bus that communicatively links the multiple SOCs.

[0013] In one or more example implementations, a system for synchronizing system resources of a multi-socket data processing system may include a plurality of SOCs. The plurality of SOCs are interconnected via a communication bus. Each SOC includes a plurality of processor cores. One SOC of the plurality of SOCs is designated as a primary SOC. The system may include a global synchronization circuit. The system may include a plurality of sideband channels coupling the global synchronization circuit to each of the plurality of SOCs. The primary SOC is configured to provide a trigger event via a first sideband channel of the plurality of sideband channels. The trigger event initiates synchronization of system resources disposed within the plurality of SOCs. In response to receiving the trigger event, the global synchronization circuit is configured to broadcast a synchronization event to each of the plurality of SOCs via a second sideband channel of the plurality of sideband channels.

[0014] In one aspect, the synchronization event is received simultaneously by each SOC of the plurality of SOCs.

[0015] In another aspect, each SOC of the plurality of SOCs is disposed in a different socket.

[0016] In another embodiment, the multiple SOCs are identical.

[0017] In another aspect, the trigger and synchronization events are communicated as part of the boot process.

[0018] In another aspect, a selected processor core of only the primary SOC provides the trigger event.

[0019] In another aspect, the synchronization event is an interrupt, and within each SOC of the plurality of SOCs, the interrupt is received by a selected processor core of the plurality of processor cores, and in response to the interrupt, the selected processor core in each SOC executes an interrupt service routine to program a system resource by writing to a control register of the system resource within the same SOC.

[0020] In another aspect, within each SOC of the plurality of SOCs, a synchronization event is received by control logic of the system resource, and in response to the synchronization event, the control logic of the system resource resets a system resource within the same SOC.

[0021] In one or more example implementations, a method of synchronizing system resources of a multi-socket data processing system can include providing a trigger event from a primary SOC to a global synchronization circuit. The primary SOC is one of a plurality of SOCs, and the trigger event is provided over a first sideband channel. The method can include broadcasting a synchronization event from the global synchronization circuit to the plurality of SOCs over a second sideband channel in response to the trigger event. The method can also include programming system resources of each SOC of the plurality of SOCs with a common value in response to the synchronization event, the programming synchronizing the system resources of the plurality of SOCs.

[0022] In one aspect, the providing, broadcasting, and programming are performed as part of a boot process for a multi-socket data processing system that includes multiple SOCs.

[0023] In another embodiment, the multiple SOCs are identical.

[0024] In another aspect, each SOC of the plurality of SOCs includes a plurality of processor cores, and further, a selected processor core of the plurality of processor cores of a selected SOC of the plurality of SOCs provides the trigger event.

[0025] In another aspect, the synchronization event is an interrupt, each SOC of the plurality of SOCs includes a plurality of processor cores, and a selected processor core of each SOC of the plurality of SOCs executes an interrupt service routine in response to the interrupt to program system resources located within the same SOC.

[0026] In another aspect, a selected processor core in each SOC of the multiple SOCs writes to control logic of a system resource in the same SOC as it executes an interrupt service routine.

[0027] In another aspect, within each SOC of the plurality of SOCs, a synchronization event is received by control logic of the system resource, and in response to the synchronization event, the control logic of the system resource resets a system resource within the same SOC.

[0028] This summary section is provided merely to introduce certain concepts, and is not intended to identify any key or essential features of the claimed subject matter. Other features of the inventive subject matter will be apparent from the accompanying drawings and the detailed description that follows.

[0029] Arrangements of the present invention are illustrated by way of example in the accompanying drawings. However, the drawings should not be construed as limiting the arrangements of the present invention to only the particular implementations shown. Various aspects and advantages will become apparent upon consideration of the following detailed description and upon reference to the drawings. [Brief description of the drawings]

[0030] [Figure 1] 1 illustrates an exemplary system that includes multiple interconnected systems-on-chips (SOCs). [Diagram 2] 2 illustrates an exemplary method for illustrating certain operational features of the system of FIG. 1 . [Diagram 3] FIG. 2 is a signal flow diagram illustrating exemplary communications between SOCs. [Figure 4] 2 illustrates another exemplary implementation of the system of FIG. 1 . [Diagram 5] 2 illustrates an exemplary method for describing certain operational features of the system of FIG. 1. [Figure 6] 4 is another signal flow diagram illustrating exemplary communications between SOCs. [Figure 7] 2 illustrates another exemplary implementation of the system of FIG. 1 . [Figure 8] 1 illustrates another exemplary system including multiple SOCs. [Figure 9] 9 illustrates an exemplary method for depicting certain operational features of the system of FIG. 8. [Figure 10] 9 is a signal flow diagram illustrating exemplary communications between the SOCs of the system of FIG. 8. [Figure 11] SUMMARY OF THE DISCLOSURE An exemplary method for synchronizing system resources in a multi-socket (MS) data processing system including a sideband channel and global synchronization circuitry is illustrated. [Figure 12] 1 illustrates an exemplary method for synchronizing system resources in a MS data processing system that does not include a sideband channel or global synchronization circuitry. [Figure 13] Illustrates an example of a multi-socket data processing system. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0031] The present disclosure relates to multiprocessing, and more particularly to synchronizing system resources in a multi-socket (MS) data processing system. The inventive arrangements may also be applied to symmetric multiprocessing (SMP) data processing systems. MS data processing systems may be implemented in which each socket includes a system-on-chip (SOC). Each SOC may include multiple processor cores.

[0032] In an MS-SMP data processing system, certain resources must be synchronized across sockets once the boot process is complete to ensure proper and error-free operation. Resources that are architecturally defined within an MS-SMP data processing system as one per SMP (e.g., one per SOC) require this synchronization. Examples of these resources include, but are not limited to, system counters, debug trace timestamp generators, etc.

[0033] For purposes of illustration, the system counter associated with each socket is free running. Although each system counter is configured to count at the same frequency, there is no guarantee that each system counter will start at the same time. If the system counters of each socket have different start times, the values ​​of the system counters will not be equal or synchronized. With the system counters of each socket not synchronized, the operating system of the MS-SMP data processing system may exhibit unpredictable behavior when managing and scheduling processes on the processor cores in the different sockets.

[0034] Debug trace timestamps are another example of a system resource that requires synchronization. When debug trace timestamps are not synchronized, the timeline of debug and / or trace events generated and stored in the system may be incorrect. This makes it very difficult to establish the true timeline of such events. Debugging activities may be hindered by inaccurate event timelines. Synchronizing debug trace timestamps improves the debug capabilities of MS-SMP systems by improving the accuracy of the event timelines, which makes the debugging task easier.

[0035] The inventive arrangements described within this disclosure provide methods, systems, and computer program products that can ensure that certain system resources of an MS data processing system and / or an MS-SMP data processing system are synchronized. Further aspects of the inventive arrangements are described below with reference to the figures.

[0036] FIG. 1 illustrates an exemplary system 100. System 100 is an example of an MS data processing system. Additionally, system 100 may be an MS-SMP data processing system. In the example of FIG. 1, system 100 includes multiple SOCs 102. System 100 may include fewer or more SOCs 102 than shown. In the example of FIG. 1, each SOC 102 is disposed in its own socket. In the example of FIG. 1, each SOC 102 includes multiple processor cores 112. Processor cores 112 may be hardwired processor cores (e.g., hardened circuits). Each processor core may be configured to execute program code. Additionally, each SOC 102 may be identical.

[0037] In another embodiment, each SOC 102 may include one or more "soft processors" formed from programmable logic. In yet another embodiment, each SOC 102 may include a combination of hardwired and soft processors. If one or more SOCs 102 implement a soft processor, the SOCs 102 that implement the soft processor should be implemented such that the soft processor has synchronized access to system resources.

[0038] In one embodiment, each of the SOCs 102 may be disposed on the same circuit board or card. In another embodiment, each of the SOCs 102 may be disposed on a different circuit board or card. In other embodiments, more than one circuit board or card may be used, with the SOCs 102 distributed across such circuit boards / cards in various combinations of one or more SOCs per circuit board. The SOCs 102 may be used as processors in a data processing system or may be accelerators included in a larger data processing system that includes a host processor or central processing unit (CPU).

[0039] Within the system 100, each socket, and therefore each SOC 102, may be identified with a socket identifier (ID) that uniquely identifies the socket, and therefore the SOC 102 disposed within the socket. Although the operating system of the MS-SMP data processing system may view each SOC equally, one SOC may be designated as the primary SOC. Thus, the SOC 102 disposed within the primary socket may be designated as the primary SOC. The SOC disposed within the primary socket is generally responsible for handling operations related to the configuration and booting of the system 100, including the other sockets. For example, the primary SOC may perform functions such as booting, power management, error, and health management of the system 100. In the example of FIG. 1, SOC 102-1 is the primary SOC 102.

[0040] The SOCs 102 are coupled to a communication bus 104. In one aspect, the communication bus 104 may be implemented as a Peripheral Component Interconnect Express (PCIe) bus. As discussed, a coherent communication link may be initiated via the bus 104. The SOCs 102 are also communicatively linked to a global synchronization circuit (GSC) 106. In the example of FIG. 1, the GSC 106 is coupled to each of the SOCs 102 via one or more sideband channels. For example, each SOC 102 may be connected to the GSC 106 via an output sideband channel 108 (e.g., 108-1, 108-2, 108-3, and 108-N). Each SOC 102 is also coupled to the GSC 106 via an input sideband channel 110 via which events and / or interrupts may be broadcast from the GSC 106. A sideband channel is a different physical signal path than that implemented with respect to communications bus 104. That is, a sideband channel may be distinct from communications bus 104.

[0041] In one or more example implementations, sideband channel 108-1 may be included, while sideband channels 108-2, 108-3, and 108-N may be omitted.

[0042] In one or more exemplary implementations in which the SOCs 102 are disposed on the same circuit board, the GSC 106 may be disposed on the same circuit board as the SOCs 102. In another exemplary implementation, the GSC 106 may be disposed on a circuit board separate and distinct from the circuit board on which the SOCs 102 are disposed. In yet another exemplary implementation in which the SOCs 102 are disposed on multiple different circuit boards, the GSC 106 may be disposed on one of the circuit boards that includes one or more of the SOCs 102. The GSC 106 may be coupled to the SOCs 102 (e.g., to a socket for the SOCs 102) via a sideband channel, via wires / traces in the circuit board, and / or by a dedicated cable.

[0043] In one or more example implementations, each SOC 102 may include a specific processor core designated to perform the synchronization-related functions described herein. In some cases, the designated processor core is referred to as a System Control Processor (SCP). The SCP typically handles the boot operations, power management, error management, and health management of the system.

[0044] In other cases, the designated processor core is referred to as a System Resource Reset Processor (SRRP). If the SOCs 102 are identical, it should be understood that the terms SCP and SRRP, to the extent such terms are used to refer to specific processor cores in the primary and / or non-primary SOCs, are used only to refer to the role of each designated processor core in the respective SOCs 102. The SCP and SRRP may be the same processor core from an architectural standpoint.

[0045] In one aspect, the SCP / SRRP is a dedicated processor core within each SOC 102 tasked with performing the operations described within this disclosure. In one or more other example implementations, the SCP / SRRP of each SOC 102 may be the same as other processor cores within the respective SOC 102, but may be designated or selected to perform the functions described herein. That is, the processor core may be the same as the SCP / SRRP that is one of multiple identical processor cores so designated in the respective SOC 102.

[0046] Each SOC 102 further includes one or more synchronized system resources (SSRs) 114. Within this disclosure, a system counter is used as an exemplary SSR 114. A system counter is a circuit that monotonically increases and does not roll over during the life of a system boot session. Processor cores of a SOC 102 may subscribe to the system counter for that SOC and use the system counters to generate timer events for each processor core that can be used by the operating system and / or applications to obtain a notion of time for scheduling, etc.

[0047] Other examples of SSR 114 may include a debug trace timestamp generator. However, it should be understood that SSR 114 is not intended to be limited to the described examples. In general, any SMP resource that is architecturally defined as one per SMP, e.g., one per SOC 102, may require synchronization and may be considered an SSR 114.

[0048] FIG. 2 illustrates an example method 200 depicting certain operational characteristics of the system 100 of FIG. 1. For purposes of illustration, SOC 102-1 is designated as the "primary" SOC in the system 100. FIG. 3 is a signal flow diagram illustrating example communications between the SOCs 102 of the system 100 of FIG. 1. FIGS. 2-3 illustrate example operations for synchronizing the SSRs 114 within the system 100, which may be performed as part of the boot process of the system 100. For purposes of illustration, the SSRs 114 are considered to be system counters. First, as mentioned, the SSRs 114 are free running and have different values ​​than the SSRs 114 because the system resources did not have a synchronized start time. Rather, one or more or all of the SSRs 114 may have begun operation at different times. The different values ​​of the SSRs 114 are represented by X1 for SSR 114-1, X2 for SSR 114-2, and X3 for SSR 114-N. N This is reflected in the values ​​stored in each, denoted as

[0049] 2-3, in block 202, the SCP 302-1, which is the SCP of the primary SOC 102-1, provides a trigger event 304 to the GSC 106. In one aspect, the trigger event 304 may be a pulse. In this example, the SCP 302-1 is configured to provide the trigger event 304 to the GSC 106 via a first sideband channel 108-1. In one or more example implementations, only the SCP 302-1, e.g., the SCP of the primary SOC, is configured to generate the trigger event 304 and forward the trigger event to the GSC 106 via the sideband channel.

[0050] In block 204, in response to the trigger event 304, the GSC 106 can broadcast a synchronization event 306 to each of the multiple SOCs 102 via the sideband channel 110. In one aspect, the synchronization event 306 is an interrupt. In response to the synchronization event 306, each SOC 102 can initiate an interrupt service routine. In block 206, the SCPs 302-1, SRRPs 302-2, 302-N in each SOC 102 are interrupted in response to receiving the synchronization event 306. An "interrupt service routine," also called an interrupt handler, is a software process that is invoked by an interrupt request, which may be generated from a hardware device, a software interrupt instruction, or a software exception. The interrupt service routine processes the request and sends a request to a specific program or piece of code to interrupt the active process being executed by the processor / processor core before the interrupt occurred. When the interrupt service routine completes, the interrupted process is resumed.

[0051] In block 208, a designated processor core in each SOC 102 programs (308) a respective SSR 114 located in the same SOC. The designated processor core may be the SCP 302-1 in the primary SOC 102-1 and the SRRPs 302-2, 302-N in the non-primary SOCs 102-2, 102-N. The SSRs 114 in each SOC 102 are synchronized as a result of the programming. For example, the programming may write the same configuration data or value to each SSR 114. In this example, in response to receiving the synchronization event 306, the SCPs 302-1, SRRPs 302-2, 302-N may execute an interrupt service routine that causes the respective processor cores to program the respective SSRs 114 located in the same SOC. Because each SOC 102 receives the synchronization event 306 broadcast over the sideband channel, each receives the synchronization event 306 simultaneously. 3, each designated processor core (e.g., SCP 302-1, SRRP 302-2, 302-N) simultaneously writes a value of 0 to the corresponding SSR 114. As a result, the SSRs 114 are synchronized. As illustrated in FIG. 3, each SSR 114 continues to operate from an initialized value, e.g., 0 in this case, and then X sync has the same value at

[0052] 4 illustrates another exemplary implementation of the system 100. In the example of FIG. 4, only SOCs 102-1 and 102-2 are shown for ease of explanation and clarity. In this example, each SOC 102 includes multiple processor cores 402, 404, 406, and 302, a system counter 408 (e.g., SSR 114), a system counter control register 410, a phase locked loop (PLL) 412, an output pin 414, and an input pin 416.

[0053] 4, the SCP 302-1 may generate a trigger event 304 that is communicated via the sideband channel 108-1 out output pin 414-1 to the GSC 106. In this example, the SRRP 302-2 may be coupled to output pin 414-2. However, the SRRP 302-2 is not part of the primary SOC 102-1 and does not generate or transmit trigger events.

[0054] In response to receiving the trigger event 304, the GSC 106 broadcasts a synchronization event 306 over the sideband channel 110. Each SOC 102 receives the synchronization event 306, e.g., an interrupt, via a corresponding input pin 416. The synchronization event 306 is provided to a designated processor core in each SOC 102. The designated processor core in the primary SOC 102 is the SCP 302-1, while the designated processor core in each non-primary SOC 102 is the SRRP 302-2.

[0055] In response to receiving the synchronization event 306, the SCP 302-1 executes an interrupt service routine that causes the SCP 302-1 to write configuration data to the system control register 410-1 and reset the system counter 408-1. In response to receiving the synchronization event 306, the SRRP 302-2 executes an interrupt service routine that causes the SRRP 302-2 to write configuration data to the system control register 410-2 and reset the system counter 408-2. Because the synchronization event 306 is broadcast to each SOC 102, each SOC 102 receives the synchronization event 306 at the same time. Therefore, the reset of the system counter 408 is performed synchronously.

[0056] The system counters 408 are free running, but each runs off of a common reference clock that runs through a PLL 412. When synchronized, each system counter 408 can count in a synchronized manner with each other system counter 408, as each is driven by a respective PLL 412 that receives the common reference clock. When synchronized, the system counters 408 continue to operate in sync with one another. Thus, when synchronized, the processor cores 402, 404, and 406 (as well as the SCPs 302-1 and SRRPs 302-2) in each SOC 102 each see the same value in either system counter 408-1 or 408-2 at any given time. That is, system counter 408-1 is synchronized with system counter 408-2.

[0057] As mentioned above, counter 408 may be a monotonically increasing counter that does not "roll over" during the life of a boot session of system 100. The processor cores of each SOC 102 subscribe to a system counter located in the same SOC to determine or obtain a notion of time and to generate timer events (e.g., per processor core) that can be used by the operating system and / or applications for scheduling purposes.

[0058] 5-7 illustrate an alternative implementation for synchronizing the SSR 114 that utilizes a dedicated sideband channel while avoiding the need for a dedicated SRRP as illustrated in FIGS.

[0059] FIG. 5 illustrates an example method 500 depicting certain operational characteristics of the system 100 of FIG. 1. For purposes of illustration, SOC 102-1 is designated as the "primary" SOC in the system 100. FIG. 6 is a signal flow diagram illustrating example communications between the SOCs 102 of the system 100 of FIG. 1. FIGS. 5-6 illustrate example operations for synchronizing the SSR 114 within the system 100, which may be performed as part of the boot process of the system 100. For purposes of illustration, the SSR 114 is considered to be a system counter. First, as mentioned, the SSR 114 is free running and has different values ​​since the SSR 114 did not have a synchronization start time. The different values ​​of the SSR 114 are X1 for SSR 114-1, X2 for SSR 114-2, and X3 for SSR 114-N. N This is reflected in the values ​​stored in each, denoted as

[0060] 5-6, in block 502, the SCP 302-1, which is the SCP of the primary SOC 102-1, provides a trigger event 304 to the GSC 106. In one aspect, the trigger event 304 may be a pulse. In this example, the SCP 302-1 is configured to provide the trigger event 304 to the GSC 106 via the first sideband channel 108-1. In one or more example implementations, only the SCP 302-1, e.g., the SCP of the primary SOC, is configured to generate the trigger event 304 and forward the trigger event to the GSC 106.

[0061] At block 504, in response to the trigger event 304, the GSC 106 can broadcast a synchronization event 306 to each of the multiple SOCs 102 via the sideband channel 110. In one aspect, the synchronization event 306 can be a pulse. In the examples of FIGS. 3-7, the synchronization event 306 does not trigger execution of an interrupt service routine. Rather, in the examples of FIGS. 3-7, the synchronization event 306 is provided directly to control logic (e.g., control logic for each SSR 114 in each SOC 102). At block 506, the control logic in each SOC 102 receives the synchronization event 306. At block 508, in response to receiving the synchronization event 306, the control logic in each SOC 102 resets the SSR 114 therein.

[0062] As a result of block 508, the SSRs 114 of each SOC 102 are synchronized as a result of the programming. For example, the programming may write the same configuration data or value to each SSR 114. Because each SOC 102 receives the synchronization event 306 broadcast over the sideband channel, each receives the synchronization event 306 at the same time. In this example, in response to receiving the synchronization event 306, the control logic of each SOC 102 simultaneously resets the SSRs 114 therein to 0, as illustrated in FIG. 6, where the value of the row corresponding to block 506 is "0". As a result, the SSRs 114 are synchronized. As illustrated in FIG. 6, each of the SSRs 114 continues to operate from an initialized value, e.g., 0 in this case, and then resets to "0". sync has the same value at

[0063] Figure 7 illustrates another exemplary implementation of the system 100 of Figure 1. In the example of Figure 7, only SOCs 102-1 and 102-2 are shown for ease of explanation and clarity. In this example, each SOC 102 includes multiple processor cores 402, 404, 406, and 302, a system counter 408 (e.g., SSR 114), control logic 702, a phase-locked loop (PLL) 412, an output pin 414, and an input pin 416.

[0064] 7, SCP 302-1 may generate trigger events 304 that are communicated via sideband channel 108-1 out output pin 414-1 to GSC 106. In this example, SCP 302-2 may be coupled to output pin 414-2. However, SCP 302-2 is not part of primary SOC 102-1 and does not generate or transmit trigger events.

[0065] In response to receiving the trigger event 304, the GSC 106 broadcasts a synchronization event 306 over the sideband channel 110. Each SOC 102 receives the synchronization event 306 via a corresponding input pin 416. The synchronization event 306 is provided directly to the control logic 702 in each respective SOC 102. In response to receiving the synchronization event 306, the control logic 702 in each SOC 102 resets a corresponding system counter 408 to 0. Because the counter control logic 702 in each SOC 102 receive the synchronization event 306 simultaneously, the resetting of the system counters 408 is performed synchronously.

[0066] The system counters 408 are free running, but each runs off of a common reference clock that runs through a PLL 412. When synchronized, each system counter 408 can count in a synchronized manner with each other system counter 408 since each is driven by a PLL 412 that receives the common reference clock. When synchronized, the system counters 408 continue to operate in sync with each other. Thus, when synchronized, the processor cores 402, 404, and 406 (as well as SCPs 302-1 and SCPs 302-2) in each SOC 102 each see the same value in either system counter 408-1 or 408-2 at any given time. That is, system counter 408-1 is synchronized with system counter 408-2.

[0067] 7, counter 408 may be a monotonically increasing counter that does not "roll over" during the life of a boot session of system 100. The processor cores of each SOC 102 subscribe to a system counter located in the same SOC to generate timer events (e.g., per processor core) that can be used by the operating system and / or applications to determine or obtain a notion of time and for scheduling.

[0068] 8-10 illustrate another alternative implementation for synchronizing SSR 114 utilizing virtualization synchronization without a dedicated sideband channel. In the example of FIG. 8-10, rather than using a sideband channel, system software is used to perform synchronization as part of the boot process of an MS-SMP data processing system. Because a dedicated sideband channel is not used, the exemplary implementation of FIG. 8-10 does not require special hardware or hardware updates for existing systems to be adopted. However, the lack of a sideband channel results in greater inter-socket communication latency than the example utilizing a dedicated sideband channel.

[0069] FIG. 8 illustrates an exemplary system 800. System 800 may be implemented substantially as described in connection with FIG. 1. In the example of FIG. 8, system 800 does not include any sideband channel or GSC 106. Each SOC 102 may be considered to have a multi-processor core architecture. For purposes of explanation, each SOC 102 may have an architecture as illustrated in the example of FIG. 7, although the particular architecture implemented in each SOC 102 is not intended to be limiting.

[0070] In the example of Figures 8-10, the SCP of the primary SOC 102 accesses the necessary resources in each individual socket of the MS-SMP data processing system. In order for the SCP of the primary SOC 102 to be able to access the system counter control registers and / or control logic for each socket of the MS-SMP data processing system, the system address map of the MS-SMP data processing system must expose the control and status registers of the MS-SMP data processing system to the SCP of the primary SOC 102. The system address map is a data structure that specifies the address ranges of various memory-mapped targets, such as components, circuit blocks, peripherals, etc., in the system. Furthermore, during the boot sequence, the synchronous operation relies on some hardware components being taken out of reset to allow the SCP of the primary SOC 102 to access such hardware components.

[0071] FIG. 9 illustrates an exemplary method 900 depicting some operational features of the system 800 of FIG. 8, e.g., an MP-SMP data processing system that does not include a sideband channel or a GSC. For purposes of illustration, SOC 102-1 is designated as the "primary" SOC in the system 800. FIG. 10 is a signal flow diagram illustrating exemplary communications between the SOCs 102 of the system 800 of FIG. 8. FIGS. 9-10 illustrate exemplary operations for synchronizing the SSR 114 within the system 800, which may be performed as part of the boot process of the system 800. For purposes of illustration, the SSR 114 is considered to be a system counter. First, as mentioned, the SSR 114 is free running and has different values ​​since the SSR 114 did not have a synchronization start time. The different values ​​of the SSR 114 are Y1 for SSR 114-1, Y2 for SSR 114-2, Y3 for SSR 114-3, and Y4 for SSR 114-N. N 8-10, the events and / or signals to be communicated are communicated via bus 104.

[0072] 9-10, in block 902, the SCP of the primary SOC 102 (e.g., SCP 302-1) issues a stop event to each SSR 114. In the example of FIG. 10, SCP 302-1 issues a stop event 1002 to the SSRs 114 in the same SOC, e.g., SSR 114-1, and to the SSRs 114 in each other SOC 102. As part of block 902, the SCP also reads the value of each SSR 114 that has been stopped. In general, the value of each SSR 114 may be determined with reference to SSR 114-1 in that each other SSR has a value that is a certain amount of delay (delay, D) after the value of SSR 114-1. SCP 302-1 may determine the delay of each SSR 114 with reference to SSR 114-1.

[0073] In block 904, the SCP of the primary SOC 102 updates the SSRs in each of the other SOCs 102. The SCP does not update the SSRs in the same SOC. Referring to Figure 10, the SCP 302-1 leaves SSR 114-1 with the same value as it originally read in block 902. The SCP 302-1 updates the value of each of the other SSRs 114-2, 114-3, 114-N by writing (1004) the value of SSR 114-1 plus a delayed value for each of the particular SSRs 114 being updated. Figure 10 illustrates the updated values ​​written by SCP 302-1 to each of SSRs 114-2, 114-3, and 114-N during block 904.

[0074] In block 906, the SCP of the primary SOC 102 sends an event signaling each of the SSRs 114 to resume operation. In the example of FIG. 10, the SCP 302-1 sends an event 1006 to each SSR to resume operation. In signaling each SSR 114 to resume operation, the signal provided from the SCP 302-1 to each respective SSR undergoes the same delay as measured in block 902. FIG. 10 illustrates that in block 908, each SSR 114 is synchronized and free running. Each SSR 114 may be clocked using a common reference clock to remain synchronized with the other SSRs 114, as discussed.

[0075] In one or more exemplary implementations, events such as stop, update, etc., as described herein and / or in connection with FIG. 9 may be implemented as register writes to, for example, control registers associated with resources being synchronized and communicated over bus 104. Register writes directed to synchronized resources in various sockets may be performed back-to-back in rapid succession. In other exemplary implementations, a dedicated broadcast function may be included to communicate events to multiple sockets such that individual writes to synchronized resources in various sockets do not need to be performed. Regardless of whether back-to-back writes or broadcast functionality is performed, delays in communicating with other sockets may be accounted for using the exemplary techniques described herein.

[0076] Table 1 illustrates exemplary states of SSR 114, eg, system counters, at various times as described in FIGS. [Table 1]

[0077] In the example of Table 1, Socket 1 corresponds to SOC 102-1, e.g., the primary SOC. Initially, each system counter is out of reset (e.g., corresponding to the boot process) and may be free-running. The operation of the system counters is not synchronized since each has a different value. The SCP of the primary SOC issues a stop event. When the system counters are stopped, the SCP reads the value of each system counter and calculates the difference between the value of the respective system counter and the value of the system counter at the primary socket. For a given counter at socket "N", this difference is D N where D N =Y1-Y N The value D N also indicates the delay between stopping the system counter for socket "N" and stopping the system counter for socket 1.

[0078] The delay may be approximated as the latency incurred by the SCP in the primary socket to access the control registers and / or control logic of a particular SSR in a particular socket, e.g., in this case, the system counter. Thus, when the SCP unfaults each system counter, the value of each system counter may be rewritten by the SCP such that the updated counter value compensates for this latency. The SCP updates the system counter of socket "N" as Y1+D N This means that when the SCP unfaults the system counter, the delay in restarting the system counter is less than the delay D N is used to ensure that the offset is cancelled out by adjustments made in the updated system counter values.

[0079] Table 2 illustrates an exemplary state of the SSR 114, eg, system counters, when the primary socket's SCP overwrites the value of the system counters and issues a resume event (eg, event 1006). [Table 2]

[0080] Although the exemplary implementations described within this disclosure are described in the context of an MS-SMP data processing system, it should be understood that the various synchronization mechanisms and techniques described herein may be extended to synchronize system resources in other data processing systems having multiple sockets that are not SMP data processing systems.

[0081] 11 illustrates an exemplary method 1100 for synchronizing system resources in an MS data processing system (system). The system may be an SMP data processing system.

[0082] At block 1102, a primary SOC of the system may provide a trigger event to the GSC 106. The primary SOC is one of the multiple SOCs, and the trigger event is provided over the first sideband channel 108-1. At block 1104, in response to the trigger event, the GSC 106 may broadcast a synchronization event to the multiple SOCs over the second sideband channel 110. At block 1106, in response to the synchronization event, system resources in each SOC of the multiple SOCs of the system are programmed with a common value. The programming synchronizes the system resources of the multiple SOCs.

[0083] Each of these and other implementations can optionally include one or more of the following features, either alone or in combination. Some example implementations include all of the following features in combination.

[0084] In one aspect, the described operations, e.g., providing, broadcasting, and programming, are performed as part of a boot process for an MS data processing system including multiple SOCs. The operations may be performed in response to system resources (e.g., SSR 114) coming out of reset.

[0085] In another embodiment, the multiple SOCs are identical.

[0086] In another aspect, each SOC of the plurality of SOCs includes a plurality of processor cores. A selected processor core of the plurality of processor cores of a selected SOC of the plurality of SOCs provides the trigger event. For example, an SCP as described herein may provide the trigger event.

[0087] In another aspect, each SOC of the plurality of SOCs includes a plurality of processor cores. The synchronization event may be an interrupt. A selected processor core of each SOC of the plurality of SOCs executes an interrupt service routine in response to the interrupt to program system resources located within the same SOC. The selected processor may be an SCP in a primary SOC and an SRRP in a non-primary SOC.

[0088] In another aspect, a selected processor core in each SOC of the plurality of SOCs writes to control registers and / or control logic of system resources within the same SOC as it executes an interrupt service routine.

[0089] In another aspect, within each SOC of the plurality of SOCs, a synchronization event is received by control logic of the system resource, and in response to the synchronization event, the control logic of the system resource in each respective SOC resets the system resource within the same SOC.

[0090] 12 illustrates an example method 1200 for synchronizing system resources in an MS data processing system (system). The system may be an SMP data processing system. In the example of FIG. 12, the system does not include a sideband channel and does not include a GSC 106.

[0091] In block 1202, the primary SOC may stop each of a plurality of system resources in the system. Each system resource is located in a different SOC of the plurality of SOCs of the multi-SOC system. The primary SOC is one of the plurality of SOCs. In block 1204, the primary SOC may write updated values ​​to the system resources of each other SOC of the plurality of SOCs while it is stopped. In block 1206, the primary SOC may begin operation (unfault) of the system resources in each SOC of the plurality of SOCs following the write.

[0092] Each of these and other implementations can optionally include one or more of the following features, either alone or in combination. Some example implementations include all of the following features in combination.

[0093] In one aspect, after stopping and before writing, the primary SOC, e.g., the SCP of the primary SOC, can read a value from each of the plurality of system resources. For each SOC of the plurality of SOCs other than the primary SOC, the SCP can determine a difference between the value read from the system resource of the primary SOC and the value read from the system resource of the other SOC. The writing includes the primary SOC, e.g., the SCP of the primary SOC writing an updated value to each of the other SOCs. The updated value written to each of the other SOCs depends on the difference between the value read from the system resource of the primary SOC and the value read from the system resource of the other SOC.

[0094] In another aspect, writing the updated values ​​and initiating the operations to each system resource of the other SOCs takes into account the latency in the primary SOC communicating with each system resource of the multiple system resources in each other SOC.

[0095] In another embodiment, the multiple SOCs are identical.

[0096] In another aspect, the stopping, writing, and starting are communicated as part of the boot process. The operations may be performed in response to system resources (e.g., SSR 114) coming out of reset.

[0097] In another embodiment, communications between the primary SOC and other SOCs are conveyed via a communications bus that communicatively links the multiple SOCs.

[0098] 13 illustrates an example implementation of an MS data processing system 1300. The data processing system 1300 may further be an SMP type system. The components of the data processing system 1300 may include, but are not limited to, multiple SOCs 1302, a memory 1304, and a bus 1306 coupling various system components including the memory 1304 to the SOC 1302. The SOC 1302 may include a processor core having any of a variety of different architectures including, but not limited to, an x86 type architecture (IA-32, IA-64, etc.), Power Architecture, an ARM processor, etc. As mentioned, in some cases, the SOC 1302 may include programmable logic that may be used to implement one or more of the SOC's multiple processor cores.

[0099] As discussed, in some cases, the MS data processing system may include a central processing unit (CPU) and may include multiple SOCs 1302 as accelerators. The accelerators may be disposed on one or more different circuit boards, cards, chassis structures, or may be implemented in other available form factors that communicatively link with the data processing system 1300 via a communication bus, such as bus 1306. In one or more exemplary implementations, the accelerators may be implemented according to any of a variety of standards and / or specifications set forth as part of the Open Compute Project (OCP) and / or OCP Accelerator Module (OAM) specifications. Additionally, it should be understood that any such accelerator may include any of a variety of different connectors and / or connector combinations for coupling to different systems and / or devices using one or more different communication protocols.

[0100] Bus 1306 represents one or more of any of a variety of communication bus structures. By way of example and not limitation, bus 1306 may be implemented as a Peripheral Component Interconnect Express (PCIe) bus. Other communication buses may be used in place of or in addition to a PCIe. Data processing system 1300 typically includes a variety of computer system readable media. Such media may include computer readable volatile and nonvolatile media and computer readable removable and non-removable media.

[0101] The memory 1304 may include computer readable media in the form of volatile memory, such as random-access memory (RAM) 1308 and / or cache memory 1310. The data processing system 1300 may also include other removable / non-removable, volatile / non-volatile computer storage media. By way of example, a storage system 1312 may be provided for reading from and writing to a non-removable, non-volatile magnetic and / or solid-state medium (not shown, typically referred to as a "hard drive"). Although not shown, a magnetic disk drive may be provided for reading from and writing to a removable, non-volatile magnetic disk (e.g., a "floppy disk"), and an optical disk drive may be provided for reading from and writing to a removable, non-volatile optical disk, such as a CD-ROM, DVD-ROM, or other optical medium. In such a case, each may be connected to the bus 1306 by one or more data media interfaces. The memory 1304 is an example of at least one computer program product.

[0102] Programs / utilities 1314 having a set (at least one) of program modules 1316 may be stored in memory 1304. The programs / utilities 1314 are executable by a processor core of SOC 1302. By way of example, program modules 1316 may represent an operating system, one or more application programs, other program modules, and program data. The program modules 1316, when executed, cause data processing system 1300, e.g., one or more CPUs (not shown) and / or SOC 1302, to perform the functions and / or methodologies of the example implementations described within this disclosure. The programs / utilities 1314 and any data items used, generated, and / or computed by data processing system 1300 are functional data structures that provide functionality when used by data processing system 1300.

[0103] For example, in one or more example implementations, SOC 1302 may be implemented as described in connection with FIG. 1 and may include a GSC (not shown). In one or more other example implementations, SOC 1302 may be implemented as described in connection with FIG.

[0104] Data processing system 1300 may include one or more input / output (I / O) interfaces 1318 communicatively linked to bus 1306. I / O interface 1318 enables data processing system 1300 to communicate with one or more external devices 1320 and / or over one or more networks, such as a local area network (LAN), a wide area network (WAN), and / or a public network (e.g., the Internet). Examples of I / O interface 1318 may include, but are not limited to, a network card, a modem, a network adapter, a hardware controller, and / or the like. Examples of external devices may also include devices that enable a user to interact with data processing system 1300 (e.g., a display, a keyboard, and / or a pointing device) and / or other devices, such as accelerator cards.

[0105] The data processing system 1300 is just one example implementation. The data processing system 1300 may be implemented as a standalone device (e.g., as a user computing device or server, as a bare metal server), in a cluster (e.g., two or more interconnected computers), or in a distributed cloud computing environment (e.g., as a cloud computing node) where tasks are performed by remote processing devices linked via a communication network. In a distributed cloud computing environment, program modules may be located in both local and remote computer system storage media, including memory storage devices. The example of FIG. 13 is not intended to suggest any limitation as to the scope of use or functionality of the example implementations described herein. The data processing system is an example of computer hardware capable of performing various operations described within this disclosure.

[0106] Data processing system 1300 may include fewer components than shown, or additional components not illustrated in FIG. 13, depending on the particular type of device and / or system being implemented. The particular operating systems and / or applications included may vary according to the device and / or system type, as well as the types of I / O devices included. Additionally, one or more of the illustrated components may be incorporated in or form a part of another component. For example, a processor may include at least some memory.

[0107] Data processing system 1300 may be operational with numerous other general purpose or special purpose computing system environments or configurations. Examples of computing systems, environments, and / or configurations that may be suitable for use with data processing system 1300 include, but are not limited to, personal computer systems, server computer systems, thin clients, thick clients, handheld or laptop devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputer systems, mainframe computer systems, and distributed cloud computing environments that include any of the above systems or devices.

[0108] Some computing environments, such as cloud computing environments and / or edge computing environments using data processing system 1300 or other suitable data processing systems, generally support an FPGA-as-a-Service (FaaS) model, in which user functions are hardware accelerated as circuit designs implemented in programmable ICs that operate under the control of a (host) data processing system. Other examples of cloud computing models are described by the National Institute of Standards and Technology (NIST), and more specifically, the Information Technology Laboratory of NIST.

[0109] The present disclosure concludes with claims defining novel features, but it is believed that the various features described within the present disclosure will be better understood from a consideration of the description in conjunction with the drawings. The processes, machines, manufactures, and any variations thereof described herein are provided for illustrative purposes. The specific structural and functional details described within the present disclosure should not be construed as limiting, but merely as a representative basis for teaching those skilled in the art to employ the described features in various ways in virtually any appropriately detailed structure. Moreover, the terms and phrases used within the present disclosure are not intended to be limiting, but rather to provide an understandable description of the described features.

[0110] For simplicity and clarity of illustration, elements shown in the figures have not necessarily been drawn to scale. For example, dimensions of some of the elements may be exaggerated relative to other elements for clarity. Further, where considered appropriate, reference numerals have been repeated among the figures to indicate corresponding, similar or like features.

[0111] As defined herein, the singular forms "a," "an," and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise.

[0112] As defined herein, the term "approximately" means close to, but not exactly, a correct or exact value or amount. For example, the term "approximately" can mean that a recited property, parameter, or value is within a given amount of the exact property, parameter, or value.

[0113] As defined herein, the terms "at least one," "one or more," and "and / or" are open-ended expressions that are both conjunctive and disjunctive in operation unless expressly stated otherwise. For example, each of the phrases "at least one of A, B, and C," "at least one of A, B, or C," "one or more of A, B, and C," "one or more of A, B, or C," and "A, B, and / or C" means A alone, B alone, C alone, a combination of A and B, a combination of A and C, a combination of B and C, or a combination of A, B, and C.

[0114] As defined herein, the term "automatically" means without human intervention. As defined herein, the term "user" means a human being.

[0115] As used herein, the term "cloud computing" refers to a computing model that facilitates convenient, on-demand network access to a shared pool of configurable computing resources, such as networks, servers, storage, applications, ICs (e.g., programmable ICs), and / or services. These computing resources may be rapidly provisioned and released with minimal administrative effort or service provider interaction. Cloud computing promotes availability and may be characterized by on-demand self-service, wide area network access, resource pooling, rapid elasticity, and measured service.

[0116] As defined herein, the term "computer-readable storage medium" means a storage medium that contains or stores program code for use by or in connection with an instruction execution system, apparatus, or device. As defined herein, a "computer-readable storage medium" is not itself a transitory propagating signal. The computer-readable storage medium can be, but is not limited to, an electronic storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination of the foregoing. The various forms of memory described herein are examples of computer-readable storage media. A non-exhaustive list of more specific examples of computer-readable storage media may include portable computer diskettes, hard disks, RAM, read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), electronically erasable programmable read-only memory (EEPROM), static random-access memory (SRAM), portable compact disc read-only memory (CD-ROM), digital versatile disk (DVD), memory sticks, floppy disks, and the like.

[0117] As defined within this disclosure, the term "data structure" refers to a physical implementation of a data model's organization of data in physical memory. Thus, a data structure is formed from specific electrical or magnetic structural elements in memory. The data structure imposes a physical organization on the data stored in memory for use by application programs executed using a processor.

[0118] As defined herein, the term "if" means "when" or "when" or "in response to" or "depending on," depending on the context. Thus, the phrase "when it is determined that" or "when [a described condition or event] is detected" may be interpreted to mean "upon determining" or "in response to determining," or "upon detecting [a described condition or event]" or "in response to detecting [a described condition or event]," depending on the context.

[0119] As defined herein, the term "in response to" and similar phrases as explained above, such as "when," "when," or "upon," means to readily respond or react to an action or event. The response or reaction is performed automatically. Thus, when a second action is performed "in response to" a first action, there is a causal relationship between the occurrence of the first action and the occurrence of the second action. The term "in response to" indicates a causal relationship.

[0120] As defined herein, "data processing system" means one or more hardware systems configured to process data, each hardware system including at least one processor programmed to initiate operations and memory.

[0121] As defined herein, the term "processor" or "processor core" means at least one circuit capable of executing instructions contained in a program code. The circuit may be an integrated circuit or may be embedded in an integrated circuit. A SOC, as described herein, may include multiple processor cores and may in some cases be referred to as a "processor."

[0122] As defined herein, the term "soft" with respect to a circuit means that the circuit is implemented in programmable logic or programmable circuitry. Thus, a "soft processor" refers to at least one circuit implemented in programmable circuitry that is capable of executing instructions contained in a program code.

[0123] As defined herein, the term "output" means storing in a physical memory element, e.g., a device, writing to a display or other peripheral output device, sending or transmitting to another system, exporting, etc.

[0124] As defined herein, the term "substantially" means that the recited characteristic, parameter, or value need not be achieved exactly, but that deviations or variations including, for example, tolerances, measurement errors, measurement accuracy limitations, and other factors known to those of ordinary skill in the art may occur in an amount that does not eliminate the effect that the characteristic is intended to provide.

[0125] Terms such as first, second, etc. may be used herein to describe various elements, and these elements should not be limited by these terms, as these terms are only used to distinguish one element from another, unless otherwise stated or the context clearly indicates otherwise.

[0126] A computer program product may include a computer-readable storage medium having computer-readable program instructions thereon for causing a processor to perform the inventive configurations described herein. Within this disclosure, the term "program code" is used interchangeably with the term "computer-readable program instructions." The computer-readable program instructions described herein may be downloaded from the computer-readable storage medium to the respective computing / processing device or over a network, such as the Internet, a LAN, a WAN, and / or a wireless network, or to an external computer or external storage device. The network may include copper transmission cables, optical transmission fiber, wireless transmission, edge devices including routers, firewalls, switches, gateway computers, and / or edge servers. A network adapter card or network interface in each computing / processing device receives the computer-readable program instructions from the network and transmits the computer-readable program instructions for storage in the computer-readable storage medium in the respective computing / processing device.

[0127] The computer readable program instructions for carrying out the operations of the inventive arrangements described herein may be either assembler instructions, instruction-set-architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, or source or object code written in any combination of one or more programming languages, including object-oriented and / or procedural programming languages. The computer readable program instructions may include state setting data. The computer readable program instructions may execute entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer, partially on a remote computer, or entirely on a remote computer or server. In the latter scenario, the remote computer may be connected to the user's computer via any type of network, including a LAN or WAN, or may be connected to an external computer (e.g., via the Internet using an Internet Service Provider). In some cases, electronic circuitry, including, for example, programmable logic circuitry, FPGAs, or PLAs, may execute the computer readable program instructions by utilizing state information of the computer readable program instructions to individualize the electronic circuitry to carry out aspects of the inventive arrangements described in this disclosure.

[0128] Certain aspects of the disclosed arrangements are described herein with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer readable program instructions, e.g., program code.

[0129] These computer readable program instructions may be provided to a processor of a computer, special purpose computer, or other programmable data processing apparatus such that the instructions, executed by the processor of the computer or other programmable data processing apparatus, result in a machine that creates means for implementing the functions / acts specified in the flowchart and / or block diagram blocks. These computer readable program instructions may also be stored on a computer readable storage medium that can direct a computer, programmable data processing apparatus, and / or other device to function in a particular manner such that the computer readable storage medium on which the instructions are stored comprises an article of manufacture that includes instructions that implement aspects of the operations specified in the flowchart and / or block diagram blocks.

[0130] The computer-readable program instructions may also be loaded into a computer, other programmable data processing apparatus, or other device to cause a sequence of operations to be performed on the computer, other programmable apparatus, or other device to generate a computer-implemented process such that the instructions, which execute on the computer, other programmable apparatus, or other device, implement the functions / acts specified in the flowchart and / or block diagram blocks.

[0131] The flowcharts and block diagrams in the Figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various aspects of the disclosed configurations. In this regard, each block in the flowchart or block diagram may represent a module, segment, or portion of instructions, which may include one or more executable instructions for implementing the specified operation.

[0132] In some alternative implementations, the actions noted in the blocks may occur out of the order noted in the figures. For example, two blocks shown in succession may be executed substantially simultaneously, or the blocks may sometimes be executed in the reverse order, depending on the functionality involved. In other examples, the blocks may be executed in generally ascending numerical order, and in still other examples, one or more blocks may be executed in various orders, with the results stored and utilized in a subsequent block or in another block that does not immediately follow. It should also be noted that each block of the block diagrams and / or flowchart illustrations, and combinations of blocks in the block diagrams and / or flowchart illustrations, may be implemented by dedicated hardware-based systems that perform the specified functions or acts, or a combination of dedicated hardware and computer instructions.

Claims

1. 1. A method for synchronizing system resources in a multi-socket data processing system, the method comprising: shutting down each of a plurality of system resources under control of a primary system on chip (SOC), each system resource being located within a different one of a plurality of SOCs of a multi-SOC system, the primary SOC being one of the plurality of SOCs; using the primary SOC to write updated values ​​to the system resources of each other SOC of the plurality of SOCs while stopped; and after said writing, initiating operation of the system resources on each SOC of said plurality of SOCs using said primary SOC.

2. after said stopping and before said writing, reading a value from each of said plurality of system resources using said primary SOC; 2. The method of claim 1, comprising: for each SOC of the plurality of SOCs other than the primary SOC, determining, by the primary SOC, a difference between a value read from the system resource of the primary SOC and the value read from the system resource of the other SOC.

3. 3. The method of claim 2, wherein the writing includes the primary SOC writing the updated value to each of the other SOCs, the updated value written to each of the other SOCs depending on the difference between the value read from the system resource of the primary SOC and the value read from the system resource of the other SOCs.

4. 2. The method of claim 1 , wherein the writing of the updated value to each system resource of the other SOC and the initiating of the operation takes into account a latency in the primary SOC communicating with the respective system resource of the plurality of system resources in the respective other SOC.

5. The method of claim 1 , wherein the multiple SOCs are the same.

6. The method of claim 1 , wherein the stopping, the writing, and the starting are communicated as part of a boot process.

7. The method of claim 1 , wherein communications between the primary SOC and the other SOCs are conveyed via a communications bus that communicatively links the plurality of SOCs.

8. 1. A system for synchronizing system resources in a multi-socket data processing system, the system comprising: a plurality of system-on-chip (SOCs), the plurality of SOCs being interconnected via a communication bus, each SOC including a plurality of processor cores, one SOC of the plurality of SOCs being designated as a primary SOC; a global synchronous circuit; a plurality of sideband channels coupling the global synchronous circuit to each of the plurality of SOCs; the primary SOC is configured to provide a trigger event over a first sideband channel of the plurality of sideband channels, the trigger event initiating synchronization of the system resources disposed within the plurality of SOCs; the global synchronization circuitry is configured to broadcast a synchronization event to each of the plurality of SOCs over a second sideband channel of the plurality of sideband channels in response to receiving the trigger event.

9. The system of claim 8 , wherein the synchronization event is received simultaneously by each SOC of the plurality of SOCs.

10. The system of claim 8 , wherein each SOC of the plurality of SOCs is disposed in a different socket.

11. The system of claim 8 , wherein the multiple SOCs are the same.

12. The system of claim 8 , wherein the trigger event and the synchronization event are communicated as part of a boot process.

13. The system of claim 8 , wherein a selected processor core of only the primary SOC provides the trigger event.

14. the synchronization event is an interrupt, and in each SOC of the plurality of SOCs, the interrupt is received by a selected processor core of the plurality of processor cores; 9. The system of claim 8, wherein the selected processor core in each SOC executes an interrupt service routine in response to the interrupt to program the system resource by writing to a control register of the system resource in the same SOC.

15. Within each SOC of the plurality of SOCs, the synchronization event is received by a control logic of the system resource; The system of claim 8 , wherein in response to the synchronization event, the control logic of the system resource resets the system resource within the same SOC.

Citation Information

Patent Citations

  • Synchronous method and apparatus for a plurality of processors

    JP1989258057A

  • Operation synchronizing circuit

    JP2004287493A

  • Timer Offset Mechanism in Virtual Machine Environment

    JP2008518367A

  • Controlling the offset of the timestamp counter (TSC) across multiple cores and threads.

    JP2012531001A

  • Balancing power between discrete components in a compute node

    US20210191490A1