Patent Application
20100058271
The present invention generally relates to the field of integrated circuits. More particularly, the present invention relates to methods, systems, and computer program products to design integrated circuits using a closed loop 1xN methodology. The integrated circuits of today commonly have hundreds of millions of transistors on a single chip, with many critical circuit features having measurements in the deep sub-micron range. As manufacturers implement more and more circuit elements in smaller and smaller silicon substrate surface areas, engineers and designers develop hardware and software tools to automate much of the integrated circuit design and manufacturing process.
The classical design methodology for designing and manufacturing integrated circuits employs an open-loop design process whereby the activities of logic capture and physical realization are separate and lack coherency. For example, a team of designers generally start with a conceptual design, develop associated logic, perform synthesis of the logic design to develop an associated physical design, and analyze the physical design for timing performance. During these phases, designers may make adjustments to the one or more circuit devices, such as changing parameters of drive strength or making power optimizations. If the register transfer level (RTL) design is not frozen, the designers may develop or change the logic, perform the synthesis to create another physical design, and analyze the new design. Until designs of the integrated circuits reach a “logic-freeze” stage, at which point synthesis stops, design improvements from physical optimizations are often lost for the iterations through the design flow. Such optimizations are often necessary for meeting design goals related to operating frequencies and power consumption.
To add to the complexity of the design process, some design implementations define specific logic functions for which there are one-for-one physical realizations. The physical realizations are often fixed entities, both logically and from the standpoint of electrical realization. For these specific design implementations, a circuit designer and/or physical designer generally need to know details of the specific design at a very early stage in the design process. Additionally, the designer must have the ability to craft the specific design. As the design progresses, the designer routinely needs to optimize the design to reduce power consumption, increase performance, and reduce the amount of the required area. Current methods of maintaining physical optimizations before a design reaches the stage of logic freeze include manual instantiation of gates and manual changes to the synthesis flow to produce predictable and repeatable results. Designers spend considerable time optimizing the circuits because the optimizations require manual intervention and constant updating for individual iterations through the design flow.
Following are detailed descriptions of embodiments depicted in the accompanying drawings. The descriptions are in such detail as to clearly communicate various aspects of the embodiments. However, the amount of detail offered is not intended to limit the anticipated variations of embodiments. On the contrary, the intention is to cover all modifications, equivalents, and alternatives of the various embodiments as defined by the appended claims. The detailed descriptions below are designed to make such embodiments obvious to a person of ordinary skill in the art.
Some embodiments comprise creating a physical design representation based on a behavioral representation of a design for an integrated circuit. The behavioral representation comprises register transfer level (RTL) hardware definition language (HDL) with a 1xN building block. The embodiments also comprise altering an element of the 1xN building block by using a logic design tool, a synthesis tool, a physical design tool, or a timing analysis tool. The embodiments further comprise generating a behavioral representation or a logical representation of the modified 1xN building block based on the physical design representation.
Further embodiments comprise an apparatus having a viewer and a 1xN compiler. The viewer may generate displays of behavioral representations of a 1xN building block for an integrated circuit design, with the behavioral representations comprising RTL definitions. The 1xN compiler may create physical design representations of the 1xN building block, create behavioral representations from the physical design representations, and generate logical representations of the 1xN building block. For the embodiments, an optimization tool, a synthesis tool, or a physical design tool may alter an element of the 1xN building block in one or more of the behavioral, logical, and physical design representations.
Even further embodiments comprise a computer program product comprising a computer readable medium having a computer readable program, wherein the computer readable program causes the computer to create physical design representations of a 1xN building block of an integrated circuit design. At least one of the physical design representations has an alteration of an element of the 1xN building block, with the alteration may have been made by an optimization tool, a synthesis tool, or a physical design tool.
Aspects of the various embodiments will become apparent upon reading the following detailed description and upon reference to the accompanying drawings in which like references may indicate similar elements:
The following is a detailed description of novel embodiments depicted in the accompanying drawings. The embodiments are in such detail as to clearly communicate the subject matter. However, the amount of detail offered is not intended to limit anticipated variations of the described embodiments. To the contrary, the claims and detailed description are to cover all modifications, equivalents, and alternatives falling within the spirit and scope of the present teachings as defined by the appended claims. The detailed descriptions below are designed to make such embodiments understandable to a person having ordinary skill in the art.
Many of the embodiments contained herein describe using 1xN building blocks to design an integrated circuit. A circuit layout or circuit design may comprise a hierarchical collection of circuit elements or cells. A building block, or macro, may generally refer to the concept that a circuit may comprise a plurality of different circuit elements or cells, wherein the cells may be divided into categories. These categories may be basic cells, level cells, and building blocks. Basic cells, which may alternatively be referred to as leaf cells, may comprise an aggregation of circuit elements, such as a collection of logic gates, and may be thought of as residing at the bottom of the design hierarchy. Basic cells may be nested together, along with other circuit elements, to form level cells at different mid-levels within the design hierarchy. The level cells, which are frequently referred to as composition cells, may be further nested together to form building blocks, alternatively referred to as macros, which may also be referred to as root cells or prime cells. Building blocks may be thought of as residing at the top of the design hierarchy. Building blocks may also be referred to as blocks or modules.
The use of a cell at a specific level in the design hierarchy may be referred to as a cell instance, or instantiation. A circuit with only one hierarchical level is said to be flat. During a design sequence a VLSI design system may need to flatten a circuit design, converting the hierarchical levels into a single level, so that certain design tools may work with the design. For instance, a circuit simulation tool or a circuit optimization tool may only be able to simulate or optimize a circuit design that has a flattened netlist. A VLSI design system may flatten, or fully instantiate, a circuit design by recursively replacing all cell instances with basic circuit elements like resistors and capacitors.
Additionally, many of the embodiments use the term “optimized” and or “tuned”. As one skilled in the art of integrated circuit design will appreciate, such terms as “tuned”, “optimal”, and “optimized” as used in the design process may refer to a portion of an integrated circuit that has been altered to comply with some design constraint. For example, a simulation tool may determine drive strength for a specific gate in the circuit, selecting a parameter for the drive strength that makes a tradeoff between two or more competing interests, such as speed versus power consumption. Consequently, such terms should not be interpreted as choosing “optimal” or “best” settings, since such settings or parameters may often be subjective in nature. Such terms should therefore be considered in the context of integrated circuit design, consistent with the manner in which the terms are used by those possessing ordinary skill in the art.
Persons having ordinary skill in the art often use various titles for persons performing different tasks of creating an integrated circuit design. For example, some refer to a design engineer as a person that performs such tasks as receiving specifications, creating behavioral and logical circuit designs, running simulations, and evaluating prototypes that are fabricated. Similarly, some refer to a test engineer as a person that develops a test plan for a circuit design based on the specification and actually performs simulations of the design and/or physical tests of a fabricated prototype. Some may refer to a computer aided design (CAD) engineer as a person that uses CAD tools, such as schematic capture, simulation, synthesis, layout, and power estimation tools to evaluate and modify the circuit design. Due to the difference in use of titles, which may vary greatly, this specification uses the terms circuit designer, logic designer, or simply designer to refer to a person that may perform any of the described tasks for creating the circuit design. In other words, the terms circuit designer, logic designer, designer may be substituted interchangeably in the specification, and should not be interpreted as referring to a person that may only perform specific tasks. To the contrary, the terms circuit designer, logic designer, designer may refer to persons that can perform any of the tasks described herein, such as the tasks of behavioral design, logical design, physical design, and timing analysis/design.
Turning now to the drawings,
VLSI design application 145 may provide a closed-loop environment that employs high-level 1xN constructs for tuning performance and power consumption. VLSI design application 145 may capture logic using both high-level RTL and references to 1xN building blocks. VLSI design application 145 may represent the 1xN building blocks via a high-level language and transform the 1xN building blocks into custom realizations using, e.g., building blocks available in a standard or custom cell library.
In various embodiments, a system may have a processor, such as processor 105, for executing program instructions of applications, such as VLSI design application 145. While executing program instructions of VLSI design application 145, computing system 100 may display images of the circuit design, or information pertaining to simulation or performance tuning, on a monitor or other computer display, such as display 110. For example, VLSI design application 145 may generate pictorial schematic elements, as well as wires connecting the various schematic elements, for a schematic in a window 115 on display 110.
Using input device 120 a circuit designer may interact with VLSI design application 145. In one or more embodiments, input device 120 may comprise a keyboard and/or a mouse, allowing the circuit designer to perform such actions as viewing different circuit sections or performing numerous circuit design activities such as performing a timing analysis, increasing the drive strength of a gate, or defining which circuit elements should not be modified during various design cycle iterations. In some embodiments input device 120 may comprise a tablet and stylus, such as a pressure-sensitive surface or a capacitance pad that recognizes hand-written characters and stylus movement. In even further embodiments input device 120 may comprise an audible input device, such as a microphone used for speech recognition. For example, in one embodiment input device 115 may allow a circuit designer to automate one or more tasks during the design flow, such as by recognizing when the circuit designer speaks the phrase “calculate Vt-hybridization values for selected 1xN building blocks”.
Display 110 and input device 120 may also allow the circuit designer to interact with VLSI design application 145 when creating, viewing, and manually modifying various 1xN building blocks 155. For example, 1xN block builder 150 may allow a circuit designer to retrieve a 1xN building block from a cell library 130, stored on a storage device 125. Storage device 125 may comprise a local solid state drive or a network hard drive, as examples, which stores standard cells and various instantiations created by a circuit designer. Upon retrieving a 1xN logic block, 1xN block builder 150 may allow a circuit designer to edit the 1xN building block by displaying configuration information for the 1xN building block in window 115. After editing the 1xN building block, VLSI design application 145 may save the edited version back to cell library 130.
In the embodiment depicted in
The product of the design concept 210 may be a flowchart, a flow graph, or pseudocode. The designer may specify the overall functionality and input-to-output mappings without developing architectural or hardware details for the circuit design. Based on the design concept 210 the circuit designer may develop a behavioral description of the circuit during logic phase 220, such as by using a high-level language to define how the circuit should work at a high level of abstraction. More specifically, the circuit designer may write a hardware description of the IC using a hardware description language (HDL), such as Verilog or VHDL (very-high speed integrated circuit HDL). As is known to those skilled in the art, the high-level code or behavioral representation may be referred to as RTL netlists. At the RTL level, the behavioral description may consist of program-like constructs, such as, variables, operators, conditionals, loops, procedures and functions.
The circuit designer may then simulate the design, or different portions of the design, by using one or more behavioral tools to verify the design works as intended. For example, the circuit designer may apply input combinations and/or sequences of data to different modules to verify that the modules respond as intended. After the hardware description of the design has been through several iterations, and exhibits some degree of operability and stability, the circuit designer may convert the HDL code to a logical representation via a synthesis phase 230. During synthesis phase 230, a VLSI design application may convert the HDL definition of the circuit, transforming the textual descriptions of the behavioral representation to logic gates. In other words, the synthesis phase 230 may transform the VHDL code into a netlist describing the hardware, which may comprise circuit elements and the wires connecting them.
The logical representation may comprise gate-level logic, such as AND, OR, NOT, NAND, and NOR gates, to name just a few. The gate-level logic may be arranged to implement the desired functionality described in the high-level HDL code. The tools that perform the conversion from the high-level description to the gate-level logic are generally called synthesis tools. For example, a set of synthesis programs from Synopsys® may perform the conversion from the high-level code to the logic-level code. As is known to those skilled in the art, the logical representation, or logic-level code, may be referred to as gate-level netlists. The logical representation may comprise a netlist of high-level macrocell components, such as adders, pipelined multipliers, arithmetic logic units, multiplexers, and register-files. Each macrocell, or building block, may have a corresponding netlist of primitive cells as well as a corresponding custom layout. This gate-level netlist of primitive cells and the layout may be generated using a silicon compiler tool, like 1xN compiler 160.
Once the high-level code is converted to gate-level code, the designer may perform additional testing via simulation and formal verification in order to verify that the gate-level netlists accurately represent the RTL netlists. In other words, one or more synthesis tools and/or behavioral tools may help the designer determine whether the logical design accurately represents what the behavioral design describes. If the logic-level code is successfully verified, a circuit designer may use one or more physical design tools to convert the logic level code into a detailed layout of structural components or elements of the circuit in physical design phase 240. The detailed layout of structural components may generally be referred to as a physical design representation. The physical design representation may comprise different forms and formats from embodiment to embodiment, as well as in the same embodiment. For example, persons having ordinary skill in the art may refer to different forms of the physical design representations using such terms as the physical realization, the hierarchical physical design (PD) netlist, the flattened netlist, the physical view, the physical placement data, the electrical realization, or simply netlist.
Two common forms of the physical design representations are the hierarchical PD netlist and the flattened netlist. As noted, the hierarchical PD netlist and the flattened netlist may represent the physical design produced during the physical design stage 240. Based on the physical design representation, the circuit designer may use one or more timing analysis tools during a timing phase 250 to evaluate the timing performance of the design. Assuming the timing performance is acceptable, the circuit designer may release the flattened netlist or the hierarchical netlist to a semiconductor fabrication facility to produce the IC.
The circuit designer may have to repeat one or more of the design activities numerous times before reaching the final physical design. For example, once the designer creates the gate-level logic in logic phase 220 and simulates the design via one or more simulation tools, the designer must often make changes to the HDL definition and create a modified version of the gate-level logic. Simulating the modified logic may uncover more design deficiencies that the circuit designer must correct. In other words, the circuit designer may have to pass through logic phase 220 numerous times.
Aside from logic changes, the circuit designer may also need to alter the design to correct problems encountered after synthesis phase 230, physical design phase 240, and timing phase 250. For example, the circuit designer may progress through the design process to the point where the design works perfectly under simulation, but finds that a routing tool cannot create a functional wiring arrangement for the design, due to rule restrictions or area constraints. Alternatively, one or more optimization tools may generate changes to the design. One such example may be incremental performance improvements, such as reduced delays, produced by iteratively feeding the results of a place and route (P&R) tool back into a synthesis tool.
During an open-loop design sequence, the circuit designer may need to exert caution to prevent having the accumulated design improvements altered or negated, due to disjointed activities of capturing logic in logic phase 220 and developing the physical design in physical design phase 240. Until the design reaches the logic-freeze stage, which halts subsequent synthesis in synthesis phase 230, physical optimizations such as tweaks made for drive strengths and power optimization may be lost during subsequent synthesis iterations. Using a closed-loop methodology offered by 1xN compilation 260, an application like VLSI design application 145 may maintain these physical optimizations without depending on disjointed tools and databases.
1xN compilation 260 may allow a VLSI design system to employ a closed-loop design process whereby physical implementation details, such as alterations due to drive strength and power optimization, may be fed back into each step of the design process. The VLSI design system may feed the design alterations back to the different design phases by using 1xN building blocks and inserting data generated by compiling the 1xN building blocks back into the design flow.
A VLSI design system may use a 1xN compiler during 1xN compilation 260 to create, modify, and maintain 1xN macros through the design process. As will be illustrated in the subsequent discussion of the various embodiments, the 1xN compiler may directly update the configuration file or files of a circuit design in a closed-loop manner. Additionally, the 1xN compiler may support direct wiring, have built-in design-for-manufacturability features, and have hierarchical wiring avoidance.
As
With a general understanding of some of the steps that may be involved in the closed-loop 1xN methodology, we return now to
While
Depending on the embodiment, computing system 100 may run a variety of different operating systems. For example, in one embodiment computing system 100 may use AIX®. In another embodiment, computing system may use Microsoft® Windows®, Linux®, or Mac OS®, as examples. Other alternative embodiments may have no operating system at all. For example, computing system 100 may comprise a state machine or microcontroller executing firmware instructions, such that no operating system is necessary.
One or more design files 135 may be stored on a storage medium of storage device 125 and accessed by computing system 100. For example, storage device 125 may comprise one or more of a variety of different mass storage devices used to store VLSI design files 135 and macro and/or cell library 130. For example storage device 125 may comprise a serial hard disk drive or a small computer system interface (SCSI) redundant array of inexpensive disks (RAID). Alternatively, storage device 125 may also comprise an optical storage device, such as a rewritable compact disc (CD) or a digital versatile disc (DVD) drive, having storage mediums of a CD and DVD, respectively. In other embodiments, storage device 125 may comprise a flash memory device, such as a universal serial bus (USB) drive. Storage device 125 may also store other files and information, such as the operating system for computer system 100.
While not shown in
In one or more alternative embodiments, computing system 100 may also work with one or more other networked computing devices to perform one or more of the different design activities. For example, in one embodiment computing system 100 may edit and compile 1xN building blocks and generate a schematic view for the IC design. However, during the synthesis, physical design, or timing analysis stages, VLSI design application 145 may transfer netlist 185 for the design to the networked computer(s). Such an arrangement may be desired, for example, where the design task requires extensive computational power that may otherwise overload computing system 100. As the above example illustrates, various embodiments of a system may comprise an almost limitless number of wired and wireless communication devices, allowing computing devices of a system to communicate with each other to share and/or process VLSI design files, wherein the computers may be located close to or remote from each other.
As described earlier, VLSI design application 145 may work with 1xN building blocks 155. 1xN building blocks 155 may comprise the means by which physical implementation details of circuit elements in 1xN building blocks 155 are stored. A building block of 1xN building blocks 155 may provide a layer of abstraction between the logical behavior of the block and the physical realization of the corresponding logic elements. In numerous embodiments, 1xN building blocks 155 may comprise one or more user-configurable text files. In alternative embodiments, the building blocks may be represented in an alternative file format. For example, the building blocks may be stored in a proprietary binary file that is encoded and decoded as a circuit designer views/edits the 1xN building blocks.
1xN compiler 160 may perform a variety of tasks in VLSI design application 145. 1xN compiler 160 may comprise a software module or other type of module that, at a high level of abstraction, converts data from the building blocks of 1xN building blocks 155 into behavioral, physical, and logical representations, or definitions. 1xN compiler 160 may convert 1xN building blocks 155 into specifically formatted implementations as required by the different design phases. For example, conversions of 1xN building blocks 155 may be represented by 1xN compilation 260 in
1xN block builder 310 may comprise a graphical user interface (GUI) routine that allows the circuit designer to create one or more building blocks and implement a desired logical function. For example, 1xN block builder 150 may allow the circuit designer to create a bank of NAND logic gates in a relatively easy and efficient manner without having to be familiar with the actual scripting language required by 1xN compiler 160.
In various embodiments, 1xN block builder 310 may comprise the initial entry point to the 1xN compilation stage, which may correspond to 1xN compilation 260 shown in
While 1xN block builder 310 may generate a default set of implementation details, the designer may be able to choose between specific implementation types and make specific customizations for 1xN building block 315, such as specific drive-strengths, Vt-hybridizations, and/or voltage island references. Once the designer has chosen a desired 1xN building block 315 to build, 1xN block builder 310 may generate a logical RTL instantiation for 1xN building block 315, as well as the physically crafted realization of 1xN building block 315. One embodiment of a window for block builder 310 is depicted in
As
1xN building blocks like 1xN building block 600 may enable a VLSI design application to perform numerous transformations, or modifications, to elements and/or cells in the building blocks during the design process. The VLSI design application or a designer may invoke timing optimizations and/or power optimizations, either directly through synthesis refinement or manually via engineering change (EC) mode. The designer or VLSI design application may change predefined 1xN building blocks. For example, the designer may transform such parameters as polarity, drive strength, and hybrid-Vt for 1xN building blocks, thereby creating custom 1xN building blocks.
Using a 1xN building block methodology to design a VLSI circuit may create a framework that allows 1xN building blocks to be utilized in different ways. For example, in some embodiments, 1xN building blocks may be configured for use with industry-standard formats. In some embodiments, 1xN building blocks may be configured to use alternative, special, or proprietary features. As will be illustrated, the 1xN building block methodology may incorporate the ability to re-form hierarchy from a previously flattened hierarchical design, allowing optimization tools that work on flat data to nonetheless enable use of 1xN building blocks.
1xN building blocks 600, 630, and 660 are exemplary building blocks, none of which may be required in any specific embodiment. 1xN building blocks 600, 630, and 660 may be the results of individual optimizations through each design phase in order to meet power, area, and timing goals for one particular latch bank and the associated timing paths for the latch bank. Various integrated circuit parameters may vary with factors such as temperature, voltage and frequency. One circuit parameter that a circuit designer may adjust is the drive strength of one or more circuit devices, such as an output driver stage. For example, the circuit designer may match the output drive strength of a signal to the circuit being driven. Another parameter that a circuit designer may adjust is Vt-hybridization. Vt-hybridization refers to an optimization technique for static power reduction while optimizing performance within a given drive strength. Worth emphasizing, in many embodiments the polarity of individual bits of a 1xN building block may also be modified by optimization routines. Additionally, the polarity of individual bits may be manually modified.
Returning back to
In some embodiments, uniquifier 320 may also be configured to selectively ignore or uniquify specific sets of 1xN building blocks 315. Selective uniquification of building blocks may be advantageous in situations where parent-child relationships do not need to be defined during the logic entry stage, which may correspond to logic phase 220. The parent-child relationships may be defined on-the-fly as the relationships are required for a design stage. Depending on the embodiment, uniquifier 320 may be configured to read netlist 330 in a mixture of different formats. For an example referring to
During the various iterations of a design, 1xN compiler 340 may generate compilation data for a number of different views 350. The different views of views 350, which may correspond to output view data 180 depicted in
1xN compiler 340 may generate compilation data for behavioral RTL as well as gate-level RTL. Additionally, 1xN compiler 160 may generate compilation data for physical views, helper scripts, and potentially numerous other views depending on the embodiment. For example, VLSI design application 145 may use 1xN compiler 160 to generate Verilog HDL behavioral RTL and corresponding Cadence™ physical views. Additionally, in numerous embodiments, 1xN compiler 160 may be configured based on the tool suite of the user, such as tools 190. In other words, an embodiment may control which languages and view types are generated. An alternative embodiment may generate VHDL behavioral RTL and corresponding Synopsys® physical views.
To illustrate in more detail the iterative logic, synthesis, physical design, and timing cycles for 1xN building blocks, we turn now to
The iterative logic cycle depicted in
1xN compiler 406 may insert checks, or verification text, in the 1xN behavioral RTL model for the parameterized features. When the design is simulated via an RTL simulation tool, the verification text may ensure that the intent of the designer, which is described in the RTL, matches the gate-level implementation defined by the 1xN behavioral model. One should note that the behavioral model for a 1xN building block may not need to be gate-specific. As a behavioral model, the model may describe the logical function using higher-level constructs. The higher-level constructs may be defined by the RTL language used. For example, the higher-level constructs for System Verilog may be different than the higher-level constructs for VHDL.
1xN compiler 406 may verify that the behavioral model, which may be represented by behavioral view 408, and the gate-level implementation are equivalent. For example, in one embodiment 1xN compiler 406 may verify that the behavioral model and the gate-level implementation are equivalent by using one or more verification tools. The verification tools may be, e.g., industry standard verification tools available in IC design systems available from Cadence™, Synopsys®, Mentor Graphics®, Magma®, or other design systems. Verifying the equivalency of the behavioral model and the gate-level implementation may help protect the designer from inadvertent changes during the logic design, as well as protect the designer from gate-level changes that might alter a function of a 1xN building block.
The behavioral model construct generated by 1xN compiler 406 may allow the circuit designer to reuse 1xN building block 404 and related logical function(s) in the design without having to worry about the gate-level tweaks that may be necessary for proper placement or timing of the specific 1xN building block. Consequently, using the generated behavioral model may also help protect the intellectual property described in the gate-level netlist because the gate-level details are not directly disclosed by the RTL listing. In other words, a hardened or physically-realized design that uses this 1xN building block methodology may be converted to a softened, or RTL-level, design and transferred to another party with less risk of exposing physical details, which may be sensitive or proprietary in nature. Generating an RTL-level design from a physically-realized design may not necessarily be limited to 1xN building blocks. 1xN compiler 406 may generate an RTL-level design from other types of hierarchical RTL elements. In other words, the 1xN methodology described herein may provide an automated closed-loop approach that may used to create RTL-level or behavioral-level design for numerous other RTL elements.
To illustrate in more detail the iterative synthesis cycle, we move to
The RTL no-touch feature may comprise a software routine of a VLSI design application. For example, RTL no-touch software may comprise a set of instructions that guides a synthesis engine, such as synthesis tool 448, preventing the engine from synthesizing a set of RTL statements. Preventing the synthesis engine from synthesizing a set of RTL statements may be helpful when the RTL for a specific section of a design has already been optimized. For example, the RTL for two sections of a design may have been optimized by some tool or by human intervention. The designer may want the two sections to remain unaltered in subsequent iterations of the synthesis engine. One may contrast the RTL no-touch feature with conventional approaches, where the synthesis engine of the conventional approaches would integrate the gate-level entities within the RTL, alter the generic nature of the RTL, and make the design technology dependent.
Various embodiments may provide an RTL no-touch feature by allowing specific sets of RTL constructs to be designated as having been previously implemented and optimized as gate-level netlists. For example, a circuit design may have twenty-five different 1xN building blocks. A circuit designer may have synthesized the circuit and tuned a portion of the circuit having ten 1xN building blocks. The circuit designer may want to add some additional circuit functions, or otherwise modify the circuit design at the behavioral level, using Verilog HDL. The circuit designer may designate the ten 1xN building blocks that have been tuned as no-touch blocks. As a consequence, any subsequent automated optimizations performed on the circuit design may tune or alter the other fifteen 1xN building blocks and the newly added circuit functions but make no more tuning or optimization changes to the ten previously tuned building blocks.
Various embodiments may implement the RTL no-touch feature by maintaining the gate-level netlists for the no-touch building blocks (element 444) separate from the behavioral level design. The embodiments may maintain separate gate-level netlists and behavioral lists to the building blocks as long as both the gate-level netlists and behavioral lists are proven and maintained to be logically equivalent, such as by using industry-available tools. The RTL no-touch feature may not be limited to 1xN building blocks. Many embodiments may implement the RTL no-touch feature for both 1xN building blocks and other types of hierarchical cells and circuit elements in a circuit design. Using the RTL no-touch feature for other types of circuit elements may allow a designer to tune one or more parts of a circuit design, which may contain both 1xN building blocks and non-1xN elements, and prevent such tuning parameters from being erased or overwritten during subsequent tuning or optimization for other portions of the circuit design. Referring to
In at least one embodiment, the RTL no-touch feature may eliminate the need for disrupting the RTL-only design environment with gate-level implementation details. The 1xN compiler may be configured to avoid generating RTL-level code for the no-touch building blocks. Preventing the generation of the RTL-level code may enable the embodiment to help protect proprietary design techniques, as well as maintain a technology-independent netlist.
Each netlist in the RTL no-touch gate-level netlist database (element 444) may have a one-to-one correspondence to a corresponding element in the behavioral RTL (element 430). The RTL no-touch gate-level netlist database (element 444) may be constructed by selective output extraction of synthesis tool 448, manual intervention (element 442), or another type of automated means. The RTL no-touch file (element 446) may designate which behavioral RTL constructs (element 430) should be replaced by the gate-level netlist counterparts. Further, the RTL no-touch file (element 446) may also specify that synthesis tool 448 should avoid further manipulation of the replaced gate-level netlist counterparts. Avoiding further manipulation of the replaced gate-level netlist counterparts may be beneficial, because the synthesis tool 448 may optimize the other gate-level constructs for better overall design performance without disturbing the previously optimized netlists. Synthesis tool 448 may then generate or create the composite design netlist (element 426), which may include the RTL no-touch gate-level constructs.
By employing a 1xN design methodology, an IC design application like VLSI design application 145 may maintain the results of optimizations performed on 1xN building blocks through a process of back-annotation. With back-annotation, a circuit designer may also perform automated optimizations of the design. In many embodiments, the circuit designer may perform the automated optimizations normally, such as the designer would in a traditional design flow. Performing the automated optimizations in a traditional manner may allow a system employing 1xN building blocks to decouple the 1xN building blocks from the tool-suite of the system. Referring back to
A premise of back-annotation may be the detection and description of optimizations via 1xN building blocks. An embodiment may perform an optimization on a design and make changes to the netlist based on the optimization results. The back-annotation feature may allow the embodiment to detect the netlist changes and then describe or create updates to the 1xN building blocks corresponding to the netlist changes.
Back-annotator 424 may read two netlists. In
Since the optimizations made to the elements in the netlists are decoupled from the 1xN building blocks, back annotation for 1xN building blocks may allow an embodiment to support numerous optimization techniques, such as optimization techniques for timing, power, or area. Additionally, an embodiment may have the capability to disable one or more optimization tools or optimization features via synthesis tool 448 or back annotator 424. The quality of the starting netlist or “seed” may determine the output quality and amount of time required for automated optimization routines. Consequently, the use of iterative cycles of synthesis and back annotation may improve runtimes and result in better-optimized designs.
Back-annotator 424 may be extendable and programmed to selectively accept or ignore a specified set of changes to a specific list of 1xN building blocks. An embodiment may program back-annotator 424 to selectively accept or ignore the specified set of changes through a list of directives, which may be similar to the use of directives by existing synthesis tools. For example, if an embodiment needs a specific 1xN bit-ordering to meet a routing requirement, the embodiment may disable placement updates by fixing cell locations in the netlist or by providing back-annotator 424 with a list of 1xN building blocks for which the placement updates should be ignored. Further, depending on the embodiment, the user may be able to select the method used for optimization-exclusion. If the tool suite in an embodiment supports optimization-exclusion, the embodiment may allow a user to configure the optimization-exclusion via synthesis tool 448. However, in another embodiment that does not support optimization-exclusion, or if the optimization-exclusion is difficult or cumbersome to use, the embodiment may allow the user to configure the optimization-exclusions via back-annotator 424.
In many embodiments, synthesis tool 448 may be hierarchy-aware. In other words, synthesis tool 448 may be configured to recognize and work with hierarchical cell elements. For embodiments that have a hierarchy-aware synthesis tool 448, 1xN building block 422 may be represented as hierarchical elements, e.g., traditional hierarchical elements. Synthesis tool 448 may manipulate the hierarchical elements to produce an optimized design. For embodiments that have hierarchy-aware synthesis tool 448, the embodiments may need to perform the timing improvements or timing optimizations using the individual gates within the hierarchical elements so that timing rules such as New Delay Rules (NDRs) are not required for each individual 1xN building block 422. Otherwise, an embodiment may need to update 1xN compiler 428 to generate NDRs for each 1xN building block 422.
In many embodiments, optimizations made to 1xN building blocks may be captured via back-annotation by back-annotator 424. Placement changes allowed on a hierarchical 1xN building block 422 may be used in the physical design phase for floorplanning updates. Because the results of synthesis and automated optimization routines may largely depend on the initial quality of the netlist, or starting seed, maintaining optimizations for 1xN building blocks through back-annotation may result in a better quality synthesis result through successive iterations.
As illustrated above for embodiments that do not have hierarchy-aware tools, the embodiments may flatten the 1xN building blocks before performing the optimizations. Embodiments that flatten the 1xN building blocks before performing the optimizations may detect and capture changes to the 1xN building blocks via back-annotation. However, flattening the design may eliminate the hierarchy established to maintain 1xN building blocks as individually tunable elements in the separate design phases. To address the eliminated hierarchy, various embodiments may recreate the hierarchy through a reassembly process.
Reassembler 432 may comprise a software module or other type of module for 1xN compilation 420 that uses netlist attributes created by 1xN compiler 428 to recreate the hierarchical design. Reassembler 432 may also comprise, e.g., reassembler 335 depicted in flowchart 300, or reassembler 175 of VLSI design application 145 shown in
Reassembler 432 may detect changes to 1xN building blocks, such as 1xN building block 422, via back-annotation. However, because the netlist may have been flattened, the changes may not be directly applied back to the hierarchical RTL and physical design netlists. Reassembler 432 may discern the hierarchical boundaries of the flattened netlist. Plus, reassembler 432 may be configured to apply the updates for the 1xN building block 422 back to the respective hierarchical physical design netlist or RTL netlist without manual intervention.
For example, reassembler 432 may use attributes created by 1xN compiler 428, which are present in the flattened netlist, to recreate the hierarchical design. The flow process for recreating the hierarchical design via attributes of the netlist is illustrated in
The physical design phase may comprise the part of the design process that captures or generates the physical attributes of the circuit design. For example, a VLSI design application may capture the physical attributes related to floor planning, cell and 1xN building block placement, routing between the cells/blocks, noise reduction, power bussing, and design rule checking, as just a few examples. Embodiments that employ only classical design methodologies generally lose many of the changes made during the physical design phase due to the iterations through the logic and/or synthesis design phases.
An embodiment employing a 1xN design methodology may automatically maintain changes made to the 1xN building blocks during the physical design phase via the mechanisms of back-annotation and reassembly.
One flow of an iterative physical design stage may comprise reading different 1xN building blocks from a netlist of physical design 468. Depending on the type of netlist (element 452), an embodiment may reassemble the netlist via reassembler 462 and/or perform back-annotation via back-annotator 454 for the individual 1xN building blocks, such as 1xN building block 456. Embodiments may employ 1xN compiler 458 to compile the reassembled and back-annotated 1xN building blocks to generate physical design view 460.
An embodiment employing a 1xN design methodology may leverage the hierarchical nature of a physical design to make many physical design tasks more efficient through the use of scripting. For example, if an integrator wants to change the relative spacing of the leaf cells in 1xN building block 456, the embodiment employing the 1xN design methodology may not require the integrator to move each individual cell manually. Instead, the embodiment may allow the integrator to update 1xN building block 456 to reflect the desired spacing. The embodiment may then allow the integrator to recompile the design via 1xN compiler 458 and re-read the relative spacing changes into the design. The hierarchical 1xN building block 456 may be rewired on-the-fly, such as when the integrator is working on the design in physical design phase 240, so that the spacing changes are ready to use once the design is re-read. Subsequently, the integrator may resolve any potential overlaps between the 1xN building block and other random-logic cells and re-route the hierarchical connections.
An embodiment may make physical design tasks more efficient through the use of scripting for numerous types of transformations. For example, various embodiments may use scripting for relative placement of cells within 1xN building blocks, relative spacing of leaf cells, bit ordering, changes of cell types used in 1xN building blocks, and DFM/redundancy-for-improved manufacturability implementations.
Embodiments employing dynamic updates to 1xN building blocks, such as 1xN building block 456, may not need to rely on pre-compiled sub-blocks that can be tiled together. Instead, 1xN compiler 458 may interpret the underlying physical structure of each cell and rewire 1xN building block 456 according to the cell placements described in 1xN building block 456. Interpreting the underlying physical structure and rewiring 1xN building block 456 may reduce turn-around-time for design optimizations and provide the designer or integrator with more physical detail information through each design phase.
Due to the nature of classical timing analysis tools, designers typically perform the timing design phase using flattened designs. Based on the results of the timing analyses, a designer may manually update the design to meet timing goals. Alternatively, the timing tool may also apply updates to the design. In other words, the timing tools may use various methods to communicate the recommended changes or updates and have the updates physically implemented in the design. Embodiments employing a 1xN design methodology may update 1xN building blocks in at least two ways. A first method may comprise updating 1xN building blocks through reassembly and back-annotation, whereby the embodiment recognizes design changes and feeds the changes or alterations back to other stages in the methodology. A second method may comprise incorporating the changes via on-the-fly editing and recompilation of 1xN building blocks. This secondary method may allow for more accurate timing information in a rapid fashion via scripting from a 1xN compiler.
In a classical design flow, the timing team may make changes to the abstract or higher-level design, communicate the changes to the physical design team, and wait until the physical design team has incorporated the changes into the physical design. This interaction between the timing team and physical design team may be relatively slow in that the updates may need to be placed and routed before the updates can be analyzed in a timing analysis. Implementing the updates via placement and routing may require days. Once the physical design is ready to be analyzed for timing performance, the timing tool may take hours or days to produce the timing data for the designer depending on the complexity of the design.
Changing 1xN building block 476 may not provide completely accurate timing results in some instances, because changes to 1xN building block 476 may affect other circuit elements, such as random logic gates. However, incorporating changes from a timing analysis using a 1xN design methodology may provide a greater degree of accuracy than a designer not using capabilities provided by a 1xN design methodology. Consequently, the designer may use the 1xN design methodology to reduce the time needed to update the physical design and meet timing goals, resulting in a quicker turnaround time for the design.
Script 480 generated from recompilation may not be limited to use in the timing design phase, nor to a specific set of transformations. Rather, 1xN compiler 478 may be configured to work with various tool sets for other phases of the design methodology and other desired transformations. For example, 1xN compiler 478 may be configured to work with placement and routing tools, DFM tools, design-rule checking tools, simulation tools, timing analysis tools, and other tools used in the logic, synthesis, physical design, and timing phases. Embodiments employing a 1xN building block methodology may provide the necessary design information for the various transformations through reassembly, back-annotation, and uniquification. Some embodiments may employ scripts for adding new 1xN building blocks to a design. Some embodiments may employ scripts for changing internal cell placement, internal rewiring, rewiring pins of a 1xN building block, and/or changing the functions for 1xN building blocks. Even further embodiments may employ scripts to delete specific 1xN building blocks from a design or remove internal cells of designated 1xN building blocks. As one skilled in the art will appreciate, scripts for 1xN building blocks may be used in a variety of different ways for a variety of different purposes. The example uses just cited are only a few of the possible uses.
The representation forms of circuit design 730 may vary from embodiment to embodiment, as well as within an embodiment. Behavioral representation 735 may comprise a high-level description of circuit design 730. For example, behavioral representation 735 may comprise one or more types of HDL code, behavioral RTL descriptions, user-configurable text files, C-language code, RTL netlists, or other high-level elements that describe various circuit devices, including 1xN building blocks, cells, and/or hierarchical cells/macros.
Logical representation 740 may comprise a gate-level description of circuit design 730. For example, logical representation 740 may describe an alternative form of circuit design 730, having such components and/or elements as AND gates, OR gates, NAND gates, and NOR gates, as well as a mixture of 1xN building blocks and other high-level macrocell components, such as adders, arithmetic logic units, multiplexers, and register-files.
Physical design representation 745 may comprise a low-level description of circuit design 730. For example, physical design representation 745 may describe an alternative form of circuit design 730, having such components and/or elements as nodes, wires, resistors, capacitors, inductors, contacts, poly, diffusion, many of which may have associated placement and dimensional data. Depending on the embodiment, physical design representation 745 may also comprise hierarchical information, such as hierarchical information of cells and 1xN building blocks. In other words, physical design representation 745 may comprise a PD netlist and/or a flattened netlist.
During the design process, the integrated circuit designer may have to use one or more tools of tool suite 750. For example, tool suite 750 may comprise a variety of synthesis and performance optimization tools, such as tools 190 in
The integrated circuit designer may have to repeatedly and iteratively use one or more of the tools of tool suite 750 before developing a final physical design representation 745 that may be fabricated. For example, the designer may use one or more logic design tools 765 that work in conjunction with 1xN compiler 720 to create 1xN building blocks in behavioral representation 735. Logic design tools 765 may comprise, e.g., cell editors, 1xN building block editors, RTL editors, and other high-level circuit design applications.
The designer may use one or more synthesis tools 770 to create logical representation 740 from behavioral representation 735. One or more of synthesis tools 770 may work in conjunction with 1xN compiler 720 when creating 1xN building blocks for logical representation 740. The designer may then analyze and/or optimize the design via one or more logic design tools 765. For example, the designer may simulate the operation of groups of elements, cells, and 1xN building blocks of logical representation 740, such as applying various sequences of bit patterns to verify that logical representation 740 behaves in a similar or equivalent fashion as behavioral representation 735 behaves. Additionally the designer may use one or more timing analysis tools 780 and logical design tools 765. For example, the designer may verify that signals propagate through sections of the circuit in required time intervals and that drive strengths are sufficient for implemented fan-outs.
The circuit designer may use physical design tools 775, which may work in conjunction with 1xN compiler 720, to convert logical representation 740 to physical representation 745. Similar to the other iterative closed-loops encountered when generating behavioral representation 735 and logical representation 740, the designer may need to make numerous changes to elements of physical design representation 745 due to results of physical design tools 775 and timing analysis tools 780. For example, the designer may discover that a routing tool cannot create a functional wiring arrangement for the design, due to rule restrictions, area constraints, and other problems encountered by physical design tools 775. Alternatively, one or more optimization tools may generate changes to the design, such as by relocating elements, selecting different pins of different cells and/or 1xN building blocks, increasing gate drive strengths, rearranging cells of 1xN building blocks, and making other types of alterations to other elements of physical design representation 745. Consequently, the designer may use apparatus 700 and 1xN compiler 720 to generate updated representations of altered the 1xN building blocks for the various representations of behavioral representation 735, logical representation 740, and physical design representation 745.
One or more elements of apparatus 700 may be in the form of hardware, software, or a combination of both hardware and software. For example, in one embodiment, one or more of the modules of apparatus 700 may comprise software instructions of an application stored in a tangible medium, executed in memory by a processor of a computer. In other words, one embodiment of apparatus 700 may comprise portions of stored code for an application, which may design integrated circuits using a closed loop 1xN methodology, such as VLSI design application 145 shown in
In alternative embodiments, one or more of the modules of apparatus 700 may comprise hardware-only modules. For example, in one embodiment 1xN compiler 720 may comprise a state machine that compiles different versions of netlists to generate the different forms of 1xN building blocks. A state machine 1xN compiler 720 may be desirable when, e.g., the designer needs to create the various representations in a rapid fashion, since a state machine compiler may be able to generate the representations considerably faster than a software-based compiler.
In some embodiments, viewer 710 and 1xN compiler 720 may both reside in a single device, such as in memory 140 of computing system 100 shown in
Alternatively, depending on the embodiment, the elements of apparatus 700 may reside in two or more devices. For example, in one alternative embodiment, viewer 710 may reside in a local workstation. In other words, the circuit designer may use a client application on a local workstation that accesses a VLSI design application located on a remote server which has 1xN compiler 720.
The arrangement of apparatus 700 may vary in different embodiments. Some embodiments may have different elements than those module depicted in
In other embodiments, apparatus 700 may include varying numbers of viewer 710 and 1xN compiler 720. One embodiment may have two or more 1xN compilers 720. Such an arrangement may be necessary to generate the different representations of the 1xN building blocks. Alternatively, numerous 1xN compilers may be necessary to interact with incompatible or differing tools of tool suite 750.
Design process 810 may include using a variety of inputs; for example, inputs from library elements 830 which may house a set of commonly used elements, circuits, and devices, including models, layouts, and symbolic representations, for a given manufacturing technology (e.g., different technology nodes, 32 nm, 45 nm, 90 nm, etc.), design specifications 840, characterization data 850, verification data 860, design rules 870, and test data files 885 (which may include test patterns and other testing information). Design process 810 may further include, for example, standard circuit design processes such as timing analysis, verification, design rule checking, place and route operations, etc. One of ordinary skill in the art of integrated circuit design can appreciate the extent of possible electronic design automation tools and applications used in design process 810 without deviating from the scope and spirit of the various embodiments.
The design structure is not limited to any specific design flow. Design process 810 preferably translates an embodiment of the invention as shown in
As illustrated in
Upon entering all of the data for the 1xN building blocks, as well as other cells and circuit elements for the circuit design, the designer may use the 1xN compiler to create logical representations from the behavioral representations (element 920). One process of compiling 1xN building blocks of the behavioral representations to create the logical representations may be shown in
The circuit designer may then use one or more circuit verification tools to simulate and/or verify that the logical representation operates or behaves like the behavioral representation. Once the circuit designer is satisfied with the performance of the logical representation, the circuit designer may use the 1xN compiler to create physical design representations based on the logical representations (element 930). For example, the circuit designer may use 1xN compiler 720 and one or more physical design tools 775 to generate physical design representation 745 based on logical representation 740. In another example using
Having a physical design representation of the design, the circuit designer may alter or change elements of the design by performing a number of optimizations or tuning operations (element 940). As an illustration, the circuit designer may use one or more tools of logic design tools 765, synthesis tools 770, physical design tools 775, and timing analysis tools 780. Examples may include creating a routing list via a routing tool, simulating one or more sections of the circuit, analyzing one or more sections using via a couple of timing analysis tools, and manually tweaking some drive strengths of numerous elements in the circuit. Each of the tools may change elements of various 1xN building blocks in the circuit.
Having modified elements of the 1xN building blocks, as well as potentially having other modified elements of the circuit, the circuit designer may generate or create behavioral, logical, and physical design representations that reflect the modified/altered circuit elements (element 960). Additionally, some alterations may require back-annotation and reassembly of hierarchical elements, including hierarchical 1xN building blocks, in order to generate the behavioral, logical, and/or physical design representations (element 950). For example, the circuit designer may perform a number of operations on the circuit using timing analysis tools 780 and a routing tool of physical design tools 775, wherein the tools operate on a flattened version of physical design representation 745 and create a modified version of physical design representation 745, which is also flattened. To recreate the hierarchical nature of the circuit and the 1xN building blocks for the various views or representations, the circuit designer may generate the representations through the processes described for
The process of using one or more tools of a tool suite to analysis, modify, tune, and/or tweak the circuit may continue many times over, wherein the circuit designer performs one or more operations of different design phases (element 970). Upon each iterative analysis and/or modification, the circuit designer may use the 1xN compiler to generated updated versions of the various representations (element 970). Referring again to
Flowchart 900 of
Another embodiment is implemented as a program product for implementing systems and methods described with reference to
Furthermore, embodiments may take the form of a computer program product accessible from a computer-usable or computer-readable medium providing program code for use by or in connection with a computer or any instruction execution system. For the purpose of describing the various embodiments, a computer-usable or computer readable medium may be any apparatus that can contain, store, communicate, propagate, or transport the program for use by or in connection with the instruction execution system, apparatus, or device.
The medium can be an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system (or apparatus or device) medium. Examples of a computer-readable medium include a semiconductor or solid state memory, magnetic tape, a removable computer diskette, a random access memory (RAM), a read-only memory (ROM), a rigid magnetic disk, and an optical disk. Current examples of optical disks include compact disk-read only memory (CD-ROM), compact disk-read/write (CD-R/W), and DVD.
A data processing system suitable for storing and/or executing program code may include at least one processor coupled directly or indirectly to memory elements through a system bus. The memory elements can include local memory employed during actual execution of the program code, bulk storage, and cache memories which provide temporary storage of at least some program code in order to reduce the number of times code is retrieved from bulk storage during execution. Input/output or I/O devices (including but not limited to keyboards, displays, pointing devices, etc.) can be coupled to the system either directly or through intervening I/O controllers.
Those skilled in the art, having the benefit of this disclosure, will realize that the present disclosure contemplates designing integrated circuits using a closed loop 1xN methodology. The form of the embodiments shown and described in the detailed description and the drawings should be taken merely as examples. The following claims are intended to be interpreted broadly to embrace all variations of the example embodiments disclosed.
Although the present disclosure and some of its advantages have been described in detail for some embodiments, one skilled in the art should understand that various changes, substitutions, and alterations can be made herein without departing from the spirit and scope of the disclosure as defined by the appended claims. Although specific embodiments may achieve multiple objectives, not every embodiment falling within the scope of the attached claims will achieve every objective. Moreover, the scope of the present application is not intended to be limited to the particular embodiments of the process, machine, manufacture, composition of matter, means, methods, and steps described in the specification. As one of ordinary skill in the art will readily appreciate from this disclosure, processes, machines, manufacture, compositions of matter, means, methods, or steps presently existing or later to be developed that perform substantially the same function or achieve substantially the same result as the corresponding embodiments described herein may be utilized. Accordingly, the appended claims are intended to include within their scope such processes, machines, manufacture, compositions of matter, means, methods, or steps.
