Bandwidth-based power management for peripheral component interconnect express devices
By dynamically adjusting the link speed and width in the PCIe link, the problem of high power consumption of PCIe links during low bandwidth use cases is solved, lower voltage levels and power consumption is achieved, and system energy efficiency is improved.
Patent Information
- Application Number
- CN202080036000.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Priority Date
- 2020-05-13
- Filing Date
- 2020-05-14
- Publication Date
- 2024-11-05
- Estimated Expiration
- 2040-05-14
AI Technical Summary
Existing PCIe links still operate at high speeds during low bandwidth use cases, resulting in the voltage domain remaining at a higher voltage level, increasing power consumption and leakage, and existing power management methods cannot effectively reduce the voltage domain of PCIe devices.
By receiving bandwidth requests from clients in the controller, dynamically adjusting PCIe link speed and link width, thereby reducing the link operation speed and voltage level, realizing bandwidth-based power management, including reducing in low throughput scenarios GEN speed, such as from GEN3 to GEN2 or GEN1, reduces the voltage level to reduce power consumption.
Effectively reduces the power consumption of PCIe links during low bandwidth use cases, and reduces leakage current and power consumption by reducing voltage levels, improving the energy efficiency of the system.
Smart Images

Figure CN113826086B_ABST
Abstract
Description
[0001] CROSS-REFERENCE TO RELATED APPLICATIONS
[0002] This application claims priority to and the benefit of Non-Provisional Patent Application No. 15 / 931,409 filed with the United States Patent and Trademark Office on May 13, 2020; Provisional Patent Application No. 62 / 851,633 filed with the United States Patent and Trademark Office on May 22, 2019; and Provisional Patent Application No. 62 / 852,925 filed with the United States Patent and Trademark Office on May 24, 2019, the entire contents of which are incorporated herein by reference as if fully set forth herein and for all applicable purposes. Technical Field
[0003] Generally speaking, aspects of the present disclosure relate to Peripheral Component Interconnect Express (PCIe) devices, and more specifically, aspects of the present disclosure relate to managing power for PCIe devices. Background Art
[0004] The system may include one or more processors (e.g., application processors) and peripherals such as wireless modems, graphics processors, displays, sensors, etc. The one or more processors may communicate with the peripherals using high-speed communication links according to a standard (i.e., protocol). One popular standard is the Peripheral Component Interconnect Express (PCIe) standard, which supports high-speed links capable of sending data at speeds of several gigabits per second. Summary of the invention
[0005] A simplified overview of one or more implementations is given below in order to provide a basic understanding of such implementations. This summary is not an exhaustive overview of all contemplated implementations, and is neither intended to identify key or important elements of all implementations nor to delineate the scope of any or all implementations. Its sole purpose is to present some concepts of one or more implementations in a simplified form as a prelude to a more detailed description that will be presented later.
[0006] One aspect relates to a system. The system includes: an interface circuit configured to provide an interface with a link, and a controller. The controller is configured to: receive one or more bandwidth requests from one or more clients; and determine at least one of a link speed and a link width for the link based on the one or more bandwidth requests.
[0007] Another aspect relates to a method. The method includes: receiving, in a controller, one or more bandwidth requests from a client for communications on a link between a link partner and the client. Additionally, the method includes: determining, in the controller, at least one of a link speed and a link width for the link based on the one or more bandwidth requests; and implementing, using the controller, a speed change in the client based on at least one of the determined link speed and link width for the link. Additionally, the method includes: sending, using the controller, a speed change request to the link partner on the link, the request being based on at least one of the determined link speed and link width for the link.
[0008] To achieve the foregoing and related ends, one or more implementations include the features fully described below and particularly pointed out in the claims. The following description and the accompanying drawings set forth in detail certain illustrative aspects of one or more implementations. However, these aspects indicate only some of the various ways in which the principles of the various implementations may be employed, and the described implementations are intended to include all such aspects and their equivalents. BRIEF DESCRIPTION OF THE DRAWINGS
[0009] Figure 1 An example of a root complex coupled to an endpoint device in accordance with certain aspects of the present disclosure is shown.
[0010] Figure 2 An example of a system including a host system and an endpoint device system according to aspects of the present disclosure is shown.
[0011] Figure 3 An exemplary implementation of a link according to certain aspects of the present disclosure is shown.
[0012] Figure 4 is a call flow diagram illustrating an example of a bandwidth-based power management method according to certain aspects of the present disclosure.
[0013] Figure 5 is a call flow diagram illustrating another example of a bandwidth-based power management method according to certain aspects of the present disclosure.
[0014] Fig. 6A An example of a lookup table mapping bandwidth requirements to link speeds is shown in accordance with certain aspects of the present disclosure.
[0015] Figure 6B An example of a lookup table that maps bandwidth requirements to link widths is shown in accordance with certain aspects of the present disclosure.
[0016] Figure 6C An example of a lookup table that maps bandwidth requirements to link speed and link width is shown in accordance with certain aspects of the present disclosure.
[0017] Figure 7 is a call flow diagram illustrating yet another example of a bandwidth-based power management method according to certain aspects of the present disclosure.
[0018] Figure 8 is a call flow diagram illustrating yet another example of a bandwidth-based power management method according to certain aspects of the present disclosure.
[0019] Fig. 9 A flow chart of an additional exemplary method for bandwidth-based power management according to aspects of the present disclosure is shown. DETAILED DESCRIPTION
[0020] The detailed description set forth below in conjunction with the accompanying drawings is intended as a description of various configurations and is not intended to represent the only configurations in which the concepts described herein may be implemented. In order to provide a thorough understanding of each concept, the detailed description includes specific details. However, it will be apparent to those skilled in the art that these concepts may be implemented without these specific details. In some instances, well-known structures and components are shown in block diagram form in order to avoid blurring such concepts.
[0021] Various aspects of the present disclosure provide bandwidth-based PCIe power management using link speed and / or link width scaling. Various aspects of the present disclosure are discussed below using examples of PCIe GEN1 to GEN 4. However, it will be appreciated that the present disclosure is not limited to these examples, and that the present disclosure may be used to provide power management for future implementations of the PCIe standard (e.g., GEN5 and above). Additionally, it should be noted that while the present disclosure is discussed with respect to PCIe links, those skilled in the art will appreciate that the underlying principles of the disclosed systems and methods may be implemented in other types of PCI links, or even in other physical serial interconnects between host and client devices.
[0022] Figure 1A specific example of a system 110 supporting PCIe GEN3 is shown, the system 110 including a PCIe root complex 115 (e.g., on a host device) and a PCIe endpoint device 120, wherein the root complex 115 and the endpoint device 120 are coupled by a PCIe link 125 operating at a GEN3 speed (capable of up to a theoretical speed of 8.0 GT / s). In this example, the link 125 is operated at the GEN3 speed even when the link 125 is being underutilized by a low-bandwidth application that can be adequately served by a GEN2 speed or lower. Operating the link 125 at the GEN3 speed may require maintaining the power supply voltage of the PCIe core and physical layer (PHY) at a higher voltage level (also referred to as a voltage corner) than at GEN2 or lower speeds. Therefore, for low-bandwidth applications, maintaining the link 125 at the GEN3 speed will maintain a higher than required power supply voltage (e.g., as shown by the box 130, the input voltage 130 at the higher corner), which will result in more leakage and faster battery drain.
[0023] The method for managing power currently specified in the PCIe specification is active state power management (ASPM), which adopts a method for reducing power based on link activity detected on a PCIe link between a root complex and an endpoint PCIe device. In this method, when data is being transmitted on a PCIe link, the link operates in an L0 power state (i.e., a link operating state). When the link is idle (e.g., in a short time interval between data bursts), the link can be transferred from the L0 state to a lower power state (e.g., L0s→L1→L1.1 / L1.2) to reduce power consumption. In this example, L0s is a low-power standby for state L0, and the L1 substate is the lowest possible active low-power state for a PCIe link. Even though the method reduces power consumption at the link level, the method does not change the voltage domain of the PCIe device. Therefore, when the link is operating at a GEN3 speed or higher PCIe speed, the voltage domain will remain at a higher voltage level (corner) for GEN3 and higher speed PCIe generations.
[0024] In the current approach, a PCIe link running at its maximum speed runs at its maximum speed even during low throughput traffic scenarios. This not only consumes more power for the PCIe controller and PHY, but also requires other systems that interface with PCIe (such as but not limited to the memory subsystem and system bus interface) to run at a higher clock frequency.
[0025] To address this issue, aspects of the present disclosure provide a coordinated power management method between a root complex and an endpoint device controller that implements dynamic PCIe link speed and / or link width scaling based on the bandwidth requirements of one or more clients. In this way, during lower bandwidth use cases, the GEN speed at which the PCIe link operates can be reduced (e.g., from GEN3 to GEN2 / GEN1), thereby allowing the voltage domain to be reduced to a lower voltage level (corner) (e.g., the lowest operating level for low throughput applications that perform well at speeds below GEN2 or GEN1). The reduced voltage level (corner) for lower bandwidth use cases reduces power consumption.
[0026] By reducing the link speed (e.g., GEN speed) and / or link width during low-throughput data traffic scenarios, aspects of the present disclosure give the subsystem an opportunity to scale down voltage levels (e.g., reduce operating levels that meet current throughput requirements on the PCIe link). The scaled-down voltage level or levels reduce power consumption (e.g., reduces leakage current during sustained low-throughput traffic or in idle use cases).
[0027] It should be noted that the system 110 may be implemented within a battery-based consumer device such as a wireless device (e.g., a mobile phone, user equipment (UE), mobile station (MS)). Further, in the context of a UE, the PCIe root complex or host device may be an application processor or other processor within the UE, and the endpoint device may be an onboard integrated circuit or client such as a wireless modem, a system on a chip (SOC), or an 802.11 WiFi radio device, as examples.
[0028] Now refer to Figure 2 An exemplary PCIe system 205 in which aspects of the present disclosure may be implemented is discussed. This will be followed by a description of a bandwidth-based power management approach in accordance with aspects of the present disclosure.
[0029] System 205 includes host system 210 and PCIe endpoint device system 250. Host system 210 may be integrated on a first chip (eg, a system on a chip), and endpoint device system 250 may be integrated on a second chip. In this example, host system 210 and endpoint device system 250 are coupled via PCIe link 285.
[0030] The host system 210 includes one or more host clients 214. Each host client 214 can be implemented on a processor that executes software that performs the functions of the host client 214 discussed herein. For examples of more than one host client 214, the host clients 214 can be implemented on the same processor or on different processors. The host system 210 also includes a host controller 212, which can perform the root complex functions specified in the PCIe specification, as further discussed below. The host controller 212 can be implemented on a processor that executes software that performs the functions of the host controller 212 discussed herein.
[0031] The host system 210 includes a PCIe interface circuit 216, a system bus interface 215, and a system memory 240. The system bus interface 215 can interface one or more host clients 214 with the host controller 212, and interface each of the one or more host clients 214 and the host controller 212 with the PCIe interface circuit 216 and the system memory 240. The PCIe interface circuit 216 provides an interface to the PCIe link 285 to the host system 210. In this regard, the PCIe interface 216 is configured to send data (e.g., from the host client 214) to the endpoint device 250 on the PCIe link 285, and receive data from the endpoint device 250 via the PCIe link 285, as discussed further below. The PCIe interface circuit 216 includes a PCIe controller 218, a digital PHY interface for PCIe architecture (PIPE) interface 220, a physical (PHY) transmit (TX) block 222, a PHY receive (RX) block 226, and a clock generator 224. PIPE interface 220 provides a parallel interface between PCIe controller 218 and PHY TX block 222 and PHY RX block 226. PCIe controller 218 (which may be implemented in hardware) may be configured to perform transaction layer, data link layer, and control flow functions specified in the PCIe specification, as discussed further below.
[0032] The host system 210 also includes an oscillator (e.g., a crystal oscillator or "XO") 230 that is configured to generate a stable reference clock signal 232. In one example, the reference clock signal 232 may have a frequency of 19.2 MHz, but is not limited to this frequency. The reference clock signal 232 is input to the clock generator 224, which generates a plurality of clock signals based on the reference clock signal 232, as discussed further below. In this regard, the clock generator 224 may include a plurality of phase-locked loops (PLLs), each of which generates a corresponding clock signal in the plurality of clock signals by multiplying the frequency of the reference clock signal 232.
[0033] The endpoint device system 250 includes one or more device clients 254. Each device client 254 can be implemented on a processor that executes software that performs the functions of the device client 254 discussed herein. For examples of more than one device client 254, the device clients 254 can be implemented on the same processor or on different processors. The endpoint device system 250 also includes a device controller 252. As discussed further below, the device controller 252 can be configured to receive bandwidth requests from one or more device clients and determine whether to change the link speed (e.g., GEN speed) and / or link width based on the bandwidth requests. The device controller 252 can be implemented on a processor that executes software that performs the functions of the device controller discussed herein.
[0034] The endpoint device system 250 includes a PCIe interface circuit 260, a system bus interface 256, and a system memory 274. The system bus interface 256 can interface one or more device clients 254 with the device controller 252, and interface each of the one or more device clients 254 and the device controller 252 with the PCI interface circuit 260 and the system memory 274. The PCIe interface circuit 260 provides an interface to a PCIe link 285 to the endpoint device system 250. In this regard, the PCIe interface circuit 260 is configured to send data (e.g., from the device client 254) to the host system 210 (also referred to as a host device) on the PCIe link 285, and receive data from the host system 210 via the PCIe link 285, as discussed further below. The PCIe interface circuit 260 includes a PCIe controller 262, a PIPE interface 264, a physical (PHY) transmit (TX) block 266, a PHY receive (RX) block 270, and a clock generator 268. PIPE interface 264 provides a parallel interface between PCIe controller 262 and PHY TX block 266 and PHY RX block 270. PCIe controller 262 (which may be implemented in hardware) may be configured to perform transaction layer, data link layer, and control flow functions specified in the PCIe specification, as discussed further below.
[0035] The endpoint device system 250 also includes an oscillator (eg, a crystal oscillator) 272 configured to generate a stable reference clock signal 273 for a system memory 274. Figure 2In the example of FIG. 1 , the clock generator 224 at the host system 210 is configured to generate an endpoint (EP) reference clock signal 287, which is forwarded by the PHY RX block 226 to the endpoint device system 250 via a differential clock line 288. At the endpoint device system 250, the RX PHY block 270 receives the EP reference clock signal 287 and forwards the EP reference clock signal 278 to the clock generator 268. The EP reference clock signal 287 may have a frequency of 100 MHz, but is not limited to this frequency. The clock generator 268 is configured to generate a plurality of clock signals based on the EP reference clock signal 287, as discussed further below. In this regard, the clock generator 268 may include a plurality of PLLs, each of which generates a corresponding clock signal in the plurality of clock signals by multiplying the frequency of the EP reference clock signal 287.
[0036] The system 205 also includes a power management integrated circuit (PMIC) 290 coupled to a battery 292 and / or another power source. The PMIC 290 is configured to convert the voltage of the battery 292 to a plurality of power supply voltages (e.g., using a switching regulator, a linear regulator, or any combination thereof). In this example, the PMIC 290 generates a voltage 242 for the oscillator 230, a voltage 244 for the PCIe controller 218, and a voltage 246 for the PHY blocks 222 and 226 and the clock generator 224. The voltages 242, 244, and 246 may be programmable, where the PMIC 290 is configured to set the voltage levels (corners) of the voltages 242, 244, and 246 according to instructions (e.g., from the host controller 212).
[0037] PMIC 290 also generates voltage 280 for oscillator 272, voltage 278 for PCIe controller 262, voltage 276 for PHY blocks 266 and 270, and clock generator 268. Voltages 280, 278, and 276 may be programmable, where PMIC 290 is configured to set the voltage levels (corners) of voltages 280, 278, and 276 according to instructions (e.g., from device controller 252). PMIC 290 may be implemented on one or more chips. Although PMIC 290 may be implemented on a single chip, PMIC 290 may be implemented on a single chip. Figure 2 290 is shown as one PMIC, but it will be appreciated that PMIC 290 may be implemented by two or more PMICs. For example, PMIC 290 may include a first PMIC for generating voltages 242, 244, and 246 and a second PMIC for generating voltages 280, 278, and 276. In this example, both the first PMIC and the second PMIC may be coupled to battery 292.
[0038] In operation, the PCIe interface circuit 216 on the host system 210 can send data from one or more host clients 214 to the endpoint device system 250 via the PCIe link 285. The data from the one or more host clients 214 can be directed to the PCIe interface 216 according to the PCIe mapping set up by the host controller 212 during initial configuration. At the PCIe interface 216, the PCIe controller 218 can perform transaction layer and data link layer functions on the data, such as packetizing the data, generating error correction codes to be sent with the data, etc. The PCIe controller 218 outputs the processed data to the PHY TX block 222 via the PIPE interface 220. The processed data includes the data from the one or more host clients 214 as well as overhead data (e.g., packet headers, error correction codes, etc.). In one example, the clock generator 224 may generate a 250 MHz clock 234 for GEN3 based on the reference clock 232, and input the 250 MHz clock 234 to the PCIe controller 218 to time the operation of the PCIe controller 218. In this example, the PIPE interface 220 may include a 32-bit parallel bus that transmits 32 bits of data in parallel to the PHY TX block in each cycle of the 250 MHz clock 234 (this translates to a transfer rate of approximately 8 GT / s).
[0039] The PHY TX block 222 serializes the parallel data from the PCIe controller 218 and drives the link 285 with the serialized data. In this regard, the PHY TX block 222 may include one or more serializers and one or more drivers. The clock generator 224 may generate a high frequency clock for one or more serializers based on the reference clock signal 232.
[0040] At the endpoint device system 250, the PHY RX block 270 receives the serialized data via the link 285 and deserializes the received data into parallel data. In this regard, the PHY RX block 270 may include one or more receivers and one or more deserializers. The clock generator 268 may generate a high frequency clock for one or more deserializers based on the EP reference clock signal 287. The PHY RX block 270 transmits the deserialized data to the PCIe controller 262 via the PIPE interface 264. The PCIe controller 262 may recover the data from the one or more host clients 214 from the deserialized data and forward the recovered data to the one or more device clients 254.
[0041] On the endpoint device system 250, the PCIe interface circuit 260 can send data from one or more device clients 254 to the host system 250 via the link 285. In this regard, the PCIe controller 262 at the PCIe interface circuit 260 can perform transaction layer and data link layer functions on the data, such as packetizing the data, generating error correction codes to be sent with the data, etc. The PCIe controller 262 outputs the processed data to the PHY TX block 266 via the PIPE interface 264. The processed data includes data from the one or more device clients 254 and overhead data (e.g., packet headers, error correction codes, etc.). In one example, the clock generator 268 can generate a 250 MHz clock for GEN3 based on the EP reference clock 287, and input the 250 MHz clock to the PCIe controller 262 to time the operation of the PCIe controller 262.
[0042] The PHY TX block 266 serializes the parallel data from the PCIe controller 262 and drives the link 285 with the serialized data. In this regard, the PHY TX block 266 may include one or more serializers and one or more drivers. The clock generator 268 may generate a high frequency clock for one or more serializers based on the EP reference clock signal 287.
[0043] At the host system 210, the PHY RX block 226 receives the serialized data via the link 285 and deserializes the received data into parallel data. In this regard, the PHY RX block 226 may include one or more receivers and one or more deserializers. The clock generator 224 may generate a high frequency clock for the one or more deserializers based on the reference clock signal 232. The PHY RX block 226 transmits the deserialized data to the PCIe controller 218 via the PIPE interface 220. The PCIe controller 218 may recover the data from the one or more device clients 254 from the deserialized data and forward the recovered data to the one or more host clients 214.
[0044] Figure 3 It is shown that according to certain aspects of the present disclosure, Figure 22. An example of a PCIe link 285 used in a system of FIG. 2. In this example, the link 285 includes a plurality of channels 310-1 to 310-n, wherein each channel includes a corresponding first differential line 312-1 to 312-n for transmitting data from the host system 210 to the endpoint device system 250, and a corresponding second differential line 315-1 to 315-n for transmitting data from the endpoint device system 250 to the host system 210. Therefore, each channel 310-1 to 310-n is bidirectional. The differential lines 312-1 to 312-n and 315-1 to 315-n can be implemented using metal traces on a substrate (e.g., a printed circuit board), wherein the host system 210 can be integrated on a first chip mounted on the substrate, and the endpoint device is integrated on a second chip mounted on the substrate. The differential lines 312-1 to 312-n and 315-1 to 315-n can also be implemented using wires, cables, etc. In this example, PHY TX block 222 may include logic for splitting data between lanes when data is sent from host system 210 to endpoint device system 250 across multiple lanes. Similarly, PHY TX block 266 may include logic for splitting data between lanes when data is sent from endpoint device system 250 to host system 210 across multiple lanes.
[0045] based on Figure 3 In the example, Figure 2 The PHY TX block 222 shown in FIG. 1 may be implemented to include a driver 320 - 1 to 320 - n for each differential line 312 - 1 to 312 - n and Figure 2 The PHY RX block 270 shown in FIG. 1 may be implemented to include a receiver 340-1 to 340-n (e.g., an amplifier) for each differential line 312-1 to 312-n. Each driver 320-1 to 320-n is configured to drive the corresponding differential line 312-1 to 312-n with data, and each receiver 340-1 to 340-n is configured to receive data from the corresponding differential line 312-1 to 312-n. In addition, in FIG. Figure 3 In the embodiment of the present invention, the PHY TX block 266 may include a driver 345-1 to 345-n for each differential line 315-1 to 315-n, and the PHY RX block 226 may include a receiver 325-1 to 325-n (e.g., an amplifier) for each differential line 315-1 to 315-n. Each driver 345-1 to 345-n is configured to drive the corresponding differential line 315-1 to 315-n with data, and each receiver 325-1 to 325-n is configured to receive data from the corresponding differential line 315-1 to 315-n.
[0046] In some aspects, the width of the link 285 is scalable. In these aspects, the width of the link 285 is scaled by controlling the number of active lanes 310-1 to 310-n. The greater the number of active lanes, the wider the width of the link 285; and the smaller the number of active lanes, the smaller the width of the link 285. In one example, the host controller 212 can configure the width of the link 285 by configuring the number of lanes 310-1 to 310-n on which the PCIe interface circuit 216 and the PCI interface circuit 260 send and / or receive data via the link 285.
[0047] In one example, the host system 210 may include a power switch circuit 350 configured to individually control power to the drivers 320-1 to 320-n and the receivers 325-1 to 325-n from the PMIC 290. In this regard, the power switch circuit 350 may couple the drivers and receivers of the active channels to the voltage 246 and decouple the drivers and receivers of the inactive channels from the voltage 246. In this example, the drivers 320 and receivers 325 of the inactive channels are powered off to save power. Thus, in this example, the number of drivers and receivers powered on scales with the width of the link 285. The power switch circuit 350 may be configured to selectively power the drivers 320-1 to 320-n and the receivers 325-1 to 325-n based on instructions from the host controller 212, wherein the host controller 212 instructs the power switch circuit 350 which drivers and receivers to power on or off (e.g., based on the current link width). For ease of illustration, in Figure 3 Not shown are separate connections or couplings between the power switch circuit 350 and the drivers 320-1 through 320-n and receivers 325-1 through 325-n.
[0048] Similarly, if Figure 2 The illustrated endpoint device system 250 may include a power switch circuit 360 that is configured to individually control power to drivers 340-1 through 340-n and receivers 345-1 through 345-n from the PMIC 290. In this regard, the power switch circuit 360 may couple the drivers and receivers of active channels to the voltage 276 and decouple the drivers and receivers of inactive channels from the voltage 276. Thus, in this example, the drivers and receivers of the inactive channels are powered down to save power. The power switch circuit 360 may be configured to selectively power the drivers 340-1 through 340-n and receivers 345-1 through 345-n based on instructions from the device controller 252, wherein the device controller 252 instructs the power switch circuit 360 which drivers and receivers to power on or off (e.g., based on the current link width). For ease of illustration, in FIG. Figure 3Not shown are separate connections or couplings between the power switch circuit 360 and the drivers 340-1 through 340-n and receivers 345-1 through 345-n.
[0049] Link 285 may support multiple link speeds. For example, link 285 may support multiple link speeds corresponding to different generations ("GEN") of the PCIe standard. In this regard, Table 1 below lists exemplary transmission speeds per channel per direction for GEN1 speeds, GEN2 speeds, GEN3 speeds, GEN4 speeds, and GEN5 speeds.
[0050] Table 1
[0051]
[0052] The exemplary transmission rates in Table 1 may be theoretical transmission rates. The actual transmission rates for one or more of these link speeds may be slightly lower than the transmission rates shown in Table 1. The transmission rate may also be expressed in Gbps. In this example, the link speed may refer to the transmission rate per channel per direction.
[0053] In the above example, the host controller 212 and the device controller 252 can negotiate a link speed (e.g., a GEN speed) and configure the PCIe interface circuits 216 and 260 to operate at the negotiated link speed (e.g., according to the PCIe hardware programming guide). In this example, the host controller 212 can set one or more of the voltages 242, 244, and 246 based on the current link speed (current GEN speed). In one example, the host controller 212 can include a table that maps each supported link speed (e.g., each supported GEN speed) to one or more corresponding voltage levels (corners). In this example, the host controller 212 can instruct the PMIC 290 to set one or more of the voltages 242, 244, and 246 provided by the PMIC 290 according to the one or more voltage levels (corners) mapped to the current link speed (e.g., the current GEN speed). The voltage level (corner) for a lower link speed may be lower than the voltage level (corner) for a higher link speed (e.g., due to looser timing requirements for the lower link speed). The host controller 212 may instruct the PMIC 290 directly or through another processor that communicates directly with the PMIC 290. Thus, in this example, the voltage level (corner) of the host system 210 scales with the link speed (eg, GEN speed).
[0054] Similarly, the device controller 252 can set one or more of the voltages 280, 278, and 276 based on the current link speed (current GEN speed). In one example, the device controller 252 can include a table that maps each supported link speed (e.g., each supported GEN speed) to one or more corresponding voltage levels (angles). In this example, the device controller 252 can instruct the PMIC 290 to set one or more of the voltages 276, 278, and 280 provided by the PMIC 290 according to the one or more voltage levels (angles) mapped to the current link speed (e.g., current GEN speed). The device controller 252 can instruct the PMIC 290 directly or through another processor that communicates directly with the PMIC 290. Therefore, in this example, the voltage level (angle) of the endpoint device system 250 scales with the link speed (e.g., GEN speed).
[0055] Figure 4 A call flow diagram of an exemplary bandwidth-based power management method 410 using link speed scaling according to aspects of the present disclosure is shown. The method 410 may also include link width scaling according to additional aspects. At 412, the device controller 252 (or also the endpoint PCIe software process) receives bandwidth requests from one or more device clients 254 (or also the endpoint PCIe client software process). Each device client 254 can generate a corresponding bandwidth request based on the bandwidth requirements of the client. Each bandwidth request can have any format. For example, each bandwidth request can indicate the bandwidth requirements of the corresponding device client 254 in bits per second of Mbps or another scale. For examples in which the endpoint device system 250 includes multiple device clients 254, the device controller 252 can receive multiple bandwidth requests from the multiple device clients 254. In this example, the device controller 252 can aggregate the bandwidth requests (e.g., aggregate the bandwidth requirements indicated in the bandwidth requests).
[0056] The device controller 252 implements a bandwidth solver 413, which can be an algorithmic process executed within the controller 252 or a hardware component of the device controller 252 or a combination thereof to determine whether to change the current link speed of the link 285 based on the bandwidth request (or in the case of multiple device clients 254, based on the aggregate bandwidth request). The bandwidth solver 413 can implement a decision process 414 to determine whether a speed change is warranted for the link 285, and also implement a scaling of the link speed or link width. The process 414 can make a determination as to whether a speed change is warranted (scaling of a reduction or increase in speed) or no change is required. The decision result is shown by box 415, where, if a change is required, a message about implementing the speed change (e.g., 416 and 418) is sent to the controller 212 and the device client 254 (a portion of the box 415 above the dashed line shown in the box). Otherwise, box 415 shows that when a speed change is not warranted, a no speed change message 418 is returned to the device client 254 (i.e., a portion of the box 415 below the dashed line). In one example, if the bandwidth request (or aggregate bandwidth request) indicates a low bandwidth requirement that can be adequately served by a lower link speed or a lower link width, the bandwidth resolver 413 can determine to reduce the link speed from the current link speed to a lower link speed (e.g., change from a GEN3 speed to a GEN2 or GEN1 speed). In another example, if the bandwidth request (or aggregate bandwidth request) indicates a high bandwidth requirement, the bandwidth resolver 413 can determine to increase the link speed from the current link speed to a higher link speed (e.g., change from a GEN1 or GEN2 speed to a GEN3 speed). As described above, the bandwidth resolver 413 can be implemented in software executed by the device controller 252, or in some hardware logic that is part of the device controller 252 or in communication with the device controller 252, or a combination thereof. An exemplary implementation of the bandwidth resolver 413 is provided below.
[0057] If the bandwidth resolver 413 determines that a link speed change is not required, the device controller 252 may send an indicator indicating no link speed change to one or more device clients at 418. If the bandwidth resolver determines a link speed change, the device controller 252 may send an indicator or message indicating a link speed change to one or more device clients 254 at 416. The indicator to the device client 254 may also indicate a new link speed or PCIe generation.
[0058] If the bandwidth resolver 413 determines that the link speed is changed, the device controller 252 sends a speed change request to the host controller 212 at 420 to change the link speed. The request 420 may indicate the new link speed. For example, if the bandwidth resolver determines to change from a GEN1 or GEN2 speed to a GEN3 speed, the request to the host controller 212 may indicate the GEN3 speed. The device controller 252 may send the request to the host controller 212 via the PCIe interface circuits 260 and 216 and the link 285.
[0059] According to further aspects, it should be noted that in the option indicated at block 419, the bandwidth resolver may also initiate the preparation of resources (e.g., system or power resources) for speed changes (including amplification). It is noted that with respect to the process in block 419, if necessary, system resources or power resources (including, but not limited to, voltage regulators or clock sources) may be amplified for higher link speed change requests. This amplification of power or system resources may be an aggregation of all previous bandwidth change requests, including the current request sent or being sent to the host and pending confirmation of link speed change completion. It is further noted that this example only shows one implementation of preparing system or power resources for uncompleted link speed change requests, but is not limited thereto. After sending the speed change request to the host (as shown at 420), in another optional aspect, the host controller 212 may be configured to prepare resources for speed changes (including changes to accommodate higher link speed change requests), as shown at block 421 (e.g., preparing system or power resources, including, but not limited to, voltage regulators and / or clock sources).
[0060] Additionally, in response to the request to change the link speed, the host controller 212 may initiate a link speed change using a link speed change implementer at 422. The link speed change implementer 422 may handle the speed change according to the PCIe hardware programming guidelines in the PCIe specification, which may include performing a link retrain for the new link speed and reconfiguring the PCIe interface circuits 216 and 260. The link speed change implementer 422 may be implemented in software executed by the host controller 212, in hardware associated with or as part of the controller 212, or a combination thereof.
[0061] When the link speed change process is complete, the host controller 212 may send an indicator indicating the speed change is complete to the device controller 252 at 424. At 426, if necessary, the host controller 212 changes or updates the voltage level (angle) of one or more of the voltages 242, 244, and 246 based on the new link speed. For example, if the new link speed is lower (e.g., changing from a GEN3 speed to a GEN2 or GEN1 speed), the host controller 212 may reduce the voltage level (angle) of one or more of the voltages 242, 244, and 246. In another example, if the new link speed is higher (e.g., changing from a GEN2 or GEN1 speed to a GEN3 speed), the host controller 212 may increase the voltage level (angle) of one or more of the voltages 242, 244, and 246. As discussed above, the host controller 212 may change the voltage level (angle) of one or more of the voltages 242, 244, and 246 by instructing the PMIC 290 to set the voltage level (angle) of one or more of the voltages 242, 244, and 246 provided by the PMIC 290 based on the new link speed. If the voltage levels of the voltages 242, 244, and 246 for the new link speed are the same as the voltage levels for the previous link speed, the device controller 252 may override these voltage levels. The voltage scaling for the new link speed at 426 may be integrated with the link speed change process performed by the link speed change implementer 422 (i.e., the voltage scaling may be part of the link speed change process).
[0062] At 428 , in response to the speed change complete indication from the host controller 212 , the device controller 252 may send an indicator 428 to the one or more device clients 254 notifying the one or more device clients 254 of the link speed change.
[0063] At 430, if necessary, the device controller 252 updates the voltage level (angle) of one or more of the voltages 276, 278, and 280 based on the new link speed. For example, if the new link speed is lower (e.g., changing from a GEN3 speed to a GEN2 or GEN1 speed), the device controller 252 may reduce the voltage level (angle) of one or more of the voltages 276, 278, and 280. In another example, if the new link speed is higher (e.g., changing from a GEN2 or GEN1 speed to a GEN3 speed), the device controller 252 may increase the voltage level (angle) of one or more of the voltages 276, 278, and 280. As discussed above, the device controller 252 may change the voltage level (angle) of one or more of the voltages 276, 278, and 280 by instructing the PMIC 290 to set the voltage level (angle) of one or more of the voltages 276, 278, and 280 provided by the PMIC 290 based on the new link speed. If the voltage levels of voltages 276, 278, and 280 for the new link speed are the same as the voltage levels for the previous link speed, then device controller 252 may override these voltage levels.
[0064] Thus, for low bandwidth use cases, the exemplary power management method 410 reduces the link speed (e.g., GEN speed), which gives the host controller 212 and the device controller 252 an opportunity to scale down one or more voltage levels of the system 205 (e.g., to reduce operating levels that meet current throughput requirements on the link 285). The scaled down one or more voltage levels reduces power consumption (e.g., reduces leakage current during sustained low throughput traffic or in idle use cases).
[0065] exist Figure 4 In the example shown, the bandwidth solver is implemented on the endpoint device side. However, it will be appreciated that the present disclosure is not limited to this example. For example, the bandwidth solver may also be implemented on the host side according to certain aspects. In this regard, Figure 5 A call flow diagram illustrating an exemplary bandwidth-based power management method 510 in which a bandwidth resolver is implemented on the host side is shown in accordance with aspects of the present disclosure.
[0066] At 512, the host controller 212 receives bandwidth requests from one or more host clients 214. Each host client 214 can generate a corresponding bandwidth request based on the bandwidth requirements of the client. Each bandwidth request can have any format (e.g., indicating the bandwidth requirements of the corresponding client in terms of Mbps or bits per second of another scale). For examples in which the host system 210 includes multiple clients 214, the host controller 212 can receive multiple bandwidth requests from the multiple clients 214. In this example, the host controller 212 can aggregate the bandwidth requests (e.g., aggregate the bandwidth requirements indicated in the bandwidth requests).
[0067] At 514, the host controller 212 implements a bandwidth resolver 513 to determine whether to change the current link speed of the link 285 based on the bandwidth request (or in the case of multiple host clients, based on the aggregate bandwidth request). For example, if the bandwidth request (or aggregate bandwidth request) indicates a low bandwidth requirement that can be adequately served at a lower link speed, the bandwidth resolver can determine to reduce the link speed from the current link speed to a lower link speed (e.g., from a GEN3 speed to a GEN2 or GEN1 speed). In another example, if the bandwidth request (or aggregate bandwidth request) indicates a high bandwidth requirement, the bandwidth resolver can determine to increase the link speed from the current link speed to a higher link speed (e.g., from a GEN1 or GEN2 speed to a GEN3 speed). An exemplary implementation of the bandwidth resolver is provided below. The bandwidth resolver 513 can be implemented in software executed by the host controller 212, in hardware associated with the host controller 212, or in some combination thereof.
[0068] If the bandwidth solver 513 determines that a link speed change is not required, the host controller 212 may send an indicator indicating no link speed change to one or more host clients at 518. If the bandwidth solver determines that a link speed change is required, the host controller 212 may send an indicator indicating a link speed change to one or more host clients 214 at 516. The indicator may also indicate a new link speed. If the bandwidth solver 513 determines that a link speed change is required, the host controller 212 may send a link speed change request to the device controller 252 at 520, which notifies the device controller 252 of the proposed link speed change. The request may indicate the new link speed. Note that after the host controller 212 sends the message in step 520, the host controller 212 may be configured to wait for a response from the device controller 252. In response to the request, the device controller 252 may send an acknowledgment (ACK) to the host controller 212, which indicates that the device controller 252 is ready to make a link speed change, as shown at 521. If the endpoint device does not support the new link speed, the device controller 524 may send a negative acknowledgement (NACK) to the host controller 212, which indicates to the host controller 212 that the endpoint device does not support the proposed link speed change request from the host controller 212. If the host controller 212 receives a NACK from the device controller 252, the host controller 212 may abort the link speed change.
[0069] In the alternative, it should also be noted that, after receiving the message 520, the device controller 252 can be configured to initiate a scaling up of the system or power resources (including but not limited to a voltage regulator or a clock source (e.g., PMIC 290 or clock generator 268)) for the higher link speed change request, as shown at block 540. Additionally, after receiving the ACK message 521, the host controller 212 can also be configured to then “scale up” the system or power resources (e.g., PMIC 290 or clock generator 224) to accommodate the higher link speed change request, as shown at block 542.
[0070] If the bandwidth resolver 513 determines to change the link speed, the host controller 212 can initiate the link speed change using a link speed change implementer at 522. The link speed change implementer 522 can handle the speed change according to the PCIe hardware programming guidelines in the PCIe specification, which can include performing link retraining and reconfiguring the PCIe interface circuits 216 and 260 for the new link speed. The link speed change implementer can be implemented in software executed by the host controller 212.
[0071] When the link speed change process is complete, the host controller 212 may send an indicator to the one or more host clients 214 at 524 notifying the one or more host clients 214 of the link speed change. The host controller 212 may also send an indicator to the device controller 252 at 526 indicating that the speed change is complete.
[0072] At 528, if necessary, the host controller 212 updates the voltage level (angle) of one or more of the voltages 242, 244, and 246 based on the new link speed. For example, if the new link speed is lower (e.g., changing from a GEN3 speed to a GEN2 or GEN1 speed), the host controller 212 may reduce the voltage level (angle) of one or more of the voltages 242, 244, and 246. In another example, if the new link speed is higher (e.g., changing from a GEN2 or GEN1 speed to a GEN3 speed), the host controller 212 may increase the voltage level (angle) of one or more of the voltages 242, 244, and 246. If the voltage levels of the voltages 242, 244, and 246 for the new link speed are the same as the voltage levels for the previous link speed, the host controller 212 may ignore these voltage levels. The voltage scaling for the new link speed at 528 may be integrated with the link speed change process performed by the link speed change implementer 522 (i.e., the voltage scaling may be part of the link speed change process).
[0073] At 530, if necessary, the device controller 252 updates the voltage levels (angles) of one or more of the voltages 276, 278, and 280 based on the new link speed. For example, if the new link speed is lower (e.g., changing from a GEN3 speed to a GEN2 or GEN1 speed), the device controller 252 may reduce the voltage levels (angles) of one or more of the voltages 276, 278, and 280. In another example, if the new link speed is higher (e.g., changing from a GEN2 or GEN1 speed to a GEN3 speed), the device controller 252 may increase the voltage levels (angles) of one or more of the voltages 276, 278, and 280. If the voltage levels of the voltages 276, 278, and 280 for the new link speed are the same as the voltage levels for the previous link speed, the device controller 252 may disregard these voltage levels.
[0074] The bandwidth solver 513 can be implemented in any of a variety of ways to convert bandwidth requirements from one or more clients (e.g., one or more host clients, one or more device clients, etc.) and convert the bandwidth requirements into one of the following link parameters: PCIe link speed only; PCIe link width only; and PCI link speed and link width. For multiple clients, the bandwidth requirement can be an aggregation of the bandwidth requirements of multiple clients. The bandwidth solver can also take into account additional parameters such as burst frequency. The above discussion Figure 4 and 5 An example is shown in which a bandwidth solver converts bandwidth requirements for one or more clients into link speeds.
[0075] Bandwidth solver 513 can be implemented in any of a variety of ways to convert bandwidth requirements into link speed and / or link width. An exemplary implementation of the bandwidth solver is discussed below. However, it will be appreciated that the bandwidth solver is not limited to these examples and can be extended to other implementations based on the system power budget.
[0076] In certain aspects, the bandwidth solver 513 may convert the bandwidth requirements to PCIe link parameters by looking up a table and using the table to determine a link speed and / or link width based on the bandwidth requirements. Fig. 6A An example of a lookup table 610 for converting bandwidth requirements (aggregate bandwidth requirements for multiple clients) to link speeds (e.g., GEN speeds) is shown. In this example, table 610 includes different bandwidths (labeled "Bandwidth 1" through "Bandwidth m") and corresponding link speeds (labeled "Link Speed 1" through "Link Speed m") for each of the bandwidths. The bandwidths may be in Mbps or another format. Thus, table 610 maps each bandwidth to a corresponding link speed. It will be appreciated that two or more bandwidths may be mapped to the same link speed. It will also be appreciated that each bandwidth entry in the table may be a range of bandwidths that map to the same link speed.
[0077] Table 610 may be pre-stored in a memory coupled to the bandwidth solver. When the bandwidth solver receives a bandwidth request from one or more clients, the bandwidth solver may convert the bandwidth request (or the aggregate bandwidth request in the case of multiple clients) to a link speed by looking up the link speed mapped to the bandwidth request in table 610. For example, if the bandwidth request (or the aggregate bandwidth request in the case of multiple clients) corresponds to bandwidth 1 in table 610, the bandwidth solver may convert the bandwidth request to link speed 1. If the corresponding link speed in table 610 is different from the current link speed, the device controller 252 or the host controller 212 may initiate a link speed change to the corresponding link speed from table 610, as discussed above.
[0078] Table 610 may be generated based on computer simulations of system 205 for various bandwidth and link speed scenarios and / or power measurements of system 205. In this example, the link speed that results in the lowest power for a particular bandwidth based on the simulation results and / or the power measurements may be mapped to the bandwidth in table 610. Thus, table 610 may be populated based on the simulation results and / or the power measurements and then stored in a memory accessible to the bandwidth solver.
[0079] Figure 6B An example of a lookup table 620 for converting bandwidth requirements (or aggregate bandwidth requirements in the case of multiple clients) to link widths is shown. In this example, table 620 includes different bandwidths (labeled "Bandwidth 1" through "Bandwidth m") and corresponding link widths for each of the bandwidths (labeled "Link Width 1" through "Link Width m"). Thus, table 620 maps each bandwidth to a corresponding link width. It will be appreciated that two or more bandwidths may be mapped to the same link width. It will also be appreciated that each bandwidth entry in the table may be a range of bandwidths that map to the same link width. In one example, the link width may be specified by the number of active lanes in link 285 that correspond to the link width. As described above with reference to Figure 3 As discussed, the greater the number of active lanes in link 285, the wider the width of link 285.
[0080] In one example, a table 620 may be generated for each supported link speed and pre-stored in a memory coupled to the bandwidth solver. Thus, in this example, each link speed may have a corresponding table 620. In this example, the bandwidth solver may use the table 620 corresponding to the current link speed.
[0081] When the bandwidth solver receives a bandwidth request from one or more clients, the bandwidth solver may convert the bandwidth request (or the aggregate bandwidth request in the case of multiple clients) to a link width by looking up the link width mapped to the bandwidth request in table 620. For example, if the bandwidth request (or the aggregate bandwidth request in the case of multiple clients) corresponds to bandwidth 1 in table 620, the bandwidth solver may convert the bandwidth request to link width 1. If the corresponding link width in table 620 is different than the current link width, the device controller 252 or the host 212 controller may initiate a link width change to the corresponding link width from table 620, as discussed further below.
[0082] Table 620 may be generated based on computer simulations of system 205 for various bandwidth and link width scenarios and / or power measurements of system 205. In this example, the link width that results in the lowest power for a particular bandwidth based on the simulation results and / or the power measurements may be mapped to a bandwidth in table 620. Thus, table 620 may be populated based on the simulation results and / or the power measurements and then stored in a memory accessible to the bandwidth solver.
[0083] For examples where the bandwidth resolver is implemented on the endpoint device side, if the bandwidth resolver determines that the link width has changed, the device controller 252 may send a link width change request to the host controller 212. In response, the host controller 212 may process the width change according to the PCIe hardware programming guidelines in the PCIe specification, which may include performing a link retrain for the new link width and reconfiguring the PCIe interface circuits 216 and 260. The host controller 212 may notify the device controller 252 when the link width change is complete.
[0084] In this example, if the link width is reduced, the host controller 212 may power down the drivers in the PHY TX block 222 and / or the receivers in the PHY RX block 226 (which correspond to the channels in the link 285 that are deactivated due to the change in link width). As discussed above, the host controller 212 may power down the selected drivers and / or receivers by sending an instruction to the power switch circuit 350 to turn off the selected drivers and / or receivers. In other words, the host controller 212 sets the number of drivers and / or receivers powered by the power switch circuit 350 based on the new link width.
[0085] Similarly, the device controller 252 can power off the drivers in the PHY TX block 266 and / or the receivers in the PHY RX block 270 that correspond to the channels in the link 285 that are deactivated due to the change in link width. As discussed above, the host controller 212 can power off the selected drivers and / or receivers by sending instructions to the power switch circuit 360 to turn off the selected drivers and / or receivers. In other words, the device controller 252 sets the number of drivers and / or receivers powered by the power switch circuit 360 based on the new link width. Therefore, the components associated with the channels that are deactivated due to the change in link width can be powered off to save power.
[0086] For examples where the bandwidth resolver is implemented on the host side, if the bandwidth resolver determines that the link width has changed, the host controller 212 can handle the width change according to the PCIe hardware programming guidelines in the PCIe specification, which may include performing a link retrain for the new link width and reconfiguring the PCIe interface circuits 216 and 260. The host controller 212 can also notify the device controller 252 of the link width change.
[0087] In this example, if the link width is reduced, the host controller 212 may power down the drivers in the PHY TX block 222 and / or the receivers in the PHY RX block 226 (which correspond to the channels in the link 285 that are deactivated due to the link width change). Similarly, the device controller 252 may power down the drivers in the PHY TX block 266 and / or the receivers in the PHY RX block 270 (which correspond to the channels in the link 285 that are deactivated due to the link width change).
[0088] Figure 6C An example of a lookup table 630 for converting bandwidth requirements to both link speeds (e.g., GEN speeds) and link widths is shown. In this example, table 630 includes different bandwidths (labeled "Bandwidth 1" through "Bandwidth m") and corresponding link speeds (labeled "Link Speed 1" through "Link Speed m") and link widths (labeled "Link Width 1" through "Link Width m") for each of the bandwidths. The bandwidths may be in Mbps or another format. Thus, table 630 maps each bandwidth to a corresponding link speed and link width. It will be appreciated that two or more bandwidths may be mapped to the same link speed and / or the same link width.
[0089] Table 630 may be pre-stored in a memory coupled to the bandwidth solver. When the bandwidth solver receives a bandwidth requirement from one or more clients, the bandwidth solver may convert the bandwidth requirement (for the case of multiple clients, the aggregate bandwidth requirement) into a link speed and a link width by looking up the link speed and link width mapped to the bandwidth requirement in table 630. If the corresponding link speed in table 630 is different from the current link speed, the device controller 252 or the host controller 212 may initiate a link speed change to the corresponding link speed from table 630, as discussed above. If the corresponding link width in table 630 is different from the current width speed, the device controller 252 or the host controller 212 may initiate a link width change to the corresponding link width from table 630, as discussed above. Therefore, the link speed may be changed, the link width may be changed, or both the link speed and the link width may be changed.
[0090] Table 630 may be generated based on computer simulations of system 205 for various bandwidth, link speed, and link width scenarios and / or power measurements of system 205. In this example, the link speed and link width that result in the lowest power for a particular bandwidth based on the simulation results and / or power measurements may be mapped to the bandwidth in table 630. Thus, table 630 may be populated based on the simulation results and / or power measurements and then stored in a memory accessible to the bandwidth solver.
[0091] In certain aspects, a power budget can be prepared by experimenting or simulating for various link configurations (e.g., various link speeds and link widths) by varying link parameters, such as, but not limited to, the following: L0 power consumption; L0s power consumption; L1 power consumption; L1ss power consumption; L0s entry time; L1 entry time; L1ss entry time; L0s exit latency; L1 exit latency; and L1S exit latency. In these aspects, the bandwidth solver can then select the link speed and link width with the lowest power consumption.
[0092] Figure 7 Another call flow diagram of an exemplary bandwidth-based power management method 700 using link speed scaling according to aspects of the present disclosure is shown. Specifically, the method 710 relates to a scenario in which the device client 254 initiates speed scaling and the speed change is also completed, processed and / or implemented by the device-side controller 252 in collaboration with the host controller 212. The method 700 may also include link width scaling according to additional aspects.
[0093] At 702, one or more device clients 254 (or also endpoint PCIe client software processes) send bandwidth requests to a device controller 254 (or also endpoint PCIe software processes). Each device client 254 may generate a corresponding bandwidth request based on the bandwidth requirements of the client, and may have any format. For example, each bandwidth request may indicate the bandwidth requirements of the corresponding device client 254 in terms of Mbps or bits per second of another scale. As another example where the endpoint device system 250 includes multiple device clients 254, the device controller 252 may receive multiple bandwidth requests from the multiple device clients 254. In this example, the device controller 252 may aggregate the bandwidth requests (e.g., aggregate the bandwidth requirements indicated in the bandwidth requests).
[0094] The device controller 252 implements a bandwidth resolver 704, which may be executed within the device controller 252 or a hardware component of the device controller 252, or a combination thereof, to determine whether to change a link (e.g., such as Figure 2 The algorithm process of calculating the current link speed of the link 285 (shown in Figure 7A). The bandwidth solver 704 can implement a decision process 706 to determine whether a speed change is warranted for the link 285, and also implement a scaling of the link speed or link width. The process 706 can make a determination whether a speed change (scaling of a reduction or increase in speed) is warranted or no change is required. The decision result is shown by box 708, where, if a change is required, a message about implementing the speed change (e.g., optional message 710 and message 712) can be sent to the host controller 212 and the device client 254 (see, for example, the portion above the dashed line shown in the box in box 708). Otherwise, box 708 shows that when a speed change is not warranted, a no speed change message 714 (i.e., the portion below the dashed line in box 708) is returned to the device client 254.
[0095] In one example, if the bandwidth request (or aggregate bandwidth request) indicates a low bandwidth requirement that can be adequately served by a lower link speed or a lower link width, the bandwidth resolver 704 can determine to reduce the link speed from the current link speed to a lower link speed (e.g., change from a GEN3 speed to a GEN2 or GEN1 speed). In another example, if the bandwidth request (or aggregate bandwidth request) indicates a high bandwidth requirement, the bandwidth resolver 704 can determine to increase the link speed from the current link speed to a higher link speed (e.g., change from a GEN1 or GEN2 speed to a GEN3 speed). As mentioned above, the bandwidth resolver 704 can be implemented in software executed by the device controller 252, or in some hardware logic that is part of the device controller 252 or in communication with the device controller 252, or a combination thereof.
[0096] If the bandwidth solver 704 determines that the link speed has changed, the device controller 252 may send an indicator or message indicating the link speed change to one or more device clients 254 at 712 by notifying the one or more device clients 254 that the speed change is in progress. The message 712 to the device client 254 may also indicate the new link speed or PCIe generation. In addition, if the bandwidth solver 704 determines that the link speed has changed, the device controller 252 may optionally send a speed change request to the host controller 212 to request a change in the link speed, as shown by message 710. The request 710 may indicate the new link speed. For example, if the bandwidth solver determines to change from a GEN1 or GEN2 speed to a GEN3 speed, the request to the host controller 212 may indicate a GEN3 speed. The device controller 252 may send the request to the host controller 212 via the PCIe interface circuits 260 and 216 and the link 285. Then, the host controller 212 can prepare the voltage (e.g., voltage angle) for the speed change, as indicated at block 716, and then send a speed change ready message 718 back to the device controller 252. In addition, a process for preparing the system or power resources for the speed change can be implemented in the device controller 252, as shown by block 719.
[0097] According to further optional aspects, it should be noted that in one option, the bandwidth solver 704 can also initiate amplification. That is, if necessary, system resources or power resources (including but not limited to voltage regulators or clock sources) can be amplified for higher link speed change requests. This amplification of power or system resources can be an aggregation of all previous bandwidth change requests, including the current request sent or being sent to the host and the pending confirmation of link speed change completion.
[0098] After the bandwidth resolver 704 determines at block 708 that a speed change is warranted, a device link speed change implementer 720 may be executed in the device controller 252 to implement the speed change for one or more device clients 254 in the device controller 252. Additionally, in response to a request to change the link speed, in some aspects the device controller 252 may be initiated by the device speed change implementer 720. The device link speed change implementer 720 may process the speed change in accordance with the PCIe hardware programming guidelines in the PCIe specification, as shown in block 722, which may include performing a link retrain and reconfiguring the PCIe interface circuits 216 and 260 for the new link speed. The device link speed change implementer 720 may be implemented in software executed by the device controller 252, in hardware associated with or as part of the controller 252, or a combination thereof.
[0099] When the device speed change process is complete, the device controller 252 may send an indicator callback message indicating the speed change is complete to the client device at 724. At 726, if necessary, the device controller 252 changes or updates the voltage level (angle) of one or more of the voltages 276, 278, or 280 based on the new device link speed. For example, if the new device link speed is lower (e.g., changing from a GEN3 speed to a GEN2 or GEN1 speed), the device controller 252 may reduce the voltage level (angle) of one or more of the voltages 276, 278, or 280. In another example, if the new device link speed is higher (e.g., changing from a GEN2 or GEN1 speed to a GEN3 speed), the device controller 252 may increase the voltage level (angle) of one or more of the voltages 276, 278, or 280. As discussed above, the device controller 252 may change the voltage level (angle) of one or more of the voltages 276, 278, or 280 by instructing the PMIC 290 to set the voltage level (angle) of one or more of the voltages 276, 278, or 280 provided by the PMIC 290 based on the new device link speed. If the voltage levels of the voltages 276, 278, or 280 for the new device link speed are the same as the voltage levels for the previous device link speed, the device controller 252 may disregard these voltage levels. The voltage scaling for the new link speed at 726 may be integrated with the device link speed change process performed by the device link speed change implementer 720 (i.e., the voltage scaling may be part of the device link speed change process).
[0100] At 728, in response to the speed change completion indication from the host controller 212, the device controller 252 may send a message 728 to the host controller 212 notifying the host controller 212 of the device speed change. At 730, if necessary, the host controller 212 updates the voltage level (angle) of one or more of the voltages 242, 244, or 246 based on the new link speed. For example, if the new link speed is lower (e.g., changing from a GEN3 speed to a GEN2 or GEN1 speed), the host controller 212 may reduce the voltage level (angle) of one or more of the voltages 242, 244, or 246. In another example, if the new link speed is higher (e.g., changing from a GEN2 or GEN1 speed to a GEN3 speed), the host controller 212 may increase the voltage level (angle) of one or more of the voltages 242, 244, or 246. As discussed above, the host controller 212 may change the voltage level (angle) of one or more of the voltages 242, 244, or 246 by instructing the PMIC 290 to set the voltage level (angle) of one or more of the voltages 242, 244, or 246 provided by the PMIC 290 based on the new device link speed. If the voltage levels of the voltages 242, 244, or 246 for the new device link speed are the same as the voltage levels for the previous link speed, the host controller 212 may disregard these voltage levels.
[0101] According to another example, Figure 8 A call flow diagram of method 800 is shown, in which a host client may initiate a bandwidth change request, and a device controller completes, implements, or performs a link speed change. At 802, a host controller 212 receives bandwidth requests from one or more host clients 214. Each host client 214 may generate a corresponding bandwidth request based on the bandwidth requirements of the host client. Each bandwidth request may have any format (e.g., indicating the bandwidth requirements of the corresponding client in bits per second of Mbps or another scale). For examples in which a host system 210 includes multiple clients 214, the host controller 212 may receive multiple bandwidth requests from the multiple clients 214. In this example, the host controller 212 may aggregate bandwidth requests (e.g., aggregate bandwidth requirements indicated in the bandwidth requests).
[0102] The host controller 212 implements a bandwidth resolver 804 to determine whether to change the current link speed of the link 285 based on the bandwidth request (or in the case of multiple host clients, based on the aggregate bandwidth request), as shown at 806. For example, if the bandwidth request (or aggregate bandwidth request) indicates a low bandwidth requirement that can be adequately served at a lower link speed, the bandwidth resolver 804 can determine to reduce the link speed from the current link speed to a lower link speed (e.g., change from a GEN3 speed to a GEN2 or GEN1 speed). In another example, if the bandwidth request (or aggregate bandwidth request) indicates a high bandwidth requirement, the bandwidth resolver 804 can determine to increase the link speed from the current link speed to a higher link speed (e.g., change from a GEN1 or GEN2 speed to a GEN3 speed). An exemplary implementation of the bandwidth resolver 804 is provided below. The bandwidth resolver 804 can be implemented in software executed by the host controller 212, in hardware associated with the host controller 212, or in some combination thereof.
[0103] If the bandwidth solver 804 determines that a link speed change is not required, the host controller 212 may send an indicator indicating no link speed change to one or more host clients at 518. If the bandwidth solver determines that a link speed change is required, the host controller 212 may send an indicator indicating a link speed change to one or more host clients 214 at 810. The indicator 810 may also indicate a new link speed. Alternatively, if the bandwidth solver 804 determines that a speed change is not required, a message 812 may be sent to the host clients 214 to indicate no speed change.
[0104] If the bandwidth solver 804 determines that the link speed has changed, the host controller 212 may optionally prepare system resources for the speed change (e.g., prepare voltage corners or clocks), as shown at block 814. Additionally, at 816, the host controller 212 may send a link speed change request to the device controller 252 requesting the proposed link speed change. The request 816 may indicate a new link speed. In response, the device controller 252 may optionally prepare system resources for the speed change (e.g., voltage corners or clocks), as shown at block 818. Additionally, the device controller may be configured to implement a device link speed change implementer 820. In one aspect, the device link speed change implementer 820 may handle the speed change according to the PCIe hardware programming guidelines, as shown at 822, which may include performing link retraining for the new link speed and reconfiguring the PCIe interface circuits 216 and 260. The link speed change implementer 820 may be implemented in software executed by the device controller 252, in hardware associated with or coupled to the device controller 252, or some combination thereof.
[0105] When the link speed change process is complete, the device controller 252 may send a speed change complete message 824 to the host controller 212 to signal the completion of the link speed change. The host controller 212 may then signal the host client 214 to call the host client delay callback function for the bandwidth / speed change request 802 to signal that the change has been completed, as shown at 826.
[0106] At 828, if necessary, the host controller 212 updates the voltage level (angle) of one or more of the voltages 242, 244, and 246 based on the new link speed. For example, if the new link speed is lower (e.g., changing from a GEN3 speed to a GEN2 or GEN1 speed), the host controller 212 may reduce the voltage level (angle) of one or more of the voltages 242, 244, and 246. In another example, if the new link speed is higher (e.g., changing from a GEN2 or GEN1 speed to a GEN3 speed), the host controller 212 may increase the voltage level (angle) of one or more of the voltages 242, 244, and 246. If the voltage levels of the voltages 242, 244, and 246 for the new link speed are the same as the voltage levels for the previous link speed, the host controller 212 may ignore these voltage levels. The voltage scaling for the new link speed at 828 may be integrated with the link speed change process performed by the device link speed change implementer 820 (i.e., the voltage scaling may be part of the link speed change process).
[0107] At 830, if necessary, the device controller 252 updates the voltage levels (angles) of one or more of the voltages 276, 278, and 280 based on the new link speed. For example, if the new link speed is lower (e.g., changing from a GEN3 speed to a GEN2 or GEN1 speed), the device controller 252 may reduce the voltage levels (angles) of one or more of the voltages 276, 278, and 280. In another example, if the new link speed is higher (e.g., changing from a GEN2 or GEN1 speed to a GEN3 speed), the device controller 252 may increase the voltage levels (angles) of one or more of the voltages 276, 278, and 280. If the voltage levels of the voltages 276, 278, and 280 for the new link speed are the same as the voltage levels for the previous link speed, the device controller 252 may disregard these voltage levels.
[0108] The bandwidth solver 704 or 804 can be implemented in any of a variety of ways to convert bandwidth requirements from one or more clients (e.g., one or more host clients, one or more device clients, etc.) and convert the bandwidth requirements into one of the following link parameters: PCIe link speed only; PCIe link width only; and PCI link speed and link width. For multiple clients, the bandwidth requirement can be an aggregation of the bandwidth requirements of multiple clients. The bandwidth solver 704 or 804 can also take into account additional parameters such as burst frequency. The above discussion Figure 7 and 8 An example is shown in which a bandwidth solver converts bandwidth requirements for one or more clients into link speeds.
[0109] Fig. 9 A flow chart of a method 900 for bandwidth-based power management for a link, such as a PCIe link, in accordance with aspects of the present disclosure is shown. In certain aspects, the method 900 implements scaling of one or more of link speed (or bandwidth) and / or link width, such as reducing the number of lanes powered (e.g., selectively powering on or off drivers 320 or 345).
[0110] If you can Fig. 9 As can be seen from the figure, method 900 includes: from a client (eg, Figure 4 and 7 The "device client" in the example 254 or Figure 5 and 8 214) receives information about the link partner (i.e., the client's link partner, such as Figure 4 Device controller 252 in or Figure 5 ) and a client (e.g., Figure 4 254 or Figure 5 214) between the link between the one or more bandwidth requests, as shown in block 902. The process of block 902 can be implemented by the device controller 252 in one example, or in executing Figure 5 The process in block 902 is implemented by the host controller 212. In addition, the process of block 902 may be implemented by software running in the device controller 252 or by hardware within the device controller 252 or coupled to the device controller 252.
[0111] The method 900 also includes the process of determining at least one of a link speed and a link width for the link based on the one or more bandwidth requests, as shown in block 904. These determinations may be implemented by the device controller 252 or software running therein, or alternatively implemented in hardware that is in communication with or part of the controller 252. In further aspects, the process of block 904 may be implemented in Figure 4 In another aspect, the process of block 904 may be implemented by the bandwidth solver 413 shown in FIG. Figure 5 , or software running thereon, or alternatively implemented in hardware that communicates with or is part of the controller 212. In another aspect, the process of block 904 may be implemented in Figure 5 This is achieved by the bandwidth solver 513 shown in FIG.
[0112] In addition, the method 900 includes: based on the determined at least one of the link speed and the link width for the link, Figure 5 The speed change is implemented in the host client in the example of , as shown in block 906. These processes can be performed by the device controller 252 (or Figure 5 212 in the example of FIG. 10 ) or software running therein, or alternatively by some dedicated hardware that communicates with or is part of the controller 252. In other aspects, the process of block 904 can be implemented by, for example, Figure 4 and 5 The bandwidth solver 413 or 513 shown is implemented in the controller 252 or 212. In addition, the process of block 906 may include coordination and / or communication with the host link speed change implementer 422 or 522.
[0113] Finally, the method 900 includes sending a speed change request to the host (or on the link) based on at least one of the determined link speed and link width for the link, as shown in block 908. According to certain aspects, the process can be implemented by the controller 252 or 212, and by the controller 252 or 212. Figure 4 and 5 The bandwidth solver 413 or 513 shown in the figure is implemented in the controller 252 or 212. In some aspects, Figure 5 In the case of, the sending in block 908 may be performed on link 285 or through an interface such as interface 215. In addition, it should be noted that the various processes in any of blocks 902, 904, 906, and 908 may include using, for example, Figure 2 PCIe interface 216 or 260 is shown.
[0114] In further aspects of method 900, it is noted that the client may include a peripheral component interconnect express (PCIe) endpoint device controller. In addition, the PCIe endpoint device controller may include a PCIe interface circuit configured to send a speed change request to the host via the link. In further aspects, the PCIe controller may implement the speed change in the client based on at least one of a link speed and a link width determined for the link.
[0115] As previously discussed, the link in method 900 can be a peripheral component interconnect express (PCIe) link. Here, it should be noted that although PCIe has been described herein, method 900 is applicable to other links. In addition, the link speed includes a PCIe link speed in a plurality of different PCIe link speeds corresponding to different PCIe generations.
[0116] The link in method 900 may also include multiple channels, and the link width corresponds to the number of activities in the multiple channels, as described above. Figure 3 As discussed. The method 900 also includes determining a link speed for the link using a lookup table that maps each of the plurality of bandwidths to a corresponding link speed in the plurality of link speeds. In further aspects, the method 900 may include determining a link width using a lookup table that maps each of the plurality of bandwidths to a corresponding link width in the plurality of link widths. In other aspects, the method 900 may utilize other algorithms based on the speed and width of the transmission to determine the optimal link parameter (e.g., link speed).
[0117] In other aspects, the method 900 includes determining a link speed for the link by determining a power consumption for each of a plurality of different link speeds that satisfies one or more bandwidth requests, and determining a link speed of the plurality of different link speeds that has the lowest power consumption. Additionally, the method 900 may include determining a link width for the link by determining a power consumption for each of a plurality of different link widths that satisfies one or more bandwidth requests, and determining a link width of the plurality of different link widths that has the lowest power consumption.
[0118] In addition, the method 900 may include: providing one of a plurality of voltages or clocks to the interface circuit, and the controller is configured to set one or more voltages or clocks for the interface circuit based on the link speed. In yet another aspect, the method 900 may include: the link includes a plurality of channels, each channel is driven by a corresponding driver of a plurality of drivers and is coupled to a power switch circuit configured to selectively power the plurality of drivers; and setting the number of the plurality of drivers to be selectively powered by the power switch circuit based on the determined link width for the link to change the link width.
[0119] It will be appreciated that the present disclosure is not limited to the exemplary terms used above to describe aspects of the present disclosure.For example, bandwidth may also be referred to as throughput, data rate, or another term.
[0120] Although aspects of the present disclosure are discussed above using the example of the PCIe standard, it will be appreciated that the present disclosure is not limited to this example and may be used with other standards.
[0121] The host client 214, host controller 212, device controller 252, and device client 254 discussed above may each be implemented using a processor configured to perform the functions described herein by executing software including code for performing the functions described herein. The software may be stored on a computer-readable storage medium, such as RAM, ROM, EEPROM, optical disk, and / or magnetic disk.
[0122] Any reference to an element using nomenclature such as "first," "second," etc. herein generally does not limit the number or order of those elements. Specifically, these nomenclatures are used herein as a convenient method of distinguishing between two or more elements or instances of an element. Thus, reference to a first element and a second element does not mean that only two elements can be used or that the first element must be before the second element.
[0123] Within this disclosure, the word "exemplary" is used to mean "serving as an example, instance, or illustration." Any implementation or aspect described herein as "exemplary" is not necessarily to be construed as preferred or advantageous over other aspects of the disclosure. Similarly, the term "aspect" does not require that all aspects of the disclosure include the discussed feature, advantage, or mode of operation. The term "coupled" is used herein to refer to a direct or indirect electrical or other communicative coupling between two structures. Additionally, the term "approximately" means within 10% of the stated value.
[0124] The previous description of the present disclosure is provided to enable any person skilled in the art to implement or use the present disclosure. It will be apparent to those skilled in the art that various modifications to the present disclosure will be apparent, and the general principles defined herein may be applied to other variations without departing from the spirit or scope of the present disclosure. Therefore, the present disclosure is not limited to the examples described herein, but is given the widest scope consistent with the principles and novel features disclosed herein.
Claims
1. A system comprising: A first chip comprising: Host device; Host controller; a first peripheral component interconnect express (PCIe) interface circuit; and a host system bus interface configured to couple the host device, the host controller, and the first PCIe interface circuit; A second chip comprising: Multiple clients; an endpoint device controller configured to control the plurality of clients; A second PCIe interface circuit; and a client system bus interface configured to couple the plurality of clients, the endpoint device controller, and the second PCIe interface circuit; and link, which couples the first PCIe interface circuit with the second PCIe interface circuit, wherein the second PCIe interface circuit comprises a serializer configured to serialize parallel transmit data received from the plurality of clients via the client system bus interface, the parallel transmit data being generated for transmission on the link, And wherein the endpoint device controller is configured to: receiving a plurality of bandwidth requests from the plurality of clients via the client system bus interface; aggregating the multiple bandwidth requests into an aggregate bandwidth request; Determining at least one of a link speed and a link width for the link based on the aggregate bandwidth request includes: determining whether a change in the link speed or the link width is needed based on the aggregate bandwidth request; and After determining the at least one of the link speed and the link width, a speed change request is sent to the host device via the second PCIe interface circuit, the link, and the first PCIe interface circuit.
2. The system according to claim 1, wherein: The endpoint device controller is a PCIe endpoint device controller.
3. The system according to claim 2, wherein: The host device comprises a PCIe host device.
4. The system according to claim 1, wherein: The endpoint device controller is configured to configure the second PCIe interface circuit to interface with the link according to the at least one of the link speed and the link width.
5. The system according to claim 1, wherein: The link comprises a PCIe link.
6. The system according to claim 5, wherein: The link speed includes one of a plurality of different PCIe link speeds corresponding to different PCIe generations.
7. The system according to claim 5, wherein: The link includes a plurality of lanes, and the link width corresponds to an amount of activity in the plurality of lanes.
8. The system according to claim 1, wherein: The endpoint device controller is configured to determine the link speed for the link using a lookup table that maps each of a plurality of bandwidths to a corresponding link speed of a plurality of link speeds.
9. The system according to claim 1, wherein: The endpoint device controller is configured to determine the link width for the link using a lookup table that maps each of a plurality of bandwidths to a corresponding link width of a plurality of link widths.
10. The system according to claim 1, wherein: The endpoint device controller is configured to determine the link speed for the link by: determining a power consumption for each of a plurality of different link speeds that satisfies the aggregate bandwidth request; as well as A link speed of the plurality of different link speeds having the lowest power consumption is determined.
11. The system according to claim 1, wherein: The endpoint device controller is configured to determine the link width for the link by: determining power consumption for each of a plurality of different link widths that satisfies the aggregate bandwidth request; and A link width of the plurality of different link widths having the lowest power consumption is determined.
12. The system of claim 1, further comprising: A power management integrated circuit PMIC configured to provide one or more voltages to the second PCIe interface circuit; as well as The endpoint device controller is configured to set the one or more voltages provided by the PMIC to the second PCIe interface circuit based on the link speed.
13. The system of claim 1, wherein: The link includes a plurality of channels; The second PCIe interface circuit includes a plurality of drivers, wherein each driver of the plurality of drivers is configured to drive a corresponding channel of the channels; The system includes a power switch circuit configured to: selectively power the plurality of drivers; and The endpoint device controller is configured to set a number of the plurality of drivers powered by the power switch circuit based on the link width.
14. The system of claim 1, wherein: The endpoint device controller is configured to change a voltage domain of the second PCIe interface circuit according to a determined link speed or link width for the link based on the aggregate bandwidth request, and the speed change request sent to the host device is configured to enable the host device to change a host device voltage domain coordinated with the voltage domain of the second PCIe interface circuit.
15. The system of claim 1, wherein: The speed change request indicates the determined link speed.
16. The system of claim 1, wherein: The second PCIe interface circuit is configured to send data from the plurality of clients to the host device over the link and through the first PCIe interface circuit through the second PCIe interface circuit.
17. The system of claim 1, wherein: The plurality of bandwidth requests includes a plurality of bandwidth requests associated with different requested bandwidths from at least one client among the plurality of clients.
18. A method comprising: receiving, in an endpoint device controller, a plurality of bandwidth requests from a plurality of clients for transmitting data using one or more of a plurality of link speeds or widths on a link between a host device and the plurality of clients, wherein the host device and a first PCIe interface circuit are included in a first chip, the host device is coupled to the link via the first PCIe interface circuit, and wherein the endpoint device controller and the plurality of clients are included in a second chip, the endpoint device controller is coupled to the link via a second PCIe interface circuit; aggregating the multiple bandwidth requests into an aggregate bandwidth request; determining, in the endpoint device controller, at least one of a link speed and a link width for the link based on the aggregate bandwidth request, comprising: determining whether a change in the link speed or the link width is required based on the aggregate bandwidth request; When the change in the link speed or the link width is required, implementing, with the endpoint device controller, a speed change based on at least one of the determined link speed and link width for the link; and sending a speed change request to the host device over the link, the speed change request being based on the at least one of the determined link speed and link width for the link, The second chip includes a client system bus interface configured to couple the plurality of clients, the endpoint device controller and the second PCIe interface circuit, and The second PCIe interface circuit is configured to serialize parallel transmission data received from the plurality of clients through the client system bus interface for transmission on the link.
19. The method according to claim 18, wherein: The endpoint device controller comprises a PCIe endpoint device controller.
20. The method according to claim 19, wherein: The PCIe endpoint device controller is configured to send the speed change request to the host device via the link and through the second PCIe interface circuit.
21. The method according to claim 20, wherein: The PCIe endpoint device controller is configured to implement the speed change based on the at least one of the determined link speed and link width for the link.
22. The method according to claim 18, wherein: The link comprises a PCIe link.
23. The method according to claim 22, wherein: The link speed includes one of a plurality of different PCIe link speeds corresponding to different PCIe generations.
24. The method of claim 18, wherein: The link includes a plurality of lanes, and the link width corresponds to an amount of activity in the plurality of lanes.
25. The method of claim 18, further comprising: The link speed for the link is determined using a lookup table that maps each of a plurality of bandwidths to a corresponding link speed of a plurality of link speeds.
26. The method of claim 18, further comprising: The link width for the link is determined using a lookup table that maps each of a plurality of bandwidths to a corresponding link width of a plurality of link widths.
27. The method of claim 18, further comprising: The link speed for the link is determined by: determining a power consumption for each of a plurality of different link speeds that satisfies the aggregate bandwidth request; as well as A link speed of the plurality of different link speeds having the lowest power consumption is determined.
28. The method of claim 18, further comprising: The link width for the link is determined by: determining a power consumption for each of a plurality of different link widths that satisfies the aggregate bandwidth request; as well as A link width of the plurality of different link widths having the lowest power consumption is determined.
29. The method of claim 18, further comprising: Providing one or more voltages or clocks to the second PCIe interface circuit; as well as The one or more voltages or clocks provided to the second PCIe interface circuit are set based on the link speed.
30. The method of claim 18, further comprising: configuring a plurality of drivers to drive the link, the link comprising a plurality of channels, each channel being driven by a respective driver of the plurality of drivers, the plurality of drivers being coupled to a power switch circuit configured to selectively power the plurality of drivers; as well as A number of the plurality of drivers to be selectively powered by the power switch circuit is set based on the link width.
31. The method of claim 18, further comprising: based on the aggregate bandwidth request, changing a voltage domain of the endpoint device controller according to the determined link speed or link width for the link; Wherein the speed change request sent to the host device is configured to enable the host device to change a host device voltage domain coordinated with the voltage domain of the endpoint device controller.
32. A system comprising: A first chip comprising: a first peripheral component interconnect express (PCIe) interface circuit; Multiple host clients; a host device controller configured to control the plurality of host clients; and a host system bus interface configured to couple the plurality of host clients, the host device controller, and the first PCIe interface circuit; A second chip comprising: Endpoint devices; Endpoint device controller; A second PCIe interface circuit; and a client system bus interface configured to couple the endpoint device, the endpoint device controller, and the second PCIe interface circuit; and link, which couples the first PCIe interface circuit with the second PCIe interface circuit, wherein the first PCIe interface circuit comprises a serializer configured to serialize parallel transmit data received from the plurality of host clients via the host system bus interface, the parallel transmit data being generated for transmission on the link, And wherein the host device controller is configured to: receiving a plurality of bandwidth requests from the plurality of host clients; aggregating the multiple bandwidth requests into an aggregate bandwidth request; Determining at least one of a link speed and a link width for the link based on the aggregate bandwidth request includes: determining whether a change in the link speed or the link width is needed based on the aggregate bandwidth request; and After determining the at least one of the link speed and the link width, a speed change notification message is sent to the endpoint device via the first PCIe interface circuitry, the link, and the second PCIe interface circuitry.
33. The system of claim 32, wherein: The endpoint device comprises a PCIe endpoint device, and the first PCIe interface circuit is configured to send the speed change notification message to the PCIe endpoint device via the link.
