Design tool for interactive connection routing during network-on-chip generation
The design tool optimizes NoC layout by employing straight-line routing and partitioning with machine learning, addressing inefficiencies in interactive recalculations and redesigns, enhancing the speed and efficiency of NoC synthesis.
Patent Information
- Application Number
- CN202410690199.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Priority Date
- 2024-05-29
- Filing Date
- 2024-05-30
- Publication Date
- 2024-12-03
AI Technical Summary
Creating an optimal Network-on-Chip (NoC) that meets performance requirements such as connectivity, latency, and area constraints is a complex and time-consuming task, requiring frequent redesign when changes occur in the chip floorplan or component additions/deletions, and existing routing algorithms are inefficient for interactive recalculations.
A design tool employing straight-line routing and partitioning the floorplan into routing areas, using a machine learning model to optimize routing paths, and incorporating macro and micro-routing algorithms to efficiently recalculate routes in real-time, while considering blocking areas and power/clock domains.
Facilitates fast and interactive NoC layout generation, reducing computational complexity and enabling real-time updates, thus improving the efficiency and speed of NoC design and synthesis.
Smart Images

Figure CN119066023A_ABST
Abstract
Description
[0001] CROSS-REFERENCE TO RELATED APPLICATIONS
[0002] This application claims the benefit of U.S. Provisional Application Serial No. 63 / 469,543, filed on May 30, 2023, entitled “INTERACTIVE WIRE ROUTING IN AFLOORPLAN by Amir CHARIF,” the entire disclosure of which is incorporated herein by reference. Technical Field
[0003] The present technology is in the field of electronic system design, and more particularly, relates to physical implementation guidelines for very fast straight-line routing of wires in a floorplan associated with a Network-on-Chip (NoC). Background Art
[0004] A multiprocessor system has been implemented in a system on chip (SoC) that communicates through a network on chip (NoC). The SoC includes instances of initiator intellectual property (IP) and target IP. Transactions are sent from the initiator to one or more targets in the form of packets using industry standard protocols. An initiator connected to the NoC sends a request transaction to a target, using an address to select the target. The NoC decodes the address and transmits the request from the initiator to the target. The target processes the transaction and sends a response transaction, which is transmitted back to the initiator through the NoC.
[0005] For a given set of performance requirements, such as connectivity and latency between sources and destinations, frequencies of various components, maximum available area for NoC logic and its associated routing (wiring), minimum throughput between sources and destinations, power requirements of the NoC, and the locations of components attached to the NoC on the floorplan, creating an optimal NoC that meets all requirements with the least amount of logic and wiring is a complex task. Creating this optimal NoC is typically the job of a chip architect or chip designer, and it is a difficult and time-consuming task.
[0006] In addition to this being a difficult task, the design of the NoC is modified whenever one of the requirements changes (such as modifying the chip floorplan, adding or deleting IP components, or modifying the expected performance). As a result, this task needs to be frequently redone during the design time of the chip using any design tool. Consider a design tool for editing the NoC topology on top of a floorplan in a graphical view. The line connecting two NoC elements needs to be routed in a straight line while avoiding blocking areas, which are also called "blocking parts". The tool should display the exact routing between the two elements connected by the network, and interactively refresh the exact routing immediately when the user / designer moves one of the elements, thereby allowing the automatic generation of the NoC layout. Automatic NoC generation (i.e., topology synthesis) also requires performing several distance calculations to generate the optimal tree. Typically, the floorplan is discretized into small square cells of a given size (i.e., resolution). There are many effective routing algorithms that can be used to calculate the routing between two cells (nodes) in the floorplan, such as the A* algorithm or the Dijkstra algorithm. According to some aspects and embodiments of the present invention, the tool uses an algorithm to access all cells of the discretized floorplan. The higher the resolution, the slower the algorithm. Therefore, what is needed is a design tool that: implements a fast routing method that is more suitable for interactively recalculating routes and distances in a floorplan and that exploits the fact that routing is always straight lines and thus requires visiting a portion of all floorplan cells or areas. Summary of the invention
[0007] According to various embodiments and aspects of the present invention, a tool is disclosed for implementing a fast routing method that is more suitable for interactively recalculating routes and distances in a floorplan. The tool utilizes straight-line routing, thereby requiring access to a portion of the entire floorplan cell or area. BRIEF DESCRIPTION OF THE DRAWINGS
[0008] Figure 1 A graphical representation showing a floorplan of a Network on Chip (NoC);
[0009] Figure 2 A device with a blocking area or blocking portion according to various aspects and embodiments of the present invention is shown. Figure 1 Layout planning;
[0010] Figure 3 A diagram showing one or more cells attached to a node in a floorplan in accordance with various aspects and embodiments of the present invention Figure 1 Layout planning;
[0011] Figure 4 FIG. 1 shows a system such as a system having routing areas defined by various constraints according to various aspects and embodiments of the present invention. Figure 1 Layout planning of the like;
[0012] Figure 5 illustrates a floorplanning scenario with regions and routing according to various aspects and embodiments of the present invention; and
[0013] Figure 6 A process used by a design tool to implement fast routing according to various aspects and embodiments of the present invention is shown. DETAILED DESCRIPTION
[0014] Various examples of the present technology illustrating various aspects and embodiments of the present invention are described below. Typically, examples can be used to describe these aspects in any combination. All statements of narration principles, aspects and embodiments and their specific examples herein are intended to encompass both their structural and functional equivalents. In addition, such equivalents are intended to include currently known equivalents and future developments (i.e., any element of the development that performs the same function regardless of the structure).
[0015] Note that, as used herein, the singular forms "a", "an", and "the" include plural referents unless the context clearly dictates otherwise. References throughout this specification to "one aspect", "aspects", "some aspects", "various aspects", or similar language mean that a particular aspect, feature, structure, or characteristic described in connection with any embodiment is included in at least one embodiment of the present invention.
[0016] The phrases "in one embodiment," "in at least one embodiment," "in an embodiment," "in certain embodiments," and similar language that appear throughout this specification may, but do not necessarily, all refer to the same embodiment or similar embodiments. In addition, the aspects and embodiments of the invention described herein are merely exemplary and should not be construed as limiting the scope or spirit of the invention as understood by those of ordinary skill in the art. The disclosed invention can be effectively made or used in any embodiment including any novel aspect described herein. All statements describing the principles, aspects, and embodiments of the invention herein are intended to cover both structural and functional equivalents thereof. Such equivalents are intended to include both currently known equivalents and equivalents developed in the future.
[0017] As used herein, a transaction may be a request transaction or a response transaction. Examples of request transactions include write requests and read requests.
[0018] As used herein, a node is defined as a distribution point or communication endpoint capable of creating, receiving and / or transmitting information through a communication path or channel. A node may refer to any of the following: a switch, a splitter, a merger, a buffer, and an adapter. As used herein, splitters and mergers are switches; not all switches are splitters or mergers. As used herein and in accordance with various aspects and embodiments of the present invention, the term "splitter" describes a switch having a single ingress port and multiple egress ports. As used herein and in accordance with various aspects and embodiments of the present invention, the term "merger" describes a switch having a single egress port and multiple ingress ports.
[0019] According to various aspects of the present invention, a system on a chip (SoC) includes multiple clock domains and multiple power domains. The clock domain is defined by all logic fed by a given clock input. The clock input is characterized by the clock frequency, which is its most important parameter. The power domain is defined by all logic that obtains power supply from the same power supply. According to various aspects of the present invention, the power supply is gated, so that a power domain can be turned on or off or isolated from other power domains. In this way, the designer provides a set of clock domain and power domain constraints as part of the initial design.
[0020] Now refer to Figure 1 and Figure 2 , according to some aspects and embodiments of the present invention, floorplan 100 and floorplan 200 are shown respectively. Floorplan 200 includes blocking region 202 and blocking region 204. Floorplan is modeled as a mesh network of floorplan nodes. The user / designer can set the number of nodes per row / column according to the resolution. Each floorplan node is connected to its four (maximum) neighboring (north, south, east, west) nodes. The blocking part of the floorplan is modeled by disconnecting the nodes classified as the blocking part from the neighboring nodes, such as Figure 2 According to some aspects and embodiments of the present invention, a floorplanning region representing a clock domain or power domain boundary is modeled by identifying nodes, such as by coloring the nodes that fall into the region. Each node can have multiple colors to represent various constraints or parameters associated with the entire system (e.g., it falls into a certain clock domain, and a certain clock tree can reach it).
[0021] Reference Figure 3, according to some aspects and embodiments of the present invention, a floorplan 300 is shown in which one or more cells are attached to a given floorplan node. According to some aspects and embodiments of the present invention, the number of cells that can be attached to a node can be determined by technical parameters (e.g., cell density) and the area of the components. According to some aspects and embodiments of the present invention, the mesh network of the floorplan 300 is an irregular mesh network. Due to disconnected nodes, a detour may be required, so it is not always possible to use a minimum routing algorithm (e.g., X, then Y) to find a route between two points. According to various aspects of the present invention, the tool utilizes routing in a mesh network by segmenting the irregular mesh network into multiple interconnected regular mesh networks.
[0022] Now refer to Figure 4 , according to some aspects and embodiments of the present invention, a floorplan 400 with 3 (area 1, area 2, area 3, area 4, area 5) is shown. A routing area is a subset of the floorplan mesh that includes at least one node and forms a regular mesh; that is, each node inside the routing area can reach every other node through the minimum path (equivalent to the Manhattan distance) without detours. According to various aspects of the present invention, inside the routing area, calculating the distance between two nodes is completed in constant time because it is equal to the Manhattan distance (dX+dY). According to various aspects of the present invention, inside the routing area, calculating the effective path between two points is performed in N steps, where N is the Manhattan distance between the two points expressed in the number of nodes. In other words, for a routing area with X columns and Y rows, in the worst case, obtaining a complete path between two points will require X+Y steps, compared to X×Y steps in the worst case when using a complete routing algorithm (e.g., Dijkstra's algorithm).
[0023] Each routing area stores a series (horizontally or vertically) of floorplan nodes that connect it to adjacent areas. A routing area can have multiple adjacent areas that can be reached through different series of nodes. According to various aspects of the present invention, for a given floorplan configuration, the routing area is calculated and stored in a memory. According to various aspects of the present invention, the tool has information about each floorplan node. The tool also knows the routing area to which each floorplan node belongs or is located. According to various aspects of the present invention, the difference is represented by a color difference and one of its colors. Therefore, each area is assigned a unique color when represented or displayed in a graphical user.
[0024] Now refer to Figure 5, shows a floorplan 400 with routing regions and global micro-routing 502 and macro-routing 504. According to various aspects of the present invention, the tool includes building a global routing algorithm in two steps using macro-routing 504 and micro-routing 502, taking advantage of the efficiency of routing (and distance calculation) within a single routing region. According to some aspects and embodiments of the present invention, when a routing region of a source node 506 and a destination node 508 is given, macro-routing 504 occurs by calculating the minimum route between the two regions including the source node 506 and the destination node 508. According to some aspects and embodiments of the present invention, the Dijkstra algorithm or A* algorithm can be used to find the shortest path between the two regions, and the shortest path will operate on the scale of routing regions rather than at the level of nodes. Starting from the starting (source) node, the cumulative distance in the shortest path algorithm is efficiently calculated, which can be done in constant time between the two regions. The result of this step is a path 504 that includes the routing region. Path 504 starts in region 1, then goes to region 2, then to region 3, and ends in region 5 where the ending (destination) node is located.
[0025] According to some aspects and embodiments of the present invention, micro-routes are represented by micro-routes 502. The tool generates micro-routes 502 using an efficient minimum path routing algorithm (such as Routing Along X and Routing Along Y) to connect each routing area covered by the macro-route 504 previously calculated by the macro-routing step until the destination node 508 is reached.
[0026] According to various aspects of the present invention, the tool handles any given partitioning of a floorplanned mesh network into routing areas. According to various aspects of the present invention, the partitioning algorithm generates routing areas that are as large as possible. The tool includes a machine learning model that allows feedback on the selection of routing areas. For example, a designer can identify routing areas that need to be updated and provide this information as feedback to further train the machine learning model. According to various aspects of the present invention, the machine learning model is updated by the tool as various routing areas are identified and later modified by the tool. In addition, the machine learning model can identify routing areas and later modify the routing areas based on the determination of macro routing or micro routing. According to various aspects of the present invention, if each routing area contains exactly one node, the complexity is the same as the classic shortest path algorithm, and the machine learning model can be trained to immediately identify and use the algorithm that is most suitable for determining routing.
[0027] According to various aspects of the present invention, instead of using a static minimum path routing algorithm (route along X and route along Y) within a given routing area, the tool uses a dynamic routing algorithm biased by any given metric. A machine learning model (such as the application of artificial intelligence (AI) in the tool) uses past modeling training and feedback to calculate a routing area that uses any given metric bias, an example of which is shown below. According to various aspects of the present invention, if each node tracks the number of links it has previously been routed through, the routing algorithm can select the neighboring point with the fewest links to reduce link congestion. According to various aspects of the present invention, if a routing algorithm is used to build a spanning tree, information about multiple destinations can be provided to enable the routing algorithm to approach as many destinations as possible at a time, rather than approaching one destination.
[0028] According to some aspects and embodiments of the present invention, the results of the macro routing step may be stored or cached to avoid repeating the determination of the macro path 504 for the macro routing multiple times in a row. If a path from one routing area to another routing area is known, it may be reused to make the routing determination faster. According to some aspects and embodiments of the present invention, if a cached routing path is calculated based on a source / destination that is far away from the new source / destination, the calculated route cannot be used because it may not be optimal.
[0029] Now refer to Figure 6, the design tool performs process 600. In step 602, the design tool receives a floorplan for the NoC along with any constraints / parameters, including blocking areas. In step 604, the NoC is analyzed or synthesized to identify routing areas. As noted, routing areas are defined and identified based on similarities of parameters and / or constraints, etc. In accordance with some aspects and embodiments of the present invention, the routing areas are divided into a set of 2×2 high-level cell arrays. Each array is separate so that routing is constrained individually within each array and no wires cross the solid line boundaries. Routing is performed in each 2×2 array to minimize congestion. In the next smaller array, the design tool forms a new set of 2×2 arrays in any high-level cell array. The process is repeated until the routing array is determined. In step 606, the tool identifies the source 506 and the destination 508, each located in a different routing area. In step 608, the tool analyzes the floorplan to determine the macro routing 504. Once the macro route is determined, then in step 610, the tool performs calculations to determine the micro route 502 that follows the macro route 504 from the source 506 to the destination 508. As noted, the tool includes a machine learning model that can detect and perform macro and micro routing analysis. According to some aspects and embodiments of the present invention, the tool receives input from the designer regarding changes or updates that need to be made to improve or change the NoC. These inputs are provided as feedback to the machine learning model to further train the machine learning model. According to some aspects and embodiments of the present invention, an adversarial machine learning model can also be used to provide feedback.
[0030] According to various aspects of the present invention, the initiator and the target are communicatively connected to the NoC. The initiator is a unit that sends a request and can be a source for routing determination. The initiator is usually configured to read and write commands. The target is a unit that serves or responds to the request and can be a destination. The target is usually configured to read and write commands. Each initiator is attached to or connected to the NoC through an NIU. The NIU attached to the initiator is called an initiator network interface unit (INIU). In addition, each target is attached to the NoC through an NIU. The NIU attached to the target is called a target network interface unit (TNIU). The main function of the NoC is to bring each request from the initiator to the desired destination target, and if the request requires or needs a response, the NoC brings the response of each target to the corresponding request initiator. The initiator and the target have many different parameters to characterize them. According to various aspects of the present invention, for each initiator and target, the clock domain and power domain to which they belong are defined. The width of the data bus they use to send, write and receive read payloads is the number of bits. In accordance with various aspects of the present invention, the width of the data bus used for the connection (communication path to / from the target) for sending, writing requests and receiving write responses is also defined. In addition, the clock and power domain definitions are references to the previously described clock and power domains present in the SoC, as described herein.
[0031] In accordance with various aspects of the present invention, an initiator is not required to be able to send requests to all targets or targets connected to the NoC. The precise definition of the targets that can receive requests from the initiator is outlined or elaborated in the connectivity table. Connectivity and traffic class labeling information can be represented as an explicit or conceptual matrix. There is a row for each initiator and a column for each target. If an initiator must be able to send traffic to a target, a traffic class label must be present at the intersection of the initiator row and the target column. If no label is present at the intersection, the tool does not require connectivity between the initiator and the target. In accordance with various aspects of the present invention, the actual format used to represent connectivity can be different, as long as each pair of initiator-target combinations has a precise definition of its traffic class, or there is no classification label if there is no connectivity. Initiator / target connections to support multiple traffic classes are within the scope of the present invention.
[0032] In the scope of the present invention, latency refers to the number of clock cycles it takes for data to travel through the network. Despite the high bandwidth (frequency), latency causes problems in the transmission network. An example of a "real-time traffic class" is video data from a camera in an autonomous car. It is unacceptable to have long delays in data propagation. If some data is lost, this will make the vehicle unsafe. For audio and / or video in an entertainment system, it will be a traffic class that is still real-time but less important. Having gaps in this data is undesirable, but not unsafe. An example of non-real-time data is data from an instrument sensor in a car. It doesn't matter if the data is delayed by a few seconds because the rate of change of the data is very slow compared to the operating speed of the SoC.
[0033] According to some aspects of the invention, no scenario is defined for the tool, in which case the tool optimizes the NoC synthesis process for physical cost, such as lowest gate cost and / or lowest wire cost.
[0034] According to various aspects of the invention, the tool processes a connectivity matrix having the following defined parameters or components:
[0035] ● Each initiator has a network interface unit
[0036] ● Each target has a network interface unit
[0037] ● Each defined traffic class creates a switch, which is called the master switch for that class,
[0038] ● Each initiator / initiator NIU is followed by a switch that divides the traffic to the different master switches that the initiator needs to reach,
[0039] ● Each target / destination NIU is preceded by a switch that merges traffic from the different master switches that are sending traffic to that target
[0040] The data width of each switch and the clock domain it belongs to are calculated using the data width of each interface attached and its clock domain as input to the tool. According to various aspects of the present invention, each step of transforming a network that is part of a NoC also performs calculations of the data width and clock domain of the newly created network elements.
[0041] According to various aspects of the present invention, using the tool, each switch of each merger in the main switch is further decomposed into a cascade of mergers, each merger in the cascade is arranged at a branch point of the merger roadmap of the attached target. The branch point of the roadmap is defined by the fact that the path is split into two or more branches. The process of decomposing the splitter in the splitter cascade retains the original splitter function, because the number of inputs of the cascade is still 1, and the number of outputs of the cascade is the same as the number of outputs of the original splitter. The process of decomposing the merger in the merger cascade retains the original merger function, because the number of outputs of the cascade is still 1, and the number of inputs of the cascade is the same as the number of inputs of the original merger. According to various aspects of the present invention, the effect of the process is to obtain a set of basic switches, which are represented by mergers and splitters, and they are physically arranged close to where actual connections are required between switches.
[0042] According to various aspects of the present invention, the tool transforms the network so as to reduce the number of wires used between switches that can be realized while maintaining performance defined in scenarios, which are minimum throughput sets required between initiators and targets. According to various aspects of the present invention, switches are clustered for performance-aware switching, and mergers and splitters that have been distributed on a roadmap are treated as ordinary switches.
[0043] According to other aspects of the present invention, an extension of the clock and power domains on the floorplan is provided, and each component is tested to ensure that it is within the boundaries of the specified clock and power domains. If the test fails, the component is moved until a suitable position is found where the test passes. Once a suitable position is found for each component, the routing of each connection between the components is completed. The routing process will find the appropriate path for the set of wires that form the connections between the components. After the routing is completed, if necessary, based on how long it takes for the signal to cover a distance of 1 mm, using the information provided about the technical capabilities, a cross-distance pipeline element is inserted on the link.
[0044] According to some aspects and embodiments of the present invention, the tool generates one or more computer files describing the generated NoC, the computer files comprising:
[0045] List of network elements and their configuration: data width, clock domain.
[0046] The location of each network element on the generated floorplan.
[0047] A collection of routes through network elements that enable connectivity. According to aspects of the invention, a route is an ordered list of network elements, one for each pair (initiator, target) and one for each pair (target, initiator). A route indicates how traffic between pairs will flow and through which elements.
[0048] According to various aspects of the present invention, the tool is used to generate metrics about the generated NoC, such as: histograms of wire length distribution, number of switches, and histograms of switch sizes.
[0049] According to another aspect of the present invention, the tool automatically inserts various adapters and buffers in the network. The tool inserts adapters based on the required adaptation between two elements with different data widths, different clock and power domains. The tool inserts buffers based on these scenarios and detected rate mismatches.
[0050] According to some aspects and embodiments, the tool can be used to ensure that multiple iterations of synthesis are completed for incremental optimization of a NoC, including situations where one of the constraints provided to the tool is information about a previous run.
[0051] After the comprehensive process is performed by the software, a result is produced in a machine-readable form, such as a computer file that captures information using a well-defined format. An example of such a format is XML, and another example of such a format is JSON. The scope of the present invention is not limited to a particular format.
[0052] In an embodiment, adaptation mechanisms for the timing implementation between a first component and a second component within a NoC include, but are not limited to, distance pipeline insertion, adding buffer stages, switching buffers to higher drivers or faster buffers, adjusting wire width, and / or indicating which wire layer(s) are being used.
[0053] It is within the scope of the present invention to physically implement different mechanisms to overcome timing issues within the synthesized topology by communicating constraints and / or guidance of the physical implementation to resolve timing issues before they occur. It is within the scope of the present invention that the physical implementation of the connection includes, but is not limited to, physically placing objects, components, and / or gates around the NoC topology so that downstream tools do not place a first connection at a sufficiently large distance from a second connection to exceed the timing constraints. It is more desirable to synthesize the network for all connections that can meet the connectivity requirements and minimize the number of gates in the synthesized network. Minimizing the number of gates in the synthesized network is important, including but not limited to enabling the NoC to adapt within the topology parameters; ensuring that timing requirements can be met; building more efficiently in a smaller network; and minimizing the total power used in the interconnection between signals.
[0054] In another embodiment, components on the driving side are replaced with components with greater driving strength to increase speed. For example, when a signal is forwarded from a first switch to a second switch, because the distance between the first switch and the second switch is too large and exceeds the timing requirement, a distance conduit can be laid between the first switch and the second switch, or parasitic elements on a specific line can be reduced by using a larger wire with lower resistance, thereby increasing the driving strength to make the signal faster.
[0055] In an embodiment, a floorplanning region is a mechanism that can include, but is not limited to, constraining components (such as switches and / or distance pipeline stages) to physical locations and / or proximity to physical locations; and / or dividing logical interconnects into multiple parts that are tuned to the size of downstream tools for proper synthesis. The physical and logical configurations of the interconnects are configured together to form logical interconnects. Floorplanning regions are used to constrain components to specific locations. The need for floorplanning regions is a response to the increasing size of networks or interconnects of SoCs being developed. In other words, the network or interconnect of the SoC has grown to be larger than the capacity of downstream implementation tools, including but not limited to adding wider interconnects and / or more interconnected components.
[0056] According to various aspects of the invention, based on data associated with signals received and transmitted by the switch, it may be more desirable for the switch to be in a particular constrained position because the performance of the switch may be adversely affected if the switch is moved outside of the constrained position.
[0057] According to various aspects of the present invention, additional constraints may be used to place the gates of components such as switches or pipeline stages as close together as possible within each individual component. A current problem with existing placement programs is that they tend to place components too far apart from each other. Within the scope of the present invention, a switch is a single component with multiple gates, each of which has a placement requirement. While the netlist facilitates the interconnection of these components, downstream tools physically place the components and form routes for connecting the specified logical connectivity.
[0058] According to various aspects of the present invention, two separate constraints of the first aspect and the second aspect may be used simultaneously. For example, if a switch is specified with a certain performance so that the components of the switch are distributed over a larger area than specified, the performance of the switch will be lower than the originally specified performance. Therefore, it will be more desirable to keep the components of the switch close together. In addition, based on a larger network, it is more desirable to keep the components within a certain area or area of the available space in the floor plan so that the entire network will perform as specified or as analyzed.
[0059] In an embodiment, a rough estimate of a given semiconductor process establishes the maximum length of a single wire to ensure that the overall timing of that particular path can be met. Additionally, moving from a first component to a second component within a given amount of time is an important aspect. Due to the length between the first component and the second component being too long, it may not be achievable to move a signal from the first component to the second component within a given amount of time. As a result, the need for a technical solution that inserts a distance pipeline level will address this technical problem compared to current technology.
[0060] According to various aspects of the invention, area estimates for first and second components are received and timing and other performance requirements are balanced in optimization of the NoC to ensure that the gates of the resulting NoC will be routable, i.e., they will fit into available space allocated to the NoC in the floorplan.
[0061] According to various aspects of the invention, power requirements of the NoC are further received, and timing and other performance requirements are balanced in the optimization of the NoC to ensure that the resulting NoC will not exceed the power requirements specified in the constraints.
[0062] According to various aspects of the invention, physical routing congestion data is also included, either as an estimate based on the NoC netlist structure or as feedback obtained from downstream implementation tools, and timing and other performance requirements are balanced in the optimization of the NoC to ensure that the resulting NoC will be routable in the available space allocated to the NoC in the floorplan.
[0063] According to some aspects of the present invention, a method for guiding the physical generation of a NoC is disclosed. The method includes: receiving at a tool including a machine learning model that is trained and receives feedback at least one constraint parameter of the NoC, the at least one constraint parameter selected from a constraint parameter group including at least one physical constraint and at least one performance constraint; using the tool to augment the physical floorplan of the NoC with information guiding the physical implementation of the NoC; and based on the at least one constraint parameter, using the physical floorplan to constrain the physical implementation of connections to locations on the physical floorplan.
[0064] Certain methods according to various aspects of the present invention may be performed by instructions stored on a non-transitory computer-readable medium. A non-transitory computer-readable medium stores code including instructions that, if executed by one or more processors, will cause a system or computer to perform the steps in the methods described herein. Non-transitory computer-readable media include: rotating magnetic disks, rotating optical disks, flash random access memory (RAM) chips, and other mechanically movable or solid-state storage media. Any type of computer-readable medium is suitable for storing code including instructions according to various examples.
[0065] Certain examples have been described herein, and it will be noted that different combinations of different components from different examples are possible. Prominent features are presented to better illustrate the examples; however, it is clear that certain features may be added, modified, and / or omitted without modifying the functional aspects of the examples described.
[0066] Various examples are methods of using the behavior of any one or a combination of machines. Method examples are complete anywhere in the world where most of the constituent steps occur. For example and in accordance with various aspects and embodiments of the present invention, IP elements or units include: processors (e.g., CPUs or GPUs), RAMs (e.g., off-chip dynamic RAMs or DRAMs), network interfaces for wired or wireless connections such as Ethernet, Wi-Fi, 3G, 4G Long Term Evolution (LTE), 5G, and other wireless interface standard radios. IP may also include various I / O interface devices as needed for different peripherals (such as touch screen sensors, geolocation receivers, microphones, speakers, Bluetooth peripherals) and USB devices (such as keyboards and mice) and others. By executing instructions stored in RAM devices, the processor performs the steps in the method as described herein.
[0067] Some examples are one or more non-transitory computer-readable media arranged to store such instructions for the methods described herein. Examples can be implemented by any machine that retains a non-transitory computer-readable medium including any necessary code. Some examples can be implemented as: a physical device such as a semiconductor chip; a hardware description language representation of the logical or functional behavior of such a device; and one or more non-transitory computer-readable media arranged to store such a hardware description language representation. The description of the principles, aspects, and embodiments described herein covers both their structural and functional equivalents. The elements described herein as being connected have an effective relationship that can be achieved by direct or indirect connection with one or more other intermediate elements.
[0068] Those skilled in the art will recognize many modifications and variations. Such modifications and variations include any relevant combination of disclosed features. The description of the principles, aspects, and embodiments described herein encompasses both their structural and functional equivalents. Elements described herein as "connected" or "communicatively connected" have an effective relationship that can be achieved by directly connecting or indirectly connecting using one or more other intervening elements. Embodiments described herein as "in communication" or "communicating" with another device, module, or element include any form of communication or link, and include an effective relationship. For example, a wired connection, wireless protocol, near field protocol, or RFID can be used to establish a communication link.
[0069] Where the terms "including," "comprising," "having," "containing," "with," or variations thereof are used in the detailed description and claims, such terms are intended to be inclusive in a manner similar to the term "comprising."
[0070] Therefore, the scope of the present invention is not intended to be limited to the exemplary embodiments shown and described herein. Rather, the scope and spirit of the present invention is embodied by the appended claims.
Claims
1. A method for guiding physical generation of a network-on-chip (NoC) using a design tool, the method comprising: receiving, at the tool, a floorplan for the NoC, wherein the floorplan includes a blocking region and a plurality of constraints for the NoC; synthesizing the NoC using the tool to determine a plurality of routing areas, wherein the routing areas are defined based on at least one constraint and include at least one node; identifying a first source in a first routing area of the plurality of routing areas and a first destination in a second routing area of the plurality of routing areas, wherein the first source and the first destination communicate using the NoC; determining a macro route from the first source to the first destination by calculating a minimum route between the first routing area and the second routing area, wherein the macro route includes at least the first routing area and the second routing area; and A micro-route is established from the first source to the first destination using a minimum path routing algorithm to connect the first source to the first destination.
2. The method according to claim 1, wherein: The macro route is determined using the Dijkstra algorithm.
3. The method according to claim 2, wherein: Determining the macro route includes finding the shortest path between each area traversed by the macro route on a scale of routing areas.
4. The method according to claim 1, wherein: The macro route is determined using the A* algorithm.
5. The method according to claim 4, wherein: Determining the macro route includes finding the shortest path between at least two areas on a scale of routing areas.
6. The method according to claim 1, wherein: The micro-route originates from the first routing area and ends at the second routing area.
7. The method according to claim 6, wherein: The micro-route traverses at least one other routing area selected from the plurality of routing areas.
8. The method according to claim 7, wherein: The micro-routing is established using along-X routing and along-Y routing through the first routing area.
9. The method according to claim 8, wherein: The micro-routing is established using along-X routing and along-Y routing through the second routing area.
10. The method of claim 1, further comprising receiving at least one block region in the floorplan at the tool.
11. A design tool for editing a network-on-chip (NoC) topology on a floorplan in a graphical view, the tool comprising: a generation module that identifies a required route between two network elements while avoiding a blocked area; a calculation module that performs distance calculations to generate an optimal tree; as well as a discretization module, wherein the discretization module divides the floorplan into cells, The calculation module uses a macro routing scheme and a micro routing scheme to calculate a route between two cells in the layout plan that meets the required routing.
12. The design tool according to claim 11, wherein: The cell is defined by a plurality of routing areas, wherein each routing area of the plurality of routing areas includes at least one constraint parameter and includes at least one node.
13. The design tool according to claim 12, wherein: The design tool determines a macro route from a first source to a first destination by calculating a minimum route between a first routing area of the plurality of routing areas and a second routing area of the plurality of routing areas.
14. The design tool according to claim 13, wherein: The design tool identifies microroutes from the first source to the first destination using a minimum path routing algorithm to connect the first source to the first destination.
