Patent Application
20140149955
The invention relates to electronic design automation, and more particularly, to methods and apparatuses for rapid checking of design rules in a circuit layout.
Advancements in process technology have impacted circuit manufacturing in at least two key ways. First, scaling of device geometry achieved through sub-wavelength lithography has facilitated packing more devices on a chip. Second, different process recipes have enabled manufacturing of heterogeneous devices with different threshold and supply voltages on the same die. A consequence of these improvements, however, has been an explosion in the number of design rules that need to be obeyed in the layout. Instead of simple width and spacing rules, modern fabrication technologies prescribe complex contextual rules that have to be obeyed for manufacturability.
The increase in the number of rules has complicated the task of creating design rule clean layouts, i.e., layouts that do not have design rule violations. Creating design rule clean layouts for digital circuit designs can be facilitated by the use of standard cell layouts as building blocks, and placement and routing tools that are extended to address the design rules.
Unfortunately, this approach usually does not work for analog, RF and custom circuit designs. Layouts for such designs are typically created manually using layout editors, and because of the number and complexity of the design rules, checking them was a laborious process.
A conventional design rule check (DRC) system requires a powerful two-dimensional geometry engine which supports geometric operations such as Boolean operations like AND, OR, NOT, XOR; sizing operations like grow/shrink horizontal/vertical/diagonal; other operations like merge, shift, flip, cut, smooth; as well as all-angle geometry for true Euclidean distance calculations. Individual rules are typically checked individually over an entire layout region. This is also true of individual rule values of same rule (e.g. a check against the minimum value for a rule, and another check against a preferred value for the same rule). Each check basically runs an independent sequence of geometry operations, and numerous passes through the layout region are required.
For example, a conventional series of operations to check a minimum spacing rule in a Manhattan only layout, might include steps of
As another example, a conventional series of operations to check a minimum width rule in a Manhattan only layout, might include steps of
So long as a good geometry engine is available, the conventional DRC techniques are simple to code, at least for simple rules. They are also flexible and powerful if the geometry engine has a scripting API for relevant geometry operations, and it is relatively straightforward to massively parallelize the DRC process among numerous CPUs.
On the other hand, it can be seen that checking even simple design rules like those above is extremely expensive computationally. Massive parallelization usually is possible only for offline checks, which typically are performed only between layout iterations. Even then they often can require hours to complete. The conventional approach also suffers from roughly linear growth of the total run time with respect to the number of rules to be checked, with multiple values for a rule counted as separate rules. This makes it very hard to reduce the total run time without turning off selected rules. The conventional approach also suffers from linear growth of run time for individual rule checks, with respect to the length of the geometry operation sequence, i.e., the complexity of the rule. The conventional approach also involves separate checks for Euclidean measurements, and also requires extensive education and training in order to optimize the performance of the customer scripts.
Recently, the march toward increasingly smaller technology nodes has pushed fabrication vendors to support double-patterning technology (DPT). If integrated circuit features on a layer are laid out too near each other for adequate separation when exposed through a lithographic mask, then the features are apportioned onto two different masks, each mask having only some of the features for the layer. During fabrication, the wafer is exposed twice, once with each mask. In this way features can be defined on the integrated circuit which are too finely spaced for them to be defined clearly and reliably by a single mask. Double-patterning technology can be extended to triple-patterning technology, in which the features for a layer are apportioned onto three different masks, and so on. Apportionment onto higher numbers of masks is possible as well, and these techniques are sometimes referred to generally as “multi-patterning technology”.
Typically the mask fabricator (which often is also the chip fabricator) uses its own algorithm to allocate shapes among the two or more masks. However, in some arrangements of layout shapes an allocation is not possible since a particular shape cannot be allocated to any of the masks without violating the spacing rules for features on that mask. In a layout layer which has not yet been split, these situations can be detected by a design rule checking algorithm which assigns different “colors” (corresponding to different lithographic masks) to any pair of shapes which are nearer to each other than the minimum multi-patterning spacing constraint for a single mask. Chains of such features are detected, each feature forcing the next feature to have a different color. In the double-patterning case, each next feature is assigned the color opposite to the former feature in the chain. In the multi-patterning case multiple colors are used, and it is sufficient that each next feature be assigned a different (also called non-conflicting) color.
If a chain loops back onto itself, then it forms a “cycle”. A cycle is not itself bad, unless there is no way to avoid two adjacent features in the chain being assigned the same color. In the double-patterning situation, if a cycle contains an even number of features, then there is no conflict because a color conflict can be avoided. The arrangement can then be allocated among the two masks without violating the single-mask spacing rules. If a cycle contains an odd number of features in the double-patterning case, however, then a coloring conflict exists and the arrangement, if not modified, cannot be allocated among the two masks without violating the single-mask spacing rules. The latter situation is sometimes referred to herein as a “DPT odd cycle violation”. (In the extension to multi-patterning, coloring conflicts are sometimes referred to herein as “multi-patterning coloring violations”.)
A number of techniques exist for handling odd cycle violations. But if none of them work, then one or more of the features must be moved away from one or more other features, in order to break the cycle. Since this solution can reduce circuit density, it would be desirable to move features only so far as needed in order to solve the problem. For a manual layout editor, this would ideally involve real time DPT odd cycle detection, so that after every movement or other edit of a shape the user can see immediately whether the violation remains. But checking for DPT odd cycle violations can be extremely time consuming by conventional methods, and prohibitively slow for such real time layout editing.
A need therefore exists for a robust solution to the problem of real time manual layout editing, in a manner that avoids double-patterning or multi-patterning violations.
Roughly described, a system enables quick and accurate depiction to a user of multi-patterning layout violations so that they may be corrected manually and in real time, and without interfering with normal manual editing process. In one embodiment, the system involves iteratively building a data structure (such as a forest structure) identifying a plurality of islands and a plurality of relationships among respective pairs of the islands, each of the relationships in a tree indicating that the pair of islands related by the relationship violates a multi-patterning spacing constraint. The system detects multi-patterning coloring violation among islands represented in the data structure, during the building of this data structure, and using the relationships that have been inserted into the data structure. The coloring violations preferably are reported to a user in the form of visual indications of the cycles among the inter-island multi-patterning candidate spacing violations, with the candidate spacing violations also themselves indicated visually and individually. Preferably the visual indications of the involved multi-patterning spacing violations are depicted with visual markers, the size of which depends upon the value of the multi-patterning spacing constraint being violated. In this way the user can see intuitively how to move the islands around, and in which directions and by what distance, in order to remove a multi-patterning spacing violation and thereby break the cycle.
Preferably, the system pre-populates a database identifying a plurality of multi-patterning candidate spacing violations before searching the database for multi-patterning coloring violations as above. At least some of the candidate violations can be detected and inserted into the database during a predecessor process of traversing the corners of shapes in a subject layout region, and for each corner, populating a layout topology database with values that depend on the respective corner locations. Preferably corner traversal is performed using scan lines scanning in at least two dimensions, and stopping only at corner positions. The layout topology database is populated with what information can be gleaned based on the current scan line. During this process, at least some types of the candidate violations are detected and inserted into the multi-patterning candidate spacing violations database.
Additional aspects of the invention are set forth herein for handling pre-coloring, pre-groups, triple patterning violation detection, and so-called “split-and-stitch” mitigation of multi-patterning violations.
The above summary of the invention is provided in order to provide a basic understanding of some aspects of the invention. This summary is not intended to identify key or critical elements of the invention or to delineate the scope of the invention. Its sole purpose is to present some concepts of the invention in a simplified form as a prelude to the more detailed description that is presented later. Particular aspects of the invention are described in the claims, specification and drawings.
The invention will be described with respect to specific embodiments thereof, and reference will be made to the drawings, in which:
The following description is presented to enable any person skilled in the art to make and use the invention, and is provided in the context of a particular application and its requirements. Various modifications to the disclosed embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be applied to other embodiments and applications without departing from the spirit and scope of the present invention. Thus, the present invention is not intended to be limited to the embodiments shown, but is to be accorded the widest scope consistent with the principles and features disclosed herein.
The EDA software design process (step 110) is itself composed of a number of steps 112-130, shown in linear fashion for simplicity. In an actual integrated circuit design process, the particular design might have to go back through steps until certain tests are passed. Similarly, in any actual design process, these steps may occur in different orders and combinations. This description is therefore provided by way of context and general explanation rather than as a specific, or recommended, design flow for a particular integrated circuit.
A brief description of the component steps of the EDA software design process (step 110) will now be provided.
System design (step 112): The designers describe the functionality that they want to implement, they can perform what-if planning to refine functionality, check costs, etc. Hardware-software architecture partitioning can occur at this stage. Example EDA software products from Synopsys, Inc. that can be used at this step include Model Architect, Saber, System Studio, and DesignWare® products.
Logic design and functional verification (step 114): At this stage, the VHDL or Verilog code for modules in the system is written and the design is checked for functional accuracy. More specifically, the design is checked to ensure that it produces correct outputs in response to particular input stimuli. Example EDA software products from Synopsys, Inc. that can be used at this step include VCS, VERA, DesignWare®, Magellan, Formality, ESP and LEDA products.
Synthesis and design for test (step 116): Here, the VHDL/Verilog is translated to a netlist. The netlist can be optimized for the target technology. Additionally, the design and implementation of tests to permit checking of the finished chip occurs. Example EDA software products from Synopsys, Inc. that can be used at this step include Design Compiler®, Physical Compiler, DFT Compiler, Power Compiler, FPGA Compiler, TetraMAX, and DesignWare® products.
Netlist verification (step 118): At this step, the netlist is checked for compliance with timing constraints and for correspondence with the VHDL/Verilog source code. Example EDA software products from Synopsys, Inc. that can be used at this step include Formality, PrimeTime, and VCS products.
Design planning (step 120): Here, an overall floor plan for the chip is constructed and analyzed for timing and top-level routing. Example EDA software products from Synopsys, Inc. that can be used at this step include Astro and Custom Designer products.
Physical implementation (step 122): The placement (positioning of circuit elements) and routing (connection of the same) occurs at this step. Example EDA software products from Synopsys, Inc. that can be used at this step include the Astro, IC Compiler, and Custom Designer products. Aspects of the invention can be performed during this step 122.
Analysis and extraction (step 124): At this step, the circuit function is verified at a transistor level, this in turn permits what-if refinement. Example EDA software products from Synopsys, Inc. that can be used at this step include AstroRail, PrimeRail, PrimeTime, and Star-RCXT products.
Physical verification (step 126): At this step various checking functions are performed to ensure correctness for: manufacturing, electrical issues, lithographic issues, and circuitry. Example EDA software products from Synopsys, Inc. that can be used at this step include the Hercules product. Aspects of the invention can be performed during this step 126 as well.
Tape-out (step 127): This step provides the “tape-out” data to be used (after lithographic enhancements are applied if appropriate) for production of masks for lithographic use to produce finished chips. Example EDA software products from Synopsys, Inc. that can be used at this step include the IC Compiler and Custom Designer families of products.
Resolution enhancement (step 128): This step involves geometric manipulations of the layout to improve manufacturability of the design. Example EDA software products from Synopsys, Inc. that can be used at this step include Proteus, ProteusAF, and PSMGen products.
Mask data preparation (step 130): This step provides mask-making-ready “tape-out” data for production of masks for lithographic use to produce finished chips. Example EDA software products from Synopsys, Inc. that can be used at this step include the CATS(R) family of products. Often this step includes partitioning or fracturing non-rectangular shaped islands into rectangles. An example method for performing such fracturing can be found in U.S. Pat. No. 8,151,236, incorporated herein by reference.
While DRC layout rules are becoming more and more complex at smaller and smaller technology nodes, most if not all of them can still be decomposed into a combination of the relationships among the edges, the corners, and the contours of shapes in the layout. Relationships “among” shapes as used herein includes relationships about a single shape as well. In embodiments herein, multiple perpendicular scan lines are used to collect all the required data in one pass, so that the combinatorial checking on the data is virtually free. The pass speed is improved even further by stopping the scan lines only at corner positions. Note that scans in multiple directions can also be combined in a particular embodiment, so that the algorithm effectively jumps from corner to corner, considering each corner only once.
In a Manhattan layout, all edges of all shapes are oriented either horizontally or vertically. In this case two scan lines would be used, one vertical (scanning horizontally) and one horizontal (scanning vertically). In each case the scan line stops only at endpoints that it encounters of the edges that are oriented perpendicularly to the scan line. The vertical scan line, for example, stops only at endpoints of horizontally oriented edges, and the horizontal scan line stops only at endpoints of vertically oriented edges. In 45 degree layouts, edges can also be oriented at a 45 degree angle or a 135 degree angle. In this case four scan lines can be used, each scanning in a direction perpendicular to a respective one of the orientations in which edges are included in the layout. While scanning the layout region in each particular direction, “corner” data structures are populated for each corner, with whatever information is easily obtainable from the edge endpoints at the corner, and from other edges that intersect the same scan line. The combined information collected from all the scan lines as they encounter the corner, is sufficient to fully populate the corner data structure.
For most of the design rules defined on derived layers, the shapes in the derived layers never need to be explicitly derived in embodiments herein. Instead, information about their shape edges is inserted into the corner data structures for physical corners that they intersect. While such information can be said to “describe” the derived shapes, only information at these physical corners need be calculated, since most of the derived layer rules are defined on corners of virtual shapes, and these corners all derive from at least one corner position of an ancestor (parent, grandparent, etc.) physical layer. The X,Y position of a derived corner might not coincide with a single physical corner, but if not, then the X position derives from one physical corner and the Y position derives from another physical corner.
Again, this can occur as part of the scan of the layout region in each particular direction. There need be no limitation on the depth or width of the derived layer graph.
Other data structures are also populated during a scan, such as information about an island (such as its area), and information about vias.
Once all the data is collected into a layout topology database, design rule checking is accomplished merely by comparing the numeric values in the layout topology database with the constraint values in the design rule data set. Unlike geometry engine approaches, the approach described herein can be performed extremely quickly, often within milliseconds, allowing for design rule checking in real time, immediately as the layout designer makes each alteration in the layout.
Moreover, since most if not all of the design rules can be framed in terms of topological relationships among edges and corners, including design rules defined on derived layers, it can be seen that the same basic information, collected during the scan, can be used in checking most if not all of the design rules, including design rules defined on derived layers, and including DPT odd cycle detection. In most embodiments, there is no need to re-scan the layout region in order to check different design rules; one scan is sufficient for collecting all the needed data. Still further, since the number of topological relationships that can be involved in checking design rules is itself limited, there is little if any additional data collection needed during the scan in order to check new and ever more complex rules. The time required to perform DRC increases less than linearly with increasing numbers of rules, and tapers off to nearly zero.
For example, if minSpacing is supported already, then minSameNetSpacing and minNotchSpacing can be supported for free (no runtime overhead). If minArea is supported already, then minRectArea can be supported for free (no runtime overhead). If 1D spacing is supported already, then 1D extension can be supported easily regardless of whether they share the same “width”. It can be seen that the more rules that are to be checked, the greater the likelihood that the next “new rule” can be supported for free or with a little extra overhead.
In step 210, the user develops a preliminary layout from a circuit design. As used herein, the term “circuit design” refers to the gate or transistor level design, before layout. The circuit design is often represented internally to the system in a netlist file. The layout is represented internally to the system in a geometry file which defines, among other things, all the shapes to be formed on each mask that will be used to expose the wafer during fabrication. The geometry file can have any of several standard formats, such as GDSII, OASIS, CREF, and so on, or it can have a non-standard format. The file describes the layout of the circuit design in the form of a mask definition for each of the masks to be generated. Each mask definition defines a plurality of polygons. At the time if
In step 212, the user views the layout on a computer monitor. The user typically selects a region of the layout for magnified viewing, so that only that region is visible on the monitor.
In step 214, the user, using a mouse or other pointing device, selects a group of one or more shapes from the visible layout region and drags them to a different location. In step 216, as the user drags the shapes, the system shows on the monitor any design rule violations in real time. In step 218, the user continues to drag the selected shapes until a position is found at which all design rule violations disappear. The user then performs the next desired editing step, which could be another drag-and-drop as in steps 214-218.
It can be seen how useful real time immediate design rule checking, enabled by the system herein, can be in manual layout or layout modification efforts.
Before discussing the methods used by an implementation of the system, it will be useful to discuss design rules in general, and how they can be represented within the system. Design rules are a set of rules that are provided by a semiconductor manufacturer, which specify minimum or maximum geometric relationships among the features of a layout. A semiconductor manufacturing process always has some variability, and the purpose of design rules is to ensure that sufficient margin is included in the layout geometries to minimize the likelihood that the variability will result in loss of yield. A set of design rules is specific to a particular semiconductor manufacturing process, so new rules are provided to designers or EDA vendors for each new process or significant process change. Despite their specificity to a particular process, there are many design rules which are similar, except for one or more numeric values, across many processes.
Design rules range from very simple to very complex. Most, however, can be framed as a set of one or more constraint parameters, and a set of one or more constraint values for the constraint parameters. (As used herein, a “parameter” is merely a slot or container for one or more values. It is not itself a value.) For example, a simple design rule is minimum edge-to-edge spacing (sometimes called minSpacing). This rule has one parameter (edge-to-edge spacing), and one value which is the minimum spacing allowed by the rule between edges in a single layer (physical or derived) of the layout. Many design rules specify more than one value for a particular parameter, such as an “absolute minimum” value and a “preferred minimum” value. As used herein, a “physical” layer is one for which geometries are specified in the geometry file.
More complex rules can have multiple parameters. An End-of-line spacing rule, for example, specifies the minimum spacing between the end of a line and its neighboring geometry. Again, this rule can be defined on any layer, including a derived layer. The constraint applies only if the width of the wire is less than a specified value, eolWidth. The constraint applies when any geometry occurs within a region defined by the minimum spacing, where the region includes the distance from each side of the wire. This distance is referred to as a lateral verification distance eolWithin. The constraint applies only if one parallel edge is within a specified rectangular region from the corners of the wire, or it applies only if two parallel edges are within a specified rectangular region from the corners of the wire. These parameters are referred to as parWithin and parSpace. The constraint applies when no parallel edges occur within the region defined by the minimum spacing, or one parallel edge occurs within the region defined by the minimum spacing, or two parallel edges. This rule has the spacing parameter itself, eolSpacing, as well as the following parameters: eolWidth, eolWithin, parWithin and parSpace.
Design rules can also specify constraints on edges in different layers. The MinDualExtension layer pair constraint, for example, specifies the minimum distance a shape on one layer must extend past a shape on a second layer. The first or second layer or both or neither can be derived layers. This rule has one parameter for extensions in the horizontal direction and another parameter for extensions in the vertical directions. This rule can also specify additional pairs of parameters, keyed by wire width. Other more complex parameters are also available for this rule, including optional parameters to qualify when the rule applies.
Design rule sets also often include area rules, such as the minimum area of an island or a hole in a layer. They can also include via rules, which specify constraints on geometric dimensions in the “cut” layer (also sometimes called the via layer), the island in the “cover” layer above the via, and the island in the “cover” layer below the via. These rules, too, can be defined on derived layers.
Derived layers are defined in the design rule sets. A derived layer can be defined by specifying a name for the derived layer, a layer number, the parent layers for the particular derived layer, and an operator for calculating the derived layer as a function of the parent layers. In one embodiment, only one or two parent layers can be specified, meaning a derived layer must have no more than two parent layers. In another embodiment, more than two parent layers can be specified. The derived layer operator can in one embodiment be a simple Boolean logic function (And, Or, XOR, or NOT). In another embodiment, the derived layer operator can be a more complex Boolean logic expression such as ‘Layer 1 AND (Layer2 OR Layer3)’. Also, the parent(s) of a derived layer need not be physical (layout) layers; one or more of them can be other derived layers, thereby allowing a nesting of derived layers. As used herein, a “parent” layer refers to an immediate parent layer. Grandparents and other ancestors (including parents) are referred to herein as “ancestor” layers.
Ultimately, each derived layer has one or more ultimate ancestor which is a physical layer, though the number of derived layers in between the particular layer and the ancestor physical layer on one side of the family tree may be different from the number on another side of the family tree. For example, a first derived layer D1 may be dependent upon physical layers P1 and P2, and a second derived layer D2 may be dependent upon D1 and physical layer P3. Then one ancestor physical layer P3 of derived layer D2 has zero derived layers in between, whereas ancestor physical layers P1 and P2 each have one derived layer in between the layer D2 and the respective physical layer P1 or P2. As used herein, the “rank” of a particular derived layer is equal to the maximum of the number of derived layers to and including the particular derived layer from each of its ancestor physical layers. The rank of a physical layer is considered to be zero, and the rank of a particular derived layer can be calculated as one plus the maximum rank of all its parent layers.
The design rules that reference derived layers are of many of the same kinds as those that reference only physical layers, and are expressed in the design rule set in the same way. For example, they can be framed as a set of one or more constraint parameters and one or more constraint values for the constraint parameters. They can include rules that apply to shapes on the particular derived layer (such as minimum edge-to-edge spacing and end-of-line spacing), as well as rules that specify constraints among different layers (such as MinDualExtension). Rules that reference more than one layer are not restricted to referencing only physical layers or only derived layers; then can also reference layers of both kinds, such as the minimum extension of a shape in a derived layer relative to a shape in a physical layer.
As a simplified example, a design rule set may include a rule that sets a minimum longitudinal spacing between transistor gate conductors. In many fabrication processes, a transistor gate conductor can be identified as the geometric intersection where a polysilicon line crosses a diffusion region. In order to establish this design rule, the design rule set may specify a derived layer G defined as L1 AND L2, where L1 is layer defining the polysilicon lines, and L2 is the layer defining the diffusion implants. In this case the derivation operator for derived layer G is “L1 AND L2”. The design rule set then specifies a design rule that references layer G, recites a parameter for specifying the minimum spacing between shapes on layer G, and recites a value for that parameter.
As used herein, a design rule “references” a particular derived layer typically by identifying the derived layer. For example, a minimum extension rule, which specifies the minimum distance by which a shape in a derived layer must extend beyond the edge of a shape in a physical layer, or vice-versa, references both layers that the rule constrains (including the derived layer) by explicitly identifying both layers (including the derived layer). Also as used herein, a derivation operator “references” a particular layer by identifying it as one of its parent layers. In the simplified example above, the derivation operator for layer G “references” both layers L1 and L2.
In an embodiment of the invention, all of the values specified by the design rules are provided to the system in the form of a design rule data set. As used herein, the term “data set” does not imply any particular organization. For example, it includes maps, multimaps, trees, as well as ordinary tables, and other data organizations as well. The term also does not necessarily imply any unity or regularity of structure. For example, two or more separate data sets, when considered together, still constitute a “data set” as that term is used herein. The terms “database” and “data structure” are also intended to have the same meaning as “data set”.
In the present embodiment, the design rule data set is sometimes referred to herein as the relationship master. A class definition for an example relationship master, in pseudo-C++, is as follows. In order to simplify the discussion, only some of the parameters are shown. A relationship_master object exists for each layer on which design rules are defined, including derived layers.
In addition to the above, relationship_master also includes a plurality of DPT rules. Among these are the following:
Note that as used herein, a “DPT spacing candidate” can be either a DPT edge-edge spacing candidate or a DPT corner-corner spacing candidate. Other types of candidates are also possible in different embodiments. A “DPT edge-edge” spacing candidate can be either a DPT end-end spacing candidate, a DPT end-side spacing candidate, or a DPT side-side spacing candidate. Again, other types of edge-edge spacing candidates are also possible in different embodiments. These terms are sometimes abbreviated herein, by omitting “spacing” or “DPT” or both.
Design Rule Checking Flow with DRC Assist
In step 310, the relationship master data set is built from a set of design rules for the target fabrication process. This can be done manually, or in some embodiments it can be automated. It is provided to the DRC system either electronically or via a computer readable medium, and it is stored accessibly to the system on a computer readable medium. As used herein, a computer readable medium is one on which information can be stored and read by a computer system. Examples include a floppy disk, a hard disk drive, a RAM, a CD, a DVD, flash memory, a USB drive, and so on. The computer readable medium may store information in coded formats that are decoded for actual use in a particular data processing system. A single computer readable medium, as the term is used herein, may also include more than one physical item, such as a plurality of CD ROMs or a plurality of segments of RAM, or a combination of several different kinds of media. A computer readable medium, as the term is used herein, excludes transitory media on which content is represented on a signal during propagation.
In step 312, the system displays on a monitor the layout or layout region selected by the user. As used herein, the term “region” refers to a portion as viewed from above, including whatever layers are pertinent. As a degenerate case, the entire layout is also a “region”. The user can manipulate (edit) objects in the layout using familiar editing commands, such as keyboard- or mouse-based behaviors recognized by the system. For example, the user can select a group of objects by clicking and dragging the mouse pointer to form a rectangle around them. Only objects on physical layers can be selected; derived layers are created only to facilitate the specification of certain error checking rules and are not visible to the user editing a layout. The user can then move the objects as a group by clicking within the rectangle and dragging it. Editing commands are recognized by the operating system and delivered to the application program by way of events in a well known manner. For example, user dragging of a group of objects might cause a series of events to be delivered to the application program, one after each movement by some number of pixels, or some number of milliseconds. The application program receives these events and determines for itself what the event represents. Step 312 can include a conventional event loop, whereby the application program repeatedly checks for new events. When it receives an event, step 312 determines that it represents a layout editing command such as user dragging of a group of shapes across the layout.
In step 313, the system determines the type of the editing command. Typically for a drag operation, the dragging behavior is preceded by one or more predecessor behaviors which indicate a predecessor command which indicates what to do in response to the dragging behavior. One predecessor command, for example, can involve selection of a group of one or more editing shapes to be moved to a different position in response to dragging behavior. Another command can involve selection of an edge or a corner, so that shape(s) will be stretched or contracted in response to dragging behavior. Yet another command, known sometimes as an array copy, can involve selection of one or more shapes to be copied, with ever increasing spacing among the copies, in response to dragging behavior. At a minimum, all of the predecessor commands relevant to the embodiment of
Initially in step 313, if the current editing command includes neither object selection nor dragging, then it is handled in a step 330. The operation of step 330 is not important for an understanding of the invention.
If the current editing command includes an object selection, then the system takes advantage of such a selection event in order to pre-calculate design rule bounds, which indicate the distance that the shapes can move before a design rule is violated. Flow passes to step 314, in which the system collects all the editing shapes, which are the ones that are being edited by the user. For a click-and-drag behavior, the editing shapes are the ones that are being moved to a different position in the layout. For a shape re-sizing command, the editing shapes are the ones being resized.
In step 315, the system populates edge speed ratios for each edge of each selected shape based on the editing command. The edge speed ratio is the ratio of edge movement distance per unit of cursor movement distance, as appropriate for the current selection command. Typically this ratio is 0 or 1, but could be anything from negative infinity to positive infinity, and need not necessarily be integral in all embodiments. Speed ratios can be better understood by reference to the illustrations in
For a derived edge, its speed ratio is determined from its ancestor physical edge(s). If all the physical ancestor edges having the same speed ratio A, then the derived edge will have the same speed ratio A. If some of the physical ancestor edges having different speed ratios, that implies that the derived geometries are about to experience a sudden change. In this case the system forces the speed ratio to zero for the derived edge. In practice, most derived edges have only a single ancestor physical edge, so these kinds of sudden derived geometry changes are infrequent. In addition, derived layers are mostly defined on lower physical layers, practically all encapsulated inside a P-CELL, so they tend to move together. That is, all the ancestor physical edges tend to have the same speed ratios. This further decreases the likelihood of sudden derived geometry changes. The ancestor physical edges are obtained from the ‘current edge group’ as described in conjunction with steps 2422 of
In a later step, the system will use the edge speed ratios to pre-calculate values indicating bounds imposed by the design rules on the distance in various directions that the selected objects can be moved. In one embodiment these design rule bounds are indicated by left, right, upward, downward and radial distances that the selected objects can be moved before one of the design rules will be violated. In another embodiment they are represented by a distance in each direction that the cursor can be moved before one of the design rules will be violated during a drag of the selected objects. This latter representation is sometimes referred to herein as a ‘slack’ distance, and is the one used in the embodiment described herein. In yet another embodiment the bounds are represented by the final bounding position in the layout to which the selected objects can be moved before one of the design rules will be violated. Other ways of indicating these bounds will be apparent. Also, in an embodiment, the set of design rules enabled to impose bounds in DRC assist mode need not include all the design rules. Fewer than all can be included in such a set in a particular embodiment.
Three types of bounds are calculated in the present embodiment: linear bounds (in the four linear directions), corner bounds and conditional bounds. Roughly described, the linear design rule bounds are calculated in the present embodiment by calculating them for each edge and successively and retaining only the most restrictive bound in each direction at each iteration. The corner and conditional bounds are applied later to further restrict the overall bounds to avoid violating any of the enabled design rules by any of the objects in the selection. Linear bounds and corner bounds need not be calculated immediately sequentially, and in present embodiment they are not. Rather, each type of bound is calculated where most cost-effective to calculate it. In fact, the calculation of vertical linear bounds is spread out during the horizontal scan to populate the corner data structures, and the calculation of horizontal linear bounds is spread out during the vertical scan to populate the corner data structures. Similarly, calculation of corner bounds is performed during the scan of corners to check corner-to-corner design rules, and the calculation for conditional bounds is performed during the check of other corner-based design rules such as the end-of-line spacing rule.
In step 316, the system collects all the surrounding shapes, which in a click-and-drag command, are the shapes near the new position of the editing shapes. Again, only real shapes, not those on derived layers, are included. A selection algorithm is used here which errs on the side of collecting more shapes than necessary, since while inclusion of additional shapes could impact performance, the exclusion of relevant shapes will impact accuracy. One efficient way to collect appropriate shapes is to create a bounding box around the editing shapes in their new position, then extend the box in all four directions by 1.5 times the worst case minimum spacing or the worst case minimum inter-layer clearance, whichever is larger. All shapes at least partially overlapping with the expanded bounding box, in any layer, are then included in the result. A conventional range search engine can be used for this step. Geometry processing is not needed.
In step 318, horizontal and vertical scan line trees sweep_x and sweep_y are built from all of the collected shapes, including both the editing shapes and the static shapes. The horizontal scan line tree sweep_x is a map of particular vertical scan lines, and will be scanned horizontally across the selected layout region, from left to right. The vertical scan line tree sweep_y is a map of particular horizontal scan lines, and will be scanned vertically across the selected layout region, from bottom to top.
Map 516 is an expansion of exit_tree 514; enter_tree 512 has the same structure and is therefore not shown in
Multimap 518 is an expansion of one of the edge_tree structures 520. The other edge_trees have the same structure and therefore are not shown in
Block 522 is an expansion of one of the edge structures 524. The other edges have the same structure and therefore are not shown in
Block 528 is an expansion of derived edge status map 526. It contains information about horizontal edges in some of the derived layers, which edges terminate at or pass through the current vertical scan line at the current vertical position on that scan line. Each entry contains the derived layer ID, in association with a status vector for the edge. The status vector for a derived edge is similar to a quadrant depth vector (discussed later), in that it contains four values indicating status in the four respective quadrants centered at the current x- and y-position. It differs from a quadrant depth vector in that each entry of the vector can contain only a 1 or a 0: a 1 indicates that the current derived layer does have an island in that quadrant, and a 0 indicates that it does not (or vice-versa). There is no need to indicate the number of superimposed shapes in the quadrant within a derived layer, because the shapes are defined in a Boolean operation: in a given location a shape is either present or absent.
It can be seen that a status vector having one ‘1’ and three ‘0’s indicates a convex corner of the derived shape, whereas a status vector having one ‘0’ and three ‘1’s indicates a concave corner of a derived shape. A status vector of (0,0,1,1), (0,1,1,0), (1,1,0,0) or (1,0,0,1) indicates a non-corner edge of a derived shape. A status vector of (0,1,0,1) or (1,0,1,0) indicates two derived shapes meeting at a common corner; a situation that will usually violate design rules. Note that only edges and corners of a derived shape appear in derived edge status maps. If the no shape at that position. If it is (1,1,1,1), then the current x/y position is inside a derived shape. In either case, no entry is made for the current derived layer in the derived edge status map 526.
It will be appreciated that each x/y position of corners on physical layers may be represented numerous times in the Sweep_x data structure. It may be represented in both Enter_tree 512 and Exit_tree 514. It might also be represented at multiple physical layers (multiple entries in block 518 all having a common y_pos key). Because many of the derived edges at a particular x/y position can be stored in a single map 526, there may be multiple edge data structures 522 which would serve as an appropriate place to store each derived edge status map 526. Different embodiments can implement different conventions on this point. In one embodiment, the status vectors for all the derived edges on a particular derived layer are inserted into an edge data structure for the “first” physical layer that is one of its physical ancestors. Preferably the “first” physical layer is defined as whichever layer is used to break a tie when vertical positions of edges are the same in the current vertical scan line. In the embodiment of
Map 616 is an expansion of exit_tree 614; enter_tree 612 has the same structure and is therefore not shown in
Multimap 618 is an expansion of one of the edge_tree structures 620. The other edge_trees have the same structure and therefore are not shown in
Block 622 is an expansion of one of the edge structures 624. The other edges have the same structure and therefore are not shown in
Block 628 is an expansion of derived edge status map 626. It contains information about vertical edges in some of the derived layers, which edges terminate at or pass through the current horizontal scan line at the current horizontal position on that scan line. Each entry contains the derived layer ID, in association with a status vector for the edge. Like for derived edge information stored in the Sweep_x data structure, the status vectors for all the vertical derived edges on a particular derived layer are inserted into an edge data structure 624 for the smallest numbered physical layer that is one of the ancestors of the particular derived layer.
As can be seen, sweep_x contains only horizontal edges (physical and derived) and sweep_y contains only vertical edges (physical and derived). Thus the scan lines in each data structure are perpendicular to the edges that will be encountered during a traversal of the structure. In an embodiment supporting diagonal edges as well, two more sweep data structures are present as well: one containing scan lines oriented parallel to one diagonal and the other containing scan lines oriented parallel to the other diagonal. Each data structure includes only edges oriented perpendicularly to its scan lines, so again, a scan line sweep of the scan lines in each structure will encounter only those edges oriented perpendicularly to the scan line.
In step 714, enter_tree is created for sweep_x. This is accomplished by, at each unique horizontal position represented in the sorted list (step 716), creating a scan line multimap (of class ‘edge_tree’) for a vertical scan line at that horizontal position (step 718). In step 720, the scan line multimap at that horizontal position is populated with all the edges (structures of class ‘edge’) in the list having left-hand endpoints at the current horizontal position.
After enter_tree has been created and populated for sweep_x, the list from step 710 is re-sorted by horizontal position of all the right-hand endpoints of the edges. Again, there may be multiple edges whose right-hand endpoints have the same horizontal position. In step 724, exit_tree is created for sweep_x. Similarly to the creation of enter_tree, this is accomplished by, at each unique horizontal position represented in the sorted list (step 726), creating a scan line multimap (of class ‘edge_tree’) for a vertical scan line at that horizontal position (step 718). In step 720, the scan line multimap at that horizontal position is populated with all the edges (structures of class ‘edge’) in the list having right-hand endpoints at the current horizontal position.
In step 814, enter_tree is created for sweep_y. This is accomplished by, at each unique vertical position represented in the sorted list (step 812), creating a scan line multimap (of class ‘edge_tree’) for a horizontal scan line at that vertical position (step 818). In step 820, the scan line multimap at that vertical position is populated with all the edges (structures of class ‘edge’) in the list having lower endpoints at the current vertical position.
After enter_tree has been created and populated for sweep_y, the list from step 810 is re-sorted by horizontal position of all the upper endpoints of the edges. Again, there may be multiple edges whose upper endpoints have the same vertical position. In step 824, exit_tree is created for sweep_y. As before, this is accomplished by, at each unique vertical position represented in the sorted list (step 822), creating a scan line multimap (of class ‘edge_tree’) for a horizontal scan line at that vertical position (step 818). In step 820, the scan line multimap at that vertical position is populated with all the edges (structures of class ‘edge’) in the list having upper endpoints at the current vertical position.
Returning now to
In step 1012, current_scan_line is updated by adding all horizontal edges having a left-hand endpoint located at the current horizontal scan position. In step 1014, the quadrant depth vector (
The quadrant depth vector indicates the number of shapes in a particular layer that border a particular edge endpoint in each of the four quadrants centered at that endpoint. In
In step 1014, the updating of the quadrant depth vector for an edge in the enter_tree (i.e. an edge whose left-hand endpoint lies on the current vertical scan line), involves incrementing the value for either quadrant I or quadrant IV by one. The value for quadrant I is incremented if the “edge against scan line?” Boolean for the edge 1112 indicates True (i.e. the edge is the bottom edge of a shape), or the value for quadrant IV is incremented if the “edge against scan line?” Boolean for the edge 1112 indicates False (i.e. the edge is the top edge of a shape). Similarly, the updating of the quadrant depth vector for an edge in the exit_tree (i.e. an edge whose right-hand endpoint lies on the current vertical scan line), involves decrementing the value for either quadrant I or quadrant IV by one. The value for quadrant I is decremented if the “edge against scan line?” Boolean for the exiting edge indicates True (i.e. the edge is the bottom edge of a shape), or the value for quadrant IV is decremented if the “edge against scan line?” Boolean for the exiting edge indicates False (i.e. the edge is the top edge of a shape). It can be seen that the quadrant depth vector increments quantities as the vertical scan line encounters shapes while moving left-to-right across the region. It decrements quantities as the scan line moves past shapes.
In step 1015, the derived layer information is populated for each derived layer having an edge intersecting the current scan line.
In step 2412, the system collects all the “related” physical layers related to any of the physical horizontal edges in current edge group. As used in
In step 2414, at the current y_position on current vertical scan line, the Boolean status vector for each related physical layer is determined and written into a map object referred to herein as a related_edge_status map. At this step, the related_edge_status map is populated only for physical layers.
In step 2416, all the derived layers derived ultimately from physical layers containing any of the edges in current edge group are collected. The system sorts these monotonically according to their rank. Thus derived layers that involve fewer derivation steps appear earlier in this collection than those that involve more derivation steps. This arrangement ensures that during a traversal of the collection in sorted order to calculate derived layer information, those derived layers which are ancestors of a child derived layer will already have been calculated by the time the child derived layer is reached.
In step 2418, a loop is begun through the derived layers in the collection, in sorted order. In step 2420, the derivation operator for the current derived layer is used to derive the Boolean status vector for the derived edge on the current derived layer at the current x/y position. The four elements of the Boolean status vector are calculated simply by applying the derivation operator for the current derived layer separately to each of the four elements of the Boolean status of the parent layer(s) referred to by the derivation operator for the current derived layer. For example, if the derivation operator for derived layer D1 is “P1 AND P2”, where P1 and P2 are physical layers, and the Boolean status vectors for P1 and P2 at the current x/y position are (a,b,c,d) and (e,f,g,h), respectively, then the Boolean status vector for the derived edge on the current derived layer at the current x/y position is calculated as (a·e, b·f, c·g, d·h).
In step 2422, the Boolean status vector just calculated for current derived layer at current y_position on current vertical scan line is inserted into the m_derived_edge_status map in the edge object 524 for the first physical edge of current edge group. In step 2424, the Boolean status vector just calculated is also inserted into the related_edge_status map for the current derived layer, at the current y_position on the current vertical scan line. This prepares the related_edge_status map in case a subsequent derived layer of higher rank refers to the current derived layer in its derivation operator.
In step 2426, it is determined whether there are any more derived layers in the collection prepared in step 2416. If so, then the system returns to step 2418 to derive the Boolean status vector for the next such derived layer. If not, then the system returns to step 2410 to address the edge group intersecting the current vertical scan line at the next y-position. If there are no more such y-positions, then step 1015 for populating the derived layer information for derived layers having an edge intersecting the current vertical scan line is complete (step 2428). It can be seen that in the embodiment of
Returning to
Note that for derived corners, the edge* objects in the above class all represent physical edges. Only the position of the edge* are useful (borrowed) for derived edges.
A ray object represents essentially an arrow with a head point and tail point. All the tail points coincide with the current corner. For Manhattan layouts the rays are either horizontal or vertical, though in 45 degree layouts it can also have either of the two diagonal orientations. The ‘ray’ class is described in a C++ like pseudocode class definition as follows:
The corner data structures developed during the scan are maintained as entries in a synchronized_corner_map structure. This structure is a map, in which the keys identify a physical or derived layer number and an x and y position on that layer, and the values are objects of class ‘corner’. Only edges on physical layers are handled in this step 1016; derived layer edges are processed in step 1019.
In step 1216, a corner data structure for the left-hand endpoint of the current edge is instantiated in synchronized_corner_map if it does not already exist. The corner data structure might already exist in synchronized_corner_map if, for example, the corner had already been encountered because of a different horizontal edge on the same layer that starts at the same point (such as the bottom edge of rectangle 1138 in
In one embodiment, all design rule checks are performed only after all scans are complete. However, the present embodiment incorporates a feature in which the system performs certain simple edge-based rule checks as part of step 1218. For example, if the current edge is a top edge and the walk upwards along the current vertical scan line meets the bottom edge of a shape in the same layer, then s_ray_y is populated in the corner data structure and the minimum spacing rule is checked as well. This check involves comparing the length of s_ray_y with the minimum spacing value in the relationship_master. If the current edge is a top edge and the walk upwards along the current vertical scan line meets the top edge of a shape in a different layer, then the minimum extension rule is checked by comparing the distance walked to the minimum extension value for the appropriate layer pair in the relationship_master. If the current edge is a bottom edge and the walk upwards along the current vertical scan line meets the top edge of a shape in the same layer, then d_ray_y is populated, and also the minimum dimension rule is checked. This check involves comparing the value of d_ray_y with the minimum dimension value in the relationship_master. If the current edge is a bottom edge and the walk upwards along the current vertical scan line meets the top edge of a shape in a different layer, then the minimum overlap rule is checked. Similar checks are performed during the walk downward from the current edge. If during the walks up and down the current vertical scan line, the distance walked exceeds the worst case limit from the relationship master, there is no design rule violation encountered and it is not necessary to populate further items in the corner data structure that would be encountered in the current walking direction.
In an embodiment which checks for DPT odd cycle variations, though not essential, it is also preferable at this point to detect edge-edge DPT candidates discernable from the above. Thus if the length of s_ray_y is less than the largest of the edge-edge DPT minimum spacing values set forth in the DPT design rules in the relationship_master, then a DPT_edge_marker object is inserted into a structure DPT_edge_marker_set. This object identifies the current edge and the edge at the far end of s_ray_y, as well as the current scanline position.
After the available corner structure information items have been populated, the system then calculates the vertical linear design rule bounds for the current entering edge in the current vertical scan line (step 1220). In order to best understand this calculation, call the current horizontal edge e1 and call the next lower horizontal entering edge in the current vertical scan line e2. Assume the current distance between them is d, and the minimum rule value is r. Depending on the particular design rule being evaluated, r could be a minimum spacing value, minimum width, minimum extension, minimum overlap, and so on. Further assume that e1 has a speed ratio (determined in step 315) of s1 and e2 has a speed ratio of s2. According to the calculation, if s1=s2, then the two edges will remain the same distance from each other vertically during the drag operation. Thus no vertical linear design rule bound is created. If s1<s2, that means e1 will move more slowly than s2 during the drag operation. If the drag has a vertically downward vector component, then the spacing between the two edges will increase during the drag and no minimum distance rule will be violated. Again, therefore, no vertical linear design rule bound is created for downward linear movement. But if the drag has a vertically upward vector component, then the spacing between the two edges will decrease during the drag and eventually violate the design rule. The maximum distance that edge e1 can be allowed to move upward is then given by (d−r), which corresponds to a maximum allowable upward cursor movement (i.e. the linear upward slack distance) of d/|s2−s1|−r. Thus a linear upward slack distance is cached for the current horizontal edge e1.
Note that vertical movement of e1 relative to the next higher horizontal entering edge in the current vertical scan line must be considered as well, and that might result in a linear downward slack distance and/or a further restriction on the linear upward slack distance. Subsequent calculations (in step 1018 for example, for exiting edges) might further restrict the distance that edge e1 can move upward or downward, thereby further restricting the linear upward or downward slack. Furthermore, slack calculations for other horizontal edges intersecting the current vertical scan line or any other vertical scan line in the vertical scan line multimap can still further restrict the distance that the cursor should be allowed to move vertically upward or downward during the drag. Since only the one most restrictive slack distance need be stored for the upward direction and one for the downward direction, each new calculation of a slack distance overwrites the previous slack distance if the new slack distance is more restrictive (smaller). This procedure also occurs for each design rule in the set of design rules enabled for DRC assist, including design rules setting a maximum rule value r rather than a minimum rule value, with always the most restrictive slack distance overwriting a less restrictive slack distance in the same direction.
After the available corner structure information items have been populated, then the system returns to step 1210 to consider the next entering edge in the current vertical scan line.
In step 1316, a corner data structure for the right-hand endpoint of the current edge is instantiated in synchronized_corner_map if it does not already exist. Again, the corner data structure might already exist in synchronized_corner_map if, for example, the corner had already been encountered because of a different horizontal edge on the same layer that ends at the same point, or as part of the vertical scan in an embodiment in which the vertical scan precedes or operates concurrently with the horizontal scan. In step 1318, the system walks upward and downward along the current vertical scan line from the current horizontal edge, populating the available corner information as it is learned. In particular, referring to the corner data structure definition above and the illustration in
In addition, preferably but not essentially, the system also in step 1318 performs the same edge-based rule checks for the exiting edges as performed and described above with respect to step 1218 for entering edges.
In an embodiment which checks for DPT odd cycle variations, though not essential, edge-edge DPT candidates discernable from the above preferably are detected in the same way as set forth above with respect to step 1218 for entering edges. Marker objects are inserted into DPT_edge_marker_set which indicate the current edge and the edge at the far end of s_ray_y, as well as the current scanline position. Note that if the new marker object duplicates one already in DPT_edge_marker_set, it is not inserted again.
After the available corner structure information items have been populated, the system then calculates additional vertical linear design rule bounds for the current exiting edge in the current vertical scan line (step 1320). These bounds are calculated similarly as set forth above with respect to step 1220, and again, only the most restrictive upward and downward vertical slack distances are retained.
After the available corner structure information items have been populated, then the system returns to step 1310 to consider the next exiting edge in the current vertical scan line.
Returning to
If there are derived edges, then in step 2514 the system begins another loop through all the derived edges in the derived edge status map in the first physical edge object at the current vertical position. In step 2516, it is determined whether the Boolean status of the current derived edge indicates a valid corner. As previously explained, so long as the Boolean status is neither all zeros nor all ones, the corner is valid. If the current derived edge is not a corner, then the edge is skipped.
If the current derived edge is a valid corner, then in step 2518, a corner data structure for the current derived edge is instantiated in synchronized_corner_map if it does not already exist. In step 2520, the system walks upward and downward along the current vertical scan line from the current derived edge, populating the available derived corner information as it is learned. In particular, as for physical layer edges, s_ray_y, tar_y and d_ray_y, as well as any others required by the design rules, are populated. Also in step 2520, like in step 1218 for physical layer edges, the system performs certain simple edge-based rule checks such as minimum spacing and minimum dimension on the current derived layer, and minimum extension and minimum overlap relative to other layers (physical or derived). All the same design rule checks are performed on the derived layers as set forth above with respect to step 1218. Additionally, like in step 1218, if during the walks up and down the current vertical scan line on the current derived layer, the distance walked exceeds the worst case limit from the relationship master, there is no design rule violation encountered and it is not necessary to populate further items in the corner data structure that would be encountered in the current walking direction.
In step 2524, the system then calculates additional vertical linear design rule bounds for the current derived edge. These bounds are calculated similarly as set forth above with respect to step 1220, and again, only the most restrictive upward and downward vertical slack distances are retained.
After the available derived corner structure information items have been populated, and the vertical design rule bounds imposed by the current derived edge have been taken into account, then the system returns to step 2514 to consider the next derived edge in the current physical edge object. If there are no more derived edges in the current physical edge object, then the system returns to step 2510 to move to the next unique vertical position in the current vertical scan line. If there are no more vertical positions represented in the current vertical scan line, then step 1019 concludes (step 2522).
Returning to
Among other things, the island data structure accumulates the following information about a particular island during the process of the horizontal scan: area of the island, area of a hole just above the island, common run lengths against other islands in the same layer and islands in other layers. For clarity of illustration, the present description will concentrate primarily on the island area as an example of island-based rule checking Reference will be made to
Roughly described, island area is accumulated during the horizontal scan by using the shape corners to divide the island into non-overlapping “island rectangles”, the area of which are easily determined from the horizontal edges represented in the current vertical scan line. In the example of
Referring to
If the current island is already represented in island_map, then effectively a vertical slice is made through the current island at the current vertical scan line; and the area of the left-adjacent rectangle is added to the area being accumulated. Accordingly, in step 2318, the height H of the left-adjacent rectangle is calculated as the distance along the current vertical scan line from the bottom edge of the current island to the top edge of the current island. This information is available in current_scan_line, because at least one of the top and bottom edges is a corner, and the y-position of the corner is available as the left- or right-hand endpoint of a horizontal edge in the current vertical scan line. The other of the top and bottom edges may also be a corner, or may be an edge that merely intersects the current vertical scan line. In either case its y-position is available as well in current_scan_line. In step 2320, the width W of the left-adjacent rectangle is calculated as the horizontal position of the current scan line minus the last scan line position at which island information was updated, which is the value in m_last_position_updated. In step 2322 the product of H and W is added to the area value for the current island.
In step 2324, as mentioned above, the value of m_last_position_updated for the new island is set equal to the x-position of the current vertical scan line. The method then returns to step 2310 for consideration of the next island having a corner on the current vertical scan line.
Once all islands having a corner on the current vertical scan line have been considered, then any two or more of such islands that are now vertically-adjacent are merged into a single island in step 2326 and their area values summed. In step 2328, any island that is now split into two, perhaps separated vertically by a newly encountered hole or notch, are split. The details of the merging and splitting operations are not important for an understanding of the invention. Note that whereas island area information is captured during the horizontal scan, it is not compared to the design rule values in the present embodiment until later. Note also that whereas
Returning to
Returning to
In step 1412, current_scan_line is updated by adding all vertical edges having a lower endpoint located at the current horizontal scan position. In step 1414, the quadrant depth vector (
In step 1415, the derived layer information is populated for each derived layer having an edge intersecting the current scan line.
In step 2612, the system collects all the “related” physical layers related to any of the physical vertical edges in current edge group. Again, as used in
In step 2614, at the current x_position on current horizontal scan line, the Boolean status vector for each related physical layer is determined and written into a related_edge_status map. At this step, the related_edge_status map is populated only for physical layers.
In step 2616, all the derived layers derived ultimately from physical layers containing any of the edges in current edge group are collected. The system sort these monotonically according to their rank, for the same reason as set forth above with respect to step 2416.
In step 2618, a loop is begun through the derived layers in the collection, in sorted order. In step 2620, the derivation operator for the current derived layer is used to derive the Boolean status vector for the derived edge on the current derived layer at the current x/y position. The four elements of the Boolean status vector are calculated by applying the derivation operator for the current derived layer separately to each of the four elements of the Boolean status of the parent layer(s) referred to by the derivation operator for the current derived layer.
In step 2622, the Boolean status vector just calculated for current derived layer at current x_position on current horizontal scan line is inserted into the m_derived_edge_status map in the edge object 524 for the first physical edge of current edge group. In step 2624, the Boolean status vector just calculated is also inserted into the related_edge_status map for the current derived layer, at the current x_position on the current horizontal scan line. This prepares the related_edge_status map in case a subsequent derived layer of higher rank refers to the current derived layer in its derivation operator.
In step 2626, it is determined whether there are any more derived layers in the collection prepared in step 2616. If so, then the system returns to step 2618 to derive the Boolean status vector for the next such derived layer. If not, then the system returns to step 2610 to address the edge group intersecting the current horizontal scan line at the next x-position. If there are no more such x-positions, then step 1415 for populating the derived layer information for derived layers having an edge intersecting the current horizontal scan line is complete (step 2628). It can be seen again that in the embodiment of
Returning to
In step 1516, a corner data structure for the left-hand endpoint of the current edge is instantiated in synchronized_corner_map if it does not already exist. In step 1518, the system walks leftward and rightward along the current horizontal scan line from the current vertical edge, populating the available corner information as it is learned. In particular, referring to the corner data structure definition above and the illustrations in
In an embodiment, certain edge-based rule checks are also performed as part of step 1518, similar to those performed in step 1218. For example, if the current edge is a right-hand edge and the walk rightward along the current horizontal scan line meets the left-hand edge of a shape in the same layer, then s_ray_x is populated in the corner data structure and the minimum spacing rule is checked as well. This check involves comparing the length of s_ray_x with the minimum spacing value in the relationship_master. If the current edge is a right-hand edge and the walk rightwards along the current horizontal scan line meets the right-hand edge of a shape in a different layer, then the minimum extension rule is checked by comparing the distance walked to the minimum extension value for the appropriate layer pair in the relationship_master. If the current edge is a left-hand edge and the walk rightwards along the current horizontal scan line meets the right-hand edge of a shape in the same layer, then d_ray_x is populated, and also the minimum dimension rule is checked. This check involves comparing the value of d_ray_x with the minimum dimension value in the relationship_master. If the current edge is a left-hand edge and the walk rightwards along the current horizontal scan line meets the right-hand edge of a shape in a different layer, then the minimum overlap rule is checked. Similar checks are performed during the walk leftward from the current edge. If during the walks leftward and rightward along the current horizontal scan line, the distance walked exceeds the worst case limit from the relationship master, there is no design rule violation encountered and it is not necessary to populate further items in the corner data structure that would be encountered in the current walking direction.
In an embodiment which checks for DPT odd cycle variations, though not essential, it is also preferable at this point to detect edge-edge DPT candidates discernable from the above. Thus if the length of s_ray_x is less than the largest of the edge-edge DPT minimum spacing values set forth in the DPT design rules in the relationship_master, then a DPT_edge_marker object is inserted into DPT_edge_marker_set as set forth with respect to step 1218. This object identifies the current edge and the edge at the far end of s_ray_x, as well as the current scanline position.
After the available corner structure information items have been populated, the system then calculates the horizontal linear design rule bounds for the current entering edge in the current horizontal scan line (step 1520). This calculation is similar to that set forth above with respect to step 1220, but will be adapted here for clarity. Call the current vertical edge e1 and call the next lower vertical entering edge in the current horizontal scan line e2. Assume the current distance between them is d, and the minimum rule value is r. Further assume that e1 has a speed ratio (determined in step 315) of s1 and e2 has a speed ratio of s2. If s1=s2, then the two edges will remain the same distance from each other horizontally during the drag operation and no horizontal linear design rule bound is created. If s1<s2, that means e1 will move more slowly than s2 during the drag operation. If the drag has a horizontally leftward vector component, then the spacing between the two edges will increase during the drag and no minimum distance rule will be violated. Again, therefore, no horizontal linear design rule bound is created for leftward linear movement. But if the drag has a horizontally rightward vector component, then the spacing between the two edges will decrease during the drag and eventually violate the design rule. The maximum distance that edge e1 can be allowed to move rightward is then given by (d−r), which corresponds to a maximum allowable rightward cursor movement (i.e. the right linear slack distance) of d/|s2−s1|−r. Thus a right linear slack distance is cached for the current vertical edge e1.
Again, horizontal movement will also be restricted by left and right linear slacks calculated for the same and other edges of the selected objects, as well as for satisfaction of other design rules. Since only the one most restrictive linear slack distances need be stored for drag vector components in each of the leftward and rightward directions, each new calculation of a slack distance overwrites the previous slack distance if the new slack distance is more restrictive (smaller). Only one linear slack distance in each direction remains.
After the available corner structure information items have been populated, then the system returns to step 1510 to consider the next entering edge in the current horizontal scan line.
In step 1616, a corner data structure for the upper endpoint of the current edge is instantiated in synchronized_corner_map if it does not already exist. Again, the corner data structure might already exist in synchronized_corner_map. In step 1618, the system walks leftward and rightward along the current horizontal scan line from the current vertical edge, populating the available corner information as it is learned. In particular, referring to the corner data structure definition above and the illustration in
In addition, preferably but not essentially, the system also in step 1618 performs similar edge-based rule checks for the exiting edges as performed and described above with respect to step 1318.
In an embodiment which checks for DPT odd cycle variations, though not essential, edge-edge DPT candidates discernable from the above preferably are detected in the same way as set forth above with respect to step 1518 for entering edges. Marker objects are inserted into DPT_edge_marker_set which indicate the current edge and the edge at the far end of s_ray_x, as well as the current scanline position. Note that if the new marker object duplicates one already in DPT_edge_marker_set, it is not inserted again.
After the available corner structure information items have been populated, the system then calculates additional horizontal linear design rule bounds for the current exiting edge in the current horizontal scan line (step 1620). These bounds are calculated similarly as set forth above with respect to step 1520, and again, only the most restrictive left and right linear slacks are retained.
After the available corner structure information items have been populated, then the system returns to step 1610 to consider the next exiting edge in the current horizontal scan line.
Returning to
If there are derived edges, then in step 2714 the system begins another loop through all the derived edges in the derived edge status map in the first physical edge object at the current horizontal position. In step 2716, it is determined whether the Boolean status of the current derived edge indicates a valid corner. As previously explained, so long as the Boolean status is neither all zeros nor all ones, the corner is valid. If the current derived edge is not a corner, then the edge is skipped.
If the current derived edge is a valid corner, then in step 2718, a corner data structure for the current derived edge is instantiated in synchronized_corner_map if it does not already exist. In step 2720, the system walks left and right along the current horizontal scan line from the current derived edge, populating the available derived corner information as it is learned. Also in step 2720, like in step 1218 for physical layer edges, the system performs certain simple edge-based rule checks such as minimum spacing and minimum dimension on the current derived layer, and minimum extension and minimum overlap relative to other layers (physical or derived). All the same design rule checks are performed on the derived layers as set forth above with respect to step 1218. Additionally, like in step 1218, if during the walks to the left and right along the current horizontal scan line on the current derived layer, the distance walked exceeds the worst case limit from the relationship master, there is no design rule violation encountered and it is not necessary to populate further items in the corner data structure that would be encountered in the current walking direction.
In step 2724, the system then calculates additional horizontal linear design rule bounds for the current derived edge in the current horizontal scan line. These bounds are calculated similarly as set forth above with respect to step 1520, and again, only the most restrictive left and right linear slacks are retained.
After the available derived corner structure information items have been populated and horizontal linear slacks have been updated, then the system returns to step 2714 to consider the next derived edge in the current physical edge object. If there are no more derived edges in the current physical edge object, then the system returns to step 2710 to move to the next unique horizontal position in the current horizontal scan line. If there are no more horizontal positions represented in the current horizontal scan line, then step 1419 concludes (step 2722).
Returning to
In step 1422, as a time saving technique, the quadrant depth vectors for each of the entering vertical edges in the current horizontal scan line are copied from the upper quadrants to the corresponding lower quadrants. In this manner the lower quadrant depth values can be incremented or decremented as the horizontal scan line moves upward, and will contain accurate values when the scan line reaches the upper endpoint of the edge. It is not necessary to update status vectors for derived edges in the same manner. In step 1424, all the exiting edges are removed from the current horizontal scan line. The routine then returns to step 1410 for the next vertical scan position.
Returning to
Corner design rule bounds are calculated as part of the check of corner-to-corner rules in step 1710. Conditional design rule bounds are calculated as part of the check of other corner-based rules in step 1712.
In step 1812, the ray map is scanned left-to-right to identify intersections of the rays. A conventional scan line algorithm can be used for this purpose.
In step 1814, it is determined whether the current ray intersection is an intersection of two space_rays. The two corners from which these space_rays extend both have to be convex, so the situation is as illustrated in
In an embodiment which checks for DPT odd cycle variations, though not essential, it is also preferable at this point to detect corner-corner DPT candidates discernable from the above. Thus in step 1817, if the two shapes are located on the same layer, the spacing 1914 is compared also to the DPT minimum corner-to-corner spacing value in relationship_master. If the distance between the corners is shorter, then a DPT_corner_marker object is inserted into DPT_edge_marker_set as set forth with respect to step 1218. A DPT_corner_marker identifies the two corners, which in turn identify the x and y positions of each corner. Note that in one embodiment, the Euclidean distance between the two corners is used whereas in another embodiment the Manhattan distance is used. In yet another embodiment the Euclidean distance is used in certain conditions and the Manhattan distance is used in other situations. In still other embodiments, yet another kind of distance measurement is used.
Returning to
If the intersecting rays are not both dimension_rays, then in step 1822 it is determined whether one is a space_ray on one layer, and the other is a dimension_ray on a different layer. Since the corner from which the space_ray extends is convex, and the corner from which the dimension_ray extends in concave, the situation is as illustrated in
Various other corner-based design rule checks can be performed within this loop as well, not shown in
A corner slack is the slack between two corners, i.e. one pair of horizontal and one pair of vertical edges. A corner slack has two linear slack values, one horizontal (either left or right) and one vertical (either up or down). These horizontal and vertical linear slack components of the corner slack are calculated in the same way that the horizontal and vertical linear slack values are calculated within step 320. However, the values calculated for corner slacks are cached separately and not merged with those calculated for the linear slacks.
In step 2014, it is determined whether the current corner is concave or convex. If it is concave, then in step 2016 the concave corner edge length rule is checked. This rule requires that at least one of the two adjacent edges forming a concave corner have at least a minimum length. This test can be performed using the same values from the corner data structure as used in step 2012 (ori_x, tar_x, ori_y and tar_y). The lengths determined for the two edges are compared to the minimum concave corner edge length value in the relationship_master.
In step 2018, the notch rule is checked. This rule requires that a ‘notch’ in an island have at least a specified minimum width. Framed in terms of corners, the rule requires that two adjacent concave corners be at least a specified distance apart. This rule need be checked for a horizontally-adjacent corner only of the horizontally-adjacent corner is concave, and need be checked for a vertically-adjacent corner only of the vertically-adjacent corner is concave. For example, in the illustration of
If in step 2014, it is determined that the current corner is convex, then in step 2020 the convex corner edge length rule is checked. This rule requires that at least one of the two adjacent edges forming a convex corner have at least a minimum length. This test can be performed using the same values from the corner data structure as used in step 2012 (ori_x, tar_x, ori_y and tar_y). The lengths determined for the two edges are compared to the minimum convex corner edge length value in the relationship_master.
In step 2022, an end-of-line spacing rule is checked. In its simplest form, this rule requires that at the end of a line, a specified minimum spacing is required to the neighboring geometry. Referring to
After all the desired rules are checked for the current corner, conditional design rule bounds are calculated in step 2026. The routine then returns to step 2010 to consider the next corner in synchronized_corner_map.
A conditional slack is the slack for a general conditional rule context. Complex design rules with multiple sub-rules, such as for example some of the design rules checked in
Returning to
In step 1715, DPT odd cycle detection is performed. In an embodiment that supports triple-patterning-technology (TPT) design rule violation detection, this can be performed as part of step 1715 as well. The embodiment of
In the embodiment of
As used herein, a “graph” is a set of nodes connected together by arcs. A “tree”, as used herein, means a connected graph with no cycles. Thus a process of “building a tree” also constitutes a process of “building a graph”, although it is a special case of a graph. A “forest”, as used herein, is a graph containing only disconnected trees, and a single tree is considered herein to be a degenerate case of a Forest. In aspects of the present invention, there is no requirement that any particular node of a tree be designated as a root node, though it is useful to do so and the embodiment herein does so.
Referring to
In step 2914, the routine begins to iterate through all the available islands represented in the three multimaps. Each new island to consider is identified at this point by a particular corner of the island, which has been designated as the “root” corner.
Referring to
If in step 3010 it is determined that candidate_corner_map is empty, then the routine next looks in corner_map. In step 3020, if corner_map is not empty, then in step 3022, the first available corner in corner_map is taken as the selected corner. As this corner was not found by way of a relationship with a previously-considered corner or edge, it is presumed (at least initially) to be disjoint from any existing tree. In step 3024, therefore, a new tree is started with its root as the island identified by the selected corner. In an embodiment which supports predefined color groups, the pre-groups value for the tree is set initially to no_color. The selected corner is then returned in step 3018 as the root corner of the first (or next) available island to consider in the iteration. No color is assigned to the island yet; it will be assigned when the island is connected to a tree.
If in step 3020 it is determined that corner_map is empty, then the routine next looks in 1st_edge_map. In step 3030, if 1st_edge_map is not empty, then in step 3032, one of the endpoints (corners) of an available edge in 1st_edge_map is taken as the selected corner. It will be recalled that 1st_edge_map is a multimap, keyed according to the head edge of the edge-edge candidates. It will also be appreciated that both endpoints of any given edge are disposed on the same island. Thus since the iteration of step 2914 is ultimately for identifying islands rather than edges or corners, step 3032 merely involves selecting one of the key edges in 1st_edge_map, and then selecting as the selected corner one of the endpoints from the selected edge. It does not matter which key edge is selected nor which endpoint of that edge is selected. Again, as the selected corner was not found by way of a relationship with a previously-considered corner or edge, it is presumed (at least initially) to be disjoint from any existing tree. In step 3024, therefore, a new tree is started with its root as the island identified by the selected corner. In an embodiment which supports predefined color groups, the pre-groups value for the tree is set initially to no_color. The selected corner is then returned in step 3018 as the root corner of the first (or next) available island to consider in the iteration. No color is assigned to the island yet; it will be assigned when the island is connected to a tree.
If in step 3030 it is determined that 1st_edge_map is empty, then the routine next looks in 2nd_edge_map. In step 3040, if 2nd_edge_map is not empty, then in step 3042, one of the endpoints (corners) of the first available edge in 2nd_edge_map is taken as the selected corner. It will be recalled that 2nd_edge_map is a multimap, keyed according to the tail edge of the edge-edge candidates. It will also be appreciated that both endpoints of any given edge are disposed on the same island. Thus since the iteration of step 2914 is ultimately for identifying islands rather than edges or corners, step 3042 merely involves selecting one of the key edges in 2nd_edge_map, and then selecting as the selected corner one of the endpoints from the selected edge. It does not matter which key edge is selected nor which endpoint of that edge is selected. Again, as the selected corner was not found by way of a relationship with a previously-considered corner or edge, it is presumed (at least initially) to be disjoint from any existing tree. In step 3034, therefore, a new tree is started with its root as the island identified by the selected corner. In an embodiment which supports predefined color groups, the pre-groups value for the tree is set initially to no_color. The selected corner is then returned in step 3018 as the root corner of the first (or next) available island to consider in the iteration. No color is assigned to the island yet; it will be assigned when the island is connected to a tree.
If in step 3040 it is determined that 2nd_edge_map is empty as well, then the routine returns NULL to indicate that the iteration is complete (step 3050).
Returning to
If the far corner is already assigned to an island (step 3214), then in step 3218 it is determined whether that island (now sometimes called the far island) has been colored. If not, then it is not yet connected to a tree. In this case, in step 3220, it is connected to the current tree. The arc for the new connection is the current corner-corner candidate, and the color assigned to the far island is whatever color is opposite of the current island. In an embodiment, assigning a color to an island can involve merely populating a color parameter in a node data structure with an integer or enum representing one of the range of available colors. The routine then returns to step 3210 to consider the next corner-corner candidate in corner_map involving the current corner.
In an embodiment that supports TPT design rule violation detection, after it is determined in step 3214 that the current far corner is already assigned to an island, then in step 3217 an identification of the island pair is inserted into a new map called TPT_island_map, which is used as described below. Additionally, in an embodiment that supports multi-patterning pre-coloring, after the newly discovered island is connected to the current tree and colored (step 3220), the current tree is checked for a multi-patterning pre-coloring violation (step 3222). Still further, in an embodiment that supports multi-patterning pre-groups, after the newly discovered island is connected to the current tree and colored (step 3220), the current tree is checked for a multi-patterning pre-grouping violation (step 3224).
Returning to step 3218, if the far island is already colored, then it is already connected to a tree. In this case, adding an arc to it would create a cycle. Thus in step 3226 it is determined whether the cycle is even or odd. If the two islands are colored oppositely, then the cycle is even and nothing need be done, connected or stored. The routine merely returns to step 3210 to consider the next corner-corner candidate in corner_map involving the current corner. If the two islands have the same color, however, then the cycle is odd. In step 3228 the odd cycle is inserted into an odd cycle collection structure for later presentation to the user. Again, there is no need to connect the new arc in the tree, so the routine merely returns to step 3210 to consider the next corner-corner candidate in corner_map involving the current corner.
A cycle is represented in the DPT odd cycle collection structure as a list of markers each identifying a DPT spacing violation that forms an element of the cycle. Each of these markers is either a corner-corner marker, in which case it identifies the specific corners of the two islands that are place too near each other; or an edge-edge marker, in which case it identifies the specific edges of the two islands that are placed too near each other. It will be appreciated that this information is sufficient to draw the marker visibly for the user.
After all the corner-corner candidates in corner_map involving the current corner have been considered (step 3210), all the corner-corner candidates keyed by the current corner are erased from corner_map (step 3230). The routine then returns (step 3232) to the contour walking process (
Referring to
If in step 3614 it was determined that the current tree has already been marked as “inconsistent”, then in step 3620 it is determined whether the pre-color of the island is the same as the color assigned to the island in step 3220. If not, then there is no pre-color violation and in step 3612 the process returns to
Referring to
As mentioned, in
If the far corner is already assigned to an island (step 3314), then in step 3318 it is determined whether that island (now sometimes called the far island) has been colored. If not, then it is not yet connected to a tree. In this case, in step 3320, it is connected to the current tree. The arc for the new connection is the current edge-edge candidate, and the color assigned to the far island is whatever color is opposite of the current island. The routine then returns to step 3310 to consider the next edge-edge candidate in 1st_edge_map involving the current edge.
In an embodiment that supports TPT design rule violation detection, after it is determined in step 3314 that the current far corner is already assigned to an island, then in step 3317 an identification of the island pair is inserted into TPT_island_map. Additionally, in an embodiment that supports multi-patterning pre-coloring, after the newly discovered island is connected to the current tree and colored (step 3320), the current tree is checked for a multi-patterning pre-coloring violation (step 3322). The method of
Returning to step 3318, if the far island is already colored, then it is already connected to a tree. In this case, adding an arc to it would create a cycle. Thus in step 3326 it is determined whether the cycle is even or odd. If the two islands are colored oppositely, then the cycle is even and nothing need be done, connected or stored. The routine merely returns to step 3310 to consider the next edge-edge candidate in 1st_edge_map involving the current edge. If the two islands have the same color, however, then the cycle is odd. In step 3328 the odd cycle is inserted into the odd cycle collection structure for later presentation to the user. Again, there is no need to connect the new arc in the tree, so the routine merely returns to step 3310 to consider the next edge-edge candidate in 1st_edge_map involving the current edge.
After all the edge-edge candidates in 1st_edge_map involving the current edge have been considered (step 3310), all the edge-edge candidates keyed by the current edge are erased from 1st_edge_map (step 3330). The routine then returns (step 3332) to the contour walking process (
Returning to
Referring to
If the far edge is already assigned to an island (step 3414), then in step 3418 it is determined whether the far island has already been colored. If not, then it is not yet connected to a tree. In this case, in step 3420, it is connected to the current tree. The arc for the new connection is the current edge-edge candidate, and the color assigned to the far island is whatever color is opposite of the current island. The routine then returns to step 3410 to consider the next edge-edge candidate in 2nd_edge_map involving the current edge.
In an embodiment that supports TPT design rule violation detection, after it is determined in step 3414 that the current far corner is already assigned to an island, then in step 3417 an identification of the island pair is inserted into TPT_island_map. Additionally, in an embodiment that supports multi-patterning pre-coloring, after the newly discovered island is connected to the current tree and colored (step 3420), the current tree is checked for a multi-patterning pre-coloring violation (step 3422). The method of
Returning to step 3418, if the far island is already colored, then it is already connected to a tree. In this case, adding an arc to it would create a cycle. Thus in step 3426 it is now determined whether the cycle that would be created by adding a new arc is even or odd. If the two islands are colored oppositely, then the cycle is even and nothing need be done, connected or stored. The routine merely returns to step 3410 to consider the next edge-edge candidate in 2nd_edge_map involving the current edge. If the two islands have the same color, however, then the cycle is odd. In step 3428 the odd cycle is inserted into the odd cycle collection structure for later presentation to the user. Again, as there is no need to connect the new arc in the tree, the routine then returns to step 3410 to consider the next edge-edge candidate in 2nd_edge_map involving the current edge.
Note that whereas arcs that would form cycles are never inserted into the tree being built in
After all the edge-edge candidates in 2nd_edge_map involving the current edge have been considered (step 3410), all the edge-edge candidates keyed by the current edge are erased from 2nd_edge_map (step 3430). The routine then returns (step 3432) to the contour walking process (
Returning to
In
Referring to
Returning to step 3818, if the loop through all the next islands C from the current odd cycle completes, and the loop did not break because one of the islands C did not form a multi-patterning candidate spacing violation with the current far island B, then far island B constitutes a “common” island which has a multi-patterning candidate spacing violation against all of the islands in the current odd cycle. In step 3824, therefore, a TPT coloring violation has been found and the current odd cycle is saved, along with the current set X, for display to the user or other reporting. Loop 3810 then continues with the next odd cycle in the odd cycle collection.
When all of the odd cycles in the odd cycle collection have been considered, the routine terminates (step 3826).
Returning to
In
Returning to
Also after step 322, in step 324 the system reports any design rule violations to the user or to another entity. In one embodiment, if the current editing command did not involve any movement of shapes, then nothing is output in this step. It will be seen however that another iteration through steps 314-322 occurs when the user begins dragging behavior, in which case any design rule violations can be reported. If reported to the user, the report can take place promptly (e.g. for real time feedback) or later (e.g. if performed as a batch job). Where the violations are reported to the user promptly, this enables the user to modify the layout to correct for the design rule violations. Whereas any form of reporting can be used, preferably the design rule violations are reported by way of visual indications on the user's monitor, as markers on the layout region itself. In an embodiment, near violations are also indicated. Marker information can be anything that can be used to render a visual indicator of the violation, but preferably it identifies a rectangle for designating the location of the violation within the layout region. In an embodiment, the rectangle is shown in a size which indicates the magnitude of the primary value of the rule being violated. This information can be very useful as it indicates graphically how much is needed to correct the violation. For near-violations, it can be a ruler indicating the current spacing. For example, if the violation is a minimum spacing violation, a rectangle might encompass the (too-small) spacing area, or a ruler disposed across the space might indicate actual spacing if it is larger than the minimum. Where violations are reported later, this can take the form of a log file or other record of the violations found. In addition, if the system proceeds directly to an automated step of correcting violations, reporting can take the form of the modified layout resulting from the corrections.
All of the design rule checks output marker information for any violation. The marker information is collected in a map structure. In step 324, the marker information is converted to visible form on the user's monitor or provided to another entity.
As shown in
Returning to step 313, if the current editing command involves dragging selected objects, then the design rule bounds that were pre-calculated during a previous traversal through steps 314-322 are used to assist the placement of the objects at the design rule boundary nearest the user's desired destination point. Note that the current drag operation may be just the next increment of a longer drag, in which case the most recent traversal through steps 314-322 (and therefore the most recent pre-calculation of design rule bounds) will have been in response to a previous drag command rather than the selection command just described. Either way, in step 328 the system determines whether the destination position of the drag event exceeds the pre-calculated design rule bounds. If not, then the system merely moves the editing objects to the target position. If so, then the system adjusts the target position to the position nearest the user's target position, but which does not violate any of the design rules in the set of design rules enabled for DRC-assist. Push-through is also performed here.
Note further that the present embodiment is intended to assist in continuous geometry changes associated with mouse drag. In the case of sudden geometry changes associated with the drag, which is possible for some of the user commands, step 328 will not do anything to block them. The operation is treated the same as push-through.
The following pseudocode describes the pertinent aspects of step 328. The starting position of the objects selected for editing is p1(x,y), and the user's destination or target position is p2(x,y):
The above pseudocode blocks the user's drag operation from exceeding any of the linear slacks, and operates by adjusting the destination position p2 to the nearest position that prevents any such violation. Next, the system checks the corner and conditional slacks, based on the adjusted destination position p2, and further adjusts p2 to avoid violating any of those. This latter process involves adjusting the destination point in two separate monotonic sequences. The first sequence is X_first, and the second sequence is Y_first. Each sequence is monotonic in the sense that no step in the sequence backtracks either horizontally or vertically. For example, for an adjustment toward the upper right, the X_first sequence is to the right, then upward, then to the right, and so on until either the destination point is reached, or both the right and upward directions are blocked. The corresponding Y_first sequence is the same, except that it begins with an adjustment upward. At each step inside the sequence, the destination point is adjusted by a small amount, and the four linear slacks (Ls, Rs, Bs, Ts) are recalculated based on adjusted positions and all the corner and conditional slacks.
Whether or not step 328 results in any adjustment of the destination position p2 as signaled by the user's drag command, the system next traverses steps 314-326 again to check the new position against the design rules and output markers if appropriate, and to pre-calculate, in advance of the next editing command, a new set of design rule bounds based on the new position p2 of the selected objects. The system then returns to step 312 to await the next editing command. This may be as simple as another slight movement of the current editing shapes being dragged across the layout region, which would result in another traversal through steps 314-326 of
The “relevant DPT edge-edge spacing rule” applied above is whichever applies of (a) the DPT minimum end-edge spacing, (b) the DPT minimum side-edge spacing, or (c) the DPT minimum side-side spacing. As previously mentioned an edge is either an “end-edge” or a “side-edge”, depending on a comparison with the width_threshold value in the DPT spacing rules, and different spacing values apply depending on whether the relationship between the two edges of a candidate is end-edge side-edge spacing, or side-side. In
Importantly, the candidate markers identify to the user not only which islands are too near each other, but also (a) which kind of violation has occurred (edge-edge or corner-corner), (b) exactly which edges and corners are involved in the violation, and (c) roughly how far the islands need to move away from each other in order to remove each of the DPT spacing violations. This helps the user to know, visually and intuitively, which islands need to be moved and in which direction, and roughly by how far, in order to remove the violation. Also importantly, the cycles are indicated among the candidates, not the islands. In particular, the broken lines 3530 and 3532 interconnect the centerpoints of the markers identifying the individual spacing violations which form the cycle, not the islands which form the cycle. This arrangement again indicates much more intuitively to the user which islands need to be moved and in which direction. That information can be more difficult to gather where the DPT odd cycle violation polygon has its vertices placed somewhere in the middle of the islands themselves.
It can be seen that the islands in the example of
Similar visual indicators to indicate violations of other design rules will be apparent to the reader. In addition, note that all of the design rule checks indicated in
In the embodiments described herein, all the corner data structures are completely populated before the corner-based rules are checked. This is the most advantageous arrangement, but some benefits of the invention can be obtained even if only some (i.e. more than one; preferably more than two) of the corner data structures are completely populated before the corner-based rules are checked. For DPT odd cycle checking, not all the data gathered for each corner is required. Therefore, while again it is most advantageous to populate at least two of the corner data structures completely before the DPT odd cycle rules are checked, some benefits of the invention can be obtained even if only the following information about such corners are populated in advance: (x,y), s_ray_x, and s_ray_y. In addition, ori_x and ori_y should be populated in advance if a contour walking technique such as in step 2916 (
Similarly, all island data structures are completely populated before the island-based rules are checked. Again, while this is the most advantageous arrangement, some benefits of the invention can be obtained even if only some (i.e. more than one; preferably more than two) of the island data structures are completely populated before the island-based rules are checked.
In the embodiments described herein, multi-patterning minimum spacing constraint violations and coloring violations formed by cycles of such violations, have been given as examples. It will be appreciated, however, that aspects of the invention also apply to other types of multi-patterning rules as well. In some embodiments, multi-patterning rules can become as complex as, or more complex than, the single-layer design rules described elsewhere herein.
The physical hardware component of network interfaces are sometimes referred to as network interface cards (NICs), although they need not be in the form of cards: for instance they could be in the form of integrated circuits
(ICs) and connectors fitted directly onto a motherboard, or in the form of macrocells fabricated on a single integrated circuit chip with other components of the computer system.
User interface input devices 2222 may include a keyboard, pointing devices such as a mouse, trackball, touchpad, or graphics tablet, a scanner, a touch screen incorporated into the display, audio input devices such as voice recognition systems, microphones, and other types of input devices. In general, use of the term “input device” is intended to include all possible types of devices and ways to input information into computer system 2210 or onto computer network 2218.
User interface output devices 2220 may include a display subsystem, a printer, a fax machine, or non-visual displays such as audio output devices. The display subsystem may include a cathode ray tube (CRT), a flat-panel device such as a liquid crystal display (LCD), a projection device, or some other mechanism for creating a visible image. The display subsystem produces the images illustrated in
Storage subsystem 2224 stores the basic programming and data constructs that provide the functionality of certain embodiments of the present invention. For example, the various modules implementing the functionality of certain embodiments of the invention may be stored in storage subsystem 2224. These software modules are generally executed by processor subsystem 2214.
Memory subsystem 2226 typically includes a number of memories including a main random access memory (RAM) 2230 for storage of instructions and data during program execution and a read only memory (ROM) 2232 in which fixed instructions are stored. File storage subsystem 2228 provides persistent storage for program and data files, and may include a hard disk drive, a floppy disk drive along with associated removable media, a CD-ROM drive, an optical drive, or removable media cartridges. The databases and modules implementing the functionality of certain embodiments of the invention may be stored by file storage subsystem 2228. The host memory 2226 contains, among other things, computer instructions which, when executed by the processor subsystem 2214, cause the computer system to operate or perform functions as described herein. As used herein, processes and software that are said to run in or on “the host” or “the computer system”, execute on the processor subsystem 2214 in response to computer instructions and data in the host memory subsystem 2226 including any other local or remote storage for such instructions and data.
Bus subsystem 2212 provides a mechanism for letting the various components and subsystems of computer system 2210 communicate with each other as intended. Although bus subsystem 2212 is shown schematically as a single bus, alternative embodiments of the bus subsystem may use multiple busses.
Computer system 2210 itself can be of varying types including a personal computer, a portable computer, a workstation, a computer terminal, a network computer, a television, a mainframe, or any other data processing system or user device. Due to the ever-changing nature of computers and networks, the description of computer system 2210 depicted in
In an embodiment, the steps set forth in the flow charts and descriptions herein are performed by a computer system having a processor such as processor subsystem 2214 and a memory such as storage subsystem 2224, under the control of software which includes instructions which are executable by the processor subsystem 2214 to perform the steps shown. The software also includes data on which the processor operates. The software is stored on a computer readable medium, which as mentioned above and as used herein, is one on which information can be stored and read by a computer system. Examples include a floppy disk, a hard disk drive, a RAM, a CD, a DVD, flash memory, a USB drive, and so on. The computer readable medium may store information in coded formats that are decoded for actual use in a particular data processing system. A single computer readable medium, as the term is used herein, may also include more than one physical item, such as a plurality of CD-ROMs or a plurality of segments of RAM, or a combination of several different kinds of media. When the computer readable medium storing the software is combined with the computer system of
Additionally, the geometry file or files storing the layout, the relationship master dataset, and the layout topology database are themselves stored on computer readable media. Such media can be distributable separately from the computer system, and form their own respective articles of manufacture. When combined with a computer system programmed with software for reading, revising, and writing the geometry files, and for design rule checking, they form yet another machine which performs the steps set forth herein.
As used herein, the “identification” of an item of information does not necessarily require the direct specification of that item of information. Information can be “identified” in a field by simply referring to the actual information through one or more layers of indirection, or by identifying one or more items of different information which are together sufficient to determine the actual item of information. In addition, the term “indicate” is used herein to mean the same as “identify”.
As used herein, a given signal, event or value is “responsive” to a predecessor signal, event or value if the predecessor signal, event or value influenced the given signal, event or value. If there is an intervening processing element, step or time period, the given signal, event or value can still be “responsive” to the predecessor signal, event or value. If the intervening processing element or step combines more than one signal, event or value, the signal output of the processing element or step is considered “responsive” to each of the signal, event or value inputs. If the given signal, event or value is the same as the predecessor signal, event or value, this is merely a degenerate case in which the given signal, event or value is still considered to be “responsive” to the predecessor signal, event or value. “Dependency” of a given signal, event or value upon another signal, event or value is defined similarly.
The foregoing description of preferred embodiments of the present invention has been provided for the purposes of illustration and description. It is not intended to be exhaustive or to limit the invention to the precise forms disclosed. Obviously, many modifications and variations will be apparent to practitioners skilled in this art. In particular, and without limitation, any and all variations described, suggested or incorporated by reference in the Background section of this patent application are specifically incorporated by reference into the description herein of embodiments of the invention. The embodiments described herein were chosen and described in order to best explain the principles of the invention and its practical application, thereby enabling others skilled in the art to understand the invention for various embodiments and with various modifications as are suited to the particular use contemplated. It is intended that the scope of the invention be defined by the following claims and their equivalents.
This application is a continuation of U.S. application Ser. No. 13/675,970, filed Nov. 12, 2012, by Scott I. Chase, Zuo Dai, Dick Liu and Ming Su, entitled “LOW-OVERHEAD MULTI-PATTERNING DESIGN RULE CHECK,” Attorney Docket No. SYNP 2079-1, which application is a continuation-in-part of U.S. application Ser. No. 12/960,086, filed Dec. 3, 2010, by Zuo Dai, Dick Liu and Ming Su, entitled “HIGH PERFORMANCE DESIGN RULE CHECKING TECHNIQUE,” Attorney Docket No. SYNP 1350-1, which applications are incorporated herein by reference in their entirety.
| Number | Date | Country | |
|---|---|---|---|
| Parent | 13675970 | Nov 2012 | US |
| Child | 14167889 | US |
| Number | Date | Country | |
|---|---|---|---|
| Parent | 12960086 | Dec 2010 | US |
| Child | 13675970 | US |
