Host system, method thereof and acceleration module

By introducing an FPGA acceleration module into the storage device, using the PCIe bus to communicate with the processor, the problem of low query efficiency of large-scale databases is solved, and acceleration instructions are executed without loading data into memory, which improves query efficiency.

CN118916310APending Publication Date: 2024-11-08SAMSUNG ELECTRONICS CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410920660.5
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Priority Date
2018-09-05
Filing Date
2019-03-04
Publication Date
2024-11-08

AI Technical Summary

Technical Problem

Existing storage devices are inefficient when handling large-scale database queries, requiring frequent loading and unloading of data, resulting in latency and waste of resources.

Method used

Using an FPGA-based acceleration module, the PCIe bus communicates with the processor and storage device to achieve the execution of acceleration instructions without loading data into memory, and the acceleration platform manager is used to process acceleration instructions to improve data processing efficiency.

Benefits of technology

Reduces data loading delay, avoids unnecessary memory space release, and improves the efficiency of large-scale database queries.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118916310A_ABST
    Figure CN118916310A_ABST
Patent Text Reader

Abstract

The invention discloses a host system. The system may include: a processor running an application; a memory storing data used by the application program; an upstream interface in communication with the processor; and a downstream interface in communication with the storage device. The system may also include an acceleration module implemented using hardware and including an acceleration platform manager to execute acceleration instructions. A storage device may include: an endpoint of the storage device to communicate with an acceleration module; a controller to manage operations of the storage device; storing application data for the application program; and the storage device acceleration platform manager is used for assisting the acceleration platform manager in executing the acceleration instruction. The processor, the acceleration module, and the storage device may communicate via a peripheral component interconnect fast bus. The acceleration module may support execution of acceleration instructions on the application data on the storage device for the application without loading the application data into the memory. A method of a host system and an acceleration module are also provided.
Need to check novelty before this filing date? Find Prior Art

Description

[0001] The present invention is a divisional application of the invention patent application with application number 201910161758.6 filed on March 4, 2019 and invention name “Host system, method and acceleration module”. Technical Field

[0002] The inventive concept relates to storage devices, and more particularly, to accelerated Solid State Drive (SSD) execution using additional hardware. Background Art

[0003] There are situations where using storage devices in a conventional manner is inefficient. For example, imagine a situation where a query needs to be run on a database. The conventional solution is to load the database into the computer's memory, execute the query on the in-memory copy of the database, and then process the results. Although this approach may be reasonable when the database is relatively small, loading a database containing thousands, millions, or more records (where the result of the query is to identify a single record in the database) is extremely inefficient. A very large amount of data needs to be moved into memory to execute the query, which is likely to displace other data already stored in memory. And then, most of the data is discarded once the query has been executed, because most of the data is not needed after the query is completed. This problem may be magnified when queries need to be executed repeatedly on the database: each query may require the database to be loaded into memory again.

[0004] There remains a need for a method for speeding up operations involving storage devices. Summary of the invention

[0005] Purpose of this Disclosure

[0006] Example embodiments of the present disclosure may provide a low-cost and low-power SSD architecture for FPGA-based acceleration.

[0007] An example embodiment provides a system, which may include: a processor that runs an application; a memory that stores data used by the application running on the processor; an upstream interface for communicating with the processor; a downstream interface for communicating with a storage device; an acceleration module that is implemented using hardware and includes an Acceleration Platform Manager (APM-F) to execute acceleration instructions; and a storage device. The storage device may include: an endpoint of the storage device for communicating with the acceleration module; a controller for managing the operation of the storage device; a storage for storing application data of the application; and a storage device Acceleration Platform Manager (APM-S) to assist the APM-F in executing acceleration instructions. The processor, the acceleration module, and the storage device may communicate via a Peripheral Component Interconnect Exchange (PCIe) bus. The acceleration module may support the execution of acceleration instructions on application data on the storage device for the application without loading the application data into the memory.

[0008] Another example embodiment provides an acceleration module implemented using hardware, the acceleration module may include: an acceleration platform manager (APM-F) to execute acceleration instructions; an upstream interface to communicate with a processor; and a downstream interface to communicate with a storage device. The storage device may include a storage device acceleration platform manager (APM-S) to assist the APM-F in executing the acceleration instructions. The acceleration module may communicate with the processor and the storage device using a peripheral component interconnect express (PCIe) bus. The acceleration module may support execution of the acceleration instructions on application data on the storage device for an application without loading the application data into a memory associated with the processor.

[0009] Yet another example embodiment provides a method that may include: receiving a peripheral component interconnect express (PCIe) transaction from a first device at an acceleration module; determining at the acceleration module whether the PCIe transaction is an acceleration instruction; processing the PCIe transaction at an acceleration platform manager (APM-F) of the acceleration module based at least in part on determining that the PCIe transaction is an acceleration instruction; and transmitting the PCIe transaction to a second device based at least in part on determining that the PCIe transaction is not an acceleration instruction. The acceleration module may support execution of the acceleration instruction on application data on a storage device for an application without loading the application data into a memory associated with a processor. The processor, the acceleration module, and the storage device may communicate using a PCIe bus.

[0010] Effects of the Disclosure

[0011] The embodiments disclosed herein may provide an acceleration module that can load data. Therefore, the embodiments disclosed herein may avoid the delay required to load data from an SSD. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Figure 1 A machine supporting an acceleration operation on a storage device according to an embodiment of the inventive concept is shown.

[0013] Figure 2 Show Figure 1 Additional details of the machine.

[0014] Figure 3 The first embodiment of the present invention is shown. Figure 1 The acceleration module and Figure 1 components of a storage device.

[0015] Figure 4 A method for accelerating the first embodiment of the present invention is shown. Figure 1 The memory usage of instructions in the system.

[0016] Figure 5 The second embodiment of the present invention is shown. Figure 1 The acceleration module and Figure 1 components of a storage device.

[0017] Figure 6 A method for accelerating the use of a second embodiment of the present invention is shown. Figure 1 The memory usage of instructions in the system.

[0018] Figure 7 The third embodiment according to the present invention is shown. Figure 1 The acceleration module and Figure 1 components of a storage device.

[0019] Figure 8 A method for accelerating the operation of the third embodiment according to the present invention is shown. Figure 1 The memory usage of instructions in the system.

[0020] Fig. 9 A fourth embodiment according to the present invention is shown. Figure 1 The acceleration module and Figure 1 components of a storage device.

[0021] Fig.10 A method for accelerating the operation of the fourth embodiment according to the present invention is shown. Figure 1 The memory usage of instructions in the system.

[0022] Fig.11A fifth embodiment according to the present invention is shown. Figure 1 The acceleration module and Figure 1 components of a storage device.

[0023] Fig.12 A method for accelerating the operation of the present invention according to a fifth embodiment of the present invention is shown. Figure 1 The memory usage of instructions in the system.

[0024] Fig.13 The sixth embodiment according to the present invention is shown. Figure 1 The acceleration module and Figure 1 components of a storage device.

[0025] Fig.14 A method for accelerating the use of a sixth embodiment according to the concept of the present invention is shown. Figure 1 The memory usage of instructions in the system.

[0026] Fig.15 A seventh embodiment according to the present invention is shown. Figure 1 The acceleration module and Figure 1 components of a storage device.

[0027] Fig.16 A method for accelerating the operation of the seventh embodiment according to the present invention is shown. Figure 1 The memory usage of instructions in the system.

[0028] Fig.17 The eighth embodiment according to the present invention is shown. Figure 1 The acceleration module and Figure 1 components of a storage device.

[0029] Fig.18 A method for accelerating the operation of an eighth embodiment according to the present invention is shown. Figure 1 The memory usage of instructions in the system.

[0030] Fig.19 An embodiment of the present invention is shown with management and Figure 1 The bridging component of the acceleration module's communication Figure 1 components of the system.

[0031] FIG. 20A to FIG. 20B The embodiment according to the present invention is shown Figure 1 processor, Figure 1 The acceleration module and Figure 1 communication between storage devices.

[0032] Fig.21 The present invention is shown in an embodiment of the present invention. Figure 1 A flowchart of an example process of an acceleration module to process a PCIe transaction.

[0033] FIG. 22A to FIG. 22C The present invention is shown in an embodiment of the present invention. Figure 1 A flowchart of a more detailed example process of an acceleration module to handle PCIe transactions.

[0034] FIG. 23A to FIG. 23B The present invention is shown in an embodiment of the present invention. Figure 1 The acceleration module determines the Figure 1 A flowchart of an example process of determining whether a PCIe transaction of a processor includes an acceleration instruction.

[0035] Fig.24 The present invention is shown in an embodiment of the present invention. Figure 1 The acceleration module determines the Figure 1 A flowchart of an example process of determining whether a PCIe transaction of a storage device includes an acceleration instruction.

[0036] Fig.25 The present invention is shown in an embodiment of the present invention. Fig.19 The first bridge component is to determine the Figure 1 A flowchart of an example process of determining whether a PCIe transaction of a processor includes an acceleration instruction.

[0037] Fig.26 The present invention is shown in an embodiment of the present invention. Fig.19 The second bridge component is used to determine the Figure 1 A flowchart of an example process of determining whether a PCIe transaction of a storage device includes an acceleration instruction.

[0038] FIG. 27A to FIG. 27C The present invention is shown in an embodiment of the present invention. Figure 1 A flow chart of an example process for a storage device to process a PCIe transaction.

[0039] FIG. 28A to FIG. 28B The present invention is shown in an embodiment of the present invention. Figure 1 storage device to determine the Figure 1 A flowchart of an example process of determining whether a PCIe transaction of an acceleration module contains an acceleration instruction.

[0040] Description of Figure Numbers

[0041] 105: Machine;

[0042] 110: processor;

[0043] 115: device driver;

[0044] 120: storage device;

[0045] 125: Application;

[0046] 130: acceleration service manager;

[0047] 135: memory controller;

[0048] 140: main memory;

[0049] 145: Acceleration module / field programmable gate array;

[0050] 205: Clock;

[0051] 210: Network connector;

[0052] 215: bus;

[0053] 220: user interface;

[0054] 225: Input / output engine;

[0055] 305, 1105, 1310, 1505: endpoints;

[0056] 310: host interface layer;

[0057] 315: Acceleration Platform Manager;

[0058] 320: Flash translation layer;

[0059] 325: Flash media;

[0060] 330: upstream port;

[0061] 335: downstream port;

[0062] 340: Acceleration platform manager;

[0063] 345: Scheduler;

[0064] 350-1, 350-2: Acceleration engine;

[0065] 355: message mailbox;

[0066] 360: downstream filter;

[0067] 365: sideband bus;

[0068] 370: SSD controller;

[0069] 405, 805, 810, 1005, 1010, 1205, 1215, 1220, 1405: block;

[0070] 410, 415: subset;

[0071] 420: downstream filter address range;

[0072] 505: upstream filter;

[0073] 605: upstream filter address range;

[0074] 705, 905, 1115, 1120: physical functions;

[0075] 710: virtual functions;

[0076] 715: VF filter, PF filter;

[0077] 1110, 1305: root ports;

[0078] 1125: configure monitor;

[0079] 1210, 1410: address mapping;

[0080] 1905, 1910: bridge components;

[0081] 2005, 2020: PCIe transactions;

[0082] 2010: Specialized commands;

[0083] 2015, 2045: acceleration instructions;

[0084] 2025: PCIe transactions, accelerated data;

[0085] 2030, 2035, 2040: Results;

[0086] 2105, 2110, 2115, 2120, 2125, 2203, 2206, 2212, 2215, 2218, 2221, 2227, 2230, 2236, 2239, 2242, 2245, 2248, 2251, 2254, 2257, 2260, 2263, 2305, 2310, 2315, 2320, 2325, 2405, 2410, 2 415, 2420, 2505, 2510, 2515, 2520, 2605, 2610, 2615, 2620, 2705, 2710, 2715, 2725, 2730, 2735, 2740, 2745, 2755, 2760, 2765, 2770, 2775, 2780, 2785, 2805, 2810, 2815, 2820, 2825: frame;

[0087] 2209, 2224, 2233, 2720, 2750: dotted lines. DETAILED DESCRIPTION

[0088] Now will be referred to in detail the embodiment of the inventive concept, the example of the embodiment is shown in the accompanying drawings. In the following detailed description, numerous specific details are set forth to achieve a thorough understanding of the inventive concept. However, it should be understood that those skilled in the art can practice the inventive concept without these specific details. In other examples, well-known methods, processes, components, circuits and networks are not described in detail to avoid unnecessary confusion of the various aspects of the embodiment.

[0089] It should be understood that although the terms first, second, etc. can be used to describe various elements in this article, these elements should not be limited by these terms. These terms are only used to distinguish one element from another element. For example, without departing from the scope of the present invention, the first module can be referred to as the second module, and similarly, the second module can be referred to as the first module.

[0090] The terms used in the description of the inventive concept herein are only for the purpose of describing specific embodiments and do not wish to limit the inventive concept. As used in the description of the inventive concept and the appended claims, unless the context clearly indicates otherwise, the singular forms "a / an" and "said" are intended to also include plural forms. It will also be understood that the term "and / or" as used herein refers to and covers any and all possible combinations of one or more of the associated listed items. It will be further understood that when used in this specification, the terms "comprises" and / or "comprising" specify the presence of stated features, integers, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or groups thereof. The components and features in the drawings may not be drawn to scale.

[0091] Embodiments of the inventive concept propose a solid-state drive (SSD) or other storage device architecture in which a field programmable gate array (FPGA) is used to accelerate certain data processing functions. The FPGA device is placed in the front of the SSD or along the side of the SSD, which provides a peripheral component interconnect express (PCIe) host interface. Since host transactions are received on the FPGA PCIe interface, those PCIe transactions are forwarded to the back-end SSD controller. The terms "SSD" and "SSD controller" are used interchangeably and have generally the same meaning unless otherwise specified. The back-end SSD implements PCIe endpoints and non-volatile memory express (NVMe) controllers. Therefore, the host directly communicates the NVMe protocol with the back-end SSD. That is, the PCIe interface from the host to the back-end SSD via the FPGA has a transparent transmission nature. The SSD performs data transfer to / from the host system memory via direct memory access (DMA). The FPGA Downstream Port (DSP) is programmed with a memory filter address range (FAR) that acts as a PCIe transaction filter. The DSP filters all PCIe transactions within the FAR window and forwards them to the logic and memory on the FPGA. All PCIe transactions that are not within the programmed FAR window belong to the host system memory and are passed directly to the host. The SSD controller uses the PCIe Vendor Defined Message (VDM) mechanism or, for example, I 2The SSD controller uses the other sideband buses of the PCIe / SMBus to program the appropriate FAR window in the FPGA. The SSD controller requests a block of address ranges from the host via the PCIe base address register (BAR). After the host BIOS has assigned the SSD controller to the requested address block, the SSD controller programs the address range subset as a FAR window in the FPGA DSP. The address range programmed in the DSP is used by the SSD controller and the FPGA to communicate with each other. That is, as the host allocates address blocks, the SSD and FPGA can share the PCIe bus with host transactions without interfering with each other or other PCIe devices in the PCIe hierarchy. Using this FAR window on the shared PCIe bus, the SSD controller can provide acceleration instructions and data to the FPGA. The FPGA or host may also use the shared PCIe bus and the address ranges mentioned above to request data for acceleration from the SSD controller. The FPGA can also use the same mechanism to provide the acceleration results back to the SSD controller. The proposed architecture and mechanism implement a low-cost and low-power solution for SSD-based application acceleration using an FPGA device.

[0092] Details of the proposed solution

[0093] The basic concept is that the FPGA and the SSD (and / or other storage devices) work together (as separate devices or merged into a single device) to communicate with the host. In the following, in the description, the storage device means the SSD. There are three traffic flows:

[0094] 1) From the host to the storage device. The communication from the host to the storage device is managed by the FPGA by forwarding all traffic from the upstream port (USP) or endpoint (EP) to the downstream port (DSP) or root port or root complex port (RP) depending on the FPGA implementation for transmission to the EP of the storage device. The FPGA may include physical functions exposed by the storage device to the host to support NVMe communication between the host and the storage device.

[0095] 2) Communication of acceleration instructions to FPGA. In some embodiments of the inventive concept, acceleration instructions are handled in the following manner: An Acceleration Service Manager (ASM) may be run on a host. The ASM may communicate with an Acceleration Platform Manager (APM), which may include components that are part of both a storage device (identified as APM-S) and an FPGA (identified as APM-F). The ASM on the host may use the NVMe protocol to tunnel acceleration instructions and related information to the SSD. The SSD then acts as an acceleration coordinator relative to the FPGA: all acceleration instructions received by the APM-S are used to provide appropriate instructions to the APM-F using a proprietary interface. An address space window is used to facilitate the proprietary interface. This address space window may be allocated within the host memory address map at the request of the storage device (partially to facilitate communication between the storage device and the host using the NVMe protocol). Any instructions using addresses in the appropriate address space may be filtered by the FPGA to be processed by the APM-F, rather than communicating directly from the host or storage device to another host or storage device. Filtering may also be performed using tags associated with the instructions or using filtering based on PCIe messages. This filtering may be performed by a filter connected to the DSP / RP that connects the FPGA to the storage device.

[0096] 3) Communication between FPGA and SSD to extract data for acceleration and processing of the data. In some embodiments of the inventive concept, acceleration is performed as follows: When the FPGA wants to extract data for accelerated processing, the FPGA can send a request using the address space allocated within the host memory address map for storage device-FPGA communication, as discussed above.

[0097] DSP filter architecture

[0098] This architecture proposes a method by which the PCIe bus between the host and the SSD can be shared with the FPGA for accelerating data processing.

[0099] Logically, the FPGA is operably placed between the host and the SSD controller. The host is connected to the upstream port (USP) of the FPGA, and the SSD is connected to the downstream port (DSP) of the FPGA. The PCIe bus used to connect to the host and / or to the SSD can be a ×4 or ×8 lane, or any other desired width. The FPGA USP port and the DSP port forward PCIe transactions (i.e., transaction layer packets (TLP)) in both directions. Examples of PCIe TLPs are configuration reads, configuration writes, memory reads, and memory writes. Therefore, the host communicates directly with the SSD. The DSP port on the FPGA has logic to filter all PCIe transactions from the SSD controller based on a programmed filter address range (FAR). The intercepted SSD controller PCIe transactions are then directed to the Accelerated Platform Manager-FPGA (APM-F) block. The APM-F module communicates with the SSD controller. The APM-F module receives data and acceleration instructions from the Accelerated Platform Manager-SSD (APM-S) firmware from the SSD controller. The APM-F module then provides the received acceleration instructions and data to the runtime (RT) scheduler. The RT scheduler then programs the appropriate acceleration engine to perform data processing.

[0100] The use of an FPGA represents one possible implementation, however implementations other than an FPGA may be used. The FPGA may be implemented within the storage device. The FPGA supports accelerated data processing that may be performed close to the storage device rather than by fetching data to host memory and then processing the data on the host. Instead of fetching data, the storage device / FPGA may receive the query and perform the processing locally.

[0101] The SSD controller uses PCIe transports to implement NVMe protocol processing logic. As part of the PCIe configuration, the SSD controller requests blocks of the host system address map for its own use. The SSD controller requests larger blocks than it normally requires to support the NVMe protocol: some or all of the extra space can be used to manage communications between the storage device and the FPGA. For example, the NVMe protocol may require (for example) a 64KB address space; then in the proposed architecture, the SSD controller may request (for example) a 10MB address block. The SSD controller uses the portion of the allocated address map to communicate with the FPGA in a host-transparent manner. The subset of the system address map reserved for SSD-FPGA communication is called the Filter Address Range (FAR). The SSD controller then programs the FAR window in the FPGA DSP. The SSD controller can use, for example, I 2The SSD controller can also use the PCIe Vendor Defined Messages (VDMs) to program the FAR windows in the FPGA.

[0102] The Host Interface Logic (HIL) module implements the NVMe protocol and communicates with the NVMe driver running on the host. The HIL module interacts with the Flash Translation Layer (FTL) to execute normal host NVMe commands. In addition, the HIL module intercepts the dedicated acceleration commands received from the host side and forwards the dedicated acceleration commands to the APM-S module. APM-S can be implemented as firmware or firmware + hardware. The APM-S module can process the dedicated acceleration commands and then prepare the acceleration instructions and data to be sent to the APM-F module on the FPGA. The APM-S module then uses the filter address range (FAR) address to send the acceleration information to the FPGA. The communication between APM-S and APM-F can be message-based. A variety of different methods can be used for this type of communication between APM-S and APM-F.

[0103] The proposed architecture and mechanism allow the SSD controller to share the host PCIe bus to achieve FPGA-based acceleration. Embodiments of the inventive concept provide a low-cost and low-power solution for application acceleration using an FPGA in an SSD.

[0104] DSP+USP filter architecture

[0105] In this architecture, the FPGA is visible to the host in an indirect manner. The communication between the FPGA and the SSD remains the same as the proposed solution 1. The SSD controller can request a larger system address space from the host. The SSD controller can divide the allocated address block into three windows. One window is used for the NVMe controller register address space. The second window is used for communication between the FPGA and the SSD, as described above. The third window is used for communication between the host and the FPGA. The host can find the location of the FPGA device based on the dedicated NVMe register. The SSD controller can broadcast the third window in a dedicated register, which can be read by the host application to know the location of the FPGA device. The SSD controller can also program the USP with the same address window so that the USP can filter the transaction. The USP can filter all transactions in the third window address space from the host and forward them to the FPGA acceleration logic. This mechanism can be used by the Acceleration Service Manager (ASM) on the host to communicate acceleration instructions and data to the FPGA.

[0106] Therefore, in some embodiments of the inventive concept, filtering may also be performed based on the business received by the FPGA from the host. That is, the host may also send accelerated instructions / data to the FPGA. Similar to the DSP / RP connected to the FPGA, the filter may also be connected to the USP / EP of the FPGA. The host may use addresses in the address space requested by the storage device. The address used by the host for host-FPGA communication may be part of the address space requested by the storage device for NVMe communication with the host (again, where the requested address space may be larger than the space required for NVMe communication), or part of a separate address space within the host memory address map (for a virtual function or for a second physical function, either of which is also exposed to the host by the storage device). It should be noted that filtering at the USP / EP and filtering at the DSP / RP may be performed using different address ranges within the host memory address map, enabling the host to send instructions to the storage device or the FPGA as needed (while still allowing the storage device to communicate with the FPGA as needed).

[0107] In embodiments of the inventive concept where a portion of the address space supports communication between the host and the FPGA, the FPGA may not be directly visible to the host. In this case, the ASM on the host can "discover" the FPGA by accessing addresses written in dedicated registers in the NVMe address space that identify the address range used for host-FPGA communication. The ASM can discover the storage device via the PCIe device table and thus know which register in the NVMe address space stores a pointer to the host-FPGA communication address space.

[0108] VF+DSP filter architecture

[0109] In this FPGA+SSD architecture, the SSD exposes one physical function (PF) and one virtual function (VF) to the host. The SSD controller is exposed via the PF. The VF is used to expose the FPGA. The PF class code may indicate a mass storage device, whereas the VF class code may be set to identify the FPGA. The SSD controller PF may request a larger system memory address block so that a subset of the memory address block may be used for communication between the FPGA and the SSD via the FPGA DSP, and the SSD controller VF may request its own memory address block for communication between the FPGA and the host via the FPGA USP.

[0110] The FPGA USP can be programmed with different memory filter address ranges and / or VF tags (FAR-USP) that can be used as PCIe transaction filters. The USP can filter all PCIe transactions within the FAR window and / or all PCIe transactions belonging to the VF and can forward them to the acceleration logic and memory on the FPGA. All PCIe transactions that are not within the programmed FAR-USP window or that do not belong to the VF belong to the SSD and can be passed directly to the SSD. The SSD controller can use the PCIe VDM mechanism or, for example, I 2 C / SMBus to program the appropriate FAR-USP window to communicate this information.

[0111] Such FAR address windows may allow ASM software running on the host to communicate with APM-F. That is, the FPGA is directly visible to the host. ASM software may use such PCIe address ranges to send acceleration orchestration logic instructions and data to the FPGA. APM-F may then provide the received acceleration instructions and data to the runtime (RT) scheduler. The RT scheduler then programs the appropriate acceleration engine to perform data processing. APM-F may also extract data from host memory or SSD storage.

[0112] PF+DSP filter architecture

[0113] This FPGA+SSD architecture is similar to the proposed solution 3, except that a second PF can be used to expose the FPGA to the host instead of using a VF. The SSD controller exposes the physical physical functions to the host. The first PF can be used for the SSD controller, and the second PF can be used for the FPGA. The base address of the second PF can be programmed in the FAR-USP in the FPGA. Therefore, the USP can filter all transactions from the host that are within the programmed address range (FAR-USP) for use with the second PF, and can forward all of them to the FPGA. This mechanism can be used by the ASM running on the host to communicate with the FPGA.

[0114] By exposing a virtual function or a (second) physical function to the host, the address space for host-FPGA communication can be requested from the host (by the storage device or by the FPGA). In the case where a virtual function or a second physical function is exposed, the filter on the USP / EP can filter based on the address range allocated for host-FPGA communication or based on the exposed virtual function or the exposed second physical function (e.g., by the filter number or some other tag). (Virtual functions require operating system support; if the second physical function is implemented / supported, then exposing the second physical function provides an alternative solution to using the virtual function.)

[0115] PF+RP filter architecture

[0116] In this FPGA SSD architecture, the FPGA PCIe ports are endpoints (EP) and root ports (RP), rather than USPs and DSPs. The difference between the USP / DSP ports and the EP / RP ports is that both the EP / RP have their own PCIe configuration space while the USP / DSP ports do not. In the architecture according to these embodiments of the inventive concept, the FPGA can expose two PFs to the host (it should be noted that the FPGA exposes its own PF to the host, rather than the SSD providing the PF / VF that exposes the FPGA). The SSD EP can be connected to the RP on the FPGA. The first FPGA EP PF can be used to connect the host directly to the SSD, while the second FPGA EP PF can be used to connect the host to the FPGA. This mechanism can be used by the ASM running on the host to communicate with the FPGA. The SSD-FPGA communication can use a portion of the address space mapping between the FPGA RP and the SSD EP. In some embodiments of the inventive concept, the first FPGA PF can request a larger address space and can map / convert the address window assigned by the BIOS to the SSD controller EP. Portions of the address space can be used for local FPGA-SSD communication. In another embodiment of the inventive concept, a portion of the address space allocated for the second FPGA EPPF may be used for communication between the FPGA and the SSD controller.

[0117] PF+ dual-port SSD architecture

[0118] In this FPGA-SSD architecture, a dual-port SSD is used in conjunction with the FPGA. In this architecture, the FPGA (again, the FPGA exposes its own PF, rather than the SSD providing the PF / VF that exposes the FPGA) can expose two PFs to the host. The SSDEP can be connected to the RP on the FPGA. The first FPGA EP PF can be used to connect the host directly to the SSD. All host transactions coming for the second FPGAEP PF can be forwarded to the FPGA acceleration logic. The ASM software running on the host can use the second FPGAEP PF to communicate with the FPGA.

[0119] For FPGA-SSD communication (for accelerated processing), the second PCIe EP on the SSD can be used. Thus, the FPGA has two RPs connected to the SSD. The first EP port of the SSD can be used to communicate with the host for normal host storage access. The second EP on the SSD can be used to pass any data needed in the FPGA for processing.

[0120] As mentioned, in embodiments of the inventive concept where the storage device may support two (or possibly more) ports, the FPGA may support two RPs to communicate with two EPs on the storage device. In such embodiments of the inventive concept, one RP on the FPGA (and its corresponding EP on the storage device) may be used to manage communications between the storage device and the host, and another RP on the FPGA (and its corresponding EP on the storage device) may be used to manage communications between the storage device and the FPGA (for accelerating instructions / data). In such embodiments of the inventive concept, the RP on the FPGA may support two address mappings (one for each RP). Thus, the address mapping of the RP used to support communications between the host and the storage device may include space allocated for NVMe commands, and another address mapping (of the RP used to manage communications of accelerated instructions / data) may be entirely dedicated to such communications. It should be noted that in such embodiments of the inventive concept, the host memory address mapping may omit any address space intended for the host to communicate accelerated instructions to the storage device, since all such instructions may be sent from the host to the FPGA (via the address space that the FPGA requests to be allocated within the host's memory address mapping for such communications). The FPGA may then process the instruction and, if necessary, forward the instruction / data to the storage device using a memory address map on a second RP that is dedicated to communication between the FPGA and the storage device.

[0121] In the case where EP / RP is used in place of USP / DSP in the FPGA, the EP / RP can also support PCIe configuration space, and the FPGA can expose its physical functions directly to the host (rather than leaving this function to the storage device). One physical function exposed by the FPGA can be used to direct communication from the host to the storage device; another physical function can be used for communication between the host and the FPGA. In such embodiments of the inventive concept, the FPGA can request an address space allocated from the host, rather than the storage device making such a request.

[0122] The EP / RP may also support its own memory mapping. Thus, the FPGA may communicate with the host using the host's memory address mapping, and the FPGA may support its own memory address mapping, which is used to communicate with the storage device. In such embodiments of the inventive concept, the FPGA may request space allocated in the host's memory address mapping to support communications from the host to the storage device (where such communications are performed via the FPGA), and additional space allocated in the host's memory address mapping to support communications from the host to the FPGA. The FPGA's memory address mapping may then include its own space allocations for communicating commands from the host to the storage device and for communicating acceleration instructions / data between the FPGA and the storage device. The FPGA may convert the address space used for communication between the host and the storage device to the address space used for communication between the FPGA and the storage device (the address spaces should be the same size).

[0123] In the case where the FPGA contains EP / RP, then the host looks directly at the FPGA. This raises the question of what PCIe performance the FPGA exposes. The FPGA should expose the same PCIe performance as the storage device. Therefore the FPGA may contain a PCIe configuration monitor that sets the FPGA's EP PCIe configuration to match the SSD controller EP PCIe configuration in the storage device. Additionally, when the host changes the PCIe configuration of the FPGA's EP, the storage device's EP PCIe configuration may be similarly modified.

[0124] Embodiments of the inventive concept may support partitioning of components / functionality into multiple separate elements as described within an FPGA as long as the overall functionality is maintained.FPGA components may be implemented using hardware, software / firmware, or a combination of the two.

[0125] exist Figure 1 , a machine 105 is shown. The machine 105 may include a processor 110. The processor 110 may be any of a variety of processors: for example, an Intel Xeon, Celeron, Itanium or Atom processor, an AMD Opteron processor, an ARM processor, etc. Although Figure 1A single processor 110 is shown in the machine 105, but the machine 105 may include any number of processors, each of which may be a single-core processor or a multi-core processor, and may be mixed in any desired combination. The processor 110 may run a device driver 115, which may support access to a storage device 120, and a different device driver may support access to other components of the machine 105. Throughout this document, the storage device 120 will be described as a solid-state drive (SSD) 120, but the storage device 120 may be any other type of storage device that supports acceleration instructions as described in the embodiments of the inventive concept below. The processor 110 may also run an application 125, which may be any application that includes acceleration instructions, and an acceleration service manager (ASM) 130, which may be used to send acceleration instructions to be executed on data stored on the storage device 120.

[0126] The machine 105 may also include a memory controller 135 that can be used to manage access to a main memory 140. The memory 140 can be any of a variety of memories, such as flash memory, dynamic random access memory (DRAM), static random access memory (SRAM), persistent random access memory (Persistent Random Access Memory), ferroelectric random access memory (FRAM), or non-volatile random access memory (NVRAM), such as magnetoresistive random access memory (MRAM). The memory 140 can also be any desired combination of different memory types.

[0127] The machine 105 may also include an acceleration module 145. The acceleration module 145 may assist the processor 110 by executing acceleration instructions on data stored on the storage device 120 as requested by the processor 110. The acceleration module 145 may be implemented using firmware alone or a combination of hardware and firmware. Throughout this document, the acceleration module 145 will be described as a field programmable gate array (FPGA) 145, however, the acceleration module 145 may be any other type of acceleration module that supports acceleration instructions as described in the embodiments of the inventive concept below. For example, the acceleration module 145 may be implemented as or using an Application-Specific Integrated Circuit (ASIC), a Graphics Processing Unit (GPU), an In-Storage Computing (ISC) capability of an SSD, or other implementation schemes.

[0128] Although Figure 1 The machine 105 is depicted as a server (which may be a stand-alone server or a rack server), but embodiments of the inventive concept may include (without limitation) any desired type of machine 105. For example, the machine 105 may be replaced with a desktop computer or a laptop computer or any other machine that may benefit from embodiments of the inventive concept. The machine 105 may also include a dedicated portable computing machine, a tablet computer, a smartphone, and other computing machines. Additionally, although Figure 1 The machine 105 is shown as including storage 120, applications 125, and ASM 130, but embodiments of the inventive concepts may have these components in separate machines: for example, the storage 120 may be installed on a server connected to the machine 105 (and the applications 125 and ASM 130) via a network connection traversing one or more networks of any type (wired, wireless, global, etc.).

[0129] regardless Figure 1 Regardless of the specific arrangement of components shown, the terms "host," "host machine," or "host processor" may also be used to describe the machine 105. This may distinguish the processor 110 from other components of the inventive concept.

[0130] exist Figure 1 Among the components, there are three business flows that the embodiments of the inventive concept are particularly concerned with (there may also be other business flows that are not relevant to the embodiments of the inventive concept):

[0131] 1) From the host to the storage device 120. The host (processor 110) may send communications to the storage device 120. In an embodiment of the inventive concept, all such services pass through the acceleration module 145 and should not be blocked from reaching the storage device 120 through the acceleration module 145. Examples of such services may include commands to read data from the storage device 120 and / or write data to the storage device 120: other commands provided by the storage device 120 may also include such services.

[0132] 2) From the ASM 130 to the acceleration module 145. The ASM 130 may request execution of certain acceleration instructions. To some extent, regardless of the specific embodiment of the inventive concept, the acceleration module 145 should receive the acceleration instructions from the ASM 130.

[0133] 3) From the acceleration module 145 to the storage device 130. In order to execute the acceleration instruction, the acceleration module 145 may need to fetch or receive data from the storage device 130.

[0134] Figure 2 Show Figure 1 Additional details of the machine. Figure 2 In the embodiment of the present invention, the machine 105 generally includes one or more processors 110, which may include a memory controller 135 and a clock 205, which may be used to coordinate the operation of the components of the device 105. The processor 110 may also be coupled to a memory 140, which may include, for example, a random access memory (RAM), a read-only memory (ROM), or other state retention media. The processor 110 may also be coupled to a storage device 120 and a network connector 210, which may be, for example, an Ethernet connector or a wireless connector. The processor 110 may also be connected to a bus 215, which may be attached to a user interface 220 and input / output interface ports, which may be managed using an input / output engine 225 and other components.

[0135] First Example Embodiment

[0136] Figure 3 The first embodiment of the present invention is shown. Figure 1 FPGA 145 and Figure 1 Components of the SSD 120. Figure 3 In FIG. 1 , processor 110, FPGA 145, and SSD 120 are shown in communication. Figure 3In the embodiment of the present invention, the processor 110, FPGA 145, and SSD 120 can communicate via a peripheral component interconnect express (PCIe) bus. The PCIe bus can use any number of lanes: typical examples are ×4 and ×8, but embodiments of the inventive concept can use any other desired number of lanes. These communications can include PCIe transactions, which can be transaction layer packets (TLPs) that encode commands using the non-volatile memory express (NVMe) protocol, but embodiments of the inventive concept can be extended to include communications using different encodings or commands in different protocols.

[0137] The SSD 120 may include an endpoint 305, a host interface layer (HIL) 310, an SSD acceleration platform manager (APM-S) 315, a flash translation layer (FTL) 320, and flash media 325. The endpoint 305 may be a logical or physical connection point at which the SSD 120 may receive and send PCIe communications. When the SSD 120 receives a PCIe transaction from the processor 110 at the endpoint 305 (via the FPGA 145), the SSD 120 may transmit the PCIe transaction to the HIL 310. The HIL 310 may then determine whether the PCIe transaction includes an acceleration instruction or does not include an acceleration instruction. If the PCIe transaction includes an acceleration instruction, the HIL may forward the PCIe transaction (or the acceleration instruction itself, unpacked from the PCIe transaction) to the APM-S 315 for processing: The APM-S 315 may be implemented using firmware alone or a combination of hardware and firmware. Otherwise, HIL 310 may pass the PCIe transaction (or unencapsulated NVMe command) to FTL 320, where FTL may Figure 1 The logical block address (LBA) used by the application 125 is converted into a physical block address (PBA) to access data stored on the flash memory medium 325.

[0138] There are essentially two different types of acceleration instructions that the APM-S 315 may process. The first type of acceleration instruction is a dedicated command from the processor 110. In a first embodiment of the inventive concept, the FPGA 145 is invisible to the processor 110: the processor 110 sends all of its communications to the SSD 120. When the processor 110 wants to execute an acceleration instruction on application data, the processor 110 may send a dedicated command to the SSD 120. The processor 110 may use NVMe commands to tunnel the dedicated command / acceleration instruction to the SSD 120. The HIL 310 may intercept such a dedicated command, which may be transmitted to the APM-S 315. The APM-S 315 may then generate an acceleration instruction in response to the dedicated command, which may be sent back to the FPGA 145 to execute the acceleration instruction. Such a dedicated command may, for example, encode a specific type of acceleration instruction to be executed and the data of the acceleration command instruction to be executed.

[0139] The second type of acceleration instruction that APM-S 315 may process will involve data. For example, FPGA 145 may not be able to directly access flash media 325, and therefore may not be able to execute the acceleration instruction without receiving the data on which the acceleration instruction is to be executed. Therefore, APM-S 315 may receive an acceleration instruction from FPGA 145 requesting the data in question. APM-S 315 may then access the requested data and pass it back to FPGA 145, thereby permitting FPGA 145 to execute the acceleration instruction.

[0140] exist Figure 3 , SSD 120 is shown to include FTL 320 and flash media 325, which are suitable for use in an SSD. If SSD 120 is replaced with an alternative storage device, then these components can be replaced with alternative components suitable for the form of the storage device. For example, if SSD 120 is replaced with a hard drive, then flash media 325 can be replaced with a hard drive platter. Additional components may also be included to support data access: Continuing with the example of a hard drive storage device, the storage device may also include a read / write head as needed.

[0141] Before FPGA 145 can intercept communications between processor 110 and SSD 120, SSD 120 can request a block of host memory addresses from processor 110. This request for a block of host memory system addresses is conventional when using PCIe transactions and can be performed at boot time or at a later time. In response, processor 110 (or Figure 1 The Basic Input / Output System (BIOS) of the machine 105 may allocate a block of host memory system addresses for use by the SSD 120. The host machine then knows that this block of host memory system addresses is not available for use by the SSD 120. Figure 1 Other devices in the machine 105 are used.

[0142] Figure 4 A method for accelerating the first embodiment of the present invention is shown. Figure 1 The memory usage of the instructions in the system. Figure 4 , SSD 120 may request a block of host memory system addresses. It should be noted that while the amount of memory needed to support NVMe communications between processor 110 and SSD 120 may be relatively small (e.g., 64KB), SSD 120 may request much larger blocks (e.g., 10MB or more). In response, processor 110 may return a block 405 of the host memory system address. One end of block 405 may be stored in a base address register (BAR) so that SSD 120 can determine block 405 based on the BAR (and using SSD 120's knowledge of the size of the requested block).

[0143] Once SSD 120 knows what addresses are in block 405, SSD 120 can partition block 405 into different areas for its use. A subset 410 of block 405 can be used for NVMe communication. Another subset 415 can be left idle. And a third subset, called downstream filter address range (FAR) 420 ("downstream" because of any filtering done downstream from processor 110), can be dedicated to communication between SSD 120 and FPGA 145. It should be noted that downstream FAR 420 can be used by SSD 120 and FPGA 145: either can use the addresses in downstream FAR 420 to indicate that the PCIe transaction in question contains an acceleration instruction.

[0144] Return to Figure 3 , if SSD 120 sends a communication using an address in downstream FAR 420, FPGA 145 may receive the communication at downstream port 335, intercept the communication, and process the communication locally rather than forwarding the communication to processor 110. Any communication received by FPGA 145 at downstream port 335 from SSD 120 that does not involve downstream FAR 420 may be transmitted by FPGA 145 to processor 110 via upstream port 330. (Any communication received by FPGA 145 from processor 110 at upstream port 330 may be automatically transmitted to SSD 120 via downstream port 335.)

[0145] FPGA 145 may be positioned between processor 110 and SSD 120 such that FPGA 145 may intercept communications between processor 110 and SSD 120. By intercepting such communications, FPGA 145 may execute acceleration instructions requested by SSD 120.

[0146] To execute the acceleration instructions, the FPGA 145 may include an upstream port 330, a downstream port 335, an FPGA acceleration platform manager (APM-F) 340, a scheduler 345, and acceleration engines 350-1 and 350-2. The upstream port 330 may be used to communicate with the processor 110; the downstream port 335 may be used to communicate with the SSD 120. The APM-F 340 is responsible for receiving any acceleration instructions intercepted by the FPGA 145. These acceleration instructions may be received as messages from the downstream port 335 using a message mailbox 355, but embodiments of the inventive concept may be extended to other mechanisms for the APM-F to receive acceleration instructions. Once the acceleration instructions are received, the APM-F 340 may process the acceleration instructions. For example, if FPGA 145 has enough information to be able to execute the acceleration instruction, APM-F 340 may pass the acceleration instruction to scheduler 345 (which may also be referred to as a "run-time scheduler"), which may then schedule the acceleration instruction using any available acceleration engines (e.g., acceleration engine 350-1 and acceleration engine 350-2). Figure 3 Two acceleration engines 350-1 and 350-2 are shown, but embodiments of the inventive concept may include any desired number of acceleration engines: Figure 3 Alternatively, if FPGA 145 requires additional information to execute an acceleration instruction (e.g., FPGA 145 requires data for an acceleration instruction to be executed), APM-F 340 may take another action (e.g., send a PCIe transaction to SSD 120) to request the necessary data.

[0147] To determine whether a particular PCIe transaction includes an acceleration instruction, the FPGA 145 may include a downstream filter 360 associated with the downstream port 335. The downstream filter 335 may identify PCIe transactions issued from the SSD 120 that may include an acceleration instruction. This may be done in any desired manner. In some embodiments of the inventive concept, the SSD 120 may include a PCIe transaction that includes an acceleration instruction. Figure 4 The downstream FAR 420 programs the downstream filter 360. Then, when the downstream filter 360 identifies the use Figure 4 When the SSD 120 receives a PCIe transaction with an address in the downstream FAR 420, the FPGA 145 may identify the PCIe transaction as containing an acceleration instruction and intercept the PCIe transaction. The SSD 120 may program the downstream filter 360 in any desired manner. For example, the SSD 120 may use, for example, an Inter-Integrated Circuit (IC). 2The SSD 120 may use a PCIe SMBus (System Management Bus) or a sideband bus 365 of the SMBus to program the downstream filter 360. Alternatively, the SSD 120 may use a PCIe Vendor Defined Message (VDM) to program the downstream filter 360. The SSD 120 may also use other mechanisms to program the downstream filter 360.

[0148] exist Figure 3 , FPGA 145 is shown to include components that enable communication with processor 110 and SSD 120: specifically, upstream port 330, downstream port 335, and downstream filter 360. Although FPGA 145 does require some mechanism for communicating with processor 110 and SSD 120, embodiments of the inventive concept can separate the communication elements from FPGA 145. For example, downstream port 335 and downstream filter 360 (the latter of which is responsible for identifying which PCIe transaction received from SSD 120 involves an acceleration instruction (and should be redirected to APM-F 340) rather than transmitted to processor 110) can be placed in the bridge component ( Figure 3 Since this bridging component will need to communicate with both FPGA 145 and SSD 120, FPGA 145 will still include downstream port 335 (or an alternative structure that enables communication with the bridging component): but downstream filter 360 can then be removed from FPGA 145.

[0149] As described above, the first embodiment of the inventive concept represents one possible combination of processor / FPGA / SSD implementations. Other implementations are also possible, and are described below as other embodiments of the inventive concept. Where there is no difference between the operations of a particular component (e.g., the operations of APM-F 340, scheduler 345, and acceleration engine 350-1 and acceleration engine 350-2), repeated descriptions of the operations of the particular component are omitted in subsequent embodiments of the inventive concept.

[0150] Second Example Embodiment

[0151] Figure 5 The second embodiment of the present invention is shown. Figure 1 FPGA 145 and Figure 1The second embodiment of the inventive concept is similar to the first embodiment of the inventive concept, except that the upstream port 330 further includes a filter: an upstream filter 505. The upstream filter 505 can filter PCIe transactions from the processor 110 (received via the upstream port 330) in a manner similar to the downstream filter 360. For example, the SSD 120 can define a filter similar to Figure 4 420, and uses the sideband bus 365, PCIe VDM, or any other mechanism to program the upstream filter 505 with the upstream FAR. Then, when the FPGA 145 receives a PCIe transaction from the processor 110 at the upstream port 330, the upstream filter 505 can check the PCIe transaction to see if it contains the address in the upstream FAR. If it does, the PCIe transaction is an accelerated instruction, and the FPGA 145 can route the PCIe transaction to the APM-F 340 for use by the processor, rather than passing the PCIe transaction to the SSD 120.

[0152] Figure 6 A method for accelerating the use of a second embodiment of the present invention is shown. Figure 1 Similar to the first embodiment of the inventive concept, the SSD 120 may request a block of host system memory addresses (which is larger than the address range required by the SSD 120 to support NVMe commands), and may receive the block 405 as a response to its BAR. The SSD 120 may then reserve a subset 410 of the blocks 405 for NVMe communications, may not use another subset 415, may reserve a third subset as a downstream FAR 420, and may reserve a fourth subset as an upstream FAR 605.

[0153] Return to Figure 5 , in the second embodiment of the inventive concept, the processor 110 still does not directly "see" the FPGA 145, because the FPGA 145 is not a discoverable device. However, the SSD 120 can inform the processor 110 of the upstream FAR 605 by programming the base address of the upstream FAR 605 in a dedicated register in the subset 410 for NVMe communication. After reading this dedicated register from the subset 410 for NVMe communication, the processor 110 may know the upstream FAR 605. The processor 110 can then send the acceleration instruction to the FPGA 145 (via the upstream port 330), instead of sending the dedicated command to the APM-S 315 of the SSD 120, which is then responsible for issuing the acceleration instruction to the FPGA 145.

[0154] exist Figure 5 In, with Figure 3, FPGA 145 is shown to include components that enable communication with processor 110 and SSD 120: specifically, upstream port 330, upstream filter 505, downstream port 335, and downstream filter 360. Figure 3 In the embodiment of the inventive concept shown in FIG. 1 , the components involved in filtering PCIe transactions can be removed from FPGA 145. Therefore, in the second embodiment of the inventive concept, the upstream port 330 and the upstream filter 505 can be placed in the first bridge component, and the downstream port 335 and the downstream filter 360 can be placed in the second bridge component, each bridge component controlling Figure 5 Alternatively, only one of these bridging components may be used (where FPGA 145 handles its own filtering for communications from another source), or a single bridging component may be used to handle all filtering for FPGA 145, regardless of the PCIe transaction source.

[0155] Third Example Embodiment

[0156] Figure 7 The third embodiment according to the present invention is shown. Figure 1 FPGA 145 and Figure 1 Components of the SSD 120 of the present inventive concept. In the third embodiment of the inventive concept, the SSD 120 includes a physical function (PF) 705 and a virtual function (VF) 710. (It should be noted that the third embodiment of the inventive concept is not intended to imply that other embodiments of the inventive concept do not include PFs and / or VFs, but that other embodiments of the inventive concept are not used in the same manner as in the third embodiment of the inventive concept.) PF 705 represents a single resource, such as a function provided by the SSD 120. VF 710 represents a function associated with a PF, but is "virtualized": that is, for a given PF, there may be more than one VF. However, VF 710 can "expose" FPGA 145 instead of representing a virtual function of SSD 120: that is, VF 710 can represent a function of FPGA 145. (VF 710 is still part of SSD 120 and not part of FPGA 145; but because VF 710 is dedicated to exposing FPGA 145, any memory addresses associated with VF 710 will not conflict with other devices.) Since PFs and VFs can be discovered by processor 110 when enumerating PCIe devices, processor 110 can indirectly discover FPGA 145 via VF 710 (even though it cannot discover it directly itself).

[0157] Figure 8 A method for accelerating the operation of the third embodiment according to the present invention is shown. Figure 14. In the embodiment of the present inventive concept, the SSD 120 may request a block of host system memory addresses (which is larger than the address range required by the SSD 120 to support NVMe commands). However, the SSD 120 may actually request two different blocks of host system memory addresses: a block 805 for the PF 705, and a block 810 for the VF 710. The downstream FAR 420 may be selected as a subset of the block 805 for the PF 705; the upstream FAR 605 may be the entire block 810 for the VF 710. (Upstream FAR 605 may be selected to be only a subset of block 810; however, since block 810 is dedicated for use by VF 710 and VF 710 may have no other purpose besides effectively exposing FPGA 145, any memory addresses in block 810 that are not used as part of upstream FAR 605 may be discarded.) Each of block 805 and block 810 has a separate BAR so that SSD 120 can be aware of the address range assigned to each block.

[0158] Return to Figure 7 , similar to the second embodiment of the inventive concept, the upstream port 330 also includes a filter: a VF filter 715. The VF filter 715 can filter PCIe transactions from the processor 110 (received via the upstream port 330) in a manner similar to the downstream filter 360. For example, the SSD 120 can use the sideband bus 365, the PCIe VDM, or any other mechanism to program the VF filter 715 with the upstream FAR 605. Then, when the FPGA 145 receives a PCIe transaction from the processor 110 at the upstream port 330, the VF filter 715 can check the PCIe transaction to see if it contains an address in the upstream FAR 605. If it does, the PCIe transaction is an accelerated instruction, and the FPGA 145 can route the PCIe transaction to the APM-F 340 for use by the processor, rather than transmitting the PCIe transaction to the SSD 120.

[0159] Alternatively, SSD 120 may program VF filter 715 with the identifier of VF 710. VF filter 715 may then check the PCIe transaction received from processor 110 at upstream port 330 to see if it includes the identifier of VF 710. If the PCIe transaction includes the identifier of VF 710, FPGA 145 may route the PCIe transaction to APM-F 340 for use by the processor, rather than passing the PCIe transaction to SSD 120.

[0160] exist Figure 7In the third embodiment of the present inventive concept, as in the previous embodiment of the present inventive concept, it is shown that the FPGA 145 includes a portion capable of implementing communication with the processor 110 and the SSD 120: specifically, the upstream port 330, the VF filter 715, the downstream port 335, and the downstream filter 360. As in the previously shown embodiment of the present inventive concept, the components involved in filtering of PCIe transactions can be removed from the FPGA 145. Therefore, in the third embodiment of the present inventive concept, the upstream port 330 and the VF filter 715 can be placed in the first bridge component, and the downstream port 335 and the downstream filter 360 can be placed in the second bridge component, each bridge component controlling Figure 7 Alternatively, only one of these bridging components may be used (where FPGA 145 handles its own filtering for communications from another source), or a single bridging component may be used to handle all filtering for FPGA 145, regardless of the PCIe transaction source.

[0161] Fourth Example Embodiment

[0162] One problem that using VF 710 to enable FPGA 145 exposes is that using VF may require support from the host operating system of processor 110. Although some operating systems support VF, not all operating systems support VF, and supporting VF involves its own complexity for the operating system. The fourth embodiment of the inventive concept solves the problem of using VF.

[0163] Fig. 9 A fourth embodiment according to the present invention is shown. Figure 1 FPGA 145 and Figure 1 120. In contrast to the third embodiment of the inventive concept, in the fourth embodiment of the inventive concept, the SSD 120 includes two PFs 705 and a PF 905. (Similar to the third embodiment of the inventive concept, the fourth embodiment of the inventive concept is not intended to imply that other embodiments of the inventive concept do not include PFs and / or VFs.) PF 705 continues to represent a single resource, such as a function provided by the SSD 120. On the other hand, PF 905 exposes FPGA 145. Similarly, since the PF can be discovered by the processor 110 when enumerating PCIe devices, the processor 110 can indirectly discover FPGA 145 via PF 905 (even if it cannot be discovered directly itself).

[0164] Fig.10 A method for accelerating the operation of the fourth embodiment according to the present invention is shown. Figure 1145. Similar to the third embodiment of the inventive concept, the SSD 120 may request two different blocks of host system memory addresses: block 1005 for the PF 705, and block 1010 for the PF 905. The downstream FAR 420 may be selected as a subset of block 1005 for the PF 705; the upstream FAR 605 may be the entire block 1010 for the PF 905. (Similarly, the upstream FAR 605 may be selected as only a subset of block 1010; but since block 1010 is dedicated to use by the PF 905 and the PF 905 may have no other purpose than to effectively expose the FPGA 145, any memory addresses in block 1010 that are not used as part of the upstream FAR 605 may be discarded.) Each of block 1005 and block 1010 has a separate BAR so that the SSD 120 can know the address range allocated for each block.

[0165] Return to Fig. 9 , similar to the third embodiment of the inventive concept, the upstream port 330 also includes a filter: PF filter 715 (PF filter 715 is almost the same as Figure 7 145 ). The PF filter 715 may filter PCIe transactions from the processor 110 (received via the upstream port 330) in a manner similar to the downstream filter 360. For example, the SSD 120 may program the PF filter 715 with the upstream FAR 605 using the sideband bus 365, the PCIe VDM, or any other mechanism. Then, when the FPGA 145 receives a PCIe transaction from the processor 110 at the upstream port 330, the PF filter 715 may check the PCIe transaction to see if it contains an address in the upstream FAR 605. If so, the PCIe transaction is an accelerated instruction, and the FPGA 145 may route the PCIe transaction to the APM-F 340 for use by the processor, rather than passing the PCIe transaction to the SSD 120.

[0166] Alternatively, SSD 120 may program PF filter 715 with the identifier of PF 905. PF filter 715 may then check the PCIe transaction received from processor 110 at upstream port 330 to see if it includes the identifier of PF 905. If the PCIe transaction includes the identifier of PF 905, FPGA 145 may route the PCIe transaction to APM-F 340 for use by the processor, rather than passing the PCIe transaction to SSD 120.

[0167] exist Fig. 9 In the fourth embodiment of the present inventive concept, as in the previous embodiment of the present inventive concept, it is shown that the FPGA 145 includes a portion capable of implementing communication with the processor 110 and the SSD 120: specifically, the upstream port 330, the PF filter 715, the downstream port 335, and the downstream filter 360. As in the previously shown embodiment of the present inventive concept, the components involved in filtering of PCIe transactions can be removed from the FPGA 145. Therefore, in the fourth embodiment of the present inventive concept, the upstream port 330 and the PF filter 715 can be placed in the first bridge component, and the downstream port 335 and the downstream filter 360 can be placed in the second bridge component, each bridge component controlling Fig. 9 Alternatively, only one of these bridging components may be used (where FPGA 145 handles its own filtering for communications from another source), or a single bridging component may be used to handle all filtering for FPGA 145, regardless of the PCIe transaction source.

[0168] Fifth Example Embodiment

[0169] As in the third embodiment of the inventive concept, the fourth embodiment of the inventive concept solves the problem of using VF. However, in order to use the fourth embodiment of the inventive concept, the SSD 120 needs to provide a PF 905 dedicated to the use of the FPGA 145. Not every SSD (or more generally, storage device) has an available PF that can be dedicated to the use of the FPGA 145. The fifth embodiment of the inventive concept provides a solution in which the SSD 120 does not need to provide multiple PFs.

[0170] Fig.11 A fifth embodiment according to the present invention is shown. Figure 1 FPGA 145 and Figure 1 Components of the SSD 120. Fig.11 In, SSD 120 returns to Figure 3 and Figure 5 The structure shown in FIG. 1 does not need to provide PF and / or VF. (Similarly, this does not mean that SSD 120 may not provide PF and / or VF, but it does not require SSD 120 to provide additional PF and / or VF.)

[0171] Compared to the first to fourth embodiments of the inventive concept, in the fifth embodiment of the inventive concept, the FPGA 145 is slightly different. Figure 3 , Figure 5 , Figure 7 as well as Fig. 9FPGA 145 may include endpoints 1105 and root ports 1110 (the term "port" may be used interchangeably with "root port"). Figure 3 , Figure 5 , Figure 7 as well as Fig. 9 The upstream port 330 and the downstream port 335 of the PCIe 1100 can be regarded as switches (which are effective transparent transmission devices), and the endpoint 1105 and the root port 1110 are terminal points for communication, which can be discovered through PCIe history. This fact means that the endpoint 1105 and the root port 1110 contain their own PCIe configuration space, see below Fig.12 However, since endpoint 1105 and root port 1110 are terminal points for communication, processor 110 and SSD 120 direct their communications to endpoint 1105 and root port 1110, respectively, rather than directing communications to each other.

[0172] Endpoint 1105 may include two PFs 1115 and PF 1120 (or alternatively, one PF and one VF: what is important is that endpoint 1105 includes two functions that can be distinguished from each other). When processor 110 sends a PCIe transaction to endpoint 1105, processor 110 may specify which PF to process. Similar to the third and fourth embodiments of the inventive concept described above, endpoint 1105 may use a tag (which is included in the PCIe transaction to identify the PF) or an address associated with the PCIe transaction (again, see below). Fig.12 APM-F 340 may be used to identify which PF to process. A PCIe transaction identifying PF 1115 may be considered to be destined for SSD 120 and may be sent by FPGA 145 to SSD 120 via root port 1110. A PCIe transaction identifying PF 1120 may be considered to contain an acceleration instruction and may be routed to APM-F 340.

[0173] The root port 1110 may include a downstream filter 360. The downstream filter 360 is similar to Figure 3 , Figure 5 , Figure 7 , Fig. 9 as well as Fig.11 The only significant difference is that the downstream filter does not filter based on the host system memory address, but rather filters based on the FPGA memory address, as shown below. Fig.12 The downstream filter 360 may be programmed with the downstream FAR 420 by the SSD 120 using the sideband bus 365, PCIe VDM, or any other desired mechanism.

[0174] FPGA 145 may also include a configuration monitor 1125. Since FPGA 145 does not replace SSD 120 but only provides additional functionality, and since FPGA 145 is inserted between processor 110 and SSD 120, it is important that processor 110 be able to view the functionality provided by SSD 120. More specifically, FPGA 145 should advertise performance that matches the PCIe configuration space of SSD 120. To this end, configuration monitor 1125 may replicate the PCIe configuration space of SSD 120, thereby providing processor 110 with the same PCIe configuration as SSD 120.

[0175] Fig.12 A method for accelerating the operation of the present invention according to a fifth embodiment of the present invention is shown. Figure 1 The memory usage of the instructions in the system. Fig.12 In FIG. 1 , SSD 120 does not request a block of host system memory addresses because SSD 120 is no longer communicating directly with processor 110. Instead, SSD 120 requests a block 1205 of FPGA memory addresses from address map 1210, which contains the configuration space for root port 1110. Block 1205 may include a subset 410 for NVMe communication with processor 110, and downstream FAR 420. Block 1205 may be identified by a BAR.

[0176] To provide a mechanism by which processor 110 can communicate with SSD 120 without FPGA 145, FPGA 145 may request a host system memory address from processor 110. In order to parallelize block 1205 as requested by SSD 120 of FPGA 145, FPGA 145 may request block 1215 for PF 1115, which should be at least as large as block 1205 (which arises because SSD 120 has requested block 1215 from processor 110). Fig.12 Block 1215 is labeled "Host-FPGA-SSD" in the figure to reflect that PCIe transactions using addresses in block 1215 are for communication between the host and SSD 120, but pass through FPGA 145. FPGA 145 can also request block 1220 for PF 1120, thereby providing a mechanism for processor 110 and APM-F 340 to communicate around acceleration instructions. Block 1215 and block 1220 can each be identified by two separate BARs. In another embodiment of the inventive concept, it is possible to use part of block 1220 as a downstream FAR 420 to facilitate communication between FPGA 145 and SSD 120.

[0177] exist Fig.11In the fifth embodiment of the inventive concept, as in the previous embodiment of the inventive concept, it is shown that the FPGA 145 includes a portion capable of implementing communication with the processor 110 and the SSD 120: specifically, the endpoint 1105, the PF 1115 and the PF 1120, the root port 1110, and the downstream filter 360. As in the previously shown embodiment of the inventive concept, the components involved in filtering of PCIe transactions can be removed from the FPGA 145. Therefore, in the fifth embodiment of the inventive concept, the endpoint 1105 and the PF 1115 and the PF 1120 can be placed in the first bridge component, and the root port 1110 and the downstream filter 360 can be placed in the second bridge component, each bridge component controlling Fig.11 Alternatively, only one of these bridging components may be used (where FPGA 145 handles its own filtering for communications from another source), or a single bridging component may be used to handle all filtering for FPGA 145, regardless of the PCIe transaction source.

[0178] Sixth Example Embodiment

[0179] The fifth embodiment of the inventive concept still relies on the downstream filter 360 to separate the acceleration instructions (between the SSD 120 and the FPGA 145) from the conventional PCIe transactions (between the processor 110 and the SSD 120). As in the sixth embodiment of the inventive concept, in the case where the SSD 120 includes a second endpoint, the downstream filter 360 can be removed.

[0180] Fig.13 The sixth embodiment according to the present invention is shown. Figure 1 FPGA 145 and Figure 1 Components of the SSD 120. Fig.13 In the example, the FPGA contains two root ports 1110 and 1305 instead of Fig.11 In the embodiment shown in FIG. 1 , only one root port 1110 is shown. Root port 1110 can be used for conventional PCIe transactions originating from processor 110 ; root port 1305 can be used for accelerated instructions and data exchanged between SSD 120 and FPGA 145 .

[0181] Since the accelerated instructions are naturally separated from the regular PCIe transactions from the processor 110 using different root ports 1110 and 1305, there is no need Figure 3 , Figure 5 , Figure 7 , Fig. 9 as well as Fig.11This fact means that SSD 120 reduces the burden of programming downstream filter 360 in FPGA 360, similar to how SSD 120 reduces the burden of programming downstream filter 360 in FPGA 360. Figure 5 , Figure 7 as well as Fig. 9 The burden of the upstream filter 505 and the upstream filter 715 (such as Fig.11 and Fig.13 The FPGA 145 in the FPGA 145 can distinguish between regular PCIe transactions and accelerated instructions based on the associated PF). The offset of such benefits is that the SSD 120 includes two endpoints 305 and endpoint 1310 to communicate with the root port 1110 and root port 1305 of the FPGA 145 respectively.

[0182] Since the SSD 120 includes two endpoints 305 and an endpoint 1310 in the sixth embodiment of the inventive concept, each of the endpoints 305 and the endpoint 1310 can request its own block of memory addresses from the FPGA 145. In addition, since each of the root port 1110 and the root port 1305 includes its own configuration space, the endpoints 305 and the endpoint 1310 of the SSD 120 can request blocks of memory addresses from different configuration spaces. Fig.14 Shows this situation.

[0183] Fig.14 A method for accelerating the use of a sixth embodiment according to the concept of the present invention is shown. Figure 1 The memory usage of the instructions in the system. Fig.14 , endpoint 305 of SSD 120 may request block 1405 from address map 1210, which contains the configuration space for root port 1110. However, since conventional PCIe transactions (between processor 110 and SSD 120) and accelerated instructions (between FPGA 145 and SSD 120) are naturally separated by the use of different root ports on FPGA 145 and different endpoints on SSD 120, SSD 120 does not need to request a block 1405 large enough to contain the downstream FAR. Therefore, without FPGA 145, block 1405 only needs to be as large as possible: that is, large enough to support NVMe communication between processor 110 and SSD 120. Block 1405 can be identified by BAR.

[0184] The endpoint 1310 of the SSD 120 can request a block of its own memory addresses from the address map 1410. However, since the root port 1305 and the endpoint 1310 are only used to exchange acceleration instructions in the sixth embodiment of the inventive concept, the entire address map 1410 can be used for such PCIe transactions: the endpoint 1310 is not required to request only a small block of the address map 1410.

[0185] As in the fifth embodiment of the inventive concept, PF 1115 may request block 1215 of host system memory addresses to manage PCIe transactions exchanged between processor 110 and SSD 120; block 1215 may be at least as large as block 1405. Similarly, PF 1120 may request block 1220 of host system memory addresses to manage acceleration instructions exchanged between processor 110 and FPGA 145. Block 1215 and block 1220 may each be identified by a BAR.

[0186] exist Fig.13 In the sixth embodiment of the inventive concept, as in the previous embodiment of the inventive concept, it is shown that the FPGA 145 includes a portion capable of implementing communication with the processor 110 and the SSD 120: specifically, the endpoint 1105, the PF 1115 and the PF 1120, and the root port 1110 and the root port 1305. As in the previously shown embodiment of the inventive concept, the components involved in filtering of PCIe transactions can be removed from the FPGA 145. Therefore, in the sixth embodiment of the inventive concept, the endpoint 1105 and the PF 1115 and the PF 1120 can be placed in the first bridge component, and the root port 1110 and the root port 1305 can be placed in the second bridge component, each bridge component controlling Fig.13 Alternatively, only one of these bridging components may be used (where FPGA 145 handles its own filtering for communications from another source), or a single bridging component may be used to handle all filtering for FPGA 145, regardless of the PCIe transaction source.

[0187] Seventh Example Embodiment

[0188] In the first six embodiments of the inventive concept, the FPGA 145 is shown using a single upstream port 330 (at Figure 5 , Figure 7 as well as Fig. 9 in) or a single endpoint 1105 (in Fig.11 and Fig.13 However, in the sixth embodiment of the inventive concept, there is no reason why the FPGA 145 should not include multiple endpoints like the SSD 120. The seventh and eighth embodiments of the inventive concept illustrate how the FPGA 145 can operate using multiple endpoints.

[0189] Fig.15 A seventh embodiment according to the present invention is shown. Figure 1 FPGA 145 and Figure 1The seventh embodiment of the inventive concept is similar to the sixth embodiment of the inventive concept, except that the FPGA 145 includes two endpoints 1105 and an endpoint 1505. Much like the endpoints 305 and 1310 of the SSD 120 can be used to distinguish between conventional host-to-SSD PCIe transactions and FPGA-to-SSD acceleration instructions, the endpoints 1105 and 1505 of the FPGA 145 can be used to distinguish between conventional host-SSD PCIe transactions and host-to-FPGA acceleration instructions. PCIe transactions received at the endpoint 1105 can be treated as conventional PCIe transactions and forwarded to the SSD 120 (via the root port 1110), whereas PCIe transactions received at the endpoint 1505 can be treated as acceleration instructions and forwarded to the APM-F 340 for processing.

[0190] Fig.16 A method for accelerating the operation of the seventh embodiment according to the present invention is shown. Figure 1 140 . For all intents and purposes, the memory usage according to the seventh embodiment of the inventive concept is consistent with the memory usage according to the sixth embodiment. The root port 1110 and the root port 1305 of the FPGA 145 each provide their own configuration space, and the endpoint 305 of the SSD 120 can request a block 1405 from the address map 1210 (as in the sixth embodiment of the inventive concept, the endpoint 1310 of the SSD 120 can request a block from the address map 1410, or the endpoint 1310 of the SSD 120 can use the entire address map 1410 for SSD-FPGA communication), and the block 1405 is identified by the BAR. The endpoint 1105 and the endpoint 1505 can request blocks 1215 and blocks 1220 of the host system memory address to manage PCIe transactions exchanged between the processor 110 and the SSD 120; the block 1215 can be at least as large as the block 1405. The block 1215 and the block 1220 can each be identified by a BAR.

[0191] exist Fig.15 In the seventh embodiment of the present inventive concept, it is shown that the FPGA 145 includes a portion capable of implementing communication with the processor 110 and the SSD 120: specifically, the endpoint 1105 and the endpoint 1505, as well as the root port 1110 and the root port 1305. As in the previously shown embodiment of the present inventive concept, the components involved in filtering of PCIe transactions can be removed from the FPGA 145. Therefore, in the seventh embodiment of the present inventive concept, the endpoint 1105 and the endpoint 1505 can be placed in the first bridge component, and the root port 1110 and the root port 1305 can be placed in the second bridge component, each bridge component controlling Fig.15Alternatively, only one of these bridging components may be used (where FPGA 145 handles its own filtering for communications from another source), or a single bridging component may be used to handle all filtering for FPGA 145, regardless of the PCIe transaction source.

[0192] Eighth Example Embodiment

[0193] Fig.17 The eighth embodiment according to the present invention is shown. Figure 1 FPGA 145 and Figure 1 The eighth embodiment of the inventive concept is similar to the fifth embodiment of the inventive concept, except that the FPGA 145 includes two endpoints 1105 and an endpoint 1505. Likewise, the endpoints 1105 and 1505 of the FPGA 145 can be used to distinguish between conventional host-SSD PCIe transactions and host-to-FPGA acceleration instructions. PCIe transactions received at the endpoint 1105 of the FPGA 145 can be treated as conventional PCIe transactions and forwarded to the SSD 120 (via the root port 1110), whereas PCIe transactions received at the endpoint 1505 of the FPGA 145 can be treated as acceleration instructions and forwarded to the APM-F 340 for processing.

[0194] Fig.18 A method for accelerating the operation of an eighth embodiment according to the present invention is shown. Figure 1 145. The memory usage of the eighth embodiment according to the inventive concept is similar to the memory usage of the fifth embodiment according to the inventive concept. With the root port 1110 being the only root port of the FPGA 145, the endpoint 305 of the SSD 120 may request the block 1205 from the address map 1210, which may include the downstream FAR 420. The endpoint 1105 and the endpoint 1505 of the FPGA 145 may then request the block 1215 and the block 1220 from the host system memory address (where the endpoint 1105 requests the block 1215 to be at least as large as the block 1205), which is identified by the BAR. The endpoint 1105 and the endpoint 1505 may request the block 1215 and the block 1220 of the host system memory address, thereby managing the PCIe transactions exchanged between the processor 110 and the SSD 120; the block 1215 may be at least as large as the block 1405. The block 1215 and the block 1220 may each be identified by two separate BARs. In yet another embodiment of the inventive concept, it is possible to use a portion of block 1220 as a downstream FAR 420 to facilitate communication between FPGA 145 and SSD 120 .

[0195] exist Fig.17In the eighth embodiment of the inventive concept, as in the previous embodiment of the inventive concept, it is shown that the FPGA 145 includes a portion capable of implementing communication with the processor 110 and the SSD 120: specifically, the endpoint 1105 and the endpoint 1505, the root port 1110, and the downstream filter 360. As in the previously shown embodiment of the inventive concept, the components involved in filtering of PCIe transactions can be removed from the FPGA 145. Therefore, in the eighth embodiment of the inventive concept, the endpoint 1105 and the endpoint 1505 can be placed in the first bridge component, and the root port 1110 and the downstream filter 360 can be placed in the second bridge component, each bridge component controlling Fig.17 Alternatively, only one of these bridging components may be used (where FPGA 145 handles its own filtering for communications from another source), or a single bridging component may be used to handle all filtering for FPGA 145, regardless of the PCIe transaction source.

[0196] As discussed above with reference to various embodiments of the inventive concept, the filtering functionality described as part of the upstream interface and / or downstream interface of FPGA 145 may be separated from FPGA 145 and handled by another component. Fig.19 A representative example of an embodiment according to the concept of the present invention is shown. Figure 1 The acceleration module 145 manipulates the bridge component of the filtering function.

[0197] exist Fig.19 , two bridge components 1905 and bridge component 1910 are shown. Bridge component 1905 can handle filtering of PCIe transactions received from processor 110, while bridge component 1910 can handle filtering of PCIe transactions received from SSD 120. Depending on whether the PCIe transaction contains an acceleration instruction, bridge component 1905 can send the PCIe transaction to FPGA 145 or SSD 120. Similarly, depending on whether the PCIe transaction contains an acceleration instruction, bridge component 1910 can send the PCIe transaction to FPGA 145 or processor 110. Fig.19 Specific implementations of bridge component 1905 and bridge component 1910 are not shown in , as the implementations are similar to those shown above as part of the upstream interface and downstream interface of FPGA 145 .

[0198] In some embodiments of the inventive concept, both the bridge component 1905 and the bridge component 1910 may be used. In other embodiments of the inventive concept, only one bridge component 1905 or bridge component 1910 is used, wherein the functionality of the other bridge component may be retained in the FPGA 145. In still other embodiments, both the bridge component 1905 and the bridge component 1910 may be included in a single component rather than as separate components.

[0199] Various embodiments of the inventive concept have been described so far, describing the data flow between processor 110, FPGA 145, and SSD 120. In the remainder of this document, all filtering functions are attributed to FPGA 145, but when filtering can be transferred to Fig.19 The bridge component 1905 and / or bridge component 1910 and how to convert to Fig.19 The bridging component 1905 and / or the bridging component 1910 should be apparent. FIG. 20A to FIG. 20B The embodiment according to the present invention is shown Figure 1 processor, Figure 1 FPGA 145 with Figure 1 The communication between the SSDs 120. Fig. 20A , a data flow according to a first embodiment of the inventive concept (and possibly other embodiments of the inventive concept) is shown. Processor 110 may send a PCIe transaction 2005 to SSD 120. PCIe transaction 2005 may include a dedicated command 2010. PCIe transaction 2005 may be transmitted to SSD 120 (via FPG 145). APM-S 315 may then generate an acceleration instruction 2015, which may be included in a PCIe transaction 2020, which SSD 120 may then send to FPGA 145. FPGA 145 and SSD 120 may also exchange acceleration data, such as Fig. 20A as shown in .

[0200] After completing the acceleration instruction 2015, the FPGA 145 may send the result 2030 back to the SSD 120, which in turn may forward the result 2030 to the processor 110 (shown as result 2035). Alternatively, the FPGA 145 may send the result 2040 directly to the processor 110, emulating the result 2035 from the SSD 120.

[0201] On the contrary, Fig. 20B In the second to eighth embodiments of the present invention, the processor 110 may directly send the acceleration instruction 2045 to the FPGA 145 as a PCIe transaction 2005. The FPGA 145 and the SSD 120 may exchange acceleration data, such as Fig. 20B Finally, FPGA 145 may send result 2040 back to processor 110.

[0202] Fig.21 A flow chart illustrating an example process for FPGA 145 to process PCIe transactions according to an embodiment of the inventive concept. Fig.21 provides a high-level view; subsequent figures provide more detailed example flow charts of the operation of FPGA 145. Fig.21 In block 2105, FPGA 145 may receive a PCIe transaction from a device. Such a PCIe transaction may be FIG. 20A to FIG. 20B 15 or PCIe transaction 2045, and the device may be the processor 110 or the SSD 120. At block 2110, the FPGA 145 may determine whether the PCIe transaction includes an acceleration instruction. At block 2115, the FPGA 145 may test to see if the PCIe transaction includes an acceleration instruction. If so, then at block 2120, the acceleration instruction may be processed by the APM-F 340; otherwise, at block 2125, the PCIe transaction may be transmitted to another device (if the PCIe transaction is received from the processor 110, then the PCIe transaction may be transmitted to the SSD 120, and vice versa). It should be noted that processing the acceleration instruction by the APM-F 340 may involve communicating with the SSD 120 to receive application data to be processed by the acceleration instruction.

[0203] FIG. 22A to FIG. 22C A flow chart showing a more detailed example process for FPGA 145 to process PCIe transactions according to an embodiment of the inventive concept. Fig.22A , at block 2203, the FPGA 145 may receive the downstream FAR 420 from the SSD 120. At block 2206, the FPGA 145 may associate the downstream FAR 420 with the downstream filter 360. It should be noted that such association may occur automatically if the SSD 120 programs the downstream FAR 420 into the downstream filter 360 via the sideband bus 365, or may require the FPGA 145 to take effective steps (e.g., when the SSD 120 sends a PCIe VDM to the FPGA 145 containing the downstream FAR 420). It should also be noted that in some embodiments of the inventive concept, blocks 2203 and 2206 may be omitted, as indicated by dashed line 2209. In some embodiments of the inventive concept, the downstream FAR 420 may be provided by the FPGA 145 itself.

[0204] At block 2212, the FPGA 145 may receive the upstream FAR 605 from the SSD 120, and at block 2215, the FPGA 145 may associate the upstream FAR 605 with the upstream filter 505. It should be noted that such association may occur automatically if the SSD 120 programs the upstream FAR 605 into the upstream filter 505 via the sideband bus 365, or may require the FPGA 145 to take active steps (e.g., when the SSD 120 sends a PCIe VDM to the FPGA 145 containing the upstream FAR 605). In some embodiments of the inventive concept, the upstream FAR 605 may be provided by the FPGA 145 itself.

[0205] Alternatively, at block 2218, FPGA 145 may receive from SSD 120 an identifier of a PF or VF to expose FPGA 145, and at block 2221, FPGA 145 may associate the PF / VF identifier with upstream filter 505. Likewise, such association may occur automatically if SSD 120 programs the PF / VF identifier into upstream filter 505 via sideband bus 365, or may require FPGA 145 to take active steps (e.g., when SSD 120 sends a PCIe VDM to FPGA 145 containing the PF / VF identifier).

[0206] It should be noted that in some embodiments of the inventive concept, box 2212 , box 2215 , box 2218 , and box 2221 may be omitted, as indicated by dashed line 2224 .

[0207] At block 2227, the configuration monitor 1125 may determine the configuration of the endpoint 305 of the SSD 120, and at block 2230, the configuration module 1125 may replicate the configuration at the endpoint 1105 of the FPGA 145, thereby presenting the same functionality as the SSD 120 to the processor 110. In some embodiments of the inventive concept, blocks 2227 and 2230 may be omitted, as indicated by dashed line 2233.

[0208] Once FPGA 145 has been properly configured, then at block 2236 ( Fig. 22B ), FPGA 145 may receive (via upstream port 330 or endpoint 1105, depending on the embodiment of the inventive concept) from processor 110 Fig. 20B PCIe transaction 2005. At block 2239, FPGA 145 may determine Fig. 20B Does PCIe transaction 2005 contain Fig. 20B If included, then at block 2242, the APM-F 340 may process Fig. 20B The acceleration instruction 2045, and at box 2245, APM-F340 may Fig. 20B The result 2040 is sent to the processor 110. Otherwise, if Fig. 20B PCIe Transaction 2005 does not include Fig. 20B If the acceleration instruction 2045 is received, then at block 2248, the FPGA 145 may (via the downstream port 335 or the root port 1110, depending on the embodiment of the inventive concept) Fig. 20B The PCIe transaction 2005 is transmitted to SSD 120.

[0209] In box 2251 ( Fig. 22C), FPGA 145 may receive (via downstream port 335 or root port 1110, depending on the embodiment of the inventive concept) Fig. 20A PCIe transaction 2020. At block 2254, FPGA 145 may determine Fig. 20A Does PCIe Transaction 2020 include Fig. 20A of accelerated instruction 2015. If Fig. 20A PCIe Transactions 2020 includes Fig. 20A The acceleration instruction 2015, then at block 2257, the APM-F 340 may process Fig. 20A The acceleration instruction 2015, and at box 2260, APM-F340 may Fig. 20A The result 2040 is sent to SSD 120. Otherwise, if Fig. 20A If the PCIe transaction 2020 does not include the acceleration instruction 2015, then at block 2263, the FPGA 145 may (depending on the embodiment of the inventive concept via the upstream port 330 or the endpoint 1105) Fig. 20A The PCIe transaction 2020 is forwarded to the processor 110.

[0210] FIG. 23A to FIG. 23B 145 is shown for determining the signal from the processor 110 according to an embodiment of the inventive concept. Fig. 20B Does PCIe transaction 2005 contain Fig. 20B A flowchart of an example process of accelerating instruction 2045. FIG. 23A to FIG. 23B Three possible tests are shown that may be used independently or collectively depending on the embodiment of the inventive concept. In embodiments of the inventive concept that use more than one test, if any of the independent tests are satisfied, then it may be determined that Fig. 20B PCIe Transaction 2005 contains Fig. 20B The acceleration instruction 2045. Fig.23A In block 2305, FPGA 145 may determine Fig. 20B The address associated with the PCIe transaction 2005 contains the address in the upstream FAR 605. At block 2310, FPGA 145 may determine Fig. 20B Whether the PCIe transaction 2005 includes an identifier of a PF or VF associated with the upstream filter 715. At block 2315, the FPGA 145 may determine whether a PCIe transaction 2005 is received at a port dedicated to accelerating instructions (e.g., endpoint 1505). Fig. 20B PCIe Transaction 2005. If any of these test results indicate Fig. 20B PCIe Transaction 2005 contains Fig. 20B The acceleration instruction 2045, then in box 2320 ( Fig. 23B), FPGA 145 knows Fig. 20B PCIe Transaction 2005 contains Fig. 20B otherwise, at block 2325, FPGA 145 knows Fig. 20B PCIe Transaction 2005 does not include Fig. 20B Acceleration instruction 2045.

[0211] Fig.24 FIG. 1 shows a method for FPGA 145 to determine the SSD 120 according to an embodiment of the present inventive concept. Fig. 20A Does PCIe Transaction 2020 include Fig. 20A A flow chart of an example process of the Accelerate Instruction 2015. Fig.24 Two possible tests are shown that can be used independently or together depending on the embodiment of the inventive concept. In embodiments of the inventive concept that use more than one test, if any of the independent tests are satisfied, then it can be determined that Fig. 20A PCIe Transactions 2020 includes Fig. 20A Accelerated instruction in 2015. Fig.24 In block 2405, FPGA 145 may determine Fig. 20A The address associated with the PCIe transaction 2020 contains the address in the downstream FAR 420. At block 2410, the FPGA may determine whether a PCIe transaction 2020 associated with the downstream FAR 4 ... Fig. 20A PCIe Transactions 2020. If any of these test results indicate Fig. 20A PCIe Transactions 2020 includes Fig. 20A The acceleration instruction 2015, then at box 2415, FPGA145 knows Fig. 20A PCIe Transactions 2020 includes Fig. 20A otherwise, at block 2420, FPGA 145 knows Fig. 20A PCIe Transaction 2020 does not include Fig. 20A Accelerated instruction 2015.

[0212] Fig.25 The present invention is shown in an embodiment of the present invention. Fig.19 The first bridge component 1905 determines the Figure 1 A flowchart of an example process of determining whether a PCIe transaction of the processor 110 includes an acceleration instruction. Fig.25 In block 2505, the first bridging component 1905 may be Figure 1The processor 110 receives the PCIe transaction. At block 2510, the first bridging component 1905 may determine whether the PCIe transaction is an acceleration instruction. If so, then at block 2515, the first bridging component 1905 may forward the PCIe transaction / acceleration instruction to Figure 1 Otherwise, at block 2520, the first bridging component 1905 may forward the PCIe transaction to Figure 1 SSD 120.

[0213] Fig.26 The present invention is shown in an embodiment of the present invention. Fig.19 The second bridge component 1910 determines the Figure 1 A flowchart of an example process for determining whether a PCIe transaction of the storage device 120 includes an acceleration instruction. Fig.26 In block 2605, the second bridging component 1910 may be Figure 1 The SSD 120 receives the PCIe transaction. At block 2610, the second bridging component 1910 may determine whether the PCIe transaction is an acceleration instruction. If so, then at block 2615, the second bridging component 1910 may forward the PCIe transaction / acceleration instruction to Figure 1 Otherwise, at block 2620, the second bridging component 1910 may forward the PCIe transaction to Figure 1 The processor 110 of FIG.

[0214] FIG. 27A to FIG. 27C A flow chart illustrating an example process for SSD 120 to process PCIe transactions according to an embodiment of the inventive concept. Fig.27A , at block 2705, the SSD 120 may request a block of memory addresses. It should be noted that the SSD 120 may request the block of memory addresses from the host system memory as in the first to fourth embodiments of the inventive concept or from the configuration space of the root port of the FPGA 145 as in the fifth to eighth embodiments of the inventive concept. At block 2710, the SSD 120 may select a subset of the block of memory addresses to use as the downstream FAR 420, and at block 2715, the SSD 120 may program the downstream filter 360 with the downstream FAR 420 using the sideband bus 365, PCIe VDM, or any other desired mechanism. It should be noted that in some embodiments of the inventive concept, blocks 2705 to 2715 may be omitted, as indicated by dashed line 2720 (dashed line 2720 also skips Fig.27B As shown in some boxes shown in .

[0215] In box 2725 ( Fig.27B), the SSD 120 may select a subset of the blocks of memory addresses to use as the upstream FAR 605, and at block 2730, the SSD 120 may program the upstream filter 505 with the upstream FAR 605 using the sideband bus 365, PCIe VDM, or any other desired mechanism. Alternatively, at block 2735, the SSD 120 may expose its own capabilities using the PF 705. Next, at block 2740, the SSD 120 may expose the FPGA 145 using the PF 905 or the VF 710, and at block 2745, the SSD 120 may program the upstream filter 505 with the identifier of the PF 905 or the VF 710 using the sideband bus 365, PCIe VDM, or any other desired mechanism. It should be noted that in some embodiments of the inventive concept, blocks 2725 to 2745 may be omitted, as indicated by dashed line 2750.

[0216] At block 2755, SSD 120 may receive a PCIe transaction from FPGA 145. Such a PCIe transaction may be Fig. 20A PCIe transaction 2005 (forwarded from processor 110 by FPGA 145), or it may be FIG. 20A to FIG. 20B Regardless of the source of the PCIe transaction, at block 2760 ( Fig.27C ), HIL 310 may determine whether the PCIe transaction includes an acceleration instruction. If so, then at block 2765, HIL 310 may forward the PCIe transaction (or unencapsulated acceleration instruction) to APM-S 315 for processing. APM-S 315 may generate a response to the acceleration instruction, which may be Fig. 20A The response may be acceleration instruction 2015 (if the PCIe transaction originates from processor 110), or it may be acceleration data 2025 (if the PCIe transaction originates from APM-F 340 of FPGA 145). In either case, at block 2770, APM-S 315 may send the response to FPGA 145.

[0217] On the other hand, if the PCIe transaction is not an accelerated instruction, then at block 2775, SSD 120 may determine whether the PCIe transaction is an accelerated instruction. Fig. 20A If yes, then at block 2780, SSD 120 may Fig. 20A The result 2035 is forwarded to the processor 110 (via the endpoint 305 of the SSD 120 and the FPGA 145). If the PCIe transaction is not Fig. 20A Result 2030 , then at box 2785 , SSD 120 can process PCIe transactions on data stored on SSD 120 normally.

[0218] FIG. 28A to FIG. 28B A flowchart illustrating an example process for SSD 120 to determine whether a PCIe transaction from FPGA 145 includes an acceleration instruction according to an embodiment of the inventive concept.

[0219] FIG. 28A to FIG. 28B Three possible tests are shown that may be used independently or collectively depending on the embodiment of the inventive concept. In embodiments of the inventive concept that use more than one test, if any of the independent tests are satisfied, then it may be determined that the PCIe transaction contains an acceleration instruction. Fig.28A In the embodiment of the present invention, at block 2805, SSD 120 may determine whether the PCIe transaction contains a special command from processor 110 (which indicates that SSD 120 should issue an acceleration instruction to FPGA 145). At block 2810, SSD 120 may determine whether the PCIe transaction originates from APM-F 340, which may be required when APM-F 340 is requesting FIG. 20A to FIG. 20B The acceleration data of APM-F 340 will be released in 2025 or Fig. 20A The result 2030 of block 2810 may occur when the PCIe transaction is sent to SSD 120. The test of block 2810 may be performed in any desired manner: for example, the PCIe transaction may include a tag to indicate that the PCIe transaction is an acceleration instruction, or the PCIe transaction may be associated with an address in downstream FAR 420. At block 2815, SSD 120 may determine whether the PCIe transaction was received at a port dedicated to acceleration instructions (e.g., endpoint 1310). If the results of any of these tests indicate that the PCIe transaction includes an acceleration instruction, then at block 2820 ( Fig.28B ), SSD 120 knows that the PCIe transaction includes an acceleration instruction; otherwise, at box 2825, SSD 120 knows that the PCIe transaction does not include an acceleration instruction.

[0220] exist Figures 21 to 28B In the accompanying drawings, some embodiments of the inventive concept are shown. However, those skilled in the art will recognize that other embodiments of the inventive concept are possible by changing the order of the blocks, by omitting blocks, or by including links not shown in the accompanying drawings. In addition, although specific operations are described as being performed by specific components, embodiments of the inventive concept may support other components that perform the described operations. All such variations of the flow chart are considered embodiments of the inventive concept, whether explicitly described or not.

[0221] Embodiments of the inventive concept provide technical advantages over the prior art. Figure 1 The acceleration module 145 is introduced to Figure 1 Machine 105, Figure 1 The processor 110 may share the Figure 1Since such commands typically involve processing large amounts of data that may subsequently be discarded, sharing the workload Figure 1 The acceleration module 145 works to avoid transferring data from Figure 1 The storage device 120 is loaded into Figure 1 of the memory 140, and avoids the possibility of needing to release Figure 1 Some space within memory 140.

[0222] Various embodiments of the inventive concept also support the use of Figure 1 Different variations of storage devices 120 can be used. Single-port and dual-port storage devices can be used, as well as supporting additional PFs and / or VFs to expose Figure 1 A storage device of the acceleration module 145.

[0223] Various embodiments of the inventive concept further support the use of processors that provide different capabilities. Figure 1 If the operating system of machine 105 supports VF, then VF can be used to expose Figure 1 If the acceleration module 145 is used; otherwise, PF can be used. Figure 1 The processor 110 can directly communicate with Figure 1 If the acceleration module 145 communicates with the accelerator module 145, an embodiment of the inventive concept supporting such communication may be used; otherwise, Figure 1 The processor 110 may send all acceleration instructions to Figure 1 The storage device 120 is reserved for Figure 1 The storage device 120 requests Figure 1 The acceleration module 145 executes the acceleration instruction.

[0224] The following discussion is intended to provide a brief general description of one or more suitable machines in which certain aspects of the inventive concepts may be implemented. One or more machines may be controlled at least in part by input from conventional input devices such as a keyboard, a mouse, and by instructions received from another machine, interaction with a virtual reality (VR) environment, biometric feedback, or other input signals. As used herein, the term "machine" is intended to broadly encompass a single machine, a virtual machine, or a system of communicatively coupled machines, virtual machines, or devices operating together. Exemplary machines include computing devices such as personal computers, workstations, servers, portable computers, handheld devices, phones, tablet computers, etc., and transportation devices such as private or public transportation, such as cars, trains, taxis, etc.

[0225] One or more machines may include embedded controllers, such as programmable or non-programmable logic devices or arrays, application specific integrated circuits (ASICs), embedded computers, smart cards, etc. One or more machines may utilize one or more connections to one or more remote machines, such as through a network interface, a modem, or other communication coupling. The machines may be interconnected by means of physical and / or logical networks, such as an intranet, the Internet, a local area network, a wide area network, etc. Those skilled in the art will understand that network communications may utilize various wired and / or wireless short-range or long-range carriers and protocols, including radio frequency (RF), satellite, microwave, Institute of Electrical and Electronics Engineers (IEEE) 802.11, Optical, infrared, cable, laser, etc.

[0226] Embodiments of the inventive concept may be described with reference to or in conjunction with associated data including functions, procedures, data structures, applications, etc., which when accessed by a machine causes the machine to perform a task or define an abstract data type or low-level hardware environment. The associated data may be stored, for example, in volatile memory and / or non-volatile memory, such as RAM, ROM, etc., or other storage devices and their associated storage media, including hard disk drives, floppy disks, optical storage devices, magnetic tapes, flash memory, memory sticks, digital video disks, biometric storage devices, etc. The associated data may be transmitted in the form of packets, serial data, parallel data, propagated signals, etc. over a transmission environment including a physical network and / or a logical network, and may be used in a compressed format or an encrypted format. The associated data may be used in a distributed environment and stored locally and / or remotely for machine access.

[0227] Embodiments of the inventive concepts may include tangible, non-transitory machine-readable media including instructions executable by one or more processors, the instructions including instructions for performing elements of the inventive concepts as described herein.

[0228] The various operations of the methods described above may be performed by any suitable means capable of performing the operations, such as various hardware and / or software components, circuits and / or modules. Software may include an ordered list of executable instructions for implementing logical functions, and may be implemented in any "processor-readable medium" for use with or in combination with an instruction execution system, device or apparatus, such as a single-core processor or a multi-core processor or a system containing a processor.

[0229] The steps of the methods or algorithms and functions described in conjunction with the embodiments disclosed herein may be implemented directly in hardware, in software modules executed by a processor, or in a combination of the two. If implemented in software, the functions may be stored as one or more instructions or codes on or transmitted via a tangible, non-transitory computer-readable medium. The software module may reside in a random access memory (RAM), a flash memory, a read-only memory (ROM), an electrically programmable ROM (EPROM), an electrically erasable programmable ROM (EEPROM), a register, a hard disk, a removable disk, a CD ROM, or any other form of storage medium known in the art.

[0230] After describing and illustrating the principles of the inventive concept with reference to the illustrated embodiments, it will be appreciated that the arrangement and details of the illustrated embodiments can be modified without departing from this principle and can be combined in any desired manner. Moreover, although the foregoing discussion focuses on specific embodiments, other configurations are contemplated. Specifically, even if expressions such as "embodiments according to the inventive concept" are used herein, these words generally represent reference to embodiment possibilities, and are not intended to limit the inventive concept to specific embodiment configurations. As used herein, these terms can refer to the same or different embodiments combined into other embodiments.

[0231] The foregoing illustrative embodiments should not be construed as limiting the inventive concepts thereof. Although several embodiments have been described, those skilled in the art will readily appreciate that many modifications to these embodiments are possible without materially departing from the novel teachings and advantages of the present disclosure. Therefore, all such modifications are intended to be included within the scope of the inventive concepts as defined in the claims.

[0232] Embodiments of the inventive concept can be extended to the following statements without limitation:

[0233] Statement 1. An embodiment of the inventive concept comprises a system comprising:

[0234] a processor, the processor executing an application program;

[0235] a memory that stores data used by applications running on the processor;

[0236] An upstream interface for communicating with the processor;

[0237] A downstream interface for communicating with a storage device;

[0238] an acceleration module implemented using hardware and including an acceleration platform manager (APM-F) to execute acceleration instructions; and

[0239] A storage device comprising:

[0240] An endpoint of the storage device, used for communicating with the acceleration module;

[0241] a controller for managing the operation of the storage device;

[0242] Storage, which is used to store application data of the application; and

[0243] The Accelerated Platform Manager for Storage (APM-S) is used to assist APM-F in executing acceleration instructions.

[0244] The processor, the acceleration module, and the storage device communicate via a peripheral component interconnect express (PCIe) bus, and

[0245] The acceleration module supports the execution of acceleration instructions on the application data on the storage device for the application program without loading the application data into the memory.

[0246] Statement 2. An embodiment of the inventive concept comprises a system according to Statement 1, said system further comprising:

[0247] A first bridging component, comprising an upstream interface, the first bridging component bridging communications between the processor and the acceleration module; and

[0248] The second bridging component includes a downstream interface, and the second bridging component bridges the communication between the acceleration module and the storage device.

[0249] Statement 3. An embodiment of the inventive concept comprises a system according to Statement 1, wherein:

[0250] Using a field programmable gate array (FPGA) to implement the acceleration module;

[0251] The acceleration module includes an upstream interface and a downstream interface; and

[0252] The storage device includes a solid state drive (SSD).

[0253] Statement 4. An embodiment of the inventive concept comprises a system according to Statement 3, wherein the APM-F and APM-S communicate using a downstream interface and an endpoint of the SSD involving application data to be used with the acceleration instructions.

[0254] Statement 5. An embodiment of the inventive concept comprises a system according to Statement 3, wherein APM-F and APM-S communicate using messages.

[0255] Statement 6. An embodiment of the inventive concept includes a system according to Statement 3, wherein the processor can send a PCIe transaction to the SSD, the PCIe transaction including a transaction layer packet (TLP) encoding a command using a non-volatile memory express (NVMe) protocol.

[0256] Statement 7. An embodiment of the inventive concept comprises a system according to Statement 3, wherein the FPGA further comprises:

[0257] Speed ​​up your engine; and

[0258] A run-time scheduler is used to schedule acceleration instructions using the acceleration engine.

[0259] Statement 8. An embodiment of the inventive concept comprises a system according to Statement 3, wherein the SSD comprises an FPGA.

[0260] Statement 9. An embodiment of the inventive concept comprises a system according to Statement 3, wherein:

[0261] The upstream interface contains the upstream port;

[0262] The downstream interface includes the downstream port;

[0263] The FPGA is operated to forward a first PCIe transaction received from the processor at the upstream port to the SSD;

[0264] The FPGA includes a downstream filter associated with the downstream port, the downstream filter being operative to intercept an acceleration instruction received from the SSD and transmit the acceleration instruction to the APM-F, the acceleration instruction being associated with a downstream filter address range (FAR); and

[0265] The FPGA is operated to forward a second PCIe transaction to the processor that is not associated with a downstream FAR received from the SSD at the downstream port.

[0266] Statement 10. An embodiment of the inventive concept comprises a system according to Statement 9, wherein the acceleration instruction is generated by APM-S.

[0267] Statement 11. An embodiment of the inventive concept includes a system according to Statement 10, wherein the SSD further includes a host interface logic (HIL) to intercept a special command received from a processor (the special command includes an acceleration instruction), and forward the special command to the APM-S to trigger the APM-S to generate the acceleration instruction.

[0268] Statement 12. An embodiment of the inventive concept comprises a system according to Statement 11, wherein the specialized command originates from an Acceleration Services Manager (ASM) running on the processor.

[0269] Statement 13. An embodiment of the inventive concept comprises a system according to Statement 9, wherein:

[0270] The SSD is operative to request blocks of host system addresses from the processor; and

[0271] The controller is operative to select a subset of the block of host system addresses as downstream FARs.

[0272] Statement 14. An embodiment of the inventive concept comprises a system according to Statement 13, wherein the controller is operative to program the downstream filter with the downstream FAR.

[0273] Statement 15. An embodiment of the inventive concept comprises a system according to Statement 14, wherein the controller is operative to program the downstream filter with the downstream FAR using the sideband bus.

[0274] Statement 16. An embodiment of the inventive concept comprises a system according to Statement 15, wherein the sideband bus is from a device comprising an inter-integrated circuit (I 2 C) A collection of buses and the System Management Bus (SMBus).

[0275] Statement 17. An embodiment of the inventive concept includes a system according to Statement 14, wherein the controller is operative to use PCIe Vendor Defined Messages (VDMs) to program the downstream filter with the downstream FAR.

[0276] Statement 18. An embodiment of the inventive concept comprises a system according to Statement 9, wherein:

[0277] APM-F is operated to send the result to APM-S via the downstream port and the endpoint of the SSD; and

[0278] The controller is operative to forward the results to the processor via an endpoint of the SSD.

[0279] Statement 19. An embodiment of the inventive concept comprises a system according to Statement 9, wherein the APM-F is operative to send the result to the processor via the upstream port.

[0280] Statement 20. An embodiment of the inventive concept comprises a system according to Statement 9, wherein:

[0281] The FPGA further includes an upstream filter associated with the upstream port, the upstream filter being operative to intercept a second acceleration instruction received from the processor and transmit the second acceleration instruction to the APM-F, the second acceleration instruction being associated with the upstream FAR; and

[0282] The FPGA is operated to forward a third PCIe transaction that is not associated with an upstream FAR received from the processor at the upstream port to the SSD.

[0283] Statement 21. An embodiment of the inventive concept comprises a system according to Statement 20, wherein the second acceleration instruction originates from an ASM running on the processor.

[0284] Statement 22. An embodiment of the inventive concept comprises a system according to Statement 20, wherein:

[0285] The SSD is operative to request blocks of host system addresses from the processor; and

[0286] The controller is operative to select a first subset of the block of host system addresses as downstream FARs and to select a second subset of the block of host system addresses as upstream FARs.

[0287] Statement 23. An embodiment of the inventive concept includes a system according to Statement 22, wherein the block of host system addresses includes a special register accessible by an ASM running on the processor, the special register identifying an upstream FAR.

[0288] Statement 24. An embodiment of the inventive concept includes a system according to Statement 22, wherein the controller is operative to program the downstream filter with the downstream FAR and to program the upstream filter with the upstream FAR.

[0289] Statement 25. An embodiment of the inventive concept includes a system according to Statement 24, wherein the controller is operative to use the sideband bus to program the downstream filter with the downstream FAR and to program the upstream filter with the upstream FAR.

[0290] Statement 26. An embodiment of the inventive concept comprises a system according to Statement 25, wherein the sideband bus is from a device comprising I 2 C-bus and SMBus.

[0291] Statement 27. An embodiment of the inventive concept includes a system according to Statement 24, wherein the controller is operative to use the PCIe VDM to program the downstream filter with the downstream FAR and to program the upstream filter with the upstream FAR.

[0292] Statement 28. An embodiment of the inventive concept comprises a system according to Statement 20, wherein:

[0293] APM-F is operated to send the result to APM-S via the downstream port and the endpoint of the SSD; and

[0294] The controller is operative to forward the results to the processor via an endpoint of the SSD.

[0295] Statement 29. An embodiment of the inventive concept comprises a system according to Statement 20, wherein the APM-F is operative to send the result to the processor via the upstream port.

[0296] Statement 30. An embodiment of the inventive concept comprises a system according to Statement 9, wherein:

[0297] The SSD includes a physical function (PF) and a virtual function (VF), the PF being operated to expose the SSD and the VF being operated to expose the FPGA;

[0298] The FPGA further includes an upstream filter associated with the upstream port, the upstream filter being operative to intercept a second acceleration instruction received from the processor and to transmit the second acceleration instruction to the APM-F; and

[0299] The FPGA is operated to forward a third PCIe transaction to the SSD that is not intercepted by the upstream filter received from the processor at the upstream port.

[0300] Statement 31. An embodiment of the inventive concept comprises a system according to Statement 30, wherein the second acceleration instruction originates from an ASM running on the processor.

[0301] Statement 32. An embodiment of the inventive concept comprises a system according to Statement 30, wherein:

[0302] The PF is operable to request a first block of host system addresses from the processor;

[0303] The controller is operative to select a first subset of the block of host system addresses as downstream FARs.

[0304] Statement 33. An embodiment of the inventive concept comprises a system according to Statement 32, wherein the controller is operative to program the downstream filter with the downstream FAR.

[0305] Statement 34. An embodiment of the inventive concept comprises a system according to Statement 33, wherein the controller is operative to program the downstream filter with the downstream FAR using the sideband bus.

[0306] Statement 35. An embodiment of the inventive concept comprises a system according to Statement 34, wherein the sideband bus is from a device comprising I 2 C-bus and SMBus.

[0307] Statement 36. An embodiment of the inventive concept includes a system according to Statement 33, wherein the controller is operative to use a PCIe VDM to program a downstream filter with a downstream FAR.

[0308] Statement 37. An embodiment of the inventive concept comprises a system according to Statement 30, wherein:

[0309] A second speed up instruction is associated with the upstream FAR; and

[0310] The upstream filter is operative to intercept a second speed-up instruction associated with the upstream FAR.

[0311] Statement 38. An embodiment of the inventive concept includes a system according to Statement 37, wherein the VF is operative to request a second block of host system addresses from the processor as an upstream FAR.

[0312] Statement 39. An embodiment of the inventive concept comprises a system according to Statement 38, wherein the controller is operative to program the upstream filter with the upstream FAR.

[0313] Statement 40. An embodiment of the inventive concept comprises a system according to Statement 39, wherein the controller is operative to program the upstream filter with the upstream FAR using the sideband bus.

[0314] Statement 41. An embodiment of the inventive concept comprises a system according to Statement 40, wherein the sideband bus is from a system comprising I 2 C-bus and SMBus.

[0315] Statement 42. An embodiment of the inventive concept includes the system of Statement 39, wherein the controller is operative to use the PCIe VDM to program the upstream filter with the upstream FAR.

[0316] Statement 43. An embodiment of the inventive concept comprises a system according to Statement 30, wherein:

[0317] The second acceleration instruction includes an identifier of the VF; and

[0318] The upstream filter is operative to intercept a second acceleration instruction associated with the identifier of the VF.

[0319] Statement 44. An embodiment of the inventive concept comprises a system according to Statement 43, wherein the controller is operative to program the upstream filter with an identifier of the VF.

[0320] Statement 45. An embodiment of the inventive concept comprises a system according to Statement 44, wherein the controller is operative to program the upstream filter with an identifier of the VF using the sideband bus.

[0321] Statement 46. An embodiment of the inventive concept comprises a system according to Statement 45, wherein the sideband bus is from a device comprising I 2 C-bus and SMBus.

[0322] Statement 47. An embodiment of the inventive concept includes a system according to Statement 44, wherein the controller is operative to use a PCIe VDM to program an upstream filter with an identifier of the VF.

[0323] Statement 48. An embodiment of the inventive concept comprises a system according to Statement 30, wherein:

[0324] APM-F is operated to send the result to APM-S via the downstream port and the endpoint of the SSD; and

[0325] The controller is operative to forward the results to the processor via an endpoint of the SSD.

[0326] Statement 49. An embodiment of the inventive concept comprises a system according to Statement 30, wherein the APM-F is operative to send the result to the processor via the upstream port.

[0327] Statement 50. An embodiment of the inventive concept comprises a system according to Statement 9, wherein:

[0328] The SSD includes a first PF and a second PF, the first PF being operated to expose the SSD and the second PF being operated to expose the FPGA;

[0329] The FPGA further includes an upstream filter associated with the upstream port, the upstream filter being operative to intercept a second acceleration instruction received from the processor and to transmit the second acceleration instruction to the APM-F; and

[0330] The FPGA is operated to forward a third PCIe transaction to the SSD that is not intercepted by the upstream filter received from the processor at the upstream port.

[0331] Statement 51. An embodiment of the inventive concept comprises a system according to Statement 50, wherein the second acceleration instruction originates from an ASM running on the processor.

[0332] Statement 52. An embodiment of the inventive concept comprises a system according to Statement 50, wherein:

[0333] The first PF is operated to request a first block of host system addresses from the processor;

[0334] The controller is operative to select a first subset of the block of host system addresses as downstream FARs.

[0335] Statement 53. An embodiment of the inventive concept comprises a system according to Statement 52, wherein the controller is operative to program the downstream filter with the downstream FAR.

[0336] Statement 54. An embodiment of the inventive concept comprises a system according to Statement 53, wherein the controller is operative to program the downstream filter with the downstream FAR using the sideband bus.

[0337] Statement 55. An embodiment of the inventive concept comprises a system according to Statement 54, wherein the sideband bus is from a device comprising I 2 C-bus and SMBus.

[0338] Statement 56. An embodiment of the inventive concept includes a system according to Statement 53, wherein the controller is operative to use a PCIe VDM to program a downstream filter with a downstream FAR.

[0339] Statement 57. An embodiment of the inventive concept comprises a system according to Statement 50, wherein:

[0340] A second speed up instruction is associated with the upstream FAR; and

[0341] The upstream filter is operative to intercept a second speed-up instruction associated with the upstream FAR.

[0342] Statement 58. An embodiment of the inventive concept includes a system according to Statement 57, wherein the second PF is operative to request a second block of host system addresses from the processor as an upstream FAR.

[0343] Statement 59. An embodiment of the inventive concept comprises a system according to Statement 58, wherein the controller is operative to program the upstream filter with the upstream FAR.

[0344] Statement 60. An embodiment of the inventive concept comprises a system according to Statement 59, wherein the controller is operative to program the upstream filter with the upstream FAR using the sideband bus.

[0345] Statement 61. An embodiment of the inventive concept comprises a system according to Statement 60, wherein the sideband bus is from a device comprising I 2 C-bus and SMBus.

[0346] Statement 62. An embodiment of the inventive concept includes the system of Statement 59, wherein the controller is operative to use the PCIe VDM to program the upstream filter with the upstream FAR.

[0347] Statement 63. An embodiment of the inventive concept comprises a system according to Statement 50, wherein:

[0348] The second acceleration instruction includes an identifier of the second PF; and

[0349] The upstream filter is operative to intercept a second accelerate instruction associated with an identifier of a second PF.

[0350] Statement 64. An embodiment of the inventive concept comprises a system according to Statement 63, wherein the controller is operative to program the upstream filter with an identifier of the second PF.

[0351] Statement 65. An embodiment of the inventive concept comprises a system according to Statement 64, wherein the controller is operative to program the upstream filter with an identifier of the second PF using the sideband bus.

[0352] Statement 66. An embodiment of the inventive concept comprises a system according to Statement 65, wherein the sideband bus is from a device comprising I 2 C-bus and SMBus.

[0353] Statement 67. An embodiment of the inventive concept includes the system of Statement 64, wherein the controller is operative to use the PCIe VDM to program the upstream filter with an identifier of the second PF.

[0354] Statement 68. An embodiment of the inventive concept comprises a system according to Statement 50, wherein:

[0355] APM-F is operated to send the result to APM-S via the downstream port and the endpoint of the SSD; and

[0356] The controller is operative to forward the results to the processor via an endpoint of the SSD.

[0357] Statement 69. An embodiment of the inventive concept comprises a system according to Statement 50, wherein the APM-F is operative to send the result to the processor via the upstream port.

[0358] Statement 70. An embodiment of the inventive concept comprises a system according to Statement 3, wherein:

[0359] The upstream interface contains the FPGA endpoint;

[0360] The downstream interface includes an FPGA root port, the FPGA root port supporting a configuration space;

[0361] The FPGA includes a first PF, a second PF, and a downstream filter associated with a corresponding FPGA root port, wherein the downstream filter is operated to intercept a first acceleration instruction received from the SSD and transmit the first acceleration instruction to the APM-F, wherein the first acceleration instruction is associated with a downstream FAR;

[0362] The FPGA is operative to request a first block of host system addresses from the processor for use with the first PF and is operative to request a second block of host system addresses from the processor for use with the second PF; and

[0363] The FPGA is operated to forward a first PCIe transaction received from a processor to the SSD, and is operated to forward a second acceleration instruction received from the processor to the APM-F, wherein the first PCIe transaction is associated with a first identifier of a first PF and the second acceleration instruction is associated with a second identifier of a second PF.

[0364] Statement 71. An embodiment of the inventive concept comprises a system according to Statement 70, wherein:

[0365] The SSD is operated to request a block of FPGA addresses from the FPGA, the block of FPGA addresses comprising a downstream FAR;

[0366] The second block of host system addresses is at least as large as the block of FPGA addresses; and

[0367] The controller is operative to select a subset of the block of FPGA addresses as downstream FARs.

[0368] Statement 72. An embodiment of the inventive concept comprises a system according to Statement 71, wherein the controller is operative to program the downstream filter with the downstream FAR.

[0369] Statement 73. An embodiment of the inventive concept comprises a system according to Statement 72, wherein the controller is operative to program the downstream filter with the downstream FAR using the sideband bus.

[0370] Statement 74. An embodiment of the inventive concept comprises a system according to Statement 73, wherein the sideband bus is from a device comprising I 2 C-bus and SMBus.

[0371] Statement 75. An embodiment of the inventive concept includes the system of Statement 72, wherein the controller is operative to use the PCIe VDM to program the downstream filter with the downstream FAR.

[0372] Statement 76. An embodiment of the inventive concept comprises a system according to Statement 70, wherein the APM-F is operative to send the result to the processor via the FPGA endpoint.

[0373] Statement 77. An embodiment of the inventive concept comprises a system according to Statement 70, wherein the FPGA further comprises a configuration monitor to replicate the performance of an endpoint of the SSD to the FPGA endpoint.

[0374] Statement 78. An embodiment of the inventive concept comprises a system according to Statement 3, wherein:

[0375] The upstream interface contains the FPGA endpoint;

[0376] The downstream interface includes a first FPGA root port and a second FPGA root port, wherein the first FPGA root port supports a first configuration space and the second FPGA root port supports a second configuration space;

[0377] The endpoint of the SSD is associated with the first FPGA root port;

[0378] The SSD further includes a second endpoint of the SSD associated with a second FPGA root port;

[0379] The FPGA includes a first PF and a second PF;

[0380] The FPGA is operative to request a first block of host system addresses from the processor for use with the first PF and is operative to request a second block of host system addresses from the processor for use with the second PF; and

[0381] The FPGA is operated to:

[0382] forwarding a first PCIe transaction received from a processor to the SSD via an endpoint of the first FPGA root port and the SSD, the first PCIe transaction being associated with a first identifier of the first PF;

[0383] forwarding a second acceleration instruction received from the processor to the APM-F, the second acceleration instruction being associated with a second identifier of the second PF;

[0384] forwarding a second PCIe transaction received from the SSD at the first FPGA root port to the processor; and

[0385] The first acceleration instruction received from the SSD at the second FPGA root port is forwarded to the APM-F.

[0386] Statement 79. An embodiment of the inventive concept comprises a system according to Statement 78, wherein the second acceleration instruction is generated by APM-S.

[0387] Statement 80. An embodiment of the inventive concept comprises a system according to Statement 78, wherein the APM-F is operative to send the result to the processor via the FPGA endpoint.

[0388] Statement 81. An embodiment of the inventive concept comprises a system according to Statement 78, wherein the FPGA further comprises a configuration monitor to replicate the performance of an endpoint of the SSD to the FPGA endpoint.

[0389] Statement 82. An embodiment of the inventive concept comprises a system according to Statement 78, wherein:

[0390] The upstream interface further comprises a second FPGA endpoint; and

[0391] The FPGA is further operated to:

[0392] forwarding a first PCIe transaction received from the processor at the FPGA endpoint to the SSD via the first FPGA root port and the endpoint of the SSD; and

[0393] The second acceleration instruction received from the processor at the second FPGA endpoint is forwarded to the APM-F.

[0394] Statement 83. An embodiment of the inventive concept comprises a system according to Statement 82, wherein the second acceleration instruction is generated by APM-S.

[0395] Statement 84. An embodiment of the inventive concept comprises a system according to Statement 82, wherein the APM-F is operated to send the result to the processor via the FPGA endpoint.

[0396] Statement 85. An embodiment of the inventive concept comprises a system according to Statement 82, wherein the FPGA further comprises a configuration monitor to replicate the performance of an endpoint of the SSD to the first FPGA endpoint.

[0397] Statement 86. An embodiment of the inventive concept comprises a system according to Statement 3, wherein:

[0398] The upstream interface includes a first FPGA endpoint and a second FPGA endpoint;

[0399] The downstream interface includes an FPGA root port, the FPGA root port supporting a configuration space;

[0400] The FPGA includes a downstream filter associated with the FPGA root port, the downstream filter being operated to intercept a first acceleration instruction received from the SSD and transmit the first acceleration instruction to the APM-F, the first acceleration instruction being associated with a downstream FAR; and

[0401] The FPGA is operated to:

[0402] forwarding a first PCIe transaction received from the processor at the FPGA endpoint to the SSD via the first FPGA root port and the endpoint of the SSD;

[0403] forwarding a second acceleration instruction received from the processor at the second FPGA endpoint to the APM-F;

[0404] forwarding a second PCIe transaction not associated with a downstream FAR received from the SSD at the first FPGA root port to the processor via the FPGA endpoint; and

[0405] The first acceleration instruction received from the SSD at the second FPGA root port is forwarded to the APM-F.

[0406] Statement 87. An embodiment of the inventive concept comprises a system according to Statement 86, wherein the second acceleration instruction is generated by APM-S.

[0407] Statement 88. An embodiment of the inventive concept comprises a system according to Statement 86, wherein:

[0408] The SSD is operated to request a block of FPGA addresses from the FPGA, the block of FPGA addresses comprising a downstream FAR;

[0409] The FPGA is operative to request from the processor a block of host system addresses for use with the first FPGA endpoint, the block of host system addresses being at least as large as the block of FPGA addresses; and

[0410] The controller is operative to select a subset of the block of FPGA addresses as downstream FARs.

[0411] Statement 89. An embodiment of the inventive concept comprises a system according to Statement 88, wherein the controller is operative to program the downstream filter with the downstream FAR.

[0412] Statement 90. An embodiment of the inventive concept comprises a system according to Statement 89, wherein the controller is operative to program the downstream filter with the downstream FAR using the sideband bus.

[0413] Statement 91. An embodiment of the inventive concept comprises a system according to Statement 90, wherein the sideband bus is from a system comprising I 2 C-bus and SMBus.

[0414] Statement 92. An embodiment of the inventive concept includes the system of Statement 89, wherein the controller is operative to use the PCIe VDM to program the downstream filter with the downstream FAR.

[0415] Statement 93. An embodiment of the inventive concept comprises a system according to Statement 86, wherein the APM-F is operative to send the result to the processor via the FPGA endpoint.

[0416] Statement 94. An embodiment of the inventive concept comprises the system of Statement 86, wherein the FPGA further comprises a configuration monitor to replicate the performance of an endpoint of the SSD to the first FPGA endpoint.

[0417] Statement 95. An embodiment of the inventive concept comprises an acceleration module implemented using hardware, the acceleration module comprising:

[0418] Accelerated Platform Manager (APM-F), used to execute acceleration instructions;

[0419] an upstream interface for communicating with a processor on which the application runs; and

[0420] A downstream interface for communicating with a storage device, wherein the storage device includes a storage device acceleration platform manager (APM-S), wherein the storage device acceleration platform manager is used to assist the APM-F in executing acceleration instructions,

[0421] wherein the acceleration module communicates with the processor and the storage device using a peripheral component interconnect express (PCIe) bus, and

[0422] The acceleration module supports executing acceleration instructions on application data on a storage device of an application program without loading the application data into a memory associated with a processor.

[0423] Statement 96. An embodiment of the inventive concept comprises an acceleration module according to Statement 95, wherein:

[0424] The acceleration module is implemented using a field programmable gate array (FPGA).

[0425] Statement 97. An embodiment of the inventive concept comprises an acceleration module according to Statement 96, wherein APM-F and APM-S communicate using a downstream interface regarding application data to be used with the acceleration instructions.

[0426] Statement 98. An embodiment of the inventive concept comprises an acceleration module according to Statement 96, wherein APM-F and APM-S communicate using messages.

[0427] Statement 99. An embodiment of the inventive concept comprises an acceleration module according to Statement 96, wherein the FPGA further comprises:

[0428] Speed ​​up your engine; and

[0429] A run-time scheduler is used to schedule acceleration instructions using the acceleration engine.

[0430] Statement 100. An embodiment of the inventive concept comprises an acceleration module according to Statement 96, wherein:

[0431] The upstream interface contains the upstream port;

[0432] The downstream interface includes the downstream port;

[0433] The FPGA is operated to forward a first PCIe transaction received from the processor at the upstream port to the memory device;

[0434] The FPGA includes a downstream filter associated with the downstream port, the downstream filter being operative to intercept an acceleration instruction received from the storage device and transmit the acceleration instruction to the APM-F, the acceleration instruction being associated with a downstream filter address range (FAR); and

[0435] The FPGA is operated to forward a second PCIe transaction to the processor that is not associated with a downstream FAR received from the memory device at the downstream port.

[0436] Statement 101. An embodiment of the inventive concept comprises the acceleration module of statement 100, wherein the downstream FAR in the downstream filter of the FPGA is programmable by a storage device.

[0437] Statement 102. An embodiment of the inventive concept comprises the acceleration module of Statement 101, wherein the downstream FAR in the downstream filter of the FPGA is programmable by the storage device via a sideband bus.

[0438] Statement 103. An embodiment of the inventive concept comprises an acceleration module according to Statement 102, wherein the sideband bus is from a device comprising an inter-integrated circuit (I 2 C) A collection of buses and the System Management Bus (SMBus).

[0439] Statement 104. An embodiment of the inventive concept comprises the acceleration module of Statement 101, wherein the downstream FAR in the downstream filter of the FPGA is programmable by the storage device using a PCIe Vendor Defined Message (VDM).

[0440] Statement 105. An embodiment of the inventive concept comprises the acceleration module of statement 100, wherein the APM-F is operative to send the result to the APM-S of the storage device via the downstream port.

[0441] Statement 106. An embodiment of the inventive concept comprises an acceleration module according to statement 100, wherein the APM-F is operative to send the result to the processor via an upstream port.

[0442] Statement 107. An embodiment of the inventive concept comprises the acceleration module according to statement 100, wherein the APM-F and the APM-S communicate using messages.

[0443] Statement 108. An embodiment of the inventive concept comprises an acceleration module according to statement 100, wherein:

[0444] The FPGA further includes an upstream filter associated with the upstream port, the upstream filter being operative to intercept a second acceleration instruction received from the processor and transmit the second acceleration instruction to the APM-F, the second acceleration instruction being associated with the upstream FAR; and

[0445] The FPGA is operated to forward a third PCIe transaction that is not associated with an upstream FAR received from the processor at the upstream port to the memory device.

[0446] Statement 109. An embodiment of the inventive concept comprises an acceleration module according to Statement 108, wherein the upstream FAR in the upstream filter of the FPGA is programmable by a storage device.

[0447] Statement 110. An embodiment of the inventive concept comprises the acceleration module of Statement 109, wherein the upstream FAR in the upstream filter of the FPGA is programmable by the memory device using a sideband bus.

[0448] Statement 111. An embodiment of the inventive concept comprises an acceleration module according to statement 110, wherein the sideband bus is from a device comprising I 2 C-bus and SMBus.

[0449] Statement 112. An embodiment of the inventive concept comprises the acceleration module of Statement 109, wherein the upstream FAR in the upstream filter of the FPGA is programmable by the storage device using the PCIe VDM.

[0450] Statement 113. An embodiment of the inventive concept includes the acceleration module of Statement 108, wherein the APM-F is operative to send the result to the APM-S of the storage device via the downstream port.

[0451] Statement 114. An embodiment of the inventive concept comprises the acceleration module of Statement 108, wherein the APM-F is operative to send the result to the processor via the upstream port.

[0452] Statement 115. An embodiment of the inventive concept comprises the acceleration module according to Statement 108, wherein the APM-F and the APM-S communicate using messages.

[0453] Statement 116. An embodiment of the inventive concept includes an acceleration module according to Statement 108, wherein the FPGA is indirectly exposed to the processor via non-volatile memory express (NVMe) registers assigned to a storage device.

[0454] Statement 117. An embodiment of the inventive concept comprises an acceleration module according to statement 100, wherein:

[0455] The FPGA is exposed by the virtual function (VF) of the storage device;

[0456] The FPGA further includes an upstream filter associated with the upstream port, the upstream filter being operative to intercept a second acceleration instruction received from the processor and to transmit the second acceleration instruction to the APM-F; and

[0457] The FPGA is operated to forward a third PCIe transaction to the memory device that is not intercepted by the upstream filter received from the processor at the upstream port.

[0458] Statement 118. An embodiment of the inventive concept comprises an acceleration module according to Statement 117, wherein:

[0459] A second speed up instruction is associated with the upstream FAR; and

[0460] The upstream filter is operative to intercept a second speed-up instruction associated with the upstream FAR.

[0461] Statement 119. An embodiment of the inventive concept comprises an acceleration module according to Statement 118, wherein the upstream FAR in the upstream filter of the FPGA is programmable by a storage device.

[0462] Statement 120. An embodiment of the inventive concept comprises the acceleration module of Statement 119, wherein the upstream FAR in the upstream filter of the FPGA is programmable by the memory device using a sideband bus.

[0463] Statement 121. An embodiment of the inventive concept comprises an acceleration module according to Statement 120, wherein the sideband bus is from a device comprising I 2 C-bus and SMBus.

[0464] Statement 122. An embodiment of the inventive concept comprises the acceleration module of Statement 119, wherein the upstream FAR in the upstream filter of the FPGA is programmable by the storage device using a PCIe VDM.

[0465] Statement 123. An embodiment of the inventive concept comprises an acceleration module according to Statement 117, wherein:

[0466] The second acceleration instruction includes an identifier of the VF; and

[0467] The upstream filter is operative to intercept a second acceleration instruction associated with the identifier of the VF.

[0468] Statement 124. An embodiment of the inventive concept comprises an acceleration module according to Statement 123, wherein an upstream filter of the FPGA is programmable by a storage device with an identifier of the VF.

[0469] Statement 125. An embodiment of the inventive concept comprises an acceleration module according to Statement 124, wherein an upstream filter of the FPGA is programmable with an identifier of the VF by a storage device using a sideband bus.

[0470] Statement 126. An embodiment of the inventive concept comprises an acceleration module according to Statement 125, wherein the sideband bus is from a device comprising I 2 C-bus and SMBus.

[0471] Statement 127. An embodiment of the inventive concept includes an acceleration module according to Statement 124, wherein an upstream filter of the FPGA is programmable with an identifier of the VF by a storage device using a PCIe VDM.

[0472] Statement 128. An embodiment of the inventive concept includes the acceleration module of Statement 117, wherein the APM-F is operative to send the result to the APM-S of the storage device via the downstream port.

[0473] Statement 129. An embodiment of the inventive concept comprises an acceleration module according to Statement 117, wherein the APM-F is operative to send the result to the processor via an upstream port.

[0474] Statement 130. An embodiment of the inventive concept comprises the acceleration module according to Statement 117, wherein the APM-F and the APM-S communicate using messages.

[0475] Statement 131. An embodiment of the inventive concept comprises an acceleration module according to statement 100, wherein:

[0476] FPGAs are exposed by physical functions (PFs) of storage devices;

[0477] The FPGA further includes an upstream filter associated with the upstream port, the upstream filter being operative to intercept a second acceleration instruction received from the processor and to transmit the second acceleration instruction to the APM-F; and

[0478] The FPGA is operated to forward a third PCIe transaction to the memory device that is not intercepted by the upstream filter received from the processor at the upstream port.

[0479] Statement 132. An embodiment of the inventive concept comprises an acceleration module according to Statement 131, wherein:

[0480] A second speed up instruction is associated with the upstream FAR; and

[0481] The upstream filter is operative to intercept a second speed-up instruction associated with the upstream FAR.

[0482] Statement 133. An embodiment of the inventive concept comprises an acceleration module according to Statement 132, wherein the upstream FAR in the upstream filter of the FPGA is programmable by a storage device.

[0483] Statement 134. An embodiment of the inventive concept comprises an acceleration module according to Statement 133, wherein an upstream FAR in an upstream filter of the FPGA is programmable by a storage device using a sideband bus.

[0484] Statement 135. An embodiment of the inventive concept comprises an acceleration module according to Statement 134, wherein the sideband bus is from a device comprising I 2 C-bus and SMBus.

[0485] Statement 136. An embodiment of the inventive concept includes the acceleration module of Statement 133, wherein the upstream FAR in the upstream filter of the FPGA is programmable by the storage device using a PCIe VDM.

[0486] Statement 137. An embodiment of the inventive concept comprises an acceleration module according to Statement 131, wherein:

[0487] The second acceleration instruction includes an identifier of the PF; and

[0488] The upstream filter is operative to intercept a second acceleration instruction associated with the identifier of the PF.

[0489] Statement 138. An embodiment of the inventive concept comprises an acceleration module according to Statement 137, wherein an upstream filter of the FPGA is programmable by a storage device with an identifier of the PF.

[0490] Statement 139. An embodiment of the inventive concept includes an acceleration module according to Statement 138, wherein an upstream filter of the FPGA is programmable with an identifier of the PF by a storage device using a sideband bus.

[0491] Statement 140. An embodiment of the inventive concept comprises an acceleration module according to Statement 139, wherein the sideband bus is from a device comprising I 2 C-bus and SMBus.

[0492] Statement 141. An embodiment of the inventive concept includes an acceleration module according to Statement 138, wherein an upstream filter of the FPGA is programmable with an identifier of the PF by a storage device using a PCIe VDM.

[0493] Statement 142. An embodiment of the inventive concept includes the acceleration module of Statement 131, wherein the APM-F is operative to send the result to the APM-S of the storage device via the downstream port.

[0494] Statement 143. An embodiment of the inventive concept comprises an acceleration module according to Statement 131, wherein the APM-F is operative to send the result to the processor via an upstream port.

[0495] Statement 144. An embodiment of the inventive concept comprises an acceleration module according to Statement 131, wherein APM-F and APM-S communicate using messages.

[0496] Statement 145. An embodiment of the inventive concept comprises an acceleration module according to Statement 96, wherein:

[0497] The upstream interface contains the FPGA endpoint;

[0498] The downstream interface includes an FPGA root port, the FPGA root port supporting a configuration space;

[0499] The FPGA includes a first PF, a second PF, and a downstream filter associated with a corresponding FPGA root port, the downstream filter being operated to intercept a first acceleration instruction received from a storage device and transmit the first acceleration instruction to the APM-F, the first acceleration instruction being associated with a downstream FAR;

[0500] The FPGA is operative to request a first block of host system addresses from the processor for use with the first PF and is operative to request a second block of host system addresses from the processor for use with the second PF; and

[0501] The FPGA is operated to forward a PCIe transaction received from a processor to a storage device, and is operated to forward a second acceleration instruction received from the processor to the APM-F, wherein the PCIe transaction is associated with a first identifier of a first PF, and the second acceleration instruction is associated with a second identifier of a second PF.

[0502] Statement 146. An embodiment of the inventive concept comprises an acceleration module according to Statement 145, wherein:

[0503] The FPGA is operated to receive a request from a storage device for a block of FPGA addresses from the FPGA, the block of FPGA addresses comprising a downstream FAR;

[0504] The FPGA is operated to allocate a block of FPGA addresses from the configuration space; and

[0505] The first block of host system addresses is at least as large as the block of FPGA addresses.

[0506] Statement 147. An embodiment of the inventive concept comprises an acceleration module according to Statement 146, wherein the downstream FAR in the downstream filter of the FPGA is programmable by a storage device.

[0507] Statement 148. An embodiment of the inventive concept comprises an acceleration module according to Statement 147, wherein a downstream FAR in a downstream filter of the FPGA is programmable by a storage device using a sideband bus.

[0508] Statement 149. An embodiment of the inventive concept comprises an acceleration module according to Statement 148, wherein the sideband bus is from a device comprising I 2 C-bus and SMBus.

[0509] Statement 150. An embodiment of the inventive concept comprises the acceleration module of Statement 147, wherein the downstream FAR in the downstream filter of the FPGA is programmable by the storage device using a PCIe VDM.

[0510] Statement 151. An embodiment of the inventive concept comprises an acceleration module according to Statement 145, wherein the APM-F is operative to send the result to the processor via the FPGA endpoint.

[0511] Statement 152. An embodiment of the inventive concept comprises an acceleration module according to Statement 145, wherein APM-F and APM-S communicate using messages.

[0512] Statement 153. An embodiment of the inventive concept comprises an acceleration module according to Statement 145, wherein the FPGA further comprises a configuration monitor to replicate the performance of an endpoint of the storage device to the FPGA endpoint.

[0513] Statement 154. An embodiment of the inventive concept comprises an acceleration module according to Statement 96, wherein:

[0514] The upstream interface contains the FPGA endpoint;

[0515] The downstream interface includes a first FPGA root port and a second FPGA root port, wherein the first FPGA root port supports a first configuration space and the second FPGA root port supports a second configuration space;

[0516] The FPGA includes a first PF and a second PF;

[0517] The FPGA is operative to request a first block of host system addresses from the processor for use with the first PF and is operative to request a second block of host system addresses from the processor for use with the second PF; and

[0518] The FPGA is operated to:

[0519] forwarding a first PCIe transaction received from a processor to a storage device via a first FPGA root port, the first PCIe transaction being associated with a first identifier of a first PF;

[0520] forwarding a second acceleration instruction received from the processor to the APM-F, the second acceleration instruction being associated with a second identifier of the second PF;

[0521] forwarding a second PCIe transaction received from the memory device at the first FPGA root port to the processor; and

[0522] The first acceleration instruction received from the storage device at the second FPGA root port is forwarded to the APM-F.

[0523] Statement 155. An embodiment of the inventive concept comprises an acceleration module according to Statement 154, wherein the APM-F is operative to send the result to the processor via the FPGA endpoint.

[0524] Statement 156. An embodiment of the inventive concept comprises an acceleration module according to Statement 154, wherein APM-F and APM-S communicate using messages.

[0525] Statement 157. An embodiment of the inventive concept comprises an acceleration module according to Statement 154, wherein the FPGA further comprises a configuration monitor to replicate the performance of an endpoint of the storage device to the FPGA endpoint.

[0526] Statement 158. An embodiment of the inventive concept comprises an acceleration module according to Statement 154, wherein:

[0527] The upstream interface further comprises a second FPGA endpoint; and

[0528] The FPGA is further operated to:

[0529] forwarding a first PCIe transaction received from the processor at the FPGA endpoint to the memory device via the first FPGA root port; and

[0530] The second acceleration instruction received from the processor at the second FPGA endpoint is forwarded to the APM-F.

[0531] Statement 159. An embodiment of the inventive concept comprises an acceleration module according to Statement 158, wherein the APM-F is operative to send the result to the processor via the FPGA endpoint.

[0532] Statement 160. An embodiment of the inventive concept comprises the acceleration module according to Statement 158, wherein the APM-F and the APM-S communicate using messages.

[0533] Statement 161. An embodiment of the inventive concept comprises the acceleration module of Statement 158, wherein the FPGA further comprises a configuration monitor to replicate the performance of an endpoint of the storage device to the first FPGA endpoint.

[0534] Statement 162. An embodiment of the inventive concept comprises an acceleration module according to Statement 96, wherein:

[0535] The upstream interface includes a first FPGA endpoint and a second FPGA endpoint;

[0536] The downstream interface includes an FPGA root port, the FPGA root port supporting a configuration space;

[0537] The FPGA includes a downstream filter associated with the FPGA root port, the downstream filter being operated to intercept a first acceleration instruction received from a storage device and transmit the first acceleration instruction to the APM-F, the first acceleration instruction being associated with a downstream FAR; and

[0538] The FPGA is operated to:

[0539] forwarding a first PCIe transaction received from the processor at the FPGA endpoint to the memory device via the first FPGA root port and the endpoint of the memory device;

[0540] forwarding a second acceleration instruction received from the processor at the second FPGA endpoint to the APM-F;

[0541] forwarding a second PCIe transaction not associated with a downstream FAR received from the memory device at the first FPGA root port to the processor via the FPGA endpoint; and

[0542] The first acceleration instruction received from the storage device at the second FPGA root port is forwarded to the APM-F.

[0543] Statement 163. An embodiment of the inventive concept comprises an acceleration module according to Statement 162, wherein:

[0544] The FPGA is operated to receive a request from a storage device for a block of FPGA addresses from the FPGA, the block of FPGA addresses comprising a downstream FAR;

[0545] The FPGA is operated to allocate a block of FPGA addresses from the configuration space; and

[0546] The FPGA is operative to request from the processor a block of host system addresses for use with the first FPGA endpoint, the block of host system addresses being at least as large as the block of FPGA addresses; and

[0547] Statement 164. An embodiment of the inventive concept comprises an acceleration module according to Statement 163, wherein the downstream FAR in the downstream filter of the FPGA is programmable by a storage device.

[0548] Statement 165. An embodiment of the inventive concept comprises an acceleration module according to Statement 164, wherein a downstream FAR in a downstream filter of the FPGA is programmable by a storage device using a sideband bus.

[0549] Statement 166. An embodiment of the inventive concept comprises an acceleration module according to Statement 165, wherein the sideband bus is from a device comprising I 2 C-bus and SMBus.

[0550] Statement 167. An embodiment of the inventive concept includes the acceleration module of Statement 164, wherein the downstream FAR in the downstream filter of the FPGA is programmable by the storage device using the PCIe VDM.

[0551] Statement 168. An embodiment of the inventive concept comprises an acceleration module according to Statement 162, wherein the APM-F is operative to send the result to the processor via the FPGA endpoint.

[0552] Statement 169. An embodiment of the inventive concept comprises an acceleration module according to Statement 162, wherein APM-F and APM-S communicate using messages.

[0553] Statement 170. An embodiment of the inventive concept comprises the acceleration module of Statement 162, wherein the FPGA further comprises a configuration monitor to replicate the performance of an endpoint of the storage device to the first FPGA endpoint.

[0554] Statement 171. An embodiment of the inventive concept comprises a first bridging component implemented using hardware, the first bridging component comprising:

[0555] an upstream interface for communicating with a processor on which the application runs; and

[0556] Downstream interface for communicating with acceleration modules and storage devices,

[0557] wherein the first bridge component communicates with the processor, the acceleration module, and the storage device using a peripheral component interconnect express (PCIe) bus, and

[0558] The downstream interface is operated to transfer the PCIe transaction from the processor to the acceleration module or the storage device depending on whether the PCIe transaction includes an acceleration instruction.

[0559] Statement 172. An embodiment of the inventive concept comprises a first bridge component according to Statement 171, wherein:

[0560] Implementing the acceleration module using a field programmable gate array (FPGA); and

[0561] The storage device includes a solid state drive (SSD).

[0562] Statement 173. An embodiment of the inventive concept comprises the first bridge component according to Statement 171, further comprising a second bridge component, the second bridge component comprising:

[0563] A second upstream interface for communicating with the processor and the acceleration module; and

[0564] A second downstream interface, for communicating with a storage device,

[0565] wherein the second bridge component communicates with the processor, the acceleration module, and the storage device using a peripheral component interconnect express (PCIe) bus, and

[0566] The second upstream interface is operated to transmit the second PCIe transaction from the storage device to the processor or the acceleration module depending on whether the second PCIe transaction includes a second acceleration instruction.

[0567] Statement 174. An embodiment of the inventive concept comprises a first bridge component according to Statement 171, wherein:

[0568] The upstream interface includes:

[0569] Upstream ports; and

[0570] an upstream filter associated with the upstream port, the upstream filter operative to identify a second acceleration instruction associated with an upstream FAR received from the processor; and

[0571] The downstream interface is operated to forward the second speed-up instruction to the acceleration module and is operated to forward a third PCIe transaction not associated with an upstream FAR received from the processor at the upstream port to the storage device.

[0572] Statement 175. An embodiment of the inventive concept comprises a first bridge component according to Statement 174, wherein an upstream FAR in an upstream filter of the first bridge component is programmable by a storage device.

[0573] Statement 176. An embodiment of the inventive concept comprises a first bridge component according to Statement 175, wherein an upstream FAR in an upstream filter of the first bridge component is programmable by a storage device using a sideband bus.

[0574] Statement 177. An embodiment of the inventive concept comprises a first bridge component according to Statement 176, wherein the sideband bus is from a device comprising I 2 C-bus and SMBus.

[0575] Statement 178. An embodiment of the inventive concept comprises the first bridge component of Statement 175, wherein the upstream FAR in the upstream filter of the first bridge component is programmable by the storage device using a PCIe Vendor Defined Message (VDM).

[0576] Statement 179. An embodiment of the inventive concept includes a first bridge component according to Statement 174, wherein the FPGA is indirectly exposed to the processor via non-volatile memory express (NVMe) registers assigned to a memory device.

[0577] Statement 180. An embodiment of the inventive concept comprises a first bridge component according to Statement 171, wherein:

[0578] The upstream interface is exposed by a virtual function (VF) of the storage device;

[0579] The upstream interface includes:

[0580] Upstream ports; and

[0581] An upstream filter is associated with the upstream port, the upstream filter being operative to identify a second acceleration instruction to the acceleration module; and

[0582] The downstream interface is operated to forward the second acceleration instruction to the acceleration module, and is operated to forward a third PCIe transaction that is not intercepted by an upstream filter received from the processor at the upstream port to the storage device.

[0583] Statement 181. An embodiment of the inventive concept comprises a first bridge component according to Statement 180, wherein:

[0584] A second speed up instruction is associated with the upstream FAR; and

[0585] The upstream filter is operative to identify a second speed-up instruction associated with the upstream FAR.

[0586] Statement 182. An embodiment of the inventive concept comprises a first bridge component according to Statement 181, wherein an upstream FAR in an upstream filter of the first bridge component is programmable by a storage device.

[0587] Statement 183. An embodiment of the inventive concept comprises the first bridge component according to Statement 182, wherein the upstream FAR in the upstream filter of the first bridge component is programmable by the storage device using a sideband bus.

[0588] Statement 184. An embodiment of the inventive concept comprises a first bridging component according to Statement 183, wherein the sideband bus is from a device comprising I 2 C-bus and SMBus.

[0589] Statement 185. An embodiment of the inventive concept includes the first bridge component of Statement 182, wherein an upstream FAR in an upstream filter of the first bridge component is programmable by a storage device using a PCIe VDM.

[0590] Statement 186. An embodiment of the inventive concept comprises a first bridge component according to Statement 180, wherein:

[0591] The second acceleration instruction includes an identifier of the VF; and

[0592] The upstream filter is operated to identify a second acceleration instruction associated with the identifier of the VF.

[0593] Statement 187. An embodiment of the inventive concept comprises a first bridge component according to Statement 186, wherein an upstream filter of the first bridge component is programmable by a storage device with an identifier of the VF.

[0594] Statement 188. An embodiment of the inventive concept comprises a first bridge component according to Statement 187, wherein an upstream filter of the first bridge component is programmable with an identifier of the VF by a storage device using a sideband bus.

[0595] Statement 189. An embodiment of the inventive concept comprises a first bridging component according to Statement 188, wherein the sideband bus is from a device comprising I 2 C-bus and SMBus.

[0596] Statement 190. An embodiment of the inventive concept includes the first bridge component of Statement 187, wherein an upstream filter of the first bridge component is programmable by a storage device using a PCIe VDM with an identifier of the VF.

[0597] Statement 191. An embodiment of the inventive concept comprises a first bridge component according to Statement 171, wherein:

[0598] The upstream interface is exposed by the physical function (PF) of the storage device;

[0599] The upstream interface includes:

[0600] Upstream ports; and

[0601] An upstream filter is associated with the upstream port, the upstream filter being operative to identify a second acceleration instruction to the acceleration module; and

[0602] The downstream interface is operated to forward the second acceleration instruction to the acceleration module, and is operated to forward a third PCIe transaction that is not intercepted by an upstream filter received from the processor at the upstream port to the storage device.

[0603] Statement 192. An embodiment of the inventive concept comprises a first bridge component according to Statement 191, wherein:

[0604] A second speed up instruction is associated with the upstream FAR; and

[0605] The upstream filter is operative to identify a second speed-up instruction associated with the upstream FAR.

[0606] Statement 193. An embodiment of the inventive concept comprises a first bridge component according to Statement 192, wherein an upstream FAR in an upstream filter of the first bridge component is programmable by a storage device.

[0607] Statement 194. An embodiment of the inventive concept comprises the first bridge component according to Statement 193, wherein the upstream FAR in the upstream filter of the first bridge component is programmable by the storage device using a sideband bus.

[0608] Statement 195. An embodiment of the inventive concept comprises a first bridging component according to Statement 194, wherein the sideband bus is from a device comprising I 2 C-bus and SMBus.

[0609] Statement 196. An embodiment of the inventive concept comprises the first bridge component of Statement 193, wherein an upstream FAR in an upstream filter of the first bridge component is programmable by a storage device using a PCIe VDM.

[0610] Statement 197. An embodiment of the inventive concept comprises a first bridge component according to Statement 191, wherein:

[0611] The second acceleration instruction includes an identifier of the PF; and

[0612] The upstream filter is operated to identify a second acceleration instruction associated with the identifier of the PF.

[0613] Statement 198. An embodiment of the inventive concept comprises a first bridge component according to Statement 197, wherein an upstream filter of the first bridge component is programmable by a storage device with an identifier of the PF.

[0614] Statement 199. An embodiment of the inventive concept comprises a first bridge component according to Statement 198, wherein an upstream filter of the first bridge component is programmable with an identifier of the PF by a storage device using a sideband bus.

[0615] Statement 200. An embodiment of the inventive concept comprises a first bridging component according to statement 199, wherein the sideband bus is from a device comprising I 2 C-bus and SMBus.

[0616] Statement 201. An embodiment of the inventive concept includes the first bridge component of Statement 198, wherein an upstream filter of the first bridge component is programmable by a storage device using a PCIe VDM with an identifier of a PF.

[0617] Statement 202. An embodiment of the inventive concept comprises a first bridging component according to statement 171, wherein:

[0618] The upstream interface includes:

[0619] Endpoints;

[0620] First PF; and

[0621] Second PF; and

[0622] The downstream interface is operated to forward a PCIe transaction received from a processor to a storage device, and is operated to forward a second acceleration instruction received from the processor to an acceleration module, wherein the PCIe transaction is associated with a first identifier of the first PF, and the second acceleration instruction is associated with a second identifier of the second PF.

[0623] Statement 203. An embodiment of the inventive concept comprises the first bridge component of statement 202, wherein the first bridge component further comprises configuring the monitor to replicate the performance of an endpoint of the storage device to the endpoint.

[0624] Statement 204. An embodiment of the inventive concept comprises a first bridging component according to statement 171, wherein:

[0625] The upstream interface includes:

[0626] a first endpoint; and

[0627] a second endpoint; and

[0628] The downstream interface is operative to forward a PCIe transaction received from the processor and associated with the first endpoint to the storage device, and is operative to forward a second acceleration instruction received from the processor and associated with the second endpoint to the acceleration module.

[0629] Statement 205. An embodiment of the inventive concept comprises the first bridge component of statement 204, wherein the first bridge component further comprises configuring the monitor to replicate the performance of the endpoint of the storage device to the first endpoint.

[0630] Statement 206. An embodiment of the inventive concept comprises a second bridging component implemented using hardware, the second bridging component comprising:

[0631] An upstream interface for communicating with the processor and the acceleration module; and

[0632] A downstream interface for communicating with a storage device;

[0633] wherein the first bridge component communicates with the processor, the acceleration module, and the storage device using a peripheral component interconnect express (PCIe) bus, and

[0634] The upstream interface is operated to transfer the PCIe transaction from the storage device to the processor or the acceleration module depending on whether the PCIe transaction includes an acceleration instruction.

[0635] Statement 207. An embodiment of the inventive concept comprises a second bridge component according to statement 206, wherein:

[0636] Implementing the acceleration module using a field programmable gate array (FPGA); and

[0637] The storage device includes a solid state drive (SSD).

[0638] Statement 208. An embodiment of the inventive concept comprises a second bridging component according to statement 206, wherein:

[0639] The downstream interface includes:

[0640] Downstream ports; and

[0641] a downstream filter associated with the downstream port, the downstream filter operative to identify an acceleration instruction associated with a downstream filter address range (FAR) received from the storage device;

[0642] The downstream interface is operated to forward the acceleration instruction to the acceleration module and is operated to forward a second PCIe transaction not associated with a downstream FAR received from the storage device at the downstream port to the processor.

[0643] Statement 209. An embodiment of the inventive concept comprises the second bridge component according to Statement 208, wherein the downstream FAR in the downstream filter of the second bridge component is programmable by a storage device.

[0644] Statement 210. An embodiment of the inventive concept comprises the second bridge component according to Statement 209, wherein the downstream FAR in the downstream filter of the second bridge component is programmable by the storage device via the sideband bus.

[0645] Statement 211. An embodiment of the inventive concept comprises the second bridge component according to statement 210, wherein the sideband bus is from a device comprising an inter-integrated circuit (I 2 C) A collection of buses and the System Management Bus (SMBus).

[0646] Statement 212. An embodiment of the inventive concept comprises the second bridging component of Statement 209, wherein the downstream FAR in the downstream filter of the second bridging component is programmable by the storage device using a PCIe Vendor Defined Message (VDM).

[0647] Statement 213. An embodiment of the inventive concept comprises a second bridging component according to statement 206, wherein:

[0648] The downstream interface includes:

[0649] Root port; and

[0650] a downstream filter associated with the root port, the downstream filter operative to identify an acceleration instruction associated with a downstream filter address range (FAR) received from a storage device;

[0651] The downstream interface is operated to forward the acceleration instruction to the acceleration module and is operated to forward a second PCIe transaction not associated with a downstream FAR received from the storage device at the downstream port to the processor.

[0652] Statement 214. An embodiment of the inventive concept comprises a second bridge component according to Statement 213, wherein a downstream FAR in a downstream filter of the second bridge component is programmable by a storage device.

[0653] Statement 215. An embodiment of the inventive concept comprises the second bridge component according to Statement 214, wherein the downstream FAR in the downstream filter of the second bridge component is programmable by the memory device using a sideband bus.

[0654] Statement 216. An embodiment of the inventive concept comprises a second bridge component according to Statement 215, wherein the sideband bus is from a device comprising I 2 C-bus and SMBus.

[0655] Statement 217. An embodiment of the inventive concept includes the second bridge component of Statement 214, wherein a downstream FAR in a downstream filter of the second bridge component is programmable by a storage device using a PCIe VDM.

[0656] Statement 218. An embodiment of the inventive concept comprises a second bridging component according to statement 206, wherein:

[0657] The downstream interface includes:

[0658] Root port; and

[0659] The second port,

[0660] The downstream interface is operated to forward a second PCIe transaction received from the storage device at the first root port to the processor, and is operated to forward an acceleration instruction received from the storage device at the second root port to the acceleration module.

[0661] Statement 219. An embodiment of the inventive concept comprises a storage device comprising:

[0662] an endpoint of the storage device for communicating with an acceleration module, the acceleration module including an acceleration platform manager (APM-F);

[0663] a controller for managing the operation of the storage device;

[0664] Storage, which is used to store application data of the application; and

[0665] The Accelerated Platform Manager for Storage (APM-S) is used to assist APM-F in executing acceleration instructions.

[0666] wherein the storage device and the acceleration module communicate using a peripheral component interconnect express (PCIe) bus, and

[0667] The acceleration module supports executing acceleration instructions on application data on a storage device of an application program without loading the application data into a memory associated with a processor.

[0668] Statement 220. An embodiment of the inventive concept comprises a storage device as recited in Statement 219, wherein the storage device comprises a solid state drive (SSD).

[0669] Statement 221. An embodiment of the inventive concept comprises a storage device according to statement 220, wherein APM-F and APM-S communicate using endpoints of the SSD involving application data to be used with the acceleration instructions.

[0670] Statement 222. An embodiment of the inventive concept comprises a storage device according to Statement 220, wherein APM-F and APM-S communicate using messages.

[0671] Statement 223. An embodiment of the inventive concept includes a storage device according to statement 220, wherein the storage device can send a PCIe transaction received from a processor to the SSD, the PCI transaction including a transaction layer packet (TLP) encoding a command using a non-volatile storage express (NVMe) protocol.

[0672] Statement 224. An embodiment of the inventive concept includes a storage device according to statement 220, wherein the SSD includes an acceleration module.

[0673] Statement 225. An embodiment of the inventive concept comprises a storage device according to statement 220, wherein:

[0674] The SSD is operated to send an acceleration instruction associated with a downstream filter address range (FAR) to the acceleration module, the first PCIe transaction being intended for APM-F; and

[0675] The SSD is operated to send a first PCIe transaction not associated with a downstream FAR to the acceleration module, the first PCIe transaction being intended for the processor.

[0676] Statement 226. An embodiment of the inventive concept comprises a storage device according to Statement 225, wherein the speed-up instruction is generated by APM-S.

[0677] Statement 227. An embodiment of the inventive concept includes a storage device according to statement 226, wherein the SSD further includes host interface logic (HIL) to intercept special commands (the special commands including acceleration instructions) and forward the special commands to the APM-S to trigger the APM-S to generate the acceleration instructions.

[0678] Statement 228. An embodiment of the inventive concept comprises a storage device according to Statement 227, wherein the dedicated command originates from an Acceleration Services Manager (ASM) running on the processor.

[0679] Statement 229. An embodiment of the inventive concept comprises a storage device according to statement 225, wherein:

[0680] The SSD is operative to request blocks of host system addresses from the processor; and

[0681] The controller is operative to select a subset of the block of host system addresses as downstream FARs.

[0682] Statement 230. An embodiment of the inventive concept comprises a storage device according to Statement 229, wherein the controller is operative to program a downstream filter of the acceleration module with a downstream FAR.

[0683] Statement 231. An embodiment of the inventive concept includes a memory device according to Statement 230, wherein the controller is operative to use a sideband bus to program a downstream filter of the acceleration module with a downstream FAR.

[0684] Statement 232. An embodiment of the inventive concept comprises a memory device according to Statement 231, wherein the sideband bus is from a device comprising an inter-integrated circuit (I 2 C) A collection of buses and the System Management Bus (SMBus).

[0685] Statement 233. An embodiment of the inventive concept includes a storage device according to Statement 230, wherein the controller is operative to use PCIe Vendor Defined Messages (VDMs) to program downstream filters of the acceleration module with downstream FARs.

[0686] Statement 234. An embodiment of the inventive concept includes a storage device according to Statement 229, wherein the controller is further operative to select a second subset of the block of host system addresses as upstream FARs.

[0687] Statement 235. An embodiment of the inventive concept includes a storage device according to Statement 234, wherein the controller is operative to store information about the upstream FAR in a dedicated register accessible by an ASM running on the processor.

[0688] Statement 236. An embodiment of the inventive concept comprises a storage device according to Statement 235, wherein the dedicated registers are within a block of host system addresses.

[0689] Statement 237. An embodiment of the inventive concept includes a storage device according to Statement 229, wherein the controller is operative to program an upstream filter of the acceleration module with an upstream FAR.

[0690] Statement 238. An embodiment of the inventive concept includes a storage device according to Statement 237, wherein the controller is operative to use a sideband bus to program an upstream filter of the acceleration module with an upstream FAR.

[0691] Statement 239. An embodiment of the inventive concept comprises a memory device according to Statement 238, wherein the sideband bus is from a device comprising an inter-integrated circuit (I 2 C) A collection of buses and the System Management Bus (SMBus).

[0692] Statement 240. An embodiment of the inventive concept includes the storage device of Statement 237, wherein the controller is operative to use a PCIe Vendor Defined Message (VDM) to program an upstream filter of the acceleration module with an upstream FAR.

[0693] Statement 241. An embodiment of the inventive concept includes a storage device according to Statement 225, wherein the SSD is operative to receive results from the APM-F via an endpoint of the SSD and is operative to forward the results to the processor via the endpoint of the SSD.

[0694] Statement 242. An embodiment of the inventive concept includes a storage device according to Statement 225, wherein the SSD includes a physical function (PF) and a virtual function (VF), the PF being operated to expose the SSD and the VF being operated to expose the acceleration module.

[0695] Statement 243. An embodiment of the inventive concept comprises a storage device according to Statement 242, wherein:

[0696] The PF is operable to request a first block of host system addresses from the processor;

[0697] The controller is operative to select a first subset of the block of host system addresses as downstream FARs.

[0698] Statement 244. An embodiment of the inventive concept comprises a storage device according to Statement 243, wherein the controller is operative to program a downstream filter of the acceleration module with a downstream FAR.

[0699] Statement 245. An embodiment of the inventive concept includes a storage device according to Statement 244, wherein the controller is operative to use a sideband bus to program a downstream filter of the acceleration module with a downstream FAR.

[0700] Statement 246. An embodiment of the inventive concept comprises a memory device according to Statement 245, wherein the sideband bus is from a device comprising I 2 C-bus and SMBus.

[0701] Statement 247. An embodiment of the inventive concept includes a storage device according to Statement 244, wherein the controller is operative to use a PCIe VDM to program a downstream filter of the acceleration module with a downstream FAR.

[0702] Statement 248. An embodiment of the inventive concept includes a storage device according to Statement 242, wherein the VF is operative to request a second block of host system addresses from the processor as an upstream FAR.

[0703] Statement 249. An embodiment of the inventive concept includes a storage device according to Statement 248, wherein the controller is operative to program an upstream filter of the acceleration module with an upstream FAR.

[0704] Statement 250. An embodiment of the inventive concept includes the memory device of Statement 249, wherein the controller is operative to program an upstream filter of the acceleration module with an upstream FAR using a sideband bus.

[0705] Statement 251. An embodiment of the inventive concept comprises a memory device according to statement 250, wherein the sideband bus is from a memory device comprising I 2 C-bus and SMBus.

[0706] Statement 252. An embodiment of the inventive concept includes the storage device of Statement 249, wherein the controller is operative to use a PCIe VDM to program an upstream filter of the acceleration module with an upstream FAR.

[0707] Statement 253. An embodiment of the inventive concept includes a storage device according to Statement 242, wherein the controller is operative to program an upstream filter of the acceleration module with an identifier of the VF.

[0708] Statement 254. An embodiment of the inventive concept includes a storage device according to Statement 253, wherein the controller is operative to use a sideband bus to program an upstream filter of the acceleration module with an identifier of the VF.

[0709] Statement 255. An embodiment of the inventive concept comprises a memory device according to Statement 254, wherein the sideband bus is from a device comprising I 2 C-bus and SMBus.

[0710] Statement 256. An embodiment of the inventive concept includes the storage device of Statement 253, wherein the controller is operative to use the PCIe VDM to program an upstream filter of the acceleration module with an identifier of the VF.

[0711] Statement 257. An embodiment of the inventive concept includes a storage device according to Statement 242, wherein the controller is operative to receive results from the APM-F via an endpoint of the SSD and is operative to forward the results to the processor via the endpoint of the SSD.

[0712] Statement 258. An embodiment of the inventive concept includes a storage device according to Statement 225, wherein the SSD includes a first PF and a second PF, the first PF being operated to expose the SSD and the second PF being operated to expose the acceleration module.

[0713] Statement 259. An embodiment of the inventive concept comprises a storage device according to Statement 258, wherein:

[0714] The first PF is operated to request a first block of host system addresses from the processor;

[0715] The controller is operative to select a first subset of the block of host system addresses as downstream FARs.

[0716] Statement 260. An embodiment of the inventive concept includes a storage device according to Statement 259, wherein the controller is operative to program a downstream filter of the acceleration module with a downstream FAR.

[0717] Statement 261. An embodiment of the inventive concept includes a storage device according to Statement 260, wherein the controller is operative to use a sideband bus to program a downstream filter of the acceleration module with a downstream FAR.

[0718] Statement 262. An embodiment of the inventive concept comprises a memory device according to Statement 261, wherein the sideband bus is from a device comprising I 2 C-bus and SMBus.

[0719] Statement 263. An embodiment of the inventive concept includes a storage device according to Statement 260, wherein the controller is operative to use a PCIe VDM to program a downstream filter of the acceleration module with a downstream FAR.

[0720] Statement 264. An embodiment of the inventive concept includes a storage device according to Statement 258, wherein the second PF is operative to request a second block of host system addresses from the processor as an upstream FAR.

[0721] Statement 265. An embodiment of the inventive concept includes a storage device according to Statement 264, wherein the controller is operative to program an upstream filter of the acceleration module with an upstream FAR.

[0722] Statement 266. An embodiment of the inventive concept includes a storage device according to Statement 265, wherein the controller is operative to use a sideband bus to program an upstream filter of the acceleration module with an upstream FAR.

[0723] Statement 267. An embodiment of the inventive concept comprises a memory device according to Statement 266, wherein the sideband bus is from a device comprising I 2 C-bus and SMBus.

[0724] Statement 268. An embodiment of the inventive concept includes a storage device according to Statement 265, wherein the controller is operative to use a PCIe VDM to program an upstream filter of the acceleration module with an upstream FAR.

[0725] Statement 269. An embodiment of the inventive concept includes a storage device according to Statement 258, wherein the controller is operative to program an upstream filter of the acceleration module with an identifier of the second PF.

[0726] Statement 270. An embodiment of the inventive concept includes the storage device of Statement 269, wherein the controller is operative to program an upstream filter of the acceleration module with an identifier of the second PF using the sideband bus.

[0727] Statement 271. An embodiment of the inventive concept comprises a memory device according to Statement 270, wherein the sideband bus is from a device comprising I 2 C-bus and SMBus.

[0728] Statement 272. An embodiment of the inventive concept includes the storage device of Statement 269, wherein the controller is operative to use the PCIe VDM to program an upstream filter of the acceleration module with an identifier of the second PF.

[0729] Statement 273. An embodiment of the inventive concept includes a storage device according to Statement 258, wherein the controller is operative to receive results from the APM-F via an endpoint of the SSD and is operative to forward the results to the processor via the endpoint of the SSD.

[0730] Statement 274. An embodiment of the inventive concept comprises a storage device according to Statement 225, wherein:

[0731] The SSD is operated to request a block of acceleration module addresses from the acceleration module; and

[0732] The controller is operative to select a subset of the blocks of acceleration module addresses as downstream FARs.

[0733] Statement 275. An embodiment of the inventive concept comprises a storage device according to Statement 274, wherein the controller is operative to program a downstream filter of the acceleration module with a downstream FAR.

[0734] Statement 276. An embodiment of the inventive concept includes a storage device according to Statement 275, wherein the controller is operative to use a sideband bus to program a downstream filter of the acceleration module with a downstream FAR.

[0735] Statement 277. An embodiment of the inventive concept comprises a memory device according to Statement 276, wherein the sideband bus is from a device comprising an inter-integrated circuit (I 2 C) A collection of buses and the System Management Bus (SMBus).

[0736] Statement 278. An embodiment of the inventive concept includes a storage device according to Statement 275, wherein the controller is operative to use PCIe Vendor Defined Messages (VDMs) to program downstream filters of the acceleration module with downstream FARs.

[0737] Statement 279. An embodiment of the inventive concept includes a storage device according to statement 220, the storage device further comprising a second endpoint of the SSD for communicating with the acceleration module, wherein the endpoint of the SSD is used to exchange communication with the processor and the second endpoint of the SSD is used to exchange communication with the APM-F.

[0738] Statement 280. An embodiment of the inventive concept comprises a method comprising:

[0739] receiving, at an acceleration module, a peripheral component interconnect express (PCIe) transaction from a first device;

[0740] Determining at the acceleration module whether the PCIe transaction is an acceleration instruction;

[0741] Based at least in part on determining that the PCIe transaction is an acceleration instruction, processing the PCIe transaction at an acceleration platform manager (APM-F) of the acceleration module; and

[0742] transmitting the PCIe transaction to the second device based at least in part on determining that the PCIe transaction is not an accelerated instruction,

[0743] wherein the acceleration module supports executing acceleration instructions on application data on a storage device of the application program without loading the application data into a memory associated with the processor, and

[0744] The processor, acceleration module and storage device communicate using the PCIe bus.

[0745] Statement 281. An embodiment of the inventive concept comprises a method according to statement 280, wherein the acceleration module is implemented using a field programmable gate array.

[0746] Statement 282. An embodiment of the inventive concept comprises a method according to statement 281, wherein:

[0747] Receiving a peripheral component interconnect express (PCIe) transaction from a first device at the FPGA includes receiving a second PCIe transaction from a memory device at a downstream port of the FPGA;

[0748] Determining at the acceleration module whether the PCIe transaction is an acceleration instruction includes determining at the FPGA whether the second PCIe transaction is associated with an address in a downstream filter address range (FAR), the downstream filter address range being associated with a downstream port of the FPGA; and

[0749] Transmitting the PCIe transaction to the second device includes using an upstream port of the FPGA to transmit the second PCIe transaction to the processor.

[0750] Statement 283. An embodiment of the inventive concept comprises a method according to statement 282, said method further comprising:

[0751] receiving a first PCIe transaction from a processor at an upstream port of the FPGA; and

[0752] A downstream processor of the FPGA is used to transmit a first PCIe transaction to a storage device.

[0753] Statement 284. An embodiment of the inventive concept comprises a method according to statement 282, said method further comprising:

[0754] receiving, at the FPGA, a downstream FAR from a memory device; and

[0755] Associate the downstream FAR with the downstream port of the FPGA.

[0756] Statement 285. An embodiment of the inventive concept comprises a method according to statement 284, wherein receiving the downstream FAR at the FPGA from the storage device comprises receiving the downstream FAR at the FPGA from the storage device via a sideband bus from a memory device comprising an inter-integrated circuit (I 2 C) A collection of buses and the System Management Bus (SMBus).

[0757] Statement 286. An embodiment of the inventive concept includes a method according to statement 284, wherein receiving, at the FPGA, a downstream FAR from a storage device includes receiving a PCIe Vendor Defined Message (VDM) from the storage device, the PCIe VDM including the downstream FAR.

[0758] Statement 287. An embodiment of the inventive concept includes the method of Statement 282, the method further comprising using a downstream port of the FPGA to send a result of the second PCIe transaction to a storage device.

[0759] Statement 288. An embodiment of the inventive concept includes the method of Statement 282, the method further comprising using an upstream port of the FPGA to send a result of the second PCIe transaction to the processor.

[0760] Statement 289. An embodiment of the inventive concept comprises a method according to statement 282, said method further comprising:

[0761] receiving a first PCIe transaction from a processor at an upstream port of the FPGA;

[0762] determining whether the first PCIe transaction is a second acceleration instruction by determining whether the first PCIe transaction is associated with a second address in an upstream FAR, the upstream FAR being associated with an upstream port of the FPGA; and

[0763] Based at least in part on determining that the first PCIe transaction is a second accelerated instruction, processing the first PCIe transaction at the APM-F of the FPGA; and

[0764] Based at least in part on determining that the first PCIe transaction is not a second acceleration instruction, the first PCIe transaction is transmitted to the storage device using a downstream port of the FPGA.

[0765] Statement 290. An embodiment of the inventive concept comprises the method of statement 289, wherein the second acceleration instruction originates from an acceleration service manager (ASM) running on the processor.

[0766] Statement 291. An embodiment of the inventive concept comprises a method according to statement 289, said method further comprising:

[0767] receiving, at the FPGA, an upstream FAR from a memory device; and

[0768] Associate the upstream FAR with the upstream port of the FPGA.

[0769] Statement 292. An embodiment of the inventive concept comprises a method according to Statement 291, wherein receiving the upstream FAR at the FPGA from the storage device comprises receiving the upstream FAR at the FPGA from the storage device via a sideband bus, the sideband bus being from a storage device comprising I 2 C-bus and SMBus.

[0770] Statement 293. An embodiment of the inventive concept includes a method according to Statement 291, wherein receiving an upstream FAR from a storage device at the FPGA includes receiving a PCIe Vendor Defined Message (VDM) from the storage device, the PCIe VDM including the upstream FAR.

[0771] Statement 294. An embodiment of the inventive concept includes the method of Statement 289, the method further comprising using an upstream port of the FPGA to send a result of the first PCIe transaction to a storage device.

[0772] Statement 295. An embodiment of the inventive concept includes the method of Statement 289, the method further comprising using an upstream port of the FPGA to send a result of the first PCIe transaction to the processor.

[0773] Statement 296. An embodiment of the inventive concept comprises the method according to statement 282, the method further comprising:

[0774] receiving a first PCIe transaction from a processor at an upstream port of the FPGA;

[0775] determining whether the first PCIe transaction is a second acceleration instruction by determining whether the first PCIe transaction is associated with a virtual function (VF) exposed by the storage device; and

[0776] Based at least in part on determining that the first PCIe transaction is a second accelerated instruction, processing the first PCIe transaction at the APM-F of the FPGA; and

[0777] Based at least in part on determining that the first PCIe transaction is not a second acceleration instruction, the first PCIe transaction is transmitted to the storage device using a downstream port of the FPGA.

[0778] Statement 297. An embodiment of the inventive concept includes the method of statement 296, wherein the second acceleration instruction originates from an acceleration service manager (ASM) running on the processor.

[0779] Statement 298. An embodiment of the inventive concept comprises the method of statement 296, wherein determining whether the first PCIe transaction is a second acceleration instruction comprises determining whether the first PCIe transaction includes a tag having an identifier of the VF.

[0780] Statement 299. An embodiment of the inventive concept comprises a method according to statement 298, said method further comprising:

[0781] Receiving, at the FPGA, an identifier of the VF from the storage device; and

[0782] Associate the VF's identifier with the FPGA's upstream port.

[0783] Statement 300. An embodiment of the inventive concept includes a method according to statement 299, wherein receiving at the FPGA an identifier of the VF from the storage device includes receiving at the FPGA an identifier of the VF from the storage device via a sideband bus from a memory device including I 2 C-bus and SMBus.

[0784] Statement 301. An embodiment of the inventive concept includes the method of statement 299, wherein receiving, at the FPGA, an identifier of the VF from a storage device includes receiving a PCIe Vendor Defined Message (VDM) from the storage device, the PCIe VDM including an identifier of the VF.

[0785] Statement 302. An embodiment of the inventive concept includes the method of Statement 296, the method further comprising using an upstream port of the FPGA to send a result of the first PCIe transaction to the processor.

[0786] Statement 303. An embodiment of the inventive concept comprises the method according to statement 282, said method further comprising:

[0787] receiving a first PCIe transaction from a processor at an upstream port of the FPGA;

[0788] determining whether the first PCIe transaction is a second acceleration instruction by determining whether the first PCIe transaction is associated with a physical function (PF) exposed by the storage device; and

[0789] Based at least in part on determining that the first PCIe transaction is a second accelerated instruction, processing the first PCIe transaction at the APM-F of the FPGA; and

[0790] Based at least in part on determining that the first PCIe transaction is not a second acceleration instruction, the first PCIe transaction is transmitted to the storage device using a downstream port of the FPGA.

[0791] Statement 304. An embodiment of the inventive concept comprises the method of statement 303, wherein the second acceleration instruction originates from an acceleration service manager (ASM) running on the processor.

[0792] Statement 305. An embodiment of the inventive concept comprises the method of statement 303, wherein determining whether the first PCIe transaction is a second acceleration instruction comprises determining whether the first PCIe transaction comprises a tag having an identifier of a PF.

[0793] Statement 306. An embodiment of the inventive concept comprises a method according to statement 305, said method further comprising:

[0794] receiving, at the FPGA, an identifier of the PF from the storage device; and

[0795] Associates the identifier of the PF with the upstream port of the FPGA.

[0796] Statement 307. An embodiment of the inventive concept includes a method according to statement 306, wherein receiving at the FPGA an identifier of the PF from the storage device includes receiving at the FPGA an identifier of the PF from the storage device via a sideband bus from a memory device including I 2 C-bus and SMBus.

[0797] Statement 308. An embodiment of the inventive concept includes a method according to statement 306, wherein receiving, at the FPGA, an identifier of the PF from the storage device includes receiving a PCIe Vendor Defined Message (VDM) from the storage device, the PCIe VDM including an identifier of the PF.

[0798] Statement 309. An embodiment of the inventive concept includes the method of statement 303, the method further comprising using a downstream port of the FPGA to send a result of the first PCIe transaction to a storage device.

[0799] Statement 310. An embodiment of the inventive concept comprises the method of statement 303, the method further comprising using an upstream port of the FPGA to send a result of the first PCIe transaction to the processor.

[0800] Statement 311. An embodiment of the inventive concept comprises a method according to statement 281, wherein:

[0801] Receiving a peripheral component interconnect express (PCIe) transaction from a first device at the FPGA includes receiving a first PCIe transaction from a processor at an endpoint of the FPGA;

[0802] Determining at the acceleration module whether the PCIe transaction is an acceleration instruction includes determining whether the first PCIe transaction includes a tag having a first identifier of a first PF of the FPGA or a second identifier of a second PF of the FPGA; and

[0803] Transmitting the PCIe transaction to the second device includes using a root port of the FPGA to transmit the first PCIe transaction to the storage device.

[0804] Statement 312. An embodiment of the inventive concept comprises a method according to statement 311, said method further comprising:

[0805] receiving a second PCIe transaction from a memory device at a root port of the FPGA;

[0806] determining whether the second PCIe transaction is a second acceleration instruction by determining at the FPGA whether the second PCIe transaction is associated with an address in a downstream FAR, the downstream FAR being associated with a root port of the FPGA;

[0807] Based at least in part on determining that the second PCIe transaction is a second accelerated instruction, processing the second PCIe transaction at the APM-F of the FPGA; and

[0808] Based at least in part on determining that the second PCIe transaction is not a second acceleration instruction, the second PCIe transaction is communicated to the processor using an endpoint of the FPGA.

[0809] Statement 313. An embodiment of the inventive concept comprises a method according to statement 312, said method further comprising:

[0810] receiving, at the FPGA, a downstream FAR from a memory device; and

[0811] Associate the downstream FAR with the root port of the FPGA.

[0812] Statement 314. An embodiment of the inventive concept comprises a method according to statement 313, wherein receiving the downstream FAR at the FPGA from the storage device comprises receiving the downstream FAR at the FPGA from the storage device via a sideband bus, the sideband bus being from a storage device comprising I 2 C-bus and SMBus.

[0813] Statement 315. An embodiment of the inventive concept includes the method of statement 313, wherein receiving at the FPGA a downstream FAR from a storage device includes receiving a PCIe Vendor Defined Message (VDM) from the storage device, the PCIe VDM including the downstream FAR.

[0814] Statement 316. An embodiment of the inventive concept comprises the method of Statement 311, wherein the second acceleration instruction originates from an acceleration service manager (ASM) running on the processor.

[0815] Statement 317. An embodiment of the inventive concept comprises the method of Statement 311, the method further comprising using an endpoint of the FPGA to send a result of the first PCIe transaction to the processor.

[0816] Statement 318. An embodiment of the inventive concept comprises a method according to statement 311, said method further comprising:

[0817] determining, using a configuration monitor of the FPGA, a configuration of an endpoint of a memory device that communicates with a root port of the FPGA; and

[0818] Use the endpoints of the FPGA to replicate the configuration of the endpoints on the storage device.

[0819] Statement 319. An embodiment of the inventive concept comprises a method according to statement 281, wherein:

[0820] Receiving a peripheral component interconnect express (PCIe) transaction from a first device at the FPGA includes receiving a first PCIe transaction from a processor at an endpoint of the FPGA;

[0821] Determining at the acceleration module whether the PCIe transaction is an acceleration instruction includes determining whether the first PCIe transaction includes a tag having an identifier of a first PF of the FPGA or a second identifier of a second PF of the FPGA; and

[0822] Transmitting the PCIe transaction to the second device includes using a first root port of the FPGA to transmit the first PCIe transaction to the storage device.

[0823] Statement 320. An embodiment of the inventive concept comprises the method according to statement 319, said method further comprising:

[0824] receiving, at the FPGA, a second PCIe transaction from the memory device;

[0825] determining whether the second PCIe transaction is received at the first root port of the FPGA or at the second root port of the FPGA;

[0826] Based at least in part on determining that the second PCIe transaction was received at the first root port of the FPGA, transmitting the second PCIe transaction to the processor using the endpoint; and

[0827] Based at least in part on determining that the second PCIe transaction was received at the second root port of the FPGA, the second PCIe transaction is processed at the APM-F of the FPGA.

[0828] Statement 321. An embodiment of the inventive concept comprises the method of statement 319, wherein the second acceleration instruction originates from an acceleration service manager (ASM) running on the processor.

[0829] Statement 322. An embodiment of the inventive concept comprises the method of Statement 319, the method further comprising using an endpoint of the FPGA to send a result of the first PCIe transaction to the processor.

[0830] Statement 323. An embodiment of the inventive concept comprises a method according to statement 319, said method further comprising:

[0831] determining, using a configuration monitor of the FPGA, a configuration of an endpoint of a memory device that communicates with a first root port of the FPGA; and

[0832] Use the endpoints of the FPGA to replicate the configuration of the endpoints on the storage device.

[0833] Statement 324. An embodiment of the inventive concept comprises a method according to statement 281, wherein:

[0834] Receiving a peripheral component interconnect express (PCIe) transaction from a first device at the FPGA includes receiving a first PCIe transaction from a processor at the FPGA;

[0835] Determining at the FPGA whether the first PCIe transaction is a speed-up instruction includes determining whether the first PCIe transaction is a second speed-up instruction by determining whether the first PCIe transaction is received from a processor at a first endpoint of the FPGA, the FPGA including a first endpoint and a second endpoint; and

[0836] Transmitting the PCIe transaction to the second device includes transmitting the first PCIe transaction to the memory device using the first root port of the FPGA, the FPGA including the first root port and the second root port.

[0837] Statement 325. An embodiment of the inventive concept comprises a method according to statement 324, said method further comprising:

[0838] receiving, at the FPGA, a second PCIe transaction from the memory device;

[0839] determining whether the second PCIe transaction is an accelerated instruction by determining whether the second PCIe transaction is received at the first root port of the FPGA or at the second root port of the FPGA;

[0840] Based at least in part on determining that the second PCIe transaction was received at the first root port of the FPGA, transmitting the second PCIe transaction to the processor using the first endpoint; and

[0841] Based at least in part on determining that the second PCIe transaction was received at the second root port of the FPGA, the second PCIe transaction is processed at the APM-F of the FPGA.

[0842] Statement 326. An embodiment of the inventive concept includes the method of Statement 324, wherein the second PCIe transaction originates from an acceleration service manager (ASM) running on the processor.

[0843] Statement 327. An embodiment of the inventive concept comprises the method of Statement 324, the method further comprising using the second endpoint of the FPGA to send a result of the first PCIe transaction to the processor.

[0844] Statement 328. An embodiment of the inventive concept comprises a method according to statement 324, said method further comprising:

[0845] determining, using a configuration monitor of the FPGA, a configuration of an endpoint of a memory device that communicates with a first root port of the FPGA; and

[0846] The configuration of the endpoint on the memory device is replicated using the first endpoint of the FPGA.

[0847] Statement 329. An embodiment of the inventive concept comprises a method according to statement 281, wherein:

[0848] Receiving a peripheral component interconnect express (PCIe) transaction from a first device at the FPGA includes receiving a first PCIe transaction from a processor at the FPGA;

[0849] Determining at the acceleration module whether the first PCIe transaction is an acceleration instruction includes determining whether the first PCIe transaction is a second acceleration instruction by determining whether the first PCIe transaction is received from a processor at a first endpoint of an FPGA, the FPGA including a first endpoint and a second endpoint; and

[0850] Transmitting the PCIe transaction to the second device includes using a root port of the FPGA to transmit the first PCIe transaction to the storage device.

[0851] Statement 330. An embodiment of the inventive concept comprises a method according to statement 329, said method further comprising:

[0852] receiving a second PCIe transaction from a memory device at a root port of the FPGA;

[0853] determining whether the second PCIe transaction is an accelerated instruction by determining at the FPGA whether the second PCIe transaction is associated with an address in a downstream FAR, the downstream FAR being associated with a root port of the FPGA;

[0854] Based at least in part on determining that the second PCIe transaction is an accelerated instruction, processing the second PCIe transaction at the APM-F of the FPGA; and

[0855] Based at least in part on determining that the second PCIe transaction is not an accelerated instruction, the second PCIe transaction is communicated to the processor using the first endpoint of the FPGA.

[0856] Statement 331. An embodiment of the inventive concept comprises a method according to statement 330, said method further comprising:

[0857] receiving, at the FPGA, a downstream FAR from a memory device; and

[0858] Associate the downstream FAR with the root port of the FPGA.

[0859] Statement 332. An embodiment of the inventive concept comprises a method according to statement 331, wherein receiving the downstream FAR at the FPGA from the storage device comprises receiving the downstream FAR at the FPGA from the storage device via a sideband bus, the sideband bus being from a storage device comprising I 2 C-bus and SMBus.

[0860] Statement 333. An embodiment of the inventive concept includes a method according to statement 331, wherein receiving, at the FPGA, a downstream FAR from a storage device includes receiving a PCIe Vendor Defined Message (VDM) from the storage device, the PCIe VDM including the downstream FAR.

[0861] Statement 334. An embodiment of the inventive concept includes the method of Statement 329, wherein the first PCIe transaction originates from an acceleration service manager (ASM) running on the processor.

[0862] Statement 335. An embodiment of the inventive concept comprises the method of Statement 329, the method further comprising using the second endpoint of the FPGA to send a result of the first PCIe transaction to the processor.

[0863] Statement 336. An embodiment of the inventive concept comprises a method according to statement 329, said method further comprising:

[0864] determining, using a configuration monitor of the FPGA, a configuration of an endpoint of a memory device that communicates with a root port of the FPGA; and

[0865] The configuration of the endpoint on the memory device is replicated using the first endpoint of the FPGA.

[0866] Statement 337. An embodiment of the inventive concept comprises a method comprising:

[0867] receiving a peripheral component interconnect express (PCIe) transaction from a processor at the first bridge component;

[0868] determining, at the first bridge component, whether the PCIe transaction is an accelerated instruction;

[0869] forwarding the PCIe transaction to an acceleration module based at least in part on determining that the PCIe transaction is an acceleration instruction; and

[0870] forwarding the PCIe transaction to a storage device based at least in part on determining that the PCIe transaction is not an accelerated instruction,

[0871] The processor, the first bridge component, the acceleration module and the storage device communicate using a PCIe bus.

[0872] Statement 338. An embodiment of the inventive concept comprises a method according to statement 337, wherein:

[0873] The acceleration module is implemented using a field programmable gate array; and

[0874] The storage device includes a solid state drive (SSD).

[0875] Statement 339. An embodiment of the inventive concept includes a method according to statement 337, wherein determining at the first bridging component whether a PCIe transaction is an accelerated instruction includes determining whether the PCIe transaction is associated with an address in an upstream FAR, wherein the upstream FAR is associated with an upstream port of the first bridging component.

[0876] Statement 340. An embodiment of the inventive concept comprises a method according to statement 339, said method further comprising:

[0877] receiving, at the first bridge component, an upstream FAR from a storage device; and

[0878] An upstream FAR is associated with an upstream port of the first bridging component.

[0879] Statement 341. An embodiment of the inventive concept comprises a method according to statement 340, wherein receiving the upstream FAR from the storage device at the first bridge component comprises receiving the upstream FAR from the storage device at the first bridge component via a sideband bus, the sideband bus being from a storage device comprising I 2 C-bus and SMBus.

[0880] Statement 342. An embodiment of the inventive concept includes the method of statement 340, wherein receiving, at the first bridge component, an upstream FAR from the storage device includes receiving a PCIe Vendor Defined Message (VDM) from the storage device, the PCIe VDM including the upstream FAR.

[0881] Statement 343. An embodiment of the inventive concept comprises the method of statement 337, wherein determining at the first bridge component whether the PCIe transaction is an accelerated instruction comprises determining whether the PCIe transaction is associated with a virtual function (VF) exposed by the storage device.

[0882] Statement 344. An embodiment of the inventive concept comprises a method according to statement 343, wherein determining whether a PCIe transaction is associated with a virtual function (VF) exposed by a storage device comprises determining whether the PCIe transaction includes a tag with an identifier of the VF.

[0883] Statement 345. An embodiment of the inventive concept comprises a method according to statement 344, said method further comprising:

[0884] receiving, at the first bridging component, an identifier of the VF from the storage device; and

[0885] An identifier of the VF is associated with an upstream port of the first bridging component.

[0886] Statement 346. An embodiment of the inventive concept includes a method according to statement 345, wherein receiving, at the first bridge component, an identifier of the VF from the storage device includes receiving, at the first bridge component, an identifier of the VF from the storage device via a sideband bus from a storage device including I 2 C-bus and SMBus.

[0887] Statement 347. An embodiment of the inventive concept includes the method of statement 345, wherein receiving, at the first bridge component, an identifier of the VF from the storage device includes receiving a PCIe Vendor Defined Message (VDM) from the storage device, the PCIe VDM including the identifier of the VF.

[0888] Statement 348. An embodiment of the inventive concept comprises the method of statement 337, wherein determining at the first bridge component whether the PCIe transaction is an accelerated instruction comprises determining whether the PCIe transaction is associated with a physical function (PF) exposed by the storage device.

[0889] Statement 349. An embodiment of the inventive concept includes a method according to statement 348, wherein determining whether a PCIe transaction is associated with a physical function (PF) exposed by a storage device includes determining whether the PCIe transaction includes a tag with an identifier of the PF.

[0890] Statement 350. An embodiment of the inventive concept comprises a method according to statement 349, said method further comprising:

[0891] receiving, at the first bridging component, an identifier of the PF from the storage device; and

[0892] An identifier of the PF is associated with an upstream port of the first bridging component.

[0893] Statement 351. An embodiment of the inventive concept includes a method according to statement 350, wherein receiving, at the first bridge component, an identifier of the PF from the storage device includes receiving, at the first bridge component, an identifier of the PF from the storage device via a sideband bus from a device including I 2 C-bus and SMBus.

[0894] Statement 352. An embodiment of the inventive concept includes the method of statement 350, wherein receiving, at the first bridging component, an identifier of the PF from the storage device includes receiving a PCIe Vendor Defined Message (VDM) from the storage device, the PCIe VDM including the identifier of the PF.

[0895] Statement 353. An embodiment of the inventive concept includes a method according to statement 337, wherein determining at the first bridge component whether a PCIe transaction is an acceleration instruction includes determining whether the PCIe transaction includes a tag having an identifier of a first PF of the first bridge component or a second identifier of a second PF of the first bridge component.

[0896] Statement 354. An embodiment of the inventive concept comprises a method according to statement 353, said method further comprising:

[0897] determining, using a configuration monitor of the first bridging component, a configuration of an endpoint of the storage device; and

[0898] The configuration of the endpoints on the storage device is replicated using the endpoints of the first bridge component.

[0899] Statement 355. An embodiment of the inventive concept includes a method according to statement 337, wherein determining at a first bridge component whether a PCIe transaction is an acceleration instruction includes determining whether the PCIe transaction is received from a processor at a first endpoint of the first bridge component, the first bridge component including a first endpoint and a second endpoint.

[0900] Statement 356. An embodiment of the inventive concept comprises a method according to statement 355, said method further comprising:

[0901] determining, using a configuration monitor of the first bridging component, a configuration of an endpoint of the storage device; and

[0902] The configuration of the endpoint on the storage device is replicated using the first endpoint of the first bridge component.

[0903] Statement 357. An embodiment of the inventive concept comprises a method comprising:

[0904] receiving a peripheral component interconnect express (PCIe) transaction from a storage device at a second bridge component;

[0905] determining, at the second bridge component, whether the PCIe transaction is an accelerated instruction;

[0906] forwarding the PCIe transaction to an acceleration module based at least in part on determining that the PCIe transaction is an acceleration instruction; and

[0907] forwarding the PCIe transaction to the processor based at least in part on determining that the PCIe transaction is not an accelerated instruction,

[0908] The processor, the second bridge component, the acceleration module and the storage device communicate using a PCIe bus.

[0909] Statement 358. An embodiment of the inventive concept comprises a method according to statement 357, wherein:

[0910] The acceleration module is implemented using a field programmable gate array; and

[0911] The storage device includes a solid state drive (SSD).

[0912] Statement 359. An embodiment of the inventive concept includes a method according to statement 357, wherein determining at the second bridging component whether a PCIe transaction is an accelerated instruction includes determining at the second bridging component whether the second PCIe transaction is associated with an address in a downstream filter address range (FAR), wherein the downstream filter address range is associated with a downstream port of the second bridging component.

[0913] Statement 360. An embodiment of the inventive concept comprises a method according to statement 359, said method further comprising:

[0914] receiving, at the second bridge component, a downstream FAR from the storage device; and

[0915] A downstream FAR is associated with a downstream port of the second bridging component.

[0916] Statement 361. An embodiment of the inventive concept comprises the method of statement 360, wherein receiving the downstream FAR from the storage device at the second bridge component comprises receiving the downstream FAR from the storage device at the second bridge component via a sideband bus from a processor comprising an inter-integrated circuit (IC) 2 C) A collection of buses and the System Management Bus (SMBus).

[0917] Statement 362. An embodiment of the inventive concept includes the method of statement 360, wherein receiving, at the second bridge component, a downstream FAR from the storage device includes receiving a PCIe Vendor Defined Message (VDM) from the storage device, the PCIe VDM including the downstream FAR.

[0918] Statement 363. An embodiment of the inventive concept includes a method according to statement 357, wherein determining at a second bridge component whether a PCIe transaction is an acceleration instruction includes determining whether the PCIe transaction is received at a second root port of the second bridge component, wherein the second bridge component includes a first root port and a second root port.

[0919] Statement 364. An embodiment of the inventive concept comprises a method comprising:

[0920] receiving, at a storage device, a first PCIe transaction from an acceleration module;

[0921] Determining whether the first PCIe transaction is an acceleration instruction;

[0922] Based at least in part on determining that the first PCIe transaction is an accelerated instruction:

[0923] generating a second PCIe transaction using an Accelerated Platform Manager for Storage (APM-S) of the storage device; and

[0924] sending a second PCIe transaction from the storage device to the acceleration module; and

[0925] performing a first PCIe transaction on data stored on a storage device based at least in part on determining that the first PCIe transaction is not an accelerated instruction,

[0926] The processor, the acceleration module, and the storage device communicate using a peripheral component interconnect express (PCIe) bus, and

[0927] The acceleration module supports executing acceleration instructions on application data on a storage device for an application running on the processor without loading the application data into a memory associated with the processor.

[0928] Statement 365. An embodiment of the inventive concept includes the method of Statement 364, wherein the storage device is a solid state drive (SSD).

[0929] Statement 366. An embodiment of the inventive concept comprises a method according to statement 365, wherein:

[0930] Receiving a first PCIe transaction from an acceleration module of a storage device includes receiving a first PCIe transaction from an acceleration module at an endpoint of the SSD;

[0931] Determining whether the first PCIe transaction is an acceleration instruction includes determining whether the first PCIe transaction includes a dedicated command from a processor or originates from an APM-F of an acceleration module;

[0932] Generating a second PCIe transaction using a storage device acceleration platform manager (APM-S) of the storage device includes generating, by the APM-S of the SSD, the second PCIe transaction in response to the first PCIe transaction; and

[0933] Sending the second PCIe transaction from the storage device to the acceleration module includes sending the second PCIe transaction from an endpoint of the SSD to the acceleration module.

[0934] Statement 367. An embodiment of the inventive concept comprises the method of Statement 366, wherein the first PCIe transaction originates from the processor and comprises a dedicated command.

[0935] Statement 368. An embodiment of the inventive concept comprises a method according to statement 366, wherein determining whether the first PCIe transaction includes a dedicated command from the processor comprises determining whether the first PCIe transaction includes a dedicated command from the processor by a host interface logic (HIL) of the SSD.

[0936] Statement 369. An embodiment of the inventive concept comprises the method of statement 368, wherein the specialized command originates from an acceleration service manager (ASM) running on the processor.

[0937] Statement 370. An embodiment of the inventive concept comprises the method according to statement 366, said method further comprising:

[0938] Request a block of host system addresses from the processor;

[0939] selecting a subset of the block of host system addresses as a downstream filter address range (FAR); and

[0940] Use the downstream FAR to program the downstream port of the acceleration module.

[0941] Statement 371. An embodiment of the inventive concept comprises the method of statement 370, wherein programming the downstream port of the acceleration module with the downstream FAR comprises programming the downstream port of the acceleration module with the downstream FAR via a sideband bus from a processor comprising an inter-integrated circuit (I 2 C) A collection of buses and the System Management Bus (SMBus).

[0942] Statement 372. An embodiment of the inventive concept includes a method according to statement 370, wherein programming the downstream port of the acceleration module with the downstream FAR includes using a PCIe vendor defined message (VDM) to program the downstream port of the acceleration module with the downstream FAR, and the PCIe VDM includes the downstream FAR.

[0943] Statement 373. An embodiment of the inventive concept comprises a method according to statement 370, said method further comprising:

[0944] selecting a second subset of the block of host system addresses as upstream FARs; and

[0945] Program the upstream port of the acceleration module with the upstream FAR.

[0946] Statement 374. An embodiment of the inventive concept comprises the method of statement 373, wherein programming the upstream port of the acceleration module with the upstream FAR comprises programming the upstream port of the acceleration module with the upstream FAR via a sideband bus from a processor comprising I 2 C-bus and SMBus.

[0947] Statement 375. An embodiment of the inventive concept includes a method according to statement 373, wherein programming the upstream port of the acceleration module with the upstream FAR includes using a PCIe vendor defined message (VDM) to program the upstream port of the acceleration module with the upstream FAR, and the PCIe VDM includes the upstream FAR.

[0948] Statement 376. An embodiment of the inventive concept comprises a method according to statement 366, said method further comprising:

[0949] Request a block of host system addresses from the processor;

[0950] selecting a subset of the block of host system addresses as downstream FARs; and

[0951] Program the root port of the acceleration module with the downstream FAR.

[0952] Statement 377. An embodiment of the inventive concept comprises the method of statement 376, wherein programming the root port of the acceleration module with the downstream FAR comprises programming the root port of the acceleration module with the downstream FAR via a sideband bus from a processor comprising I 2 C-bus and SMBus.

[0953] Statement 378. An embodiment of the inventive concept includes a method according to statement 376, wherein programming the root port of the acceleration module with a downstream FAR includes using a PCIe vendor defined message (VDM) to program the root port of the acceleration module with the downstream FAR, the PCIe VDM including the downstream FAR.

[0954] Statement 379. An embodiment of the inventive concept comprises the method according to statement 366, said method further comprising:

[0955] receiving a result of the first PCIe transaction from the acceleration module at an endpoint of the SSD; and

[0956] The results of the first PCIe transaction are forwarded to the processor using the endpoint of the SSD.

[0957] Statement 380. An embodiment of the inventive concept comprises the method according to statement 366, said method further comprising:

[0958] Providing a physical function (PF) for exposing the SSD; and

[0959] Provides virtual functions (VFs) that expose acceleration modules.

[0960] Statement 381. An embodiment of the inventive concept comprises the method according to Statement 380, the method further comprising programming an upstream port of the acceleration module with an identifier of the VF.

[0961] Statement 382. An embodiment of the inventive concept comprises the method according to statement 366, said method further comprising:

[0962] providing a first PF to expose the SSD; and

[0963] A second PF is provided to expose the acceleration module.

[0964] Statement 383. An embodiment of the inventive concept includes the method according to Statement 382, ​​the method further comprising programming an upstream port of the acceleration module with an identifier of the second PF.

[0965] Statement 384. An embodiment of the inventive concept comprises a method according to statement 365, wherein:

[0966] Receiving a first PCIe transaction from an acceleration module of a storage device includes receiving a first PCIe transaction from an acceleration module at an endpoint of the SSD;

[0967] Determining whether the first PCIe transaction is an acceleration instruction includes determining whether the first PCIe transaction is received by an SSD at a second endpoint, the SSD including the second endpoint and the first endpoint;

[0968] Generating a second PCIe transaction using a storage device acceleration platform manager (APM-S) of the storage device includes generating, by the APM-S of the SSD, the second PCIe transaction in response to the first PCIe transaction; and

[0969] Sending the second PCIe transaction from the storage device to the acceleration module includes sending the second PCIe transaction from the second endpoint of the SSD to the acceleration module.

[0970] Statement 385. An embodiment of the inventive concept comprises an article of manufacture comprising a non-transitory storage medium having stored thereon instructions that when executed by a machine cause the following operations:

[0971] receiving, at an acceleration module, a peripheral component interconnect express (PCIe) transaction from a first device;

[0972] Determining at the acceleration module whether the PCIe transaction is an acceleration instruction;

[0973] Based at least in part on determining that the PCIe transaction is an acceleration instruction, processing the PCIe transaction at an acceleration platform manager (APM-F) of the acceleration module; and

[0974] transmitting the PCIe transaction to the second device based at least in part on determining that the PCIe transaction is not an accelerated instruction,

[0975] wherein the acceleration module supports execution of acceleration instructions on application data on a storage device for an application program without loading the application data into a memory associated with the processor, and

[0976] The processor, acceleration module and storage device communicate using the PCIe bus.

[0977] Statement 386. An embodiment of the inventive concept comprises an article of manufacture according to Statement 385, wherein the acceleration module is implemented using a field programmable gate array.

[0978] Statement 387. An embodiment of the inventive concept comprises an article according to Statement 386, wherein:

[0979] Receiving a peripheral component interconnect express (PCIe) transaction from a first device at the FPGA includes receiving a second PCIe transaction from a memory device at a downstream port of the FPGA;

[0980] Determining at the acceleration module whether the PCIe transaction is an acceleration instruction includes determining at the FPGA whether the second PCIe transaction is associated with an address in a downstream filter address range (FAR), the downstream filter address range being associated with a downstream port of the FPGA; and

[0981] Transmitting the PCIe transaction to the second device includes using an upstream port of the FPGA to transmit the second PCIe transaction to the processor.

[0982] Statement 388. An embodiment of the inventive concept includes an article of manufacture according to Statement 387, the non-transitory storage medium having stored thereon other instructions that when executed by a machine cause the following operations:

[0983] receiving a first PCIe transaction from a processor at an upstream port of the FPGA; and

[0984] A downstream processor of the FPGA is used to transmit a first PCIe transaction to a storage device.

[0985] Statement 389. An embodiment of the inventive concept includes an article of manufacture according to Statement 387, the non-transitory storage medium having stored thereon other instructions that when executed by a machine cause the following operations:

[0986] receiving, at the FPGA, a downstream FAR from a memory device; and

[0987] Associate the downstream FAR with the downstream port of the FPGA.

[0988] Statement 390. An embodiment of the inventive concept comprises an article of manufacture according to Statement 389, wherein receiving at the FPGA a downstream FAR from a storage device comprises receiving at the FPGA a downstream FAR from the storage device via a sideband bus from a memory device comprising an inter-integrated circuit (I 2 C) A collection of buses and the System Management Bus (SMBus).

[0989] Statement 391. An embodiment of the inventive concept comprises an article of manufacture according to Statement 389, wherein receiving, at the FPGA, a downstream FAR from a storage device comprises receiving a PCIe Vendor Defined Message (VDM) from the storage device, the PCIe VDM comprising the downstream FAR.

[0990] Statement 392. An embodiment of the inventive concept includes an article of manufacture according to Statement 387, the non-transitory storage medium having stored thereon other instructions that when executed by a machine cause the following operations: using a downstream port of the FPGA to send a result of the second PCIe transaction to a storage device.

[0991] Statement 393. An embodiment of the inventive concept comprises an article of manufacture according to Statement 387, the non-transitory storage medium having stored thereon other instructions that when executed by a machine cause the following operations: using an upstream port of the FPGA to send a result of the second PCIe transaction to the processor.

[0992] Statement 394. An embodiment of the inventive concept comprises an article of manufacture according to Statement 387, said non-transitory storage medium having stored thereon other instructions that when executed by a machine cause the following operations:

[0993] receiving a first PCIe transaction from a processor at an upstream port of the FPGA;

[0994] determining whether the first PCIe transaction is a second acceleration instruction by determining whether the first PCIe transaction is associated with a second address in an upstream FAR, the upstream FAR being associated with an upstream port of the FPGA; and

[0995] Based at least in part on determining that the first PCIe transaction is a second accelerated instruction, processing the first PCIe transaction at the APM-F of the FPGA; and

[0996] Based at least in part on determining that the first PCIe transaction is not a second acceleration instruction, the first PCIe transaction is transmitted to the storage device using a downstream port of the FPGA.

[0997] Statement 395. An embodiment of the inventive concept comprises an article of manufacture according to Statement 394, wherein the second acceleration instruction originates from an acceleration service manager (ASM) running on the processor.

[0998] Statement 396. An embodiment of the inventive concept comprises an article of manufacture according to Statement 394, the non-transitory storage medium having stored thereon other instructions that when executed by a machine cause the following operations:

[0999] receiving, at the FPGA, an upstream FAR from a memory device; and

[1000] Associate the upstream FAR with the upstream port of the FPGA.

[1001] Statement 397. An embodiment of the inventive concept comprises an article of manufacture according to Statement 396, wherein receiving at the FPGA an upstream FAR from a storage device comprises receiving at the FPGA an upstream FAR from the storage device via a sideband bus from a storage device comprising I 2 C-bus and SMBus.

[1002] Statement 398. An embodiment of the inventive concept comprises an article of manufacture according to statement 396, wherein receiving, at the FPGA, an upstream FAR from a storage device comprises receiving a PCIe Vendor Defined Message (VDM) from the storage device, the PCIe VDM comprising the upstream FAR.

[1003] Statement 399. An embodiment of the inventive concept includes an article of manufacture according to Statement 394, the non-transitory storage medium having stored thereon other instructions that when executed by a machine cause the following operations: using an upstream port of the FPGA to send a result of the first PCIe transaction to a storage device.

[1004] Statement 400. An embodiment of the inventive concept comprises the article of manufacture of statement 394, the non-transitory storage medium having stored thereon other instructions that when executed by a machine cause the following operations: using an upstream port of the FPGA to send a result of the first PCIe transaction to a processor.

[1005] Statement 401. An embodiment of the inventive concept comprises an article of manufacture according to Statement 387, said non-transitory storage medium having stored thereon other instructions that when executed by a machine cause the following operations:

[1006] receiving a first PCIe transaction from a processor at an upstream port of the FPGA;

[1007] determining whether the first PCIe transaction is a second acceleration instruction by determining whether the first PCIe transaction is associated with a virtual function (VF) exposed by the storage device; and

[1008] Based at least in part on determining that the first PCIe transaction is a second accelerated instruction, processing the first PCIe transaction at the APM-F of the FPGA; and

[1009] Based at least in part on determining that the first PCIe transaction is not a second acceleration instruction, the first PCIe transaction is transmitted to the storage device using a downstream port of the FPGA.

[1010] Statement 402. An embodiment of the inventive concept comprises an article of manufacture as recited in statement 401, wherein the second acceleration instruction originates from an acceleration service manager (ASM) executing on the processor.

[1011] Statement 403. An embodiment of the inventive concept comprises an article of manufacture according to statement 401, wherein determining whether the first PCIe transaction is a second acceleration instruction comprises determining whether the first PCIe transaction includes a tag having an identifier of the VF.

[1012] Statement 404. An embodiment of the inventive concept comprises an article of manufacture according to statement 403, said non-transitory storage medium having stored thereon other instructions that when executed by a machine cause the following operations:

[1013] Receiving, at the FPGA, an identifier of the VF from the storage device; and

[1014] Associate the VF's identifier with the FPGA's upstream port.

[1015] Statement 405. An embodiment of the inventive concept comprises an article of manufacture according to statement 404, wherein receiving at the FPGA an identifier of the VF from the storage device comprises receiving at the FPGA an identifier of the VF from the storage device via a sideband bus from a memory device comprising I 2 C-bus and SMBus.

[1016] Statement 406. An embodiment of the inventive concept includes the article of manufacture of statement 404, wherein receiving, at the FPGA, an identifier of the VF from a storage device includes receiving a PCIe Vendor Defined Message (VDM) from the storage device, the PCIe VDM including an identifier of the VF.

[1017] Statement 407. An embodiment of the inventive concept comprises an article of manufacture according to statement 401, the non-transitory storage medium having further instructions stored thereon that when executed by a machine cause the following operations: using an upstream port of the FPGA to send a result of the first PCIe transaction to a processor.

[1018] Statement 408. An embodiment of the inventive concept comprises an article of manufacture according to Statement 387, said non-transitory storage medium having stored thereon other instructions that when executed by a machine cause the following operations:

[1019] receiving a first PCIe transaction from a processor at an upstream port of the FPGA;

[1020] determining whether the first PCIe transaction is a second acceleration instruction by determining whether the first PCIe transaction is associated with a physical function (PF) exposed by the storage device; and

[1021] Based at least in part on determining that the first PCIe transaction is a second accelerated instruction, processing the first PCIe transaction at the APM-F of the FPGA; and

[1022] Based at least in part on determining that the first PCIe transaction is not a second acceleration instruction, the first PCIe transaction is transmitted to the storage device using a downstream port of the FPGA.

[1023] Statement 409. An embodiment of the inventive concept comprises an article of manufacture as recited in Statement 408, wherein the second acceleration instruction originates from an acceleration service manager (ASM) executing on the processor.

[1024] Statement 410. An embodiment of the inventive concept comprises the article of manufacture of statement 408, wherein determining whether the first PCIe transaction is a second acceleration instruction comprises determining whether the first PCIe transaction comprises a tag having an identifier of a PF.

[1025] Statement 411. An embodiment of the inventive concept comprises an article of manufacture according to statement 410, said non-transitory storage medium having stored thereon other instructions that when executed by a machine cause the following operations:

[1026] receiving, at the FPGA, an identifier of the PF from the storage device; and

[1027] Associates the identifier of the PF with the upstream port of the FPGA.

[1028] Statement 412. An embodiment of the inventive concept comprises an article of manufacture according to statement 411, wherein receiving at the FPGA an identifier of the PF from the storage device comprises receiving at the FPGA an identifier of the PF from the storage device via a sideband bus from a memory device comprising I 2 C-bus and SMBus.

[1029] Statement 413. An embodiment of the inventive concept comprises an article of manufacture according to statement 411, wherein receiving, at the FPGA, an identifier of the PF from a storage device comprises receiving a PCIe Vendor Defined Message (VDM) from the storage device, the PCIe VDM comprising an identifier of the PF.

[1030] Statement 414. An embodiment of the inventive concept comprises an article of manufacture according to statement 408, the non-transitory storage medium having stored thereon other instructions that when executed by a machine cause the following operations: using an upstream port of the FPGA to send a result of the first PCIe transaction to a processor.

[1031] Statement 415. An embodiment of the inventive concept comprises an article according to statement 386, wherein:

[1032] Receiving a peripheral component interconnect express (PCIe) transaction from a first device at the FPGA includes receiving a first PCIe transaction from a processor at an endpoint of the FPGA;

[1033] Determining at the acceleration module whether the PCIe transaction is an acceleration instruction includes determining whether the first PCIe transaction includes a tag having a first identifier of a first PF of the FPGA or a second identifier of a second PF of the FPGA; and

[1034] Transmitting the PCIe transaction to the second device includes using a root port of the FPGA to transmit the first PCIe transaction to the storage device.

[1035] Statement 416. An embodiment of the inventive concept comprises an article of manufacture according to Statement 415, said non-transitory storage medium having stored thereon other instructions that when executed by a machine cause the following operations:

[1036] receiving a second PCIe transaction from a memory device at a root port of the FPGA;

[1037] determining whether the second PCIe transaction is a second acceleration instruction by determining at the FPGA whether the second PCIe transaction is associated with an address in a downstream FAR, the downstream FAR being associated with a root port of the FPGA;

[1038] Based at least in part on determining that the second PCIe transaction is a second accelerated instruction, processing the second PCIe transaction at the APM-F of the FPGA; and

[1039] Based at least in part on determining that the second PCIe transaction is not a second acceleration instruction, the second PCIe transaction is communicated to the processor using an endpoint of the FPGA.

[1040] Statement 417. An embodiment of the inventive concept comprises an article of manufacture according to Statement 416, said non-transitory storage medium having stored thereon other instructions that when executed by a machine cause the following operations:

[1041] receiving, at the FPGA, a downstream FAR from a memory device; and

[1042] Associate the downstream FAR with the root port of the FPGA.

[1043] Statement 418. An embodiment of the inventive concept comprises an article of manufacture according to Statement 417, wherein receiving at the FPGA a downstream FAR from a storage device comprises receiving at the FPGA a downstream FAR from the storage device via a sideband bus from a storage device comprising I 2 C-bus and SMBus.

[1044] Statement 419. An embodiment of the inventive concept comprises an article of manufacture as recited in Statement 417, wherein receiving, at the FPGA, a downstream FAR from a storage device comprises receiving a PCIe Vendor Defined Message (VDM) from the storage device, the PCIe VDM comprising the downstream FAR.

[1045] Statement 420. An embodiment of the inventive concept comprises the article of manufacture of statement 415, wherein the second acceleration instruction originates from an acceleration service manager (ASM) executing on the processor.

[1046] Statement 421. An embodiment of the inventive concept comprises an article of manufacture according to statement 415, a non-transitory storage medium having stored thereon other instructions that when executed by a machine cause the following operations: using an endpoint of the FPGA to send a result of the first PCIe transaction to a processor.

[1047] Statement 422. An embodiment of the inventive concept comprises an article of manufacture according to Statement 415, said non-transitory storage medium having stored thereon other instructions which when executed by a machine cause the following operations:

[1048] determining, using a configuration monitor of the FPGA, a configuration of an endpoint of a memory device that communicates with a root port of the FPGA; and

[1049] Use the endpoints of the FPGA to replicate the configuration of the endpoints on the storage device.

[1050] Statement 423. An embodiment of the inventive concept comprises an article according to statement 386, wherein:

[1051] Receiving a peripheral component interconnect express (PCIe) transaction from a first device at the FPGA includes receiving a first PCIe transaction from a processor at an endpoint of the FPGA;

[1052] Determining at the acceleration module whether the PCIe transaction is an acceleration instruction includes determining whether the first PCIe transaction includes a tag having a first identifier of a first PF of the FPGA or a second identifier of a second PF of the FPGA; and

[1053] Transmitting the PCIe transaction to the second device includes using a first root port of the FPGA to transmit the first PCIe transaction to the storage device.

[1054] Statement 424. An embodiment of the inventive concept comprises an article of manufacture according to statement 423, said non-transitory storage medium having stored thereon other instructions that when executed by a machine cause the following operations:

[1055] receiving, at the FPGA, a second PCIe transaction from the memory device;

[1056] determining whether the second PCIe transaction is received at the first root port of the FPGA or at the second root port of the FPGA;

[1057] Based at least in part on determining that the second PCIe transaction was received at the first root port of the FPGA, transmitting the second PCIe transaction to the processor using the endpoint; and

[1058] Based at least in part on determining that the second PCIe transaction was received at the second root port of the FPGA, the second PCIe transaction is processed at the APM-F of the FPGA.

[1059] Statement 425. An embodiment of the inventive concept comprises an article of manufacture as recited in Statement 423, wherein the second acceleration instruction originates from an acceleration service manager (ASM) executing on the processor.

[1060] Statement 426. An embodiment of the inventive concept comprises an article of manufacture according to statement 423, a non-transitory storage medium having stored thereon other instructions that when executed by a machine cause the following operations: using an endpoint of the FPGA to send a result of the first PCIe transaction to a processor.

[1061] Statement 427. An embodiment of the inventive concept comprises an article of manufacture according to Statement 423, said non-transitory storage medium having stored thereon other instructions that when executed by a machine cause the following operations:

[1062] determining, using a configuration monitor of the FPGA, a configuration of an endpoint of a memory device that communicates with a first root port of the FPGA; and

[1063] Use the endpoints of the FPGA to replicate the configuration of the endpoints on the storage device.

[1064] Statement 428. An embodiment of the inventive concept comprises an article according to statement 386, wherein:

[1065] Receiving a peripheral component interconnect express (PCIe) transaction from a first device at the FPGA includes receiving a first PCIe transaction from a processor at the FPGA;

[1066] Determining at the FPGA whether the first PCIe transaction is a speed-up instruction includes determining whether the first PCIe transaction is a second speed-up instruction by determining whether the first PCIe transaction is received from a processor at a first endpoint of the FPGA, the FPGA including a first endpoint and a second endpoint; and

[1067] Transmitting the PCIe transaction to the second device includes transmitting the first PCIe transaction to the memory device using the first root port of the FPGA, the FPGA including the first root port and the second root port.

[1068] Statement 429. An embodiment of the inventive concept comprises an article of manufacture according to Statement 428, said non-transitory storage medium having stored thereon other instructions that when executed by a machine cause the following operations:

[1069] receiving, at the FPGA, a second PCIe transaction from the memory device;

[1070] determining whether the second PCIe transaction is an accelerated instruction by determining whether the second PCIe transaction is received at the first root port of the FPGA or at the second root port of the FPGA;

[1071] Based at least in part on determining that the second PCIe transaction was received at the first root port of the FPGA, transmitting the second PCIe transaction to the processor using the first endpoint; and

[1072] Based at least in part on determining that the second PCIe transaction was received at the second root port of the FPGA, the second PCIe transaction is processed at the APM-F of the FPGA.

[1073] Statement 430. An embodiment of the inventive concept comprises the article of manufacture of statement 428, wherein the second PCIe transaction originates from an acceleration service manager (ASM) running on the processor.

[1074] Statement 431. An embodiment of the inventive concept comprises an article of manufacture according to statement 428, the non-transitory storage medium having further instructions stored thereon that when executed by a machine cause the following operations: using the second endpoint of the FPGA to send a result of the first PCIe transaction to the processor.

[1075] Statement 432. An embodiment of the inventive concept comprises an article of manufacture according to Statement 428, said non-transitory storage medium having stored thereon other instructions which when executed by a machine cause the following operations:

[1076] determining, using a configuration monitor of the FPGA, a configuration of an endpoint of a memory device that communicates with a first root port of the FPGA; and

[1077] The configuration of the endpoint on the memory device is replicated using the first endpoint of the FPGA.

[1078] Statement 433. An embodiment of the inventive concept comprises an article according to statement 386, wherein:

[1079] Receiving a peripheral component interconnect express (PCIe) transaction from a first device at the FPGA includes receiving a first PCIe transaction from a processor at the FPGA;

[1080] Determining at the acceleration module whether the first PCIe transaction is an acceleration instruction includes determining whether the first PCIe transaction is a second acceleration instruction by determining whether the first PCIe transaction is received from a processor at a first endpoint of an FPGA, the FPGA including a first endpoint and a second endpoint; and

[1081] Transmitting the PCIe transaction to the second device includes using a root port of the FPGA to transmit the first PCIe transaction to the storage device.

[1082] Statement 434. An embodiment of the inventive concept comprises an article of manufacture according to statement 433, said non-transitory storage medium having stored thereon other instructions that when executed by a machine cause the following operations:

[1083] receiving a second PCIe transaction from a memory device at a root port of the FPGA;

[1084] determining whether the second PCIe transaction is an accelerated instruction by determining at the FPGA whether the second PCIe transaction is associated with an address in a downstream FAR, the downstream FAR being associated with a root port of the FPGA;

[1085] Based at least in part on determining that the second PCIe transaction is an accelerated instruction, processing the second PCIe transaction at the APM-F of the FPGA; and

[1086] Based at least in part on determining that the second PCIe transaction is not an accelerated instruction, the second PCIe transaction is communicated to the processor using the first endpoint of the FPGA.

[1087] Statement 435. An embodiment of the inventive concept comprises an article of manufacture according to statement 434, said non-transitory storage medium having stored thereon other instructions that when executed by a machine cause the following operations:

[1088] receiving, at the FPGA, a downstream FAR from a memory device; and

[1089] Associate the downstream FAR with the root port of the FPGA.

[1090] Statement 436. An embodiment of the inventive concept comprises an article of manufacture according to Statement 435, wherein receiving at the FPGA a downstream FAR from a storage device comprises receiving at the FPGA a downstream FAR from the storage device via a sideband bus from a storage device comprising I 2 C-bus and SMBus.

[1091] Statement 437. An embodiment of the inventive concept comprises an article of manufacture according to statement 435, wherein receiving, at the FPGA, a downstream FAR from a storage device comprises receiving a PCIe Vendor Defined Message (VDM) from the storage device, the PCIe VDM comprising the downstream FAR.

[1092] Statement 438. An embodiment of the inventive concept comprises an article of manufacture as recited in Statement 433, wherein the first PCIe transaction originates from an Acceleration Services Manager (ASM) running on the processor.

[1093] Statement 439. An embodiment of the inventive concept comprises an article of manufacture according to statement 433, the non-transitory storage medium having stored thereon other instructions that when executed by a machine cause the following operations: using the second endpoint of the FPGA to send the result of the first PCIe transaction to the processor.

[1094] Statement 440. An embodiment of the inventive concept comprises an article of manufacture according to statement 433, said non-transitory storage medium having stored thereon other instructions that when executed by a machine cause the following operations:

[1095] determining, using a configuration monitor of the FPGA, a configuration of an endpoint of a memory device that communicates with a root port of the FPGA; and

[1096] The configuration of the endpoint on the memory device is replicated using the first endpoint of the FPGA.

[1097] Statement 441. An embodiment of the inventive concept comprises an article of manufacture comprising a non-transitory storage medium having stored thereon instructions that, when executed by a machine, cause the following operations:

[1098] receiving, at a storage device, a first PCIe transaction from an acceleration module;

[1099] Determining whether the first PCIe transaction is an acceleration instruction;

[1100] Based at least in part on determining that the first PCIe transaction is an accelerated instruction:

[1101] generating a second PCIe transaction using an Accelerated Platform Manager for Storage (APM-S) of the storage device; and

[1102] sending a second PCIe transaction from the storage device to the acceleration module; and

[1103] performing a first PCIe transaction on data stored on a storage device based at least in part on determining that the first PCIe transaction is not an accelerated instruction,

[1104] The processor, the acceleration module, and the storage device communicate using a peripheral component interconnect express (PCIe) bus, and

[1105] The acceleration module supports executing acceleration instructions on application data on a storage device for an application running on the processor without loading the application data into a memory associated with the processor.

[1106] Statement 442. An embodiment of the inventive concept comprises an article of manufacture according to Statement 441, wherein the storage device is a solid state drive (SSD).

[1107] Statement 443. An embodiment of the inventive concept comprises an article according to statement 442, wherein:

[1108] Receiving a first PCIe transaction from an acceleration module of a storage device includes receiving a first PCIe transaction from an acceleration module at an endpoint of the SSD;

[1109] Determining whether the first PCIe transaction is an acceleration instruction includes determining whether the first PCIe transaction includes a dedicated command from a processor or originates from an APM-F of an acceleration module;

[1110] Generating a second PCIe transaction using a storage device acceleration platform manager (APM-S) of the storage device includes generating, by the APM-S of the SSD, the second PCIe transaction in response to the first PCIe transaction; and

[1111] Sending the second PCIe transaction from the storage device to the acceleration module includes sending the second PCIe transaction from an endpoint of the SSD to the acceleration module.

[1112] Statement 444. An embodiment of the inventive concept comprises an article of manufacture as recited in Statement 443, wherein the first PCIe transaction originates from a processor and comprises a dedicated command.

[1113] Statement 445. An embodiment of the inventive concept comprises an article of manufacture according to statement 443, wherein determining whether the first PCIe transaction includes a dedicated command from the processor comprises determining whether the first PCIe transaction includes a dedicated command from the processor by a host interface logic (HIL) of the SSD.

[1114] Statement 446. An embodiment of the inventive concept comprises an article of manufacture as recited in Statement 445, wherein the specialized command originates from an Acceleration Services Manager (ASM) running on the processor.

[1115] Statement 447. An embodiment of the inventive concept comprises an article of manufacture according to Statement 443, said non-transitory storage medium having stored thereon other instructions that when executed by a machine cause the following operations:

[1116] Request a block of host system addresses from the processor;

[1117] selecting a subset of the block of host system addresses as a downstream filter address range (FAR); and

[1118] Use the downstream FAR to program the downstream port of the acceleration module.

[1119] Statement 448. An embodiment of the inventive concept comprises an article of manufacture according to statement 447, wherein programming the downstream port of the acceleration module with the downstream FAR comprises programming the downstream port of the acceleration module with the downstream FAR via a sideband bus from a processor comprising an inter-integrated circuit (I 2 C) A collection of buses and the System Management Bus (SMBus).

[1120] Statement 449. An embodiment of the inventive concept includes an article of manufacture according to statement 447, wherein programming a downstream port of an acceleration module with a downstream FAR includes using a PCIe vendor defined message (VDM) to program the downstream port of the acceleration module with a downstream FAR, wherein the PCIe VDM includes the downstream FAR.

[1121] Statement 450. An embodiment of the inventive concept comprises an article of manufacture according to Statement 447, said non-transitory storage medium having stored thereon other instructions that when executed by a machine cause the following operations:

[1122] selecting a second subset of the block of host system addresses as upstream FARs; and

[1123] Program the upstream port of the acceleration module with the upstream FAR.

[1124] Statement 451. An embodiment of the inventive concept comprises an article of manufacture according to statement 450, wherein programming the upstream port of the acceleration module with the upstream FAR comprises programming the upstream port of the acceleration module with the upstream FAR via a sideband bus from a processor comprising I 2 C-bus and SMBus.

[1125] Statement 452. An embodiment of the inventive concept includes an article of manufacture according to statement 450, wherein programming an upstream port of an acceleration module with an upstream FAR includes using a PCIe vendor defined message (VDM) to program the upstream port of the acceleration module with an upstream FAR, wherein the PCIe VDM includes the upstream FAR.

[1126] Statement 453. An embodiment of the inventive concept comprises an article of manufacture according to Statement 443, said non-transitory storage medium having stored thereon other instructions that when executed by a machine cause the following operations:

[1127] Request a block of host system addresses from the processor;

[1128] selecting a subset of the block of host system addresses as downstream FARs; and

[1129] Program the root port of the acceleration module with the downstream FAR.

[1130] Statement 454. An embodiment of the inventive concept comprises an article of manufacture according to statement 453, wherein programming the root port of the acceleration module with the downstream FAR comprises programming the root port of the acceleration module with the downstream FAR via a sideband bus from a processor comprising I 2 C-bus and SMBus.

[1131] Statement 455. An embodiment of the inventive concept includes an article of manufacture according to statement 453, wherein programming the root port of the acceleration module with a downstream FAR includes using a PCIe vendor defined message (VDM) to program the root port of the acceleration module with the downstream FAR, the PCIe VDM including the downstream FAR.

[1132] Statement 456. An embodiment of the inventive concept comprises an article of manufacture according to Statement 443, said non-transitory storage medium having stored thereon other instructions that when executed by a machine cause the following operations:

[1133] receiving a result of the first PCIe transaction from the acceleration module at an endpoint of the SSD; and

[1134] The results of the first PCIe transaction are forwarded to the processor using the endpoint of the SSD.

[1135] Statement 457. An embodiment of the inventive concept comprises an article of manufacture according to Statement 443, said non-transitory storage medium having stored thereon other instructions which when executed by a machine cause the following operations:

[1136] Providing a physical function (PF) for exposing the SSD; and

[1137] Provides virtual functions (VFs) that expose acceleration modules.

[1138] Statement 458. An embodiment of the inventive concept comprises an article of manufacture according to statement 457, the non-transitory storage medium having stored thereon other instructions that when executed by a machine cause the following operations: programming an upstream port of the acceleration module with an identifier of the VF.

[1139] Statement 459. An embodiment of the inventive concept comprises an article of manufacture according to Statement 443, said non-transitory storage medium having stored thereon other instructions which when executed by a machine cause the following operations:

[1140] providing a first PF to expose the SSD; and

[1141] A second PF is provided to expose the acceleration module.

[1142] Statement 460. An embodiment of the inventive concept comprises an article of manufacture according to Statement 459, the non-transitory storage medium having stored thereon other instructions that when executed by a machine cause the following operations: programming an upstream port of the acceleration module with an identifier of the second PF.

[1143] Statement 461. An embodiment of the inventive concept comprises an article according to Statement 442, wherein:

[1144] Receiving a first PCIe transaction from an acceleration module of a storage device includes receiving a first PCIe transaction from an acceleration module at an endpoint of the SSD;

[1145] Determining whether the first PCIe transaction is an acceleration instruction includes determining whether the first PCIe transaction is received by an SSD at a second endpoint, the SSD including the second endpoint and the first endpoint;

[1146] Generating a second PCIe transaction using a storage device acceleration platform manager (APM-S) of the storage device includes generating, by the APM-S of the SSD, the second PCIe transaction in response to the first PCIe transaction; and

[1147] Sending the second PCIe transaction from the storage device to the acceleration module includes sending the second PCIe transaction from the second endpoint of the SSD to the acceleration module.

[1148] Statement 462. An embodiment of the inventive concept comprises an article of manufacture comprising a non-transitory storage medium having stored thereon instructions that when executed by a machine cause the following operations:

[1149] receiving a peripheral component interconnect express (PCIe) transaction from a processor at the first bridge component;

[1150] determining, at the first bridge component, whether the PCIe transaction is an accelerated instruction;

[1151] forwarding the PCIe transaction to an acceleration module based at least in part on determining that the PCIe transaction is an acceleration instruction; and

[1152] forwarding the PCIe transaction to a storage device based at least in part on determining that the PCIe transaction is not an accelerated instruction,

[1153] The processor, the first bridge component, the acceleration module and the storage device communicate using a PCIe bus.

[1154] Statement 463. An embodiment of the inventive concept comprises an article according to Statement 462, wherein:

[1155] The acceleration module is implemented using a field programmable gate array; and

[1156] The storage device includes a solid state drive (SSD).

[1157] Statement 464. An embodiment of the inventive concept comprises an article of manufacture according to statement 462, wherein determining at the first bridging component whether a PCIe transaction is an accelerated instruction comprises determining whether the PCIe transaction is associated with an address in an upstream FAR, wherein the upstream FAR is associated with an upstream port of the first bridging component.

[1158] Statement 465. An embodiment of the inventive concept comprises an article of manufacture according to statement 464, said non-transitory storage medium having stored thereon other instructions which when executed by a machine cause the following operations:

[1159] receiving, at the first bridge component, an upstream FAR from a storage device; and

[1160] An upstream FAR is associated with an upstream port of the first bridging component.

[1161] Statement 466. An embodiment of the inventive concept comprises an article of manufacture according to Statement 465, wherein receiving the upstream FAR from the storage device at the first bridge component comprises receiving the upstream FAR from the storage device at the first bridge component via a sideband bus from a storage device comprising I 2 C-bus and SMBus.

[1162] Statement 467. An embodiment of the inventive concept comprises the article of manufacture of statement 465, wherein receiving, at the first bridging component, an upstream FAR from the storage device comprises receiving a PCIe Vendor Defined Message (VDM) from the storage device, the PCIe VDM comprising the upstream FAR.

[1163] Statement 468. An embodiment of the inventive concept comprises an article of manufacture according to statement 462, wherein determining, at the first bridge component, whether the PCIe transaction is an accelerated instruction comprises determining whether the PCIe transaction is associated with a virtual function (VF) exposed by the storage device.

[1164] Statement 469. An embodiment of the inventive concept comprises an article of manufacture as recited in Statement 468, wherein determining whether a PCIe transaction is associated with a virtual function (VF) exposed by a storage device comprises determining whether the PCIe transaction includes a tag having an identifier of the VF.

[1165] Statement 470. An embodiment of the inventive concept comprises an article of manufacture according to Statement 469, said non-transitory storage medium having stored thereon other instructions that when executed by a machine cause the following operations:

[1166] receiving, at the first bridging component, an identifier of the VF from the storage device; and

[1167] An identifier of the VF is associated with an upstream port of the first bridging component.

[1168] Statement 471. An embodiment of the inventive concept comprises an article of manufacture according to statement 470, wherein receiving, at the first bridge component, an identifier of the VF from the storage device comprises receiving, at the first bridge component, an identifier of the VF from the storage device via a sideband bus from a storage device comprising I 2 C-bus and SMBus.

[1169] Statement 472. An embodiment of the inventive concept includes an article of manufacture according to statement 470, wherein receiving, at the first bridging component, an identifier of the VF from the storage device includes receiving a PCIe Vendor Defined Message (VDM) from the storage device, the PCIe VDM including an identifier of the VF.

[1170] Statement 473. An embodiment of the inventive concept comprises an article of manufacture according to statement 462, wherein determining at the first bridge component whether the PCIe transaction is an accelerated instruction comprises determining whether the PCIe transaction is associated with a physical function (PF) exposed by the storage device.

[1171] Statement 474. An embodiment of the inventive concept comprises an article of manufacture as recited in Statement 473, wherein determining whether a PCIe transaction is associated with a physical function (PF) exposed by a storage device comprises determining whether the PCIe transaction comprises a tag having an identifier of the PF.

[1172] Statement 475. An embodiment of the inventive concept comprises an article of manufacture according to statement 474, said non-transitory storage medium having stored thereon other instructions that when executed by a machine cause the following operations:

[1173] receiving, at the first bridging component, an identifier of the PF from the storage device; and

[1174] An identifier of the PF is associated with an upstream port of the first bridging component.

[1175] Statement 476. An embodiment of the inventive concept comprises an article of manufacture according to Statement 475, wherein receiving, at the first bridge component, an identifier of the PF from the storage device comprises receiving, at the first bridge component, an identifier of the PF from the storage device via a sideband bus from a processor comprising I 2C-bus and SMBus.

[1176] Statement 477. An embodiment of the inventive concept includes an article of manufacture according to statement 475, wherein receiving, at the first bridging component, an identifier of the PF from the storage device includes receiving a PCIe Vendor Defined Message (VDM) from the storage device, the PCIe VDM including an identifier of the PF.

[1177] Statement 478. An embodiment of the inventive concept includes an article according to statement 462, wherein determining at the first bridge component whether a PCIe transaction is an acceleration instruction includes determining whether the PCIe transaction includes a tag having an identifier of a first PF of the first bridge component or a second identifier of a second PF of the first bridge component.

[1178] Statement 479. An embodiment of the inventive concept comprises an article of manufacture according to Statement 478, said non-transitory storage medium having stored thereon other instructions that when executed by a machine cause the following operations:

[1179] determining, using a configuration monitor of the first bridging component, a configuration of an endpoint of the storage device; and

[1180] The configuration of the endpoints on the storage device is replicated using the endpoints of the first bridge component.

[1181] Statement 480. An embodiment of the inventive concept comprises an article of manufacture according to statement 462, wherein determining at a first bridge component whether a PCIe transaction is an acceleration instruction comprises determining whether the PCIe transaction is received from a processor at a first endpoint of the first bridge component, the first bridge component comprising a first endpoint and a second endpoint.

[1182] Statement 481. An embodiment of the inventive concept comprises an article of manufacture according to statement 480, said non-transitory storage medium having stored thereon other instructions that when executed by a machine cause the following operations:

[1183] determining, using a configuration monitor of the first bridging component, a configuration of an endpoint of the storage device; and

[1184] The configuration of the endpoint on the storage device is replicated using the first endpoint of the first bridge component.

[1185] Statement 482. An embodiment of the inventive concept comprises an article of manufacture comprising a non-transitory storage medium having stored thereon instructions that when executed by a machine cause the following operations:

[1186] receiving a peripheral component interconnect express (PCIe) transaction from a storage device at a second bridge component;

[1187] determining, at the second bridge component, whether the PCIe transaction is an accelerated instruction;

[1188] forwarding the PCIe transaction to an acceleration module based at least in part on determining that the PCIe transaction is an acceleration instruction; and

[1189] forwarding the PCIe transaction to the processor based at least in part on determining that the PCIe transaction is not an accelerated instruction,

[1190] The processor, the second bridge component, the acceleration module and the storage device communicate using a PCIe bus.

[1191] Statement 483. An embodiment of the inventive concept comprises an article according to Statement 482, wherein:

[1192] The acceleration module is implemented using a field programmable gate array; and

[1193] The storage device includes a solid state drive (SSD).

[1194] Statement 484. An embodiment of the inventive concept includes an article according to statement 482, wherein determining at the second bridging component whether a PCIe transaction is an accelerated instruction includes determining at the second bridging component whether the second PCIe transaction is associated with an address in a downstream filter address range (FAR), wherein the downstream filter address range is associated with a downstream port of the second bridging component.

[1195] Statement 485. An embodiment of the inventive concept comprises an article of manufacture according to Statement 484, said non-transitory storage medium having stored thereon other instructions that when executed by a machine cause the following operations:

[1196] receiving, at the second bridge component, a downstream FAR from the storage device; and

[1197] A downstream FAR is associated with a downstream port of the second bridging component.

[1198] Statement 486. An embodiment of the inventive concept comprises an article of manufacture according to Statement 485, wherein receiving the downstream FAR from the storage device at the second bridge component comprises receiving the downstream FAR from the storage device at the second bridge component via a sideband bus from a processor comprising an inter-integrated circuit (I 2 C) A collection of buses and the System Management Bus (SMBus).

[1199] Statement 487. An embodiment of the inventive concept comprises the article of manufacture of statement 485, wherein receiving, at the second bridging component, a downstream FAR from the storage device comprises receiving a PCIe Vendor Defined Message (VDM) from the storage device, the PCIe VDM comprising the downstream FAR.

[1200] Statement 488. An embodiment of the inventive concept includes an article according to statement 482, wherein determining at a second bridging component whether a PCIe transaction is an acceleration instruction includes determining whether the PCIe transaction is received at a second root port of the second bridging component, the second bridging component including a first root port and a second root port.

[1201] Therefore, in view of various permutations of the embodiments described herein, this detailed description and the accompanying materials are intended to be illustrative only and should not be taken as limiting the scope of the inventive concept. Therefore, what is claimed for the inventive concept is all such modifications that fall within the scope and spirit of the appended claims and their equivalents.

Claims

1. A host system, It is characterized in that include: a processor, the processor executing an application program; a memory storing data used by the application running on the processor; an upstream port for communicating with the processor; a downstream port for communicating with a storage device; an acceleration module implemented using hardware and including an acceleration platform manager to execute acceleration instructions; and The storage device comprises: An endpoint of the storage device is used to communicate with the acceleration module; a physical function operable to expose the storage device; a second function, operable to expose the acceleration module; a controller for managing the operation of the storage device; Storage, for storing application data of the application; as well as A storage device acceleration platform manager is used to assist the acceleration platform manager in executing acceleration instructions. a downstream filter associated with the downstream port, the downstream filter being operative to intercept an acceleration instruction received from the storage device and to transmit the acceleration instruction to the acceleration platform manager, the acceleration instruction being associated with a downstream filter address range; as well as an upstream filter associated with the upstream port, the upstream filter being operative to intercept a second acceleration instruction received from the processor and to transmit the second acceleration instruction to the acceleration platform manager, wherein the processor, the acceleration module and the storage device communicate via a peripheral component interconnect express bus, and The acceleration module supports executing the acceleration instruction on the application data on the storage device for the application program without loading the application data into the memory.

2. The host system according to claim 1, It is characterized in that Also includes: A first bridging component, comprising the upstream port and the upstream filter, the first bridging component bridging the communication between the processor and the acceleration module; as well as The second bridge component includes the downstream port and the downstream filter, and the second bridge component bridges the communication between the acceleration module and the storage device.

3. The host system according to claim 1, Features: Implementing the acceleration module using a field programmable gate array; The acceleration module includes the upstream port, the upstream filter, the downstream port, and the downstream filter; and The storage device includes a solid state drive.

4. The host system according to claim 3, Features: The physical function is operable to request a first block of host system addresses from the processor; The controller is operative to select a first subset of a first block of the host system addresses as the downstream filter address range.

5. The host system according to claim 4, It is characterized in that The controller is operative to program the downstream filter with the downstream filter address range.

6. The host system according to claim 3, It is characterized in that The acceleration platform manager is operative to send results to the processor via the upstream port.

7. The host system according to claim 3, Features: The solid state drive further includes a virtual function operative to expose the field programmable gate array; The second acceleration instruction includes an identifier of the virtual function; and The upstream filter is operative to intercept the second acceleration instruction associated with the identifier of the virtual function.

8. The host system according to claim 3, Features: The solid state drive also includes a second physical function operative to expose the field programmable gate array; The second acceleration instruction includes an identifier of the second physical function; and The upstream filter is operative to intercept the second acceleration instruction associated with the identifier of the second physical function.

9. An acceleration module implemented using hardware, It is characterized in that include: An acceleration platform manager for executing acceleration instructions; An upstream port for communicating with a processor on which an application program is running; as well as a downstream port for communicating with a storage device, wherein the storage device includes a storage device acceleration platform manager to assist the acceleration platform manager in executing the acceleration instruction; a downstream filter associated with the downstream port, the downstream filter being operative to intercept an acceleration instruction received from the storage device and to transmit the acceleration instruction to the acceleration platform manager, the acceleration instruction being associated with a downstream filter address range; as well as an upstream filter associated with the upstream port, the upstream filter being operative to intercept a second acceleration instruction received from the processor and to transmit the second acceleration instruction to the acceleration platform manager, wherein the acceleration module communicates with the processor and the storage device using a peripheral component interconnect express bus, wherein the acceleration module supports executing the acceleration instructions on the application data on the storage device for the application program without loading the application data into a memory associated with the processor, and The acceleration module is exposed through the function of the storage device.

10. The acceleration module according to claim 9, Features: The acceleration module is implemented using a field programmable gate array.

11. The acceleration module according to claim 10, It is characterized in that The acceleration platform manager is operative to send results to the processor via the upstream port.

12. The acceleration module according to claim 10, Features: The field programmable gate array is exposed through a virtual function of the storage device; The second acceleration instruction includes an identifier of the virtual function; and The upstream filter is operative to intercept the second acceleration instruction associated with the identifier of the virtual function.

13. The acceleration module according to claim 10, Features: The field programmable gate array is exposed through a physical function of the storage device; The second acceleration instruction includes an identifier of the physical function; and The upstream filter is operative to intercept the second acceleration instruction associated with the identifier of the physical function.

14. A method of a host system, include: receiving a first peripheral component interconnect express transaction from a processor at an upstream port at an acceleration module; determining whether the first PCI fast transaction is a second acceleration instruction by determining whether the first PCI fast transaction is associated with a function of a storage device that exposes the acceleration module; processing the first PCI express transaction at an acceleration manager of the acceleration module based at least in part on determining that the first PCI express transaction is the second speed up instruction; and Based at least in part on determining that the first PCI fast transaction is not the second speed up instruction, transmitting the first PCI fast transaction to the memory device using a downstream port of the speed up module, wherein the acceleration module supports execution of acceleration instructions on the application data on the storage device for an application program without loading the application data into a memory associated with the processor, and The processor, the acceleration module and the storage device communicate via a peripheral component interconnect express bus.

15. The method according to claim 14, It is characterized in that The acceleration module is implemented using a field programmable gate array.

16. The method according to claim 15, It is characterized in that Also includes: The result of the first peripheral component interconnect fast transaction is sent to the processor using the upstream port of the field programmable gate array.

17. The method according to claim 15, It is characterized in that Determining whether the first PCI fast transaction is a second acceleration instruction includes determining whether the first PCI fast transaction is a second acceleration instruction by determining whether the first PCI fast transaction is associated with a virtual function of the storage device exposing the acceleration module.

18. The method according to claim 17, It is characterized in that Determining whether the first PCI fast transaction is a second speed-up instruction includes determining whether the first PCI fast transaction includes a tag having an identifier of the virtual function.

19. The method according to claim 15, It is characterized in that Determining whether the first PCI fast transaction is a second speed-up instruction includes determining whether the first PCI fast transaction is a second speed-up instruction by determining whether the first PCI fast transaction is associated with a physical function of the storage device exposing the acceleration module.

20. The method according to claim 19, It is characterized in that Determining whether the first PCI fast transaction is a second speed-up instruction includes determining whether the first PCI fast transaction includes a tag having an identifier of the physical function.