WO2006041833A2 - Transaction-based system and method for abstraction of hardware designs - Google Patents

Transaction-based system and method for abstraction of hardware designs Download PDF

Info

Publication number
WO2006041833A2
WO2006041833A2 PCT/US2005/035663 US2005035663W WO2006041833A2 WO 2006041833 A2 WO2006041833 A2 WO 2006041833A2 US 2005035663 W US2005035663 W US 2005035663W WO 2006041833 A2 WO2006041833 A2 WO 2006041833A2
Authority
WO
WIPO (PCT)
Prior art keywords
fsm
transaction
fsms
abstracted
state
Prior art date
Application number
PCT/US2005/035663
Other languages
French (fr)
Other versions
WO2006041833A3 (en
Inventor
James C. Wilson
Original Assignee
Nusym Technologies, Inc.
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Nusym Technologies, Inc. filed Critical Nusym Technologies, Inc.
Publication of WO2006041833A2 publication Critical patent/WO2006041833A2/en
Publication of WO2006041833A3 publication Critical patent/WO2006041833A3/en

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/30—Circuit design
    • G06F30/32—Circuit design at the digital level
    • G06F30/33—Design verification, e.g. functional simulation or model checking
    • G06F30/3323—Design verification, e.g. functional simulation or model checking using formal methods, e.g. equivalence checking or property checking
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • G06F30/23—Design optimisation, verification or simulation using finite element methods [FEM] or finite difference methods [FDM]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/30—Circuit design
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/30—Circuit design
    • G06F30/32—Circuit design at the digital level
    • G06F30/33—Design verification, e.g. functional simulation or model checking
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/30—Circuit design
    • G06F30/32—Circuit design at the digital level
    • G06F30/33—Design verification, e.g. functional simulation or model checking
    • G06F30/3308—Design verification, e.g. functional simulation or model checking using simulation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/30—Circuit design
    • G06F30/36—Circuit design at the analogue level
    • G06F30/367—Design verification, e.g. using simulation, simulation program with integrated circuit emphasis [SPICE], direct methods or relaxation methods

Definitions

  • This invention relates generally to systems and methods for simulation and verification of electronic hardware designs. More particularly, this invention relates to systems and methods for performing abstraction to permit more efficient simulation, and more specifically relates to systems, methods and techniques for temporal abstraction in which hardware operation is redefined in terms of transactions.
  • Abstraction is, in simplified terms, a technique for masking irrelevant detail so that the key elements of a concept or other construct may be more clearly appreciated.
  • abstraction is useful because it reduces the amount of work required to perform the simulation, thereby allowing more simulation to be done in a given amount of time.
  • Abstractions can be classified into four basic types: structural, behavioral, data, and temporal. Structural abstraction removes unneeded structure, for example, gates or registers in the case of hardware design. Behavioral abstraction removes functionality. Data abstraction reduces the range of data values possible. Temporal abstraction removes unneeded time steps during the simulation.
  • Transaction-based abstraction is a form of temporal abstraction.
  • hardware is assumed to process transactions, which is defined to mean that the hardware accepts a request, processes the request, and returns a response, and that the hardware can process such requests infinitely often.
  • a simulation of the unabstracted design causes each transaction to start at a first time step and end one or more time steps later.
  • the design is abstracted such that, a simulation of the device causes every transaction to occur in exactly one abstracted time step.
  • FSMs finite-state machines
  • select signals representing FSMs that have wait states indicating the FSM waits for either a request or response
  • Signals may also require abstraction as a result of the transaction-based abstraction of the present invention. For example, if a circuit processing transactions requires four cycles, it may have four pipeline stages, each stage being defined by a different signal. After abstraction, these four pipeline stages will be compressed to one stage. Therefore, the signal definitions for each stage need to be modified to account for this difference.
  • signals can be classified based on the abstracted design. Signals can be classified as either transient, temporary, or persistent. Transient signals are those that implement the transaction-processing FSMs that define the transaction boundaries. Persistent state-holding elements (e.g., registers, RAMs, etc.) are those signals whose values are set in one transaction and accessed in another. Temporary signals are all other signals.
  • transaction-based abstraction enables signals to be classified easily according to the above criteria.
  • Transient signals are identified during the abstraction process as the transaction-processing FSMs that define the transaction boundaries.
  • all combinational signals are temporary since they are set and used in the same time step and all transactions are at least one time step in length.
  • State-holding signals may also be classified as temporary if, after abstraction, they become combinational in the abstracted design. All signals that remain as state-holding signals after abstraction are persistent.
  • An additional feature of the present invention is that it allows a comprehensive test plan to be automatically generated for a device. This is beneficial because it relieves the user from performing this task manually.
  • a test plan comprises a set of conditions called coverage points that are derived based on a signal classification created using transaction- based analysis to identify the architecturally-visible state of the design.
  • Figure 1 illustrates an overall transaction-based abstraction process according to certain embodiments of the invention
  • FIG. 2 illustrates, in state diagram form, an exemplary finite state machine ("FSM") which represents a device that processes transactions according to certain embodiments of the invention
  • FIG. 3 illustrates an FSM abstraction process according to certain embodiments of the invention
  • Figure 4 illustrates, in state diagram form, two exemplary interrelated FSM's which interact, representing two devices which similarly interact in processing transactions, according to certain embodiments of the invention
  • Figure 5 illustrates, in table form, state labels for the two exemplary FSM's shown in Figure 4, together with state labels for the composite FSM, according to certain embodiments of the invention
  • Figure 6 illustrates, in state diagram form, an abstracted FSM from the two FSM's of Figure 4 according to certain embodiments of the invention
  • Figure 7 illustrates a signal abstraction process according to certain embodiments of the invention
  • Figure 8 illustrates a process for computing an abstracted word terminal for a given signal, delay, and composite FSM edge (u,v) according to certain embodiments of the invention
  • Figure 9 illustrates an example of signal abstraction for the FSM of Figure 2, showing the three phases of the abstraction process with signal abstractions labeled with the appropriate state from the FSM shown in Figure 2, according to certain embodiments of the invention.
  • Figure 10 illustrates a flow for creating a test plan automatically from a design description according to certain embodiments of the invention.
  • transaction-based abstraction is a form of temporal abstraction.
  • hardware is assumed to process transactions, which is defined to mean that the hardware accepts a request, processes the request, and returns a response, and that the hardware can process such requests infinitely often.
  • a simulation of the unabstracted design causes each transaction to start at a first time step and end one or more time steps later.
  • the design is abstracted such that, a simulation of the device causes every transaction to occur in exactly one abstracted time step.
  • Figure 1 illustrates an overall transaction-based abstraction process according to certain embodiments of the invention, which can include three main phases.
  • FSMs finite state machines
  • a heuristic technique can be used to identify finite-state machines (FSMs) in the design that process transactions, from which the transaction boundaries can be identified. Heuristics can filter signals that can be used to determine transaction boundaries. Possible heuristics include:
  • select signals representing FSMs that have wait states indicating the FSM waits for either a request or response
  • the steps for identifying a set of signals representing transaction-processing FSMs include:
  • Transaction- processing FSMs can have the following characteristics:
  • a wait state is a state that transitions to itself based on some condition.
  • FIG. 2 illustrates, in state diagram form, an exemplary finite state machine ("FSM") which represents a device that processes transactions according to certain embodiments of the invention
  • the exemplary FSM consists of four states, s ⁇ , si, s2, and s3.
  • a transition consists of a current state, u, next state, v, and guard, g. If the FSM is in state u at time t and condition g is true, it will transition to state v in the next time step. Transitions are indicated as edges in the state transition graph with edge labels indicating the guard condition g.
  • Unlabelled edges represent unconditional transitions. Assignments indicated by a box attached to an edge in a graph indicate signals that are set as a function of the FSM.
  • State sO is a wait state because it has an edge to itself labeled "-start", indicating that the FSM should wait in that state until a start signal occurs. Once started, the FSM proceeds through the other states and then loops back to the wait state, at which time it indicates completion of the transaction by asserting the "done" signal. The transaction boundaries occur when the FSM transitions out of the wait state at the beginning and back to the wait state at the end.
  • the extracted FSMs are combined to form a composite FSM.
  • Composition of FSMs is done by enumerating all combinations of states and edges in the individual extracted FSMs.
  • Composite edges are formed by computing the product of all edge transition conditions for individual FSMs. Since FSM inputs may be functions of one or more other FSMs being composed, this may cause some edges or states to be unreachable if the computed product condition is false.
  • an abstract FSM is computed in which states in the abstract FSM correspond to sets of states in the composite FSMs.
  • Figure 3 illustrates this FSM abstraction process according to certain embodiments of the invention. As shown in the first step of Figure 3, all wait states in the individual FSMs are given a unique label and all other states are given a common label.
  • the set of abstract states corresponds to the unique labels produced using the above procedure for the composite FSM.
  • An edge exists between two abstract states if there exists an edge between two corresponding states in the composite FSM.
  • the transition condition for this edge is the OR of all composite transition conditions for this abstract edge.
  • Figure 4 illustrates, in state diagram form, two exemplary interrelated FSM' s which interact, representing two devices which similarly interact in processing transactions, according to certain embodiments of the invention.
  • FSM A drives transactions to FSM B, which waits for a transaction start, processes the transaction, and then signals done.
  • FSM A waits for the done indication before starting the next transaction.
  • state transition aO to al causes signal "start" to be set to 1.
  • FSM B is in state b ⁇ , it transitions to bO if signal "start” is 0, indicated by the edge guard " ⁇ start", otherwise, when “start” is 1, it transitions to state bl.
  • Figure 5 illustrates, in table form, state labels for the two exemplary FSM' s shown in Figure 4, together with state labels for the composite FSM, according to certain embodiments of the invention.
  • composite states ⁇ a2,b2 ⁇ and ⁇ a2,b3 ⁇ have the same composite label.
  • these two composite states will be a single abstract state as shown in the abstract FSM in Figure 6 computed from the two FSMs in Figure 5.
  • Figure 6 illustrates, in state diagram form, an abstracted FSM from the two FSM' s of Figure 4 according to certain embodiments of the invention.
  • a signal definition function specifies the value of a signal over time as a function of other signals and time.
  • the goal of signal abstraction is to modify the signal definition such that it is a function of abstract time, f .
  • composition refers to the substitution of a function defining a signal for a signal reference in an expression.
  • Composing the definition of signal "c” into the expression for signal "a” is done in two steps. First, the expression for signal “a” refers to " 1 C(M)” indicating that this is the value of signal "c” at time M. However, the definition for "c” is at time t.
  • Figure 7 illustrates a signal abstraction process according to certain embodiments of the invention.
  • the first step of Figure 7 is done by performing a breadth-first search of the composite state graph.
  • Each step of this traversal specifies a tuple (u,g,v) where u is the current state, v is the next state, and g is a condition that must be true for the state transition to occur.
  • Values u and v correspond to assignments to the signals that make up the individual signals of the composite state.
  • the expression g contains references to composite state signals and other signals.
  • Signal values are represented by expressions comprising operators such as add, and, etc., ite operators, constant terminals, and word te ⁇ ninals.
  • the ite operator specifies three subexpressions, an if condition, then assignment, and else assignment and represents an if-then-else (ite) assignment.
  • Word terminals specify the name of a signal and a delay. Since signal values are functions of time, a word terminal with delay k specifies the value of the signal at time t-k. In at least some embodiments, the present invention only considers cases in which the word terminal delay is 0 or 1. Operator nodes have the usual semantics.
  • the abstract boundary state is defined to be state a_v of a transaction boundary as defined above. For example, in Figure 2, the edge from s3 to sO would be considered a transaction boundary even if s3 and sO mapped to the same abstract state, and, therefore, state sO is defined to be the abstract boundary state.
  • FIG. 8 illustrates a process for computing an abstracted word terminal for a given signal, delay, and composite FSM edge (u,v) according to certain embodiments of the invention.
  • the composite FSM may have states with multiple incoming edges. Since the result of the above process produces an expression for each signal for each state, multiple expressions will be produced for states that have multiple incoming edges and will need to be combined. Typically, multiple incoming edges occur due to multiple outgoing edges somewhere else in the state graph. Multiple outgoing edges are differentiated by different edge guard conditions. A guard condition is just another expression that is a function of time. These guards also need to be abstracted as a function of the composite state using the exact same parsing and modification as described above. Further, guard expressions are propagated along states and edges.
  • the signal will be processed and stored in an array indexed by the signal and abstract state, u', corresponding to state u. Since multiple composite states may map to a single abstract state, this pass is typically performed in composite state order. That is, state u is typically processed before state v if (u,g,v) is an edge in the composite FSM and (u,g,v) is not a transaction boundary.
  • This step modifies the expression produced in the previous step before storing it in the array.
  • the modifications also only affect word terminals marked for composition in the previous pass.
  • a marked word terminal specifies a signal p and state w. There are two possible modifications:
  • the word terminal is composed by replacing the word terminal with the abstracted value of signal p associated with state w that was computed in the previous step.
  • the resulting expression is recursively traversed so that multiple compositions may occur.
  • state w is an abstract boundary state
  • the word terminal is replaced with a word terminal specifying word p with delay t'-l (delay of 1 relative to the abstract clock).
  • the resulting array of signal values for each abstract state has expressions that are functions of signals at abstract boundaries only.
  • the last step of Figure 7 combines these into a single assignment for each signal. Given a signal s with values stored in the array for abstract states a_u and a_v, these can be combined by performing an if-then-else in which the if condition is the abstract state being equal to state a_u for example.
  • Figure 9 illustrates an example of signal abstraction for the FSM of Figure 2, showing the three phases of the abstraction process with signal abstractions labeled with the appropriate state from the FSM shown in Figure 2, according to certain embodiments of the invention.
  • the composite FSM is the FSM given in Figure 2 and a signal, s, is defined as specified at the top of Figure 9.
  • signal s is abstracted for each edge in the composite FSM.
  • the resulting expressions are stored for each state s ⁇ -s3 of the FSM.
  • the value of the signal is abstracted by substituting the state value in the expression in s and marking signal s by specifying a composite state.
  • the signal for each state is modified by substituting the appropriate expression for each marked word. These substitutions are done in state order, sO - si — s2 - s3 and after substitution, the result is stored in the corresponding abstract state array entry. These expressions are shown in the text box marked "pass 2". Each expression is the result of composition and all states are labeled with abstract state aO with the corresponding composite state in parentheses. Since there is only a single abstract state, the final value is just the value of the array for s in state aO. Since the expression for s3 was created last, this is the final value of the abstracted value of signal s.
  • the unabstracted version of signal s is a function of itself, which means that it is a state-holding element. Furthermore, it is a function of another state-holding element, the FSM. With no other context for this signal, it has a fairly complex definition. After abstraction, signal s is no longer a function of itself or the FSM and is not a state- holding element. We know that the abstraction represents the function of this signal for one transaction. After abstraction, it is much more obvious what the meaning of this signal is. Also, because the definition is simpler, it can be more efficiently simulated. The ability to make signal functions easier to understand and to simplify them illustrate one of the many advantages of transaction-based abstraction. 4) Signal Classification
  • signals are classified according to their use in processing transactions. There are three types of signals:
  • Transient signals consist of the transaction-processing finite-state machines (FSMs) that control the transaction by detecting when a transaction starts, controlling the processing of the transaction, and then signaling its completion.
  • FSMs transaction-processing finite-state machines
  • Temporary signals are those that are used only within a single transaction. A signal that is used only within a transaction must be set before being read within the transaction. Because it is set before being used within a transaction, the value of a temporary signal in one transaction cannot affect the operation of other transactions.
  • Persistent signals are those that are set in one transaction and read in another.
  • signals are classified by performing transaction-based analysis of the design.
  • Each of the signal types is determined by the abstraction process as follows:
  • Transient signals are the signals corresponding to the transaction-processing FSMs identified by the abstraction heuristics.
  • a test plan is a set of coverage points that must be exercised in a design.
  • a coverage point can be either:
  • a range of values for a signal for example, a 4-bit wide signal that can have values in the range 0-15.
  • Coverage points are typically defined in terms of a simulation model, which comprises a device plus a test case and other models, that may be used to compute the results of the simulation.
  • a coverage point is considered covered during simulation if 1) it becomes true at some point during simulation if it is a Boolean expression, or 2) the signal takes on a value from the specified range at some point during simulation.
  • a test plan can be automatically generated by identifying the architecturally visible state in the design and then creating coverage for signals in the design.
  • the architecturally-visible state can be identified by classifying all signals in the design as either transient, persistent or temporary, as discussed above using transaction-based abstraction analysis.
  • Figure 10 illustrates a flow for creating a test plan automatically from a design description according to certain embodiments of the invention.
  • the input to the test plan generation process consists of a design description, which, in some embodiments consists of a description written in a hardware description language (HDL) such as Verilog.
  • the first step is to determine the architecturally visible state and to classify the signals.
  • the second step is then to coverage points for the classified signals based on the information computed in the first step using techniques described previously herein and output this as a test plan.
  • HDL hardware description language
  • Each transient signal represents a transaction-processing finite state machine (FSM).
  • FSM finite state machine
  • a finite state machine consists of a set of states and transitions between states. Transitions may be conditional, specified by a Boolean condition.
  • Transaction-processing FSMs implement the protocol that defines a transaction. It is necessary to verify all possible timings of a protocol. Since the FSM must handle all possible cases, to completely test the protocol requires exercising every transition in the FSM.
  • temporary signals will comprise the majority of signals. These primarily are data signals, not control signals. Therefore, collectively, they represent a large number of bits. Normally, with data, it is desirable to cover a large number of values.
  • Temporary signals are normally defined as a set of assignments, each of which is enabled under different conditions defined by Boolean expressions over all signals. Assume that a temporary signal has a set of assignments with an associated condition, represented as (c h gi) where e, is an assignment expression and g, is the Boolean guard expression. The set of coverage points for a temporary signal is the set of guards, g ⁇ plus the range of all possible values for the signal based on its defined width (number of bits).
  • Persistent signals may be data or control. For cases in which persistent signals are data values, they can generate the same coverage points as temporary signals. If not, it is desirable to cover as many different values as possible since these potentially represent different states. Also, if they do represent state, persistent signals often interact with other persistent signals. Therefore, the coverage points for persistent state to be measured are the ranges of values of persistent signal values.
  • the set of persistent signals can be partitioned such that persistent signals that are not transitively dependent on each other are put in different partitions. Coverage points for persistent state are then created as ranges of combinations of values for signals within a partition.
  • the set of coverage points generated by this analysis forms a comprehensive test plan for a simulated device. This is beneficial because it is automatically generated, relieving the user from the tedious chore of creating a test plan manually.

Abstract

Transaction-based hardware abstraction involves a process of using heuristic techniques to identify finite-state machines (FSMs) in a hardware design that process transactions, from which transaction boundaries can be identified. Signals can be abstracted as a result of the transaction-based abstraction. Signals can be classified based on the abstracted design as either transient, temporary, or persistent. Transaction-based abstraction allows a comprehensive test plan to be automatically generated for a device. A test plan comprises a set of conditions called coverage points that are derived based on a signal classification created using transaction-based analysis to identify the architecturally-visible state of the design.

Description

TRANSACTION-BASED SYSTEM AND METHOD FOR ABSTRACTION OF HARDWARE DESIGNS
Cross-Reference to Related Applications
[0001] This application claims the benefit of priority from U.S. Provisional Patent Application No. 60/615,892, filed October 4, 2004 and entitled "Transaction-Based System and Method for Abstraction of Hardware Designs", which is fully incorporated herein by reference.
Background of the Invention
Field of the Invention
[0002] This invention relates generally to systems and methods for simulation and verification of electronic hardware designs. More particularly, this invention relates to systems and methods for performing abstraction to permit more efficient simulation, and more specifically relates to systems, methods and techniques for temporal abstraction in which hardware operation is redefined in terms of transactions.
Description of the Art
[0003] Abstraction is, in simplified terms, a technique for masking irrelevant detail so that the key elements of a concept or other construct may be more clearly appreciated. In the context of verification of electronic hardware designs using simulation, abstraction is useful because it reduces the amount of work required to perform the simulation, thereby allowing more simulation to be done in a given amount of time. Abstractions can be classified into four basic types: structural, behavioral, data, and temporal. Structural abstraction removes unneeded structure, for example, gates or registers in the case of hardware design. Behavioral abstraction removes functionality. Data abstraction reduces the range of data values possible. Temporal abstraction removes unneeded time steps during the simulation.
Summary Of The Invention
[0004] Transaction-based abstraction is a form of temporal abstraction. In the present invention, hardware is assumed to process transactions, which is defined to mean that the hardware accepts a request, processes the request, and returns a response, and that the hardware can process such requests infinitely often. A simulation of the unabstracted design causes each transaction to start at a first time step and end one or more time steps later. In accordance with the present invention, the design is abstracted such that, a simulation of the device causes every transaction to occur in exactly one abstracted time step.
[0005] To achieve the desired effectiveness of transaction-based abstraction, the boundaries of a transaction, that is, the starting and ending time steps of each transaction, must be correctly determined. In accordance with the present invention, a heuristic technique is used to identify finite-state machines (FSMs) in the design that process transactions, from which the transaction boundaries can be identified. Heuristics filter signals that determine transaction boundaries. Possible heuristics include:
• select signals that look like FSMs that may process transactions;
• select signals representing FSMs that have wait states indicating the FSM waits for either a request or response;
• select signals representing FSMs that have state transition loops indicating that the FSM can process transactions infinitely often; and
• allow the user to manually specify signals and conditions that determine transaction boundaries.
[0006] Signals may also require abstraction as a result of the transaction-based abstraction of the present invention. For example, if a circuit processing transactions requires four cycles, it may have four pipeline stages, each stage being defined by a different signal. After abstraction, these four pipeline stages will be compressed to one stage. Therefore, the signal definitions for each stage need to be modified to account for this difference.
[0007] In accordance with the present invention, signals can be classified based on the abstracted design. Signals can be classified as either transient, temporary, or persistent. Transient signals are those that implement the transaction-processing FSMs that define the transaction boundaries. Persistent state-holding elements (e.g., registers, RAMs, etc.) are those signals whose values are set in one transaction and accessed in another. Temporary signals are all other signals.
[0008] In accordance with the present invention, transaction-based abstraction enables signals to be classified easily according to the above criteria. Transient signals are identified during the abstraction process as the transaction-processing FSMs that define the transaction boundaries. In the unabstracted design, all combinational signals are temporary since they are set and used in the same time step and all transactions are at least one time step in length. State-holding signals may also be classified as temporary if, after abstraction, they become combinational in the abstracted design. All signals that remain as state-holding signals after abstraction are persistent.
[0009] An additional feature of the present invention is that it allows a comprehensive test plan to be automatically generated for a device. This is beneficial because it relieves the user from performing this task manually. A test plan comprises a set of conditions called coverage points that are derived based on a signal classification created using transaction- based analysis to identify the architecturally-visible state of the design.
Brief Description of the Drawings
[0010] These and other aspects and features of the present invention will become apparent to those ordinarily skilled in the art upon review of the following description of specific embodiments of the invention in conjunction with the accompanying figures, wherein:
[0011] Figure 1 illustrates an overall transaction-based abstraction process according to certain embodiments of the invention;
[0012] Figure 2 illustrates, in state diagram form, an exemplary finite state machine ("FSM") which represents a device that processes transactions according to certain embodiments of the invention;
[0013] Figure 3 illustrates an FSM abstraction process according to certain embodiments of the invention;
[0014] Figure 4 illustrates, in state diagram form, two exemplary interrelated FSM's which interact, representing two devices which similarly interact in processing transactions, according to certain embodiments of the invention;
[0015] Figure 5 illustrates, in table form, state labels for the two exemplary FSM's shown in Figure 4, together with state labels for the composite FSM, according to certain embodiments of the invention;
[0016] Figure 6 illustrates, in state diagram form, an abstracted FSM from the two FSM's of Figure 4 according to certain embodiments of the invention; [0017] Figure 7 illustrates a signal abstraction process according to certain embodiments of the invention;
[0018] Figure 8 illustrates a process for computing an abstracted word terminal for a given signal, delay, and composite FSM edge (u,v) according to certain embodiments of the invention;
[0019] Figure 9 illustrates an example of signal abstraction for the FSM of Figure 2, showing the three phases of the abstraction process with signal abstractions labeled with the appropriate state from the FSM shown in Figure 2, according to certain embodiments of the invention; and
[0020] Figure 10 illustrates a flow for creating a test plan automatically from a design description according to certain embodiments of the invention.
Detailed Description Of The Invention
[0021] The present invention will now be described in detail with reference to the drawings, which are provided as illustrative examples of the invention so as to enable those skilled in the art to practice the invention. The figures and examples below are not meant to limit the scope of the invention.
[0022] In a general sense, transaction-based abstraction is a form of temporal abstraction. In the present invention, hardware is assumed to process transactions, which is defined to mean that the hardware accepts a request, processes the request, and returns a response, and that the hardware can process such requests infinitely often. A simulation of the unabstracted design causes each transaction to start at a first time step and end one or more time steps later. In accordance with the present invention, the design is abstracted such that, a simulation of the device causes every transaction to occur in exactly one abstracted time step.
[0023] Figure 1 illustrates an overall transaction-based abstraction process according to certain embodiments of the invention, which can include three main phases.
• Select a set of transaction-processing finite state machines (FSMs) that determine the transaction boundaries.
• Extract the transaction-processing FSMs, compose them, and compute an abstracted FSM corresponding to the composed FSM after abstraction. • Abstract all signals in the design based on the computed abstract FSM. Each of these three phases will is discussed in further detail below.
1) Selecting a Set of Transaction-Processing FSMs
[0024] To achieve the desired effectiveness of transaction-based abstraction, the boundaries of a transaction, that is, the starting and ending time steps of each transaction, must be correctly deteπnined. In accordance with the present invention, a heuristic technique can be used to identify finite-state machines (FSMs) in the design that process transactions, from which the transaction boundaries can be identified. Heuristics can filter signals that can be used to determine transaction boundaries. Possible heuristics include:
• select signals that look like FSMs that may process transactions;
• select signals representing FSMs that have wait states indicating the FSM waits for either a request or response;
• select signals representing FSMs that have state transition loops indicating that the FSM can process transactions infinitely often; and
• allow the user to manually specify signals and conditions that determine transaction boundaries.
[0025] In certain embodiments of the present invention, the steps for identifying a set of signals representing transaction-processing FSMs include:
• Identify the subset of signals that can be used to determine FSMs.
• Eliminate from this list those FSMs that do not process transactions. Transaction- processing FSMs can have the following characteristics:
• They have at least one wait state, where a wait state is a state that transitions to itself based on some condition.
• There is at least one loop in the FSM, where a loop is a path in the FSM state graph that starts in some state and ends at the same state.
• One of the states contained within the loop path is a wait state. The resulting set of FSMs is the transaction-processing FSM set. [0026] Figure 2 illustrates, in state diagram form, an exemplary finite state machine ("FSM") which represents a device that processes transactions according to certain embodiments of the invention As shown in Figure 2, the exemplary FSM consists of four states, sθ, si, s2, and s3. According to the present invention, a transition consists of a current state, u, next state, v, and guard, g. If the FSM is in state u at time t and condition g is true, it will transition to state v in the next time step. Transitions are indicated as edges in the state transition graph with edge labels indicating the guard condition g. Unlabelled edges represent unconditional transitions. Assignments indicated by a box attached to an edge in a graph indicate signals that are set as a function of the FSM. State sO is a wait state because it has an edge to itself labeled "-start", indicating that the FSM should wait in that state until a start signal occurs. Once started, the FSM proceeds through the other states and then loops back to the wait state, at which time it indicates completion of the transaction by asserting the "done" signal. The transaction boundaries occur when the FSM transitions out of the wait state at the beginning and back to the wait state at the end.
2) Computing the Abstract FSM
[0027] After selecting and extracting the transaction-processing FSMs from the design, the extracted FSMs are combined to form a composite FSM. Composition of FSMs is done by enumerating all combinations of states and edges in the individual extracted FSMs. Composite edges are formed by computing the product of all edge transition conditions for individual FSMs. Since FSM inputs may be functions of one or more other FSMs being composed, this may cause some edges or states to be unreachable if the computed product condition is false.
[0028] In accordance with the present invention, an abstract FSM is computed in which states in the abstract FSM correspond to sets of states in the composite FSMs. Figure 3 illustrates this FSM abstraction process according to certain embodiments of the invention. As shown in the first step of Figure 3, all wait states in the individual FSMs are given a unique label and all other states are given a common label.
[0029] Next, all composite states in the composite FSM are labeled based on the labels created for the individual FSMs. The goal of labeling the composite states in this second step of Figure 3 is to delineate states that have different wait conditions since different wait conditions may indicate different transactions. One way to accomplish this to label states in the composite FSM with the combination of individual FSM labels corresponding to the component of the composite state. If a composite state has no component that is labeled as a wait state, give this composite state a unique label.
[0030] After all composite states have been labeled, then, if all composite states have no component that is labeled a wait state, re-label all composite states with the same label instead of giving each a unique label (since there is no wait state). In this way, a single pass through the FSM indicates a single transaction. In this example, all states would be given the same label.
[0031] The set of abstract states corresponds to the unique labels produced using the above procedure for the composite FSM. An edge exists between two abstract states if there exists an edge between two corresponding states in the composite FSM. The transition condition for this edge is the OR of all composite transition conditions for this abstract edge.
[0032] As an example, Figure 4 illustrates, in state diagram form, two exemplary interrelated FSM' s which interact, representing two devices which similarly interact in processing transactions, according to certain embodiments of the invention. In this example, FSM A drives transactions to FSM B, which waits for a transaction start, processes the transaction, and then signals done. FSM A waits for the done indication before starting the next transaction. In FSM A, state transition aO to al causes signal "start" to be set to 1. When FSM B is in state bθ, it transitions to bO if signal "start" is 0, indicated by the edge guard "~start", otherwise, when "start" is 1, it transitions to state bl. Since "start" is set to 1 in FSM A state aO, the only valid composite transition from composite state {aθ,bθ} is to composite state {al,bl}. Similarly, FSM B controls signal "done", setting it to 1 as indicated by the assignment attached to transitions b2->bθ and b3->bθ ("done" is assumed to be set to 0 for all other transitions). The wait state of FSM A, a2, is dependent on the "done" signal produced by FSM B and either waits when "done" is 0, indicated by the edge labeled "~done", or continues to state aO when "done" is 1.
[0033] Figure 5 illustrates, in table form, state labels for the two exemplary FSM' s shown in Figure 4, together with state labels for the composite FSM, according to certain embodiments of the invention. Note that composite states {a2,b2} and {a2,b3} have the same composite label. In the abstract, or composite, FSM, these two composite states will be a single abstract state as shown in the abstract FSM in Figure 6 computed from the two FSMs in Figure 5. Figure 6 illustrates, in state diagram form, an abstracted FSM from the two FSM' s of Figure 4 according to certain embodiments of the invention.
[0034] When a FSM or composite FSM has a loop but no wait states, abstraction is still possible by simply selecting some state arbitrarily as the transaction boundary state. One way to do this is to select the initial state as the boundary state if the loop is through the initial state, otherwise a boundary state can be selected by traversing the FSM state graph from the initial state and selecting the first loop state encountered to be the boundary state. Since no state is a wait state, instead of each composite state getting a unique label, all states are given the same label. The loop will be abstracted to a single state since states with common labels are combined together during abstraction. The result of this is that abstracted transactions will be one time step in duration since a single abstracted state corresponds to a complete transaction.
3) Abstracting Signals
[0035] A signal definition function specifies the value of a signal over time as a function of other signals and time. For example, the definition of signal "a" may be "a(t) = b(t) + c(t- 1)" which says that at time t, the value of "a" is given by adding the value of signal "b" at time t to the value of signal "c" at time t-1. The goal of signal abstraction, in accordance with the present invention, is to modify the signal definition such that it is a function of abstract time, f .
[0036] Since time steps are abstracted away, it is possible that signals are functions of time steps that do not exist after abstraction. In this case it is necessary to abstract the signal such that it is a function only of time steps that correspond to abstracted time steps. For example, assume that a simulation of a device contains the signal assignment, "a(t) = b(t) + c(t-l)", and that the first transaction starts at time 0 and ends at time 15, which corresponds to abstract time 0. In accordance with the present invention, the signal value for signal "a" at abstract time 0 is defined as "a(15) = b(15) + c(14)", which can be abstracted by substituting the appropriate abstract time, 0, for 15 in the expression.
[0037] However, there is no abstract time step in the defined signal example above that corresponds to time step 14, thus, this signal cannot be abstracted as is. To solve this problem, signals are composed such that they are functions of time steps that have corresponding abstract time steps. In this context, composition refers to the substitution of a function defining a signal for a signal reference in an expression. For example, suppose that signal "c" is defined as "c(t) = c(t-l)+l". Composing the definition of signal "c" into the expression for signal "a" is done in two steps. First, the expression for signal "a" refers to "1C(M)" indicating that this is the value of signal "c" at time M. However, the definition for "c" is at time t. Therefore, t-1 must be substituted for t in the expression for "c" to yield the definition "c(t-l) = c(t-2)+l". Substituting this into the expression for "a" yields "a(t) = b(t) + c(t-2) + 1". Composing signal "c" 15 times using this procedure in the expression for "a" produces "a(t) = b(t) + c(t-16) + 15", which now can be abstracted as "a(t') = b(t') + c(t'-l) + 15". The term "c(t'-l)" arises because, for the first transaction, the transaction length is 16 and therefore, the unabstracted term, "c(t-16)" refers to the value in the previous transaction.
[0038] It will also be appreciated that different transactions may have different abstractions for the same signal. In the example above, if the second transaction is only five cycles long, for example, the composition of "c" need only be done five times, yielding the abstracted expression "a(t') = b(t') + c(t'-l) + 5. Different length transactions are indicated by different labels in the composite FSM. Therefore, an abstracted signal definition is needed for each unique label. These different signal definitions are then combined by selecting the correct one as a function of the computed abstract state corresponding to the labeled composite state. During simulation, as the abstract FSM transitions to different states, this enables the different abstract functions for each signal such that each signal is computed correctly for each abstract FSM state. The complete process for performing signal abstraction is described below. This process ensures that the computed signal values for each abstract time step will correspond to the values at the equivalent normal time steps.
[0039] In summary, the signal abstraction algorithm is as shown in Figure 7. Figure 7 illustrates a signal abstraction process according to certain embodiments of the invention. The first step of Figure 7 is done by performing a breadth-first search of the composite state graph. Each step of this traversal specifies a tuple (u,g,v) where u is the current state, v is the next state, and g is a condition that must be true for the state transition to occur. Values u and v correspond to assignments to the signals that make up the individual signals of the composite state. The expression g contains references to composite state signals and other signals.
[0040] Signal values are represented by expressions comprising operators such as add, and, etc., ite operators, constant terminals, and word teπninals. The ite operator specifies three subexpressions, an if condition, then assignment, and else assignment and represents an if-then-else (ite) assignment. Word terminals specify the name of a signal and a delay. Since signal values are functions of time, a word terminal with delay k specifies the value of the signal at time t-k. In at least some embodiments, the present invention only considers cases in which the word terminal delay is 0 or 1. Operator nodes have the usual semantics.
[0041] In the exemplary implementation described herein, BFS-based abstraction depends on whether the current edge (u,g,v) is a transaction boundary or not. For each pair of states, u,v, that defines a transition, there is a mapping from these states to abstract states in the abstract FSM. Let these abstract states be a_u and a_v. A transaction boundary is defined if a_u != a_v or if this edge is looping back from the last state to first state of the composite FSM. The abstract boundary state is defined to be state a_v of a transaction boundary as defined above. For example, in Figure 2, the edge from s3 to sO would be considered a transaction boundary even if s3 and sO mapped to the same abstract state, and, therefore, state sO is defined to be the abstract boundary state.
[0042] The signal function definition is parsed and modified only if a word terminal specifying signal s and a delay d are encountered for edge (u,v). The process for determining the abstracted word terminal according to certain embodiments is shown in Figure 8. Figure 8 illustrates a process for computing an abstracted word terminal for a given signal, delay, and composite FSM edge (u,v) according to certain embodiments of the invention.
[0043] The composite FSM may have states with multiple incoming edges. Since the result of the above process produces an expression for each signal for each state, multiple expressions will be produced for states that have multiple incoming edges and will need to be combined. Typically, multiple incoming edges occur due to multiple outgoing edges somewhere else in the state graph. Multiple outgoing edges are differentiated by different edge guard conditions. A guard condition is just another expression that is a function of time. These guards also need to be abstracted as a function of the composite state using the exact same parsing and modification as described above. Further, guard expressions are propagated along states and edges. For example, if (u,g,v) is an edge with g' being the abstracted expression for g, and edge (v,h,w) is also an edge, then g' will be propagated to edge (v,h,w) because of the common state v. The effective guard for edge (v,h,w) is the expression "g' & h'" where h' is the abstracted value of h which is anded with the previously abstracted value g'. [0044] If a word terminal is marked for composition, this composition will occur the second step of Figure 7. This second step iterates over all composite states and signals. Each abstracted signal value for each composite state is processed in this pass. For signal s and composite state u, the signal will be processed and stored in an array indexed by the signal and abstract state, u', corresponding to state u. Since multiple composite states may map to a single abstract state, this pass is typically performed in composite state order. That is, state u is typically processed before state v if (u,g,v) is an edge in the composite FSM and (u,g,v) is not a transaction boundary.
[0045] This step modifies the expression produced in the previous step before storing it in the array. The modifications also only affect word terminals marked for composition in the previous pass. A marked word terminal specifies a signal p and state w. There are two possible modifications:
• If state w is not an abstract boundary state, the word terminal is composed by replacing the word terminal with the abstracted value of signal p associated with state w that was computed in the previous step. The resulting expression is recursively traversed so that multiple compositions may occur.
• If state w is an abstract boundary state, the word terminal is replaced with a word terminal specifying word p with delay t'-l (delay of 1 relative to the abstract clock).
[0046] The resulting array of signal values for each abstract state has expressions that are functions of signals at abstract boundaries only. The last step of Figure 7 combines these into a single assignment for each signal. Given a signal s with values stored in the array for abstract states a_u and a_v, these can be combined by performing an if-then-else in which the if condition is the abstract state being equal to state a_u for example. The resulting expression will consist of a series of ite expressions, such as: s (f ) = ite (a_fsm = sθ , array [s] [ sθ] , ite (a_f sm = si , array [s ] [ si ] ,
[0047] As an example of signal abstraction, Figure 9 illustrates an example of signal abstraction for the FSM of Figure 2, showing the three phases of the abstraction process with signal abstractions labeled with the appropriate state from the FSM shown in Figure 2, according to certain embodiments of the invention. Assume the composite FSM is the FSM given in Figure 2 and a signal, s, is defined as specified at the top of Figure 9. In the first pass, signal s is abstracted for each edge in the composite FSM. The resulting expressions are stored for each state sθ-s3 of the FSM. In the first pass, the value of the signal is abstracted by substituting the state value in the expression in s and marking signal s by specifying a composite state. The abstracted expression for signal s for each state sθ, si, s2, s3, is shown in Figure 9 in the box labeled "pass 1". Each word terminal is of the form "s(sx)", where "s" is the signal name and "sx" indicates a state (sθ, si, s2, s3), and is marked for composition in the next step. In this example, s is reset to 0 if the FSM is in state sO and, otherwise, increments. Thus, the abstract value for state sO is 0. For each of the other states, the expression references s(t-l), which results in the word terminal being marked with the appropriate state. Note that signal g, which exists in the FSM also needs to be abstracted. For clarity, the abstracted value is simply represented as g\ however, it also would go through this process.
[0048] In the second pass, the signal for each state is modified by substituting the appropriate expression for each marked word. These substitutions are done in state order, sO - si — s2 - s3 and after substitution, the result is stored in the corresponding abstract state array entry. These expressions are shown in the text box marked "pass 2". Each expression is the result of composition and all states are labeled with abstract state aO with the corresponding composite state in parentheses. Since there is only a single abstract state, the final value is just the value of the array for s in state aO. Since the expression for s3 was created last, this is the final value of the abstracted value of signal s.
[0049] In this example, the unabstracted version of signal s is a function of itself, which means that it is a state-holding element. Furthermore, it is a function of another state-holding element, the FSM. With no other context for this signal, it has a fairly complex definition. After abstraction, signal s is no longer a function of itself or the FSM and is not a state- holding element. We know that the abstraction represents the function of this signal for one transaction. After abstraction, it is much more obvious what the meaning of this signal is. Also, because the definition is simpler, it can be more efficiently simulated. The ability to make signal functions easier to understand and to simplify them illustrate one of the many advantages of transaction-based abstraction. 4) Signal Classification
[0050] It is useful during verification of hardware to classify signals into different types that have different verification goals. In particular, sequential logic requires different verification techniques than combinational logic. Normally, the user must perform this classification manually. In large circuits, this is a tedious task.
[0051] As discussed previously herein, one aspect of the present invention is that signals are classified according to their use in processing transactions. There are three types of signals:
• Transient
• Temporary
• Persistent
[0052] Transient signals consist of the transaction-processing finite-state machines (FSMs) that control the transaction by detecting when a transaction starts, controlling the processing of the transaction, and then signaling its completion.
[0053] Temporary signals are those that are used only within a single transaction. A signal that is used only within a transaction must be set before being read within the transaction. Because it is set before being used within a transaction, the value of a temporary signal in one transaction cannot affect the operation of other transactions.
[0054] Persistent signals are those that are set in one transaction and read in another.
[0055] In accordance with the present invention, signals are classified by performing transaction-based analysis of the design. Each of the signal types is determined by the abstraction process as follows:
• Transient signals are the signals corresponding to the transaction-processing FSMs identified by the abstraction heuristics.
• Persistent signals are the sequential signals left after abstraction.
• Temporary signals are all other signals.
[0056] The rationale behind this is that since after abstraction, each transaction is compressed into one abstract time step, all sequential signals that are used within only a single transaction become combinational after abstraction. Any signal that is sequential after abstraction must persist between transaction. The definition of transient signals is precisely the definition of transaction-processing FSMs defining transactions in the abstraction process. Therefore, this process produces a signal classification.
a) Automatically Creating a Test Plan Based on a Signal Classification
[0057] A test plan is a set of coverage points that must be exercised in a design. A coverage point can be either:
• a Boolean expression over all the signals in a design that must be true, or
• a range of values for a signal, for example, a 4-bit wide signal that can have values in the range 0-15.
[0058] Coverage points are typically defined in terms of a simulation model, which comprises a device plus a test case and other models, that may be used to compute the results of the simulation. A coverage point is considered covered during simulation if 1) it becomes true at some point during simulation if it is a Boolean expression, or 2) the signal takes on a value from the specified range at some point during simulation.
[0059] In accordance with the present invention, a test plan can be automatically generated by identifying the architecturally visible state in the design and then creating coverage for signals in the design. In some embodiments the architecturally-visible state can be identified by classifying all signals in the design as either transient, persistent or temporary, as discussed above using transaction-based abstraction analysis..
[0060] Figure 10 illustrates a flow for creating a test plan automatically from a design description according to certain embodiments of the invention. As shown in Figure 10, the input to the test plan generation process consists of a design description, which, in some embodiments consists of a description written in a hardware description language (HDL) such as Verilog. The first step, as shown in Figure 10, is to determine the architecturally visible state and to classify the signals. The second step is then to coverage points for the classified signals based on the information computed in the first step using techniques described previously herein and output this as a test plan.
[0061] Signals are classified as transient, temporary, or persistent, where persistent signals can be considered the architecturally visible state and transient and temporary signals are the non-architecturally visible. The testing strategies for architecturally visible and non- architecturally visible state are different. Therefore, the coverage points for different classifications are different. For each of these classifications, coverage points are generated as follows:
(i) Transient signals
[0062] Each transient signal represents a transaction-processing finite state machine (FSM). A finite state machine consists of a set of states and transitions between states. Transitions may be conditional, specified by a Boolean condition. Transaction-processing FSMs implement the protocol that defines a transaction. It is necessary to verify all possible timings of a protocol. Since the FSM must handle all possible cases, to completely test the protocol requires exercising every transition in the FSM. Let (u,g,v) be a transition in a transaction-processing FSM where u and v are states and g is a Boolean expression over all signals be defined as if, at time t, the value of the FSM signal at time t-1 was u, and the value at time t is v, and g is true at time t, then this transition has been taken at time t. Therefore the coverage point is the expression: s (t-l ) == u && s (t) == v && g (t)
[0063] There will be one coverage point for each transition in the FSM. (ii) Temporary signals
[0064] In a typical device, temporary signals will comprise the majority of signals. These primarily are data signals, not control signals. Therefore, collectively, they represent a large number of bits. Normally, with data, it is desirable to cover a large number of values. Temporary signals are normally defined as a set of assignments, each of which is enabled under different conditions defined by Boolean expressions over all signals. Assume that a temporary signal has a set of assignments with an associated condition, represented as (chgi) where e, is an assignment expression and g, is the Boolean guard expression. The set of coverage points for a temporary signal is the set of guards, gϊ plus the range of all possible values for the signal based on its defined width (number of bits).
(iii) Persistent signals
[0065] Persistent signals may be data or control. For cases in which persistent signals are data values, they can generate the same coverage points as temporary signals. If not, it is desirable to cover as many different values as possible since these potentially represent different states. Also, if they do represent state, persistent signals often interact with other persistent signals. Therefore, the coverage points for persistent state to be measured are the ranges of values of persistent signal values.
[0066] As a refinement of this, the set of persistent signals can be partitioned such that persistent signals that are not transitively dependent on each other are put in different partitions. Coverage points for persistent state are then created as ranges of combinations of values for signals within a partition.
[0067] The set of coverage points generated by this analysis forms a comprehensive test plan for a simulated device. This is beneficial because it is automatically generated, relieving the user from the tedious chore of creating a test plan manually.
[0068] Although the present invention has been particularly described with reference to embodiments thereof, it should be readily apparent to those of ordinary skill in the art that various changes, modifications, substitutes and deletions are intended within the form and details thereof, without departing from the spirit and scope of the invention. Accordingly, it will be appreciated that in numerous instances some features of the invention will be employed without a corresponding use of other features. Further, those skilled in the art will understand that variations can be made in the number and arrangement of inventive elements illustrated and described in the above figures. It is intended that the scope of the appended claims include such changes and modifications.

Claims

ClaimsWhat is claimed is:
1. A method for transaction-based abstraction of a hardware design comprising the steps of: selecting one or more transaction-processing finite state machines (FSMs), wherein the one or more transaction-processing FSMs determine one or more transaction boundaries, a transaction boundary being a starting and ending time step of a transaction; extracting the one or more transaction-processing FSMs; and obtaining an abstracted FSM based upon the one or more extracted transaction- processing FSMs.
2. The method of claim 1, wherein the step of obtaining includes the steps of: composing the one or more extracted transaction-processing FSMs; and computing the abstracted FSM corresponding to the composed one or more FSMs.
3. The method of claim 1, wherein the step of selecting includes the steps of: recognizing signals of the hardware design that indicate FSM candidates; and identifying the one or more transaction-processing FSMs from the FSM candidates.
4. The method of claim 3, wherein the step of recognizing includes using one or more heuristic techniques selected from a group of heuristic techniques to filter the signals to determine the one or more transaction boundaries.
5. The method of claim 4, wherein the group of heuristic techniques includes: selecting signals that look like FSMs and may process transactions; selecting signals representing FSMs that have wait states indicating the FSM waits for either a request or a response; selecting signals representing FSMs that have state transition loops indicating that the
FSM can process transactions infinitely often; and allowing a user to manually specify signals and conditions that determine transaction boundaries.
6. The method of claim 3, wherein identifying the one or more transaction-processing FSMs includes recognizing one or more identifying characteristics selected from a group of identifying characteristics of the FSMs.
7. The method of claim 6, wherein the group of identifying characteristics includes: at least one wait state, wherein a wait state is a state that transitions to itself based on a transition condition; at least one loop, wherein a loop is a path that starts in a first state and ends in that same state; and a combination of at least one loop with at least one wait state.
8. The method of claim 2, wherein the step of composing includes the steps of: enumerating combinations of states and edges of the extracted FSMs; and forming composite edges by computing the product of edge transaction conditions for the extracted FSMs.
9. The method of claim 2, wherein the step of computing includes the steps of: providing a label to each wait state of each extracted FSM; providing a common label to each non-wait state of each extracted FSM; labeling each composite state of the composed FSM based on the labels provided to each wait and non-wait state of each extracted FSM; and if no composite state of the composed FSM has a wait state label, re-labeling the composite states of the composed FSM with a single label.
10. The method of claim 1, further comprising the step of: abstracting signals in the hardware design based on the abstracted FSM.
11. The method of claim 10, wherein the step of abstracting signals includes the steps of: creating one or more abstracted signal functions associated with composite states of the composed FSM; creating abstracted expressions by combining abstracted signal functions from different composite states that are associated with the an abstracted state of the abstracted FSM; and combining the abstracted expressions to form a single expression per signal over abstracted states of the abstracted FSM.
12. The method of claim 10, further comprising the step of automatically generating a coverage test plan using the abstracted FSM and abstracted signals.
13. The method of claim 12, further comprising the step of simulating the hardware design using coverage test plan.
14. The method of claim 10, further comprising the step of simulating the hardware design based on the abstracted FSM and using the abstracted signals
15. A method for transaction-based abstraction of a hardware design comprising the steps of: recognizing signals of the hardware design that indicate finite state machine (FSM) candidates; identifying one or more transaction-processing FSMs from the FSM candidates, wherein the one or more transaction-processing FSMs determine one or more transaction boundaries, a transaction boundary being a starting and ending time step of a transaction; extracting the one or more transaction-processing FSMs; composing the one or more extracted transaction-processing FSMs; computing an abstracted FSM corresponding to the composed one or more FSMs; and abstracting signals in the hardware design based on the abstracted FSM.
16. The method of claim 15, wherein identifying the one or more transaction-processing FSMs includes recognizing one or more identifying characteristics of the FSMs.
17. The method of claim 16, wherein the step of composing includes the steps of: enumerating combinations of states and edges of the extracted FSMs; and forming composite edges by computing the product of edge transaction conditions for the extracted FSMs.
18. The method of claim 17, wherein the step of computing includes the steps of: providing a label to each wait state of each extracted FSM; providing a common label to each non-wait state of each extracted FSM; labeling each composite state of the composed FSM based on the labels provided to each wait and non-wait state of each extracted FSM; and if no composite state of the composed FSM has a wait state label, re-labeling the composite states of the composed FSM with a single label.
19. The method of claim 18, further comprising the step of simulating the hardware design based on the abstracted FSM and using the abstracted signals
20. A transaction-based hardware design verification system comprising: a simulator, wherein the simulator uses one or more abstracted finite state machines (FSMs) and associated abstracted signals such that transactions of the one or more abstracted FSMs occur in one abstracted time step.
21. The system of claim 20, further including an automatically generated coverage test plan derived from the one or more abstracted FSMs and the associated abstracted signals.
PCT/US2005/035663 2004-10-04 2005-10-04 Transaction-based system and method for abstraction of hardware designs WO2006041833A2 (en)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
US61589204P 2004-10-04 2004-10-04
US60/615,892 2004-10-04

Publications (2)

Publication Number Publication Date
WO2006041833A2 true WO2006041833A2 (en) 2006-04-20
WO2006041833A3 WO2006041833A3 (en) 2007-11-15

Family

ID=36148835

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/US2005/035663 WO2006041833A2 (en) 2004-10-04 2005-10-04 Transaction-based system and method for abstraction of hardware designs

Country Status (2)

Country Link
US (2) US7996200B2 (en)
WO (1) WO2006041833A2 (en)

Families Citing this family (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US7512531B1 (en) * 2002-12-30 2009-03-31 Daniel Shia Method and apparatus for specifying reactive systems
WO2009002447A1 (en) 2007-06-21 2008-12-31 Gen-Probe Incorporated Instrument and receptacles for use in performing processes
US20110029331A1 (en) * 2009-08-03 2011-02-03 Bart Parker Method of processing structure damage insurance claims
US8799850B2 (en) * 2009-10-29 2014-08-05 Synopsys, Inc. Simulation-based design state snapshotting in electronic design automation
US8473883B2 (en) 2011-08-03 2013-06-25 International Business Machines Corporation Abstraction for arrays in integrated circuit models
US8667436B1 (en) * 2013-03-01 2014-03-04 Xilinx, Inc. Object identification in an electronic circuit design
US9311444B1 (en) * 2014-07-10 2016-04-12 Sandia Corporation Integrated circuit test-port architecture and method and apparatus of test-port generation
GB201511175D0 (en) 2015-06-25 2015-08-12 Mclaren Applied Technologies Ltd Analysing physical systems

Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20040054500A1 (en) * 2002-05-14 2004-03-18 Cadence Designs Systems Method and mechanism for improved performance analysis in transaction level models

Family Cites Families (16)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5953519A (en) * 1995-06-12 1999-09-14 Fura; David A. Method and system for generating electronic hardware simulation models
US6421815B1 (en) * 1998-01-09 2002-07-16 Synopsys, Inc. Method and apparatus for optimized partitioning of finite state machines synthesized from hierarchical high-level descriptions
US6917909B1 (en) * 1998-05-18 2005-07-12 Lev A. Markov Facilitating guidance provision for an architectural exploration based design creation process
US6212667B1 (en) * 1998-07-30 2001-04-03 International Business Machines Corporation Integrated circuit test coverage evaluation and adjustment mechanism and method
US6237101B1 (en) * 1998-08-03 2001-05-22 International Business Machines Corporation Microprocessor including controller for reduced power consumption and method therefor
US6675138B1 (en) * 1999-06-08 2004-01-06 Verisity Ltd. System and method for measuring temporal coverage detection
US7356786B2 (en) * 1999-11-30 2008-04-08 Synplicity, Inc. Method and user interface for debugging an electronic system
US6810372B1 (en) * 1999-12-07 2004-10-26 Hewlett-Packard Development Company, L.P. Multimodal optimization technique in test generation
US6859770B2 (en) * 2000-11-30 2005-02-22 Hewlett-Packard Development Company, L.P. Method and apparatus for generating transaction-based stimulus for simulation of VLSI circuits using event coverage analysis
JP3965068B2 (en) * 2001-04-18 2007-08-22 インターナショナル・ビジネス・マシーンズ・コーポレーション Integrated test case language for verifying hardware designs
US7007251B2 (en) * 2003-11-12 2006-02-28 International Business Machines Corporation Database mining system and method for coverage analysis of functional verification of integrated circuit designs
US7260798B2 (en) * 2003-12-29 2007-08-21 Mentor Graphics Corporation Compilation of remote procedure calls between a timed HDL model on a reconfigurable hardware platform and an untimed model on a sequential computing platform
US7213225B2 (en) * 2004-05-28 2007-05-01 International Business Machines Corporation Method, system and program product for specifying and using register entities to configure a simulated or physical digital system
US7389490B2 (en) * 2004-07-29 2008-06-17 International Business Machines Corporation Method, system and program product for providing a configuration specification language supporting selective presentation of configuration entities
US7555689B2 (en) * 2005-06-28 2009-06-30 Dhiraj Goswami Generating responses to patterns stimulating an electronic circuit with timing exception paths
US8239791B2 (en) * 2008-06-09 2012-08-07 International Business Machines Corporation Method of designing multi-state restore circuitry for restoring state to a power managed functional block

Patent Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20040054500A1 (en) * 2002-05-14 2004-03-18 Cadence Designs Systems Method and mechanism for improved performance analysis in transaction level models

Also Published As

Publication number Publication date
WO2006041833A3 (en) 2007-11-15
US7996200B2 (en) 2011-08-09
US20090012771A1 (en) 2009-01-08
US8020123B2 (en) 2011-09-13
US20060085172A1 (en) 2006-04-20

Similar Documents

Publication Publication Date Title
US8020123B2 (en) Transaction-based system and method for abstraction of hardware designs
US7159198B1 (en) System and method for identifying design efficiency and effectiveness parameters for verifying properties of a circuit model
US7383166B2 (en) Verification of scheduling in the presence of loops using uninterpreted symbolic simulation
US7895552B1 (en) Extracting, visualizing, and acting on inconsistencies between a circuit design and its abstraction
US6553514B1 (en) Digital circuit verification
US7587690B1 (en) Method and system for global coverage analysis
US4692921A (en) Method for generating verification tests
US6564365B1 (en) Method of simultaneously displaying schematic and timing data
US6993470B2 (en) Method of evaluating test cases in a simulation environment by harvesting
US7603647B2 (en) Recognition of a state machine in high-level integrated circuit description language code
CN107040433A (en) A kind of automated testing method and system
US20020144218A1 (en) Method and device for verification of VLSI designs
JP3366475B2 (en) How to evaluate data processing speed
JP2001522084A (en) Method of generating ISA simulator and assembler from machine language description
JP2003036278A (en) Fmea system
US6754878B1 (en) Recognizing structure information from a netlist
US8050904B2 (en) System and method for circuit symbolic timing analysis of circuit designs
CN108875374A (en) Malice PDF detection method and device based on document node type
US7689399B1 (en) Automatic extraction of design properties
Wang et al. Automatic test case generation for Estelle
US6185518B1 (en) Method and system for logic design constraint generation
Santos et al. A design methodology for the formal specification and verification of hypermedia documents
Miller et al. Saving space by fully exploiting invisible transitions
US5818850A (en) Speed coverage tool and method
Fleischhack et al. Towards Compositional Verification of SDL Systems

Legal Events

Date Code Title Description
AK Designated states

Kind code of ref document: A2

Designated state(s): AE AG AL AM AT AU AZ BA BB BG BR BW BY BZ CA CH CN CO CR CU CZ DE DK DM DZ EC EE EG ES FI GB GD GE GH GM HR HU ID IL IN IS JP KE KG KM KP KR KZ LC LK LR LS LT LU LV LY MA MD MG MK MN MW MX MZ NA NG NI NO NZ OM PG PH PL PT RO RU SC SD SE SG SK SL SM SY TJ TM TN TR TT TZ UA UG US UZ VC VN YU ZA ZM ZW

AL Designated countries for regional patents

Kind code of ref document: A2

Designated state(s): BW GH GM KE LS MW MZ NA SD SL SZ TZ UG ZM ZW AM AZ BY KG KZ MD RU TJ TM AT BE BG CH CY CZ DE DK EE ES FI FR GB GR HU IE IS IT LT LU LV MC NL PL PT RO SE SI SK TR BF BJ CF CG CI CM GA GN GQ GW ML MR NE SN TD TG

121 Ep: the epo has been informed by wipo that ep was designated in this application
NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase