Probe filter retention based low power state

The data fabric in SOCs employs a power state control circuit to gate clock signals to probe filters, addressing the inefficiencies in power saving modes by maintaining power and reducing idle consumption.

JP7592200B2Active Publication Date: 2024-11-29ADVANCED MICRO DEVICES INC
View PDF 4 Cites 0 Cited by

Patent Information

Application Number
JP2023577544
Authority / Receiving Office
JP · JP
Patent Type
Patents
Current Assignee / Owner
Priority Date
2021-06-24
Filing Date
2022-06-07
Publication Date
2024-11-29
Estimated Expiration
2042-06-07

AI Technical Summary

Technical Problem

The challenge in implementing chip-wide low power states in complex system-on-chips (SOCs) arises from the diverse operations of peripheral controllers requiring real-time bandwidth, hindering the realization of efficient power saving modes, especially when CPU and GPU are idle but peripheral controllers remain active.

Method used

A data fabric within the SOC is designed with a power state control circuitry that gates the clock signal to probe filters while maintaining power, entering a retained low power state upon completion of pending lookups, and inhibiting new lookups based on specified conditions.

Benefits of technology

This approach allows for significant power conservation in the data fabric by clock-gating probe filters without power gating, maintaining data integrity and reducing idle power consumption in the SOC.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 0007592200000001
    Figure 0007592200000001
  • Figure 0007592200000002
    Figure 0007592200000002
  • Figure 0007592200000003
    Figure 0007592200000003
Patent Text Reader

Abstract

The data fabric routes requests between a plurality of requesters and a plurality of responders. The data fabric includes a crossbar router, a coherent slave controller coupled to the crossbar router, and a probe filter coupled to the coherent slave controller for tracking a state of a cached line of memory. The power state control circuitry operates in response to detecting any of a plurality of specified conditions to cause the probe filter to enter a retention low power state in which a clock signal to the probe filter is gated while power to the probe filter is maintained. Entry into the retention low power state is performed when all in-flight probe filter lookups are completed.
Need to check novelty before this filing date? Find Prior Art

Description

[Background technology]

[0001] Computer systems utilize a variety of peripheral components for different input / output and communication functions. A system-on-a-chip (SOC) combines data processing circuitry, such as a central processing unit (CPU) core and a graphics processing unit (GPU), with peripheral controllers and memory interfaces on a single integrated circuit chip, and is well suited for portable, battery-powered operation. For example, a SOC may incorporate a display controller, an image signal processor (ISP), and other peripheral controllers on the SOC to enable the input and output of information to and from the computer system. In such large, complex SOCs, devices typically transfer data between resources, such as memory, by routing accesses through a large on-chip routing circuitry or "data fabric."

[0002] The diversity of circuits on a typical SOC presents a problem for implementing power saving modes. Many of the peripheral controllers operate in real time, are isochronous, and require a certain amount of bandwidth from the shared memory resource over a period of time. This requirement has prevented the realization of chip-wide low power states. For example, when the CPU cores and GPU are idle, the peripheral controllers may still be active and receiving data that can be stored in a buffer, but the data must be sent to memory before the buffer is full.

[0003] On the other hand, CPUs and GPUs tend to have periods of high activity separated by idle periods. If the idle periods are expected to be short, the data fabric, memory controller, and memory will typically remain active and therefore can quickly service requests from the peripheral controller to transfer more data. [Brief description of the drawings]

[0004] [Figure 1]FIG. 1 is a block diagram of an Advanced Configuration and Power Interface (ACPI) model for computer system power states as known in the prior art. [Diagram 2] 1 is a block diagram of a data processing system according to some embodiments. [Diagram 3] 3 is a block diagram of the data fabric of FIG. 2 in accordance with some embodiments. [Figure 4] 5 is a flowchart illustrating a process for entering a retained low power state in accordance with an illustrative embodiment. [Diagram 5] 5 is a flowchart illustrating a process for exiting the retained low power state process of FIG. 4. [Figure 6] 6 is a flowchart illustrating another process for entering a retained low power state in accordance with another illustrative embodiment. [Figure 7] 6 is a flowchart illustrating another process for exiting and entering a retained low power state in accordance with an illustrative embodiment. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0005] In the following description, the use of the same reference numbers in different figures indicates similar or identical items. Unless otherwise noted, the word "coupled" and its related verb forms include both direct and indirect electrical connections by means known in the art, and unless otherwise noted, any description of a direct connection also refers to an alternative embodiment using a suitable form of indirect electrical connection.

[0006] The data processor includes a plurality of requesters, a plurality of responders, and a data fabric for routing requests between the plurality of requesters and the plurality of responders, the data fabric including a crossbar router, a coherent slave controller connected to the crossbar router, a probe filter connected to the coherent slave controller, and power state control circuitry. The probe filter tracks a state of a cached line of memory. The power state control circuitry is operable, in response to detecting any of a plurality of specified conditions, to cause the probe filter to enter a retention low power state in which a clock signal to the probe filter is gated while power to the probe filter is maintained. Entry into the retention low power state is performed when all outstanding probe filter lookups are completed.

[0007] The method includes routing requests between a plurality of requesters and a plurality of responders in a data fabric of a data processor. A state of a cached line of memory is tracked in a probe filter coupled to the data fabric. A condition suitable for a probe-filter retaining low power state is detected, and in response, a coherent slave controller is instructed to inhibit or disable new probe-filter lookups. A clock signal provided to the probe filter is gated.

[0008] The power state control circuit controls the power state of the data fabric. The power state control circuit includes a power state control circuitry, a first clock gating circuit, and a second clock gating circuit. The power state control circuitry is operable to cause a probe filter of the data fabric to enter a retention low power state in response to detecting any of a plurality of specified conditions, in which a clock signal to the probe filter is gated while power to the probe filter is maintained. The first clock gating circuit is coupled to the power state control circuitry for selectively gating a first clock signal to the probe filter. The second clock gating circuit is coupled to the power state control circuitry for selectively gating a second clock signal to the data fabric. Entering the retention low power state is performed when all ongoing probe filter lookups are completed.

[0009] 1 is a block diagram of an Advanced Configuration and Power Interface (ACPI) model for computer system power states 100 as known in the prior art. The ACPI model 100 is part of the ACPI specification developed by various computer system, software and component vendors to establish an industry standard interface for device configuration and power management of both devices and entire computer systems.

[0010] The ACPI model 100 defines a hierarchy of states for various devices within a computer system organized around a set of global system states including a G0 (S0) state 110, known as a "Working" state, a G1 state 120, known as a "Sleeping" state, a G2 (S5) state 130, known as a "Soft Off" state, and a G3 state 140, known as a "Mechanical Off" state. In addition, the ACPI model 100 includes a state 150, known as a "Legacy" state, that corresponds to existing Basic Input / Output System (BIOS) power management interfaces that correspond to BIOS operation before an ACPI-compatible operating system is loaded. The ACPI model 100 also defines a set of Central Processing Unit (CPU) states 160 and a set of device states or "D" states 170 for devices such as modems, hard disk drives (HDDs), and compact disc read-only memories (CDROMs).

[0011] The ACPI model 100 is well known and most aspects will not be described further, however certain aspects relating to central processing unit power states are relevant to understanding the present disclosure and will be described here.

[0012] In the G0 (S0) state 110, peripheral devices can dynamically change their power states and a user can select various performance and power characteristics of the system through a user interface so that software systems can optimize the computer for desired performance or battery life. In this state, the computer system responds in real time to external events. From the G0 (S0) state, the computer system can transition to the G1 state 120, which includes various power-saving sleep states or "S states," under the control of system firmware, such as the BIOS, and return to an operational state in response to a wake event.

[0013] In the G0(S0) state 110, the CPU can operate in a variety of states including a "C0" state, in which the processor executes instructions, and various low power or "idle" states, labeled "C1" through "Cn." In the C0 state, the CPU can operate in various performance states or "P-states," and the operating frequency can be increased or throttled to support the needs of currently running application programs using dynamic voltage and frequency scaling (DVFS). In the idle states C1 through Cn, the CPU is idle and does not execute instructions, but has various combinations of power consumption and wake-up latency, with lower power consumption being traded off for longer wake-up latency. Thus, for example, the C1 state has the highest power consumption but the shortest wake-up latency, while the Cn state has the lowest power consumption but the longest wake-up latency.

[0014] Using ACPI Model 100, each particular computer system can support different combinations of P-states and C-states appropriate to the application environment. Achieving a better tradeoff between reduced power consumption and low latency in ways not specified by ACPI Model 100 remains a desirable goal.

[0015] 2 is a block diagram of a data processing system 200, according to some embodiments. Data processing system 200 includes a data processor in the form of a SOC 210 and external memory in the form of low power double data rate synchronous dynamic random access memories (LPDDRx SDRAMs) 273 and 283. Many other components of an actual data processing system would typically be present, but are not relevant to understanding the present disclosure and are not shown in FIG. 2 for ease of illustration.

[0016] SOC 210 generally includes system management unit (SMU) 211, system management network (SMN) 212, central processing unit (CPU) core complex 220, labeled "CCX," graphics controller 230, labeled "GFX," real-time client subsystem 240, memory / client subsystem 250, data fabric 260, memory channels 270 and 280, and peripheral component interface express (PCIe) subsystem 290. As will be appreciated by those skilled in the art, SOC 210 may not have all of these elements present in all embodiments and may further have additional elements included therein.

[0017] SMU 211 is bidirectionally connected to the major components in SOC 210 via SMN 212. SMN 212 forms the control fabric for SOC 210. SMU 211 is a local controller that controls the operation of resources on SOC 210 and synchronizes communication between them. SMU 211 manages the power-up sequencing of the various processors on SOC 210 and controls multiple off-chip devices via reset, enable, and other signals. SMU 211 includes one or more clock sources (not shown), such as a phase locked loop (PLL), to provide clock signals to each of the components of SOC 210. SMU 211 can also receive measured power consumption values ​​from the CPU cores in CPU core complex 220 and graphics controller 230 to manage power for the various processors and other functional blocks and determine appropriate P-states.

[0018] CPU core complex 220 includes a set of CPU cores, each of which is bidirectionally connected to SMU 211 via SMN 212. Each CPU core may be a single core that shares only a last-level cache with the other CPU cores, or may be combined with some, but not all, of the other cores in a cluster.

[0019] Graphics controller 230 is bidirectionally connected to SMU 211 via SMN 212. Graphics controller 230 is a high-performance graphics processing unit capable of performing graphics operations such as vertex processing, fragment processing, shading, and texture blending in a highly integrated and parallel manner. Graphics controller 230 requires periodic access to external memory to perform its operations. In the embodiment shown in FIG. 2, graphics controller 230 shares a common memory subsystem with the CPU cores in CPU core complex 220, an architecture known as a unified memory architecture. Because SOC 210 includes both a CPU and a GPU, it is also referred to as an accelerated processing unit (APU).

[0020] The real-time client subsystem 240 includes a set of real-time clients, such as representative real-time clients 242 and 243, and a memory management hub 241, labeled "MM HUB." Each real-time client is bidirectionally connected to the SMU 211 via the SMN 212, and to the memory management hub 241. A real-time client may be any type of peripheral controller that requires the periodic movement of data, such as an image signal processor (ISP), an audio coder-decoder (CODEC), a display controller that renders and rasterizes objects generated by the graphics controller 230 for display on a monitor, etc.

[0021] Memory / client subsystem 250 includes a set of memory elements or peripheral controllers, such as representative memory / client devices 252 and 253, and a system and input / output hub, labeled "SYSHUB / IOHUB," 251. Each memory / client device is bidirectionally connected to SMU 211 via SMN 212, and bidirectionally connected to system and input / output hub 251. A memory / client device is a circuit that stores or requires access to data on a non-periodic basis, such as non-volatile memory, static random access memory (SRAM), external disk controllers such as a Serial Advanced Technology Attachment (SATA) interface controller, a Universal Serial Bus (USB) controller, a system management hub, etc.

[0022] Data fabric 260 is the interconnect that controls the flow of traffic in SOC 210. Data fabric 260 is bidirectionally connected to SMU 211 via SMN 212, and bidirectionally connected to CPU core complex 220, graphics controller 230, memory management hub 241, and system and I / O hub 251. Data fabric 260 includes a crossbar switch for routing memory mapped access requests and responses between any of the various devices in SOC 210. The data fabric also includes a system memory map defined by the basic input / output system (BIOS) to determine the destination of memory accesses based on the system configuration, and buffers for each virtual connection.

[0023] Memory channels 270 and 280 are circuits that control the transfer of data between external memories 273 and 283. Memory channel 270 is formed by memory controller 271 and physical interface circuit 272, labeled "PHY", connected to external memory 273. Memory controller 271 is bidirectionally connected to SMU 211 via SMN 212, and has an upstream port and a downstream port bidirectionally connected to data fabric 260. Physical interface circuit 272 has an upstream port bidirectionally connected to memory controller 271, and a downstream port bidirectionally connected to external memory 273. Similarly, memory channel 280 is formed by memory controller 281 and physical interface circuit 282 connected to external memory 283. Memory controller 281 is bidirectionally connected to SMU 211 via SMN 212, and has an upstream port and a downstream port bidirectionally connected to data fabric 260. The physical interface circuit 282 has an upstream port bidirectionally connected to the memory controller 281 and a downstream port bidirectionally connected to the external memory 283 .

[0024] Peripheral Component Interface Express (PCIe) subsystem 290 includes PCIe controller 291 and PCIe physical interface circuit 292. PCIe controller 291 is bidirectionally connected to SMU 211 via SMN 212 and has an upstream port bidirectionally connected to system and input / output hub 251 and a downstream port. PCIe physical interface circuit 292 has an upstream port bidirectionally connected to PCIe controller 291 and a downstream port bidirectionally connected to a PCIe fabric, not shown in FIG. 2. The PCIe controller can form a PCIe root complex of a PCIe system for connection to a PCIe network including PCIe switches, routers and devices.

[0025] In operation, SOC 210 integrates on a single chip a complex combination of computing and storage devices, including CPU core complex 220 and graphics controller 230. Most of these controllers are well known and will not be described further. SOC 210 also implements various system monitoring and power saving functions, including P-state elevation and throttling, and entering and exiting C-states, under the control of SMU 211 and / or operating system software.

[0026] For example, SOC 210 controls the entry and exit of CPU core complex 220 and graphics controller 230 into and out of various idle states based on idle periods during which no instructions are being executed. As is known, when CPU core complex 220 and graphics controller 230 all enter an idle state, other components of SOC 210 may also be placed into appropriate idle states to save power. In one example, CPU core complex 220 supports two idle states, including a shallow idle state (C1) and a deep idle state (C6) state. According to ACPI mode, the C1 state is the lightest C state and may simply involve clock gating and reduced power supply voltages that allow SOC 210 to remain in that state but also return to an active state with very low latency. On the other hand, the C6 state is a very deep C state that reduces power consumption by gating various clock signals and power supply voltages, but also has a very long exit latency.

[0027] In some known systems, other components of SOC 210 may also be placed in various idle states to correspond to the idle states of the data processing elements. In one particular example, when each CPU and GPU enters the C1 state, the data fabric and memory system remain fully operational, but consume significant power when they are not generally needed. However, when each CPU and GPU enters the C6 state, data fabric 260 is also placed in a corresponding idle power state using power gating or clock gating. Before entering this idle power state, the data fabric places the external memory in a self-refresh state and the memory controller and PHY in a similar low power state.

[0028] 3 is a block diagram of data fabric 260 of FIG. 2 according to some embodiments. Data fabric 260 includes a set of master ports 310, a set of slave ports 320, a crossbar router 330, and a power controller circuit 350.

[0029] Master ports 310 include a port for each device capable of generating memory access requests, including representative coherent master ports 311 and 312, and representative master ports 313 and 314. Each of ports 311, 312, 313, 314 is adapted to be bidirectionally connected to a corresponding requesting device and has a bidirectional downstream port, as shown in FIG. 2. Each master port has an internal FIFO buffer for temporarily storing accesses. Coherent master ports 311 and 312 connect in the upstream direction to coherent requestors, such as CCX 220, GFX 230, or other CPU or GPU requestors, in various embodiments. Master ports 313 and 314 connect in the upstream direction to primary non-coherent requestors, such as a display controller or multimedia controller, via MM hub 241 and IOHUB 251, respectively.

[0030] Slave ports 320 include a port for each device capable of responding to memory access requests, including representative slave port 321, and representative coherent slave ports 322 and 323. Each of slave ports 321, 322, 323 has a bidirectional upstream port and is adapted to be bidirectionally connected to a corresponding responding device as shown in FIG. 2. Each slave port has an internal FIFO buffer for temporarily storing accesses. Each of coherent slave ports 322 and 323 is bidirectionally connected to a respective probe filter 360 to track the state of cached lines of memory. Each coherent slave port includes a coherent slave controller for the port.

[0031] Crossbar router 330 has a bidirectional connection to the downstream bidirectional connection of master ports 310 and a bidirectional connection to the upstream bidirectional connection of slave ports 320. Crossbar router 330 is the heart of data fabric 260, routing accesses between master ports and corresponding slave ports, thereby creating virtual connections between masters and slaves based on the destination address of the requester.

[0032] Power controller circuit 350 includes local C-state controller 351, clock gating circuit 352, power gate® circuit 353, and probe filter clock gate 345. Local C-state controller 351 has a first input for receiving a signal labeled “SOC STATE”, a first output connected to probe filter clock gate 354, a second output connected to coherent slave ports 322 and 323 carrying a signal labeled “Flushed”, a third output connected to coherent slave ports 322 and 323 carrying a signal labeled “Disable”, a fourth output connected to coherent slave ports 322 and 323 carrying a signal labeled “Wake”, and fifth and sixth outputs. Clock gating circuit 352 has an input for receiving a data fabric clock signal labeled “FCLK”, a control input connected to a fifth output of local C-state controller 351, and an output for selectively providing a gated clock signal to data fabric 260. Power Gate® circuit 353 has an input for receiving a data fabric power supply voltage labeled “VDD”, a control input connected to the sixth output of local C-state controller 351, and an output for providing a selectively gated power supply voltage to data fabric 260.

[0033] Probe filter clock gate 354 has an input (not shown) that receives the clock signal FCLK from clock gating circuit 352, and an output connected to probe filter 360. Probe filter clock gate 354 selectively operates under control of C-state controller 351 to gate the clock signal to probe filter 360 to provide a retain low power state. In the retain low power state, the clock signal to probe filter 354 is gated but power is maintained to the probe filter, as described further below.

[0034] Generally, data fabric 260 routes memory access requests and responses between requesters (CPU cores, GPUs, real-time client controllers, etc.) and responders (external memory via memory controllers and PHYs, on-chip memory, etc.). In some scenarios, data fabric 260 implements a retained low power state in which the probe filter is clock gated but retains power. This retained low power state is used in conjunction with C-states in which the probe filter 360 is not power gated, but instead clock gated. Caches for CCX 220, such as level 3 caches, are flushed in some scenarios in which this retained low power state is employed, but not in other scenarios.

[0035] In response to detecting any of a number of specified conditions, power controller circuit 350 is operable to cause the probe filter to enter a retaining low power state in which the clock signal to the probe filter is gated while power to the probe filter 360 is maintained. Entering the retaining low power state involves coherent slave controllers 322 and 323 inhibiting new probe filter lookups and signaling the power state control circuitry when all outstanding probe filter lookups are completed.

[0036] In this embodiment, C-state controller 351 and probe-filter clock gate 354 include power state control circuitry that operates to provide a retention low power state for the probe-filter. This state is used at selected times during the C-state of this embodiment and at selected times when data fabric 260 is active. Use of this state allows probe-filters 360 to conserve power when not in use by gating their clock signals, but maintain their power supply to retain probe-filter data. The power state control circuitry of C-state controller 351 is operable to cause probe-filter 360 to enter the retention low power state in response to detecting any of a number of specified conditions. As described further below, the process of entering the retention low power state includes a coherent slave controller, such as one in coherent slave port 322 or 323, signaling the power state control circuitry by deasserting a "wake" signal to inhibit new probe-filter lookups and when all outstanding probe-filter lookups have completed.

[0037] In this embodiment, each of three specified conditions can cause probe filter 360 to enter a holding low power state. These include a first state in which all requesters that are probe filter clients have flushed their caches, a second state in which APU 200 or other host data processor or data fabric 260 has entered an idle power state, and a third state in which the data processor has entered a low power state in which only the primary non-coherent requester is active. In some embodiments, only one or two of these conditions are supported.

[0038] 4 is a flowchart 400 illustrating a process for entering a retained low power state, according to an example embodiment. The illustrated process is suitable for implementation by data fabric 260 under control of local C-state controller 351 or other suitable data fabric circuitry including probe filter clock gating functionality and suitable power state controller power state control circuitry similar to local C-state controller 351, etc.

[0039] The illustrated process begins at block 402 with a local power state controller detecting a specified condition where probe filter client requesters flush their caches and enter an idle state. In some embodiments, the probe filter clients include only CPU requesters or only CPU and GPU requesters. In this embodiment, detecting the specified condition includes receiving a signal from a system level power state controller.

[0040] In response, in block 404, the local power state controller asserts a "flush" signal (FIG. 3) to signal the coherent slave controller to begin entering the probe filter retention low power state (entry into low power state retention). In the illustrated scenario, this signal serves as a command for the coherent slave controller to inhibit probe filter lookups. Although multiple coherent slave controllers are described in this example, in some embodiments the process is used with only one coherent slave controller. Upon receiving this signal, in block 406, the coherent slave controller begins inhibiting new probe filter lookups. Because the cache has been flushed, in block 405, the coherent slave controller can treat all requests to memory addresses that were previously cacheable as cache misses and not generate a probe filter lookup.

[0041] In block 408, the coherent slave controller completes all probe filter lookups that were in progress when the command to inhibit new probe filter lookups was received in block 404. Once these in-progress lookups are complete, in block 410, the coherent slave controller deasserts the wake signal to signal to the power state controller that the in-progress lookups are complete.

[0042] Upon receiving the deasserted wake signal, the local power state controller waits at least a delay time of the round trip signaling delay from the local power state controller to the coherent slave controller, block 412. The local power state controller then signals the coherent slave controller to disable all probe filter lookups by asserting a disable signal, block 414. It also places the probe filter in a retention low power state by activating clock gate 354 to gate the clock signal to the probe filter. In the retention low power state, the probe filter power is maintained and the probe filter maintains the cache line state data held therein.

[0043] Figure 5 is a flow chart 500 illustrating a process for exiting the retained low power state of Figure 4. The flush, disable and wake signals (Figure 3) used to control entry into and exit from the retained low power state allow entry into and exit from the retained low power state in a variety of scenarios depending on which of several specified conditions prompt entry into the state. The process of Figure 5 is used for exit when the idle condition of block 402 (Figure 4) causes entry into the retained low power state.

[0044] At block 502, the process begins in response to a CPU interrupt causing the CPU to exit an idle state. In this embodiment, a SOC STATE signal provides this status to the local C-state controller 351. In other embodiments, a similar signal is provided to the local power state controller. In response to receiving this signal, the local power state controller restores the clock signal to the probe filter and waits for the round trip signaling delay time from the local power state controller to the coherent slave controller.

[0045] The local power state controller then signals the coherent slave controller to resume servicing the probe filter lookup by deasserting the disable signal in block 506. Finally, as part of the CPU process of exiting the idle state, it reconnects to the data fabric by opening the coherent master port. In response to this connection, the local power state controller deasserts the Flushed signal to the coherent slave controller, causing the coherent slave controller to resume performing the probe filter lookup.

[0046] 6 is a flowchart 600 of another process for entering a retained low power state, according to another example embodiment. The illustrated process is suitable for implementation using data fabric 260 under control of local C-state controller 351 or other suitable data fabric circuitry.

[0047] In block 602, the process detects a condition for entering a low power state, such as the C state described above, during an idle state of the APU that does not include a flushed cache for the CPU. The process for entering the retained low power state differs from the process of FIG. 4 because the CPU caches have not been flushed.

[0048] In block 604, the local power state controller determines whether the wake signal is asserted. If so, indicating that the coherent slave controller is processing a probe filter lookup, the process proceeds to block 605, where it temporarily refrains from entering the hold low power state and waits for the wake signal to deassert. If the wake signal is not asserted in block 604 (or is deasserted in block 605), the process proceeds to block 606, where it signals the coherent slave controller to disable the probe filter lookup by asserting a disable signal.

[0049] The local power state controller then waits for the round trip signaling delay time from the local power state controller to the coherent slave controller in block 608. Following this wait, it again checks whether the wake signal is asserted in block 610. If not, the process proceeds to block 612. If so, the local power state controller deasserts the disable signal in block 611 and aborts the process of entering a holding low power state to wait for the wake signal to deassert in block 613. The data fabric is then placed in a low power state in block 614, such as the C state described above.

[0050] The illustrated sequence allows the data fabric to move from the idle state to the C state, while having the added advantage of allowing coherent requests to be serviced if they arise during the transition process.

[0051] Figure 7 is a flow chart 700 illustrating another process for entering and exiting a retained low power state, according to an exemplary embodiment. The illustrated process is suitable for implementation with data fabric 260 under control of local C-state controller 351 or other suitable data fabric circuitry. Preferably, the illustrated process is supported in combination with the process of Figure 6 to provide the ability to provide probe filter lookup for coherent requests from stuttering clients. Support for the enter process of Figure 4 and the exit process of Figure 5 are also advantageously used in combination with the processes of Figure 6, Figure 7, or both.

[0052] Referring to flowchart 700, the process begins at block 702 when the data fabric exits an idle state to service stuttering client data traffic. Stuttering clients are considered to be the primary non-coherent requestors on the data fabric because the majority of their requests are not coherent requests. At the start of this scenario, the CPU clients are idle and only the stuttering clients are active. This state can be detected by detecting a disconnection condition on the fabric data port coupled to the requestor that is the probe filter client. Traffic from stuttering clients is typically non-coherent, although a small percentage of traffic from such clients may be coherent. These clients are therefore referred to as primarily non-coherent. Because such coherent requests are rare, the probe filter can exit a hold low power state to service such requests without incurring any detrimental additional latency performance penalty.

[0053] The coherent slave controller receives a coherent request while the probe filter is in the hold low power state, at block 704. In response to the coherent request, the coherent slave controller asserts a wake signal to the local power state controller, at block 706.

[0054] Then, in block 708, the local power state controller restores the clock signal to the probe filter, for example, by operating clock gate 354 (FIG. 3). Then, a round trip signaling delay from the local power state controller to the coherent slave controller is waited for. After waiting, in block 710, the local power state controller resumes generating and servicing probe filter lookups for coherent requests by signaling the coherent slave controller to deassert the disable signal. The sequence of blocks 702-710 is also suitable for exiting a held low power state entered according to the process of FIG. 6.

[0055] In block 712, the local power state controller determines whether the request is from a stuttering client. This determination is preferably based on the state of the data fabric and the port on which the request is received. If not, the process proceeds to block 713, where the probe filter is maintained in an operational state to service the probe filter lookup. If so, the process proceeds to block 714, where the coherent slave services the coherent request from the stuttering client, generates the probe filter lookup, and receives the results from the probe filter. Once the stutter client has completed the data transfer for the current cycle of stuttering, in block 716, the stutter client informs the local power state controller that the stutter data transfer is complete.

[0056] In response, in block 718, the local power state controller signals the coherent slave controller to disable the probe filter lookup by asserting a disable signal and gates the clock signal to the probe filter to place the probe filter in a hold low power state. Then, in block 720, the local power state controller causes the data fabric to re-enter the idle state until the next cycle of the stutter client data transfer.

[0057] As can be appreciated, if the stuttering client does not make any coherent requests, as shown in block 704, as is often the case during the C state described above, the probe filter is not required to exit the holding low power state, and further power savings are realized.

[0058] SOC 210 or any portion thereof, such as data fabric 260 of FIG. 2, may be described or represented by a computer-accessible data structure in the form of a database or other data structure that may be read by a program and used directly or indirectly to manufacture an integrated circuit. For example, this data structure may be a behavioral or register-transfer level (RTL) description of the hardware functionality in a high-level design language (HDL) such as Verilog or VHDL. The description may be read by a synthesis tool that may synthesize the description to generate a netlist that includes a list of gates from a synthesis library. The netlist includes a set of gates that also represent the functionality of the hardware that comprises the integrated circuit. The netlist may then be placed and routed to generate a data set that describes the geometric shapes to be applied to a mask. The mask may then be used in various semiconductor manufacturing processes to manufacture the integrated circuit. Alternatively, the database on the computer-accessible storage medium may be a netlist (with or without a synthesis library) or a data set, or Graphic Data System (GDS) II data, as desired.

[0059] Although specific embodiments have been described, various modifications to these embodiments will be apparent to those skilled in the art. The various techniques used in conjunction with the retained low power state for the probe filter disclosed herein may be used independently or in conjunction with other techniques. Furthermore, different techniques and circuits may be used to detect the condition for entering the retained low power state.

[0060] Therefore, the appended claims are intended to cover all modifications of the disclosed embodiments that fall within the scope of the disclosed embodiments.

Claims

1. 1. A data processor comprising: Multiple requesters; Multiple responders, a data fabric for routing requests between the plurality of requesters and the plurality of responders; The data fabric comprises: A crossbar router; a coherent slave controller coupled to the crossbar router; a probe filter coupled to the coherent slave controller for tracking a state of a cached line of memory; and power state control circuitry for causing the probe-filter to enter a retention low power state in response to conditions suitable for the probe-filter retention low power state in which a clock signal to the probe-filter is gated while power to the probe-filter is maintained; entering the holding low power state is performed when all in-flight probe filter lookups are completed. Data processor.

2. the coherent slave controller inhibits or disables new probe filter lookups and signaling to the power state control circuitry when all in-flight probe filter lookups are completed; 2. The data processor of claim 1.

3. The conditions for qualifying for the probe-filter retained low power state include a condition in which all requesters that are probe-filter clients have flushed their caches and entered an idle power state.

2. The data processor of claim 1.

4. the conditions for qualifying for the probe-filter retained low power state include the condition that the data fabric has entered an idle low power state; 2. The data processor of claim 1.

5. the conditions for qualifying for the probe filter retained low power state include a condition in which the data processor has entered a low power state in which only a primary non-coherent requestor is active; 2. The data processor of claim 1.

6. detecting that the data processor has entered a low power state in which only a primary non-coherent requestor is active includes detecting a disconnect condition of a fabric data port coupled to a requestor that is a probe filter client.

5. The data processor of claim 4.

7. the coherent slave controller sends a wake signal to the power state control circuitry upon receiving a coherent cacheable request from a primary non-coherent requestor; the power state control circuitry causes the probe filter to exit the retained low power state in response to receiving the wake signal.

5. The data processor of claim 4.

8. Suitable conditions for each probe-filter-preserving low power state include two or more of: all requesters that are probe-filter clients flush their caches; the data processor enters an idle power state; and the data processor enters a low power state in which only primary non-coherent requesters are active.

2. The data processor of claim 1.

9. A probe filter client, the requester, includes a central processing unit (CPU) core; the primary non-coherent requestor comprises a display controller or a graphics controller; 8. The data processor of claim 7.

10. 1. A method comprising: Routing requests between a plurality of requesters and a plurality of responders in a data fabric of a data processor; tracking a state of a cached line of memory in a probe filter coupled to the data fabric; Detecting conditions suitable for a probe-filter-preserving low power state and responsively instructing the coherent slave controllers to inhibit or disable new probe-filter lookups; gating a clock signal provided to the probe filter. method.

11. receiving a signal from the coherent slave controller indicating that all probe filter lookups that were in process when the command to the coherent slave controller was received are completed; The method of claim 10.

12. The conditions for qualifying for the probe-filter retained low power state include all requesters that are probe-filter clients flushing their caches and entering an idle power state. The method of claim 10.

13. the conditions for qualifying for the probe filter retained low power state include the data processor entering a low power state in which only a primary non-coherent requestor is active; The method of claim 10.

14. detecting that the data processor has entered a low power state in which only a primary non-coherent requestor is active includes detecting a disconnect condition of a plurality of data ports of the data fabric coupled to a probe filter client.

14. The method of claim 13.

15. transmitting a wake signal from the coherent slave controller in response to receiving a coherent cacheable request from a primary non-coherent requestor while in the probe filter retaining low power state; in response to receiving the wake signal, causing the probe-filter to exit the probe-filter-holding low power state and execute the coherent cacheable request.

14. The method of claim 13.

16. 1. A power state control circuit for controlling a power state of a data fabric, comprising: power state control circuitry operable to cause a probe-filter of the data fabric to enter a retention low power state in which a clock signal to the probe-filter is gated while power to the probe-filter is maintained in response to conditions suitable for the probe-filter retention low power state; a first clock gating circuit coupled to the power state control circuitry for selectively gating a first clock signal to the probe filter; a second clock gating circuit coupled to the power state control circuitry for selectively gating a second clock signal to the data fabric; the coherent slave controller inhibits or disables new probe filter lookups and signaling to the power state control circuitry when all of the in-flight probe filter lookups are completed; Power state control circuit.

17. The conditions for qualifying for the probe-filter retained low power state include a condition in which all requesters that are probe-filter clients have flushed their caches and entered an idle power state.

20. The power state control circuit of claim 16.

18. the conditions for qualifying for the probe-filter retained low power state include the condition that the data fabric has entered an idle low power state; 20. The power state control circuit of claim 16.

19. a condition for qualifying for the probe filter retained low power state includes a host data processor entering a low power state in which only a primary non-coherent requester is connected to the data fabric; 20. The power state control circuit of claim 16.

Citation Information

Patent Citations

  • Data processing device and method for transferring workload between a source processing circuit and a destination processing circuit.

    JP2013521556A

  • Low power management for multi-node systems

    JP2021507412A

  • Power Management of Cache Duplicate Tags

    US20170010655A1

  • Multiprocessor control device, multiprocessor control method, and multiprocessor control circuit

    WO2008152790A1