Key comparison diagnostic method based on dual-isomorphic logic circuits

Abstract

This invention relates to the field of logic circuit testing and fault diagnosis technology, specifically a critical comparison diagnosis method based on dual heterogeneous logic circuits. This method acquires state change records, critical output records, and intermediate logic records formed by two heterogeneous logic circuits under the same input event sequence. Based on the input support set, output influence set, state transition conditions, and stability interval, it converts the physical nodes on both sides into common functional propositions and establishes semantic correspondences. It constructs local causal cones using input event identifiers and determines causal completion markers. After both sides' corresponding propositions are completed, a critical comparison is performed. Based on the difference recovery position, propagation direction, and the consistent state of the preceding propositions, it distinguishes between propagation beat differences and irreversible logic differences. Finally, it determines the minimum first difference boundary through propagation verification and dependency pruning. This invention can reduce misjudgments caused by normal propagation differences and locate critical output inconsistencies to the functional proposition region where the earliest irreversible divergence occurs.

Description

Technical Field

[0001] This invention relates to the field of logic circuit testing and fault diagnosis technology, specifically a key comparison and diagnosis method based on dual heterogeneous logic circuits. Background Technology

[0002] In existing technologies, for reliability verification scenarios of critical logic functions, dual-path logic circuit comparison and diagnostic techniques are typically employed. This involves setting up two sets of logic circuits that implement the target function and comparing the output results of the two logic circuits under the same input conditions to determine if there are any anomalies in the logic execution process. Conventional dual-path comparison methods usually design the two sets of logic circuits with different implementations, reducing the probability of a single design flaw affecting both results simultaneously by increasing the differences in logic structure. During operation, the system collects the state values ​​or output values ​​of the two logic circuits according to a preset clock cycle and compares the logic results at the corresponding positions. When inconsistency is detected between the outputs of the two logic circuits, an anomaly diagnosis process is triggered. To accommodate differences between different logic implementations, some technical solutions further introduce signal delay compensation, time window matching, or state mapping mechanisms to adjust for the propagation cycle differences between the two logic circuits, enabling comparison of logic states under different execution cycles. This type of method mainly relies on fixed timing relationships or preset mapping relationships to complete the comparison process. The comparison objects are usually concentrated on critical output signals, some state registers, or predefined logic nodes, and the diagnostic results are generally used to determine whether the dual outputs are consistent.

[0003] As the complexity of implementing dual heterogeneous logic circuits increases, the two sets of logic circuits may differ significantly in terms of internal state encoding, logic partitioning, condition judgment order, and signal propagation paths. Traditional comparison methods based on fixed node correspondences are ill-suited to complex heterogeneous logic structures. Some existing solutions establish a logic node mapping table to associate corresponding nodes in the two heterogeneous logic circuits and then perform state comparisons based on the mapping results. However, the correspondence between nodes usually relies on manual configuration during the design phase, and the mapping relationship needs to be re-maintained when the logic implementation changes or the number of intermediate states increases. Some diagnostic methods use output backtracking, tracing upstream nodes that may affect the output after a critical output anomaly is detected to determine the source of the fault. However, because the internal propagation paths of heterogeneous logic circuits differ, simply tracing back based on the output results can easily confuse normal propagation delays and state temporary storage processes with actual logic errors. Some methods reduce false alarms caused by timing differences by expanding the comparison time range, but a larger comparison window reduces the ability to detect short-term logic deviations, making it difficult to identify early anomalies that may affect critical results in subsequent propagation.

[0004] The main problem with existing critical comparison diagnostic methods for heterogeneous logic circuits lies in the difficulty of distinguishing between normal state propagation differences and actual logic anomalies under different logic implementations. Because the two heterogeneous logic circuits have different state transition paths and internal logic organization methods, the same functional state may correspond to different intermediate logic behaviors. When comparing at fixed periods or fixed nodes, intermediate states that have not yet completed logic propagation are easily misjudged as abnormal. Extending the waiting time to avoid misjudgment may hide logic differences that have already occurred and propagated to the critical output. Furthermore, after a true logic deviation occurs, traditional methods typically only obtain inconsistent final output results and cannot determine the location of the earliest irreversible difference based on input events, logical dependencies, and state transition processes. This results in a lack of ability to locate the first abnormal logic boundary in the diagnostic process. Therefore, a critical comparison diagnostic method is needed that can combine the correspondence between heterogeneous logic functions, causal completion states, and difference propagation processes to identify actual logic differences and determine their location under different logic implementations. Summary of the Invention

[0005] The purpose of this invention is to provide a key comparison diagnostic method based on dual heterogeneous logic circuits, which can solve the problems in the background art mentioned above.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows: A key comparison diagnostic method based on dual heterogeneous logic circuits includes: acquiring state change records, key output records, and observable intermediate logic records formed by the first heterogeneous logic circuit and the second heterogeneous logic circuit under the same input event sequence; Based on the input support set, output influence set, state transition conditions, and stability interval of each logical quantity, the physical nodes on both sides are converted into common functional propositions and semantic correspondences are established. Using input event identifiers as a unified index, local causal cones pointing to each key comparison item are constructed along logical dependencies, and causal completion markers are determined for common functional propositions; The critical comparison is performed when both sides of the corresponding common functional proposition have reached the causal completion state. The differences in comparison are then jointly assessed by subsequent recovery, propagation direction, and consistency of the preceding propositions to obtain differences in propagation rhythm or irreversible logical differences. For the irreversible logical differences, forward propagation verification and backward dependency pruning are performed alternately to determine the minimum first difference boundary.

[0007] Preferably, establishing the semantic correspondence includes: dividing the state change records on both sides according to the arrival order of the input events, extracting an input support set for each logical quantity consisting of necessary input conditions, exclusionary input conditions and preceding states, and forming an output influence sequence based on the key outputs that can be reached after the logical quantity changes and the order of propagation. The input support set, the output influence sequence, the state transition condition, and the stability interval are encoded as functional descriptors; Candidate pairs are generated between the two function descriptors, and forward consistency checks from input to output and reverse consistency checks from critical output to input are performed respectively. Retain candidate pairs that pass both-way checks and do not have one-to-many conflicts with other candidate pairs that pass both-way checks, and mark logic quantities that fail checks as non-directly corresponding propositions.

[0008] Preferably, constructing the local causal cone and determining the causal completion marker includes: configuring a consistent event identifier across both sides for each input event, and incorporating state transition nodes triggered by the same input event and belonging to the same key comparison item into the corresponding local causal cone; A partial order hierarchy is generated for nodes within the cone based on the logical dependency direction, and the conditions for the completion of the corresponding common functional proposition are defined as: the preceding proposition has obtained a stable value, all necessary migration branches have been selected, and there are no pending upstream updates for the key comparison item. Record the completion positions of the common functional propositions on both sides at each partial order level; When one side meets the completion condition first, its stable proposition value is frozen, and the other side continues to receive records until the corresponding proposition on the other side reaches the same partial order level, at which point the freeze is lifted and the key comparison is performed. If subsequent input events enter the same local causal cone, the propagation branch of the previous input event that has not yet been completed is separated from the subsequent input event and marked.

[0009] Preferably, the joint determination and determination of the minimum first difference boundary includes: starting from the earliest common functional proposition where inconsistency occurs, recording the difference polarity, duration event interval, subsequent recovery position, downstream affected proposition set and the consistency state of the preceding proposition to form a difference propagation description; Downstream propositions that inherit differential polarity or are triggered by differential migration deviation are marked layer by layer along the local causal cone, and a recovery boundary is generated when the differential disappears and both sides re-enter the same functional state. Differences that do not cross the key comparison item and are eliminated within the recovery boundary are classified as propagation beat differences, while differences that cross the key comparison item, change subsequent migration branches, or do not exist at the recovery boundary are classified as irreversible logic differences. In response to the aforementioned irreversible logical differences, reverse deletion is performed on the preceding branches of the affected propositions that do not have dependent paths, and forward marking and reverse deletion are repeated on the remaining branches until the set of the first difference propositions no longer shrinks.

[0010] Preferably, for the non-directly corresponding proposition, according to the input support set and key output influence range that can be unified to the same functional meaning on both sides, multiple logical quantities with continuous dependency relationship are selected from the state change record on the same side, and a composite migration path is synthesized according to the migration order of the multiple logical quantities. The starting preconditions, ending stable state, exclusion conditions within the path, and the state change intervals crossed by the composite migration path are encoded as composite function descriptors. On the other side, a single logical quantity or another composite migration path is searched as a composite candidate corresponding item, and event order maintenance verification, state reachability verification and critical output boundary verification are performed on the composite candidate corresponding item. Composite candidate items that simultaneously satisfy all three checks and whose internal intermediate states do not participate in other key comparison items are registered as common functional propositions, and the internal logic quantities of the path are limited to subordinate observations of the common functional propositions.

[0011] Preferably, the separation marker includes: establishing an event lineage identifier for each input event entering the same local causal cone, and associating the state transition node with the earliest-arrived event lineage that has not yet closed, based on the necessary input support set of the state transition node; When the feedback path causes the same state transition node to be driven by multiple event families, the preceding proposition versions of each event family are retained, and a composite version containing all unclosed families is generated at the merging node. Maintain a set of unresolved upstream updates based on the propagation position of each event lineage in the partial order hierarchy. When all necessary branches constituting the current common functional proposition have been completed and there are no unresolved upstream updates belonging to the same event lineage, close the corresponding event lineage. When performing the key comparison, stable proposition values ​​with the same event spectrum identifier and the same preceding proposition version are selected from both sides, and the state change records of other unclosed event spectrums are retained in the local causal cone.

[0012] Preferably, the generation of the recovery boundary includes: establishing a propagation marker for each inconsistent common functional proposition, carrying the difference polarity, the source first difference proposition and the identifier of the input event to which it belongs; copying the propagation marker along each downstream branch of the local causal cone; and recording the branch selection change caused by the propagation marker at the state transition node. When multiple downstream branches obtain the same proposition value again at the confluence proposition, check the current state of the confluence propositions on both sides, the selected migration branches, and the set of pending upstream updates respectively; The converging proposition is registered as the recovery boundary only when the current state, the selected migration branch, and the set of pending upstream updates are all consistent, and there are no more pending downstream branches carrying the source first difference proposition in the subsequent partial order level; For converging propositions with the same proposition value but different migration branches or pending upstream update sets, the propagation marker is retained and propagated to subsequent key comparison terms.

[0013] Preferably, after the set of first difference propositions no longer shrinks, the common functional propositions with the comparison differences in the remaining dependency branches and their dependency edges are used to form a difference dependency subgraph. From the difference-dependency subgraph, select common functional propositions that are inconsistent with themselves and whose corresponding predecessors are consistent as boundary candidate propositions, and calculate the set of affected key comparison items that can be reached by each boundary candidate proposition, as well as the set of path interception from the entry proposition with no inconsistent predecessors in the difference-dependency subgraph to the affected key comparison items. Delete the boundary candidate propositions whose affected key comparison item set is completely covered by upstream boundary candidate propositions and do not have exclusive path interception relationships; The retained boundary candidate propositions are used to form proposition antichains that do not have upstream or downstream inclusion relationships according to their logical dependencies, and the proposition antichains and their incoming edge preceding propositions are jointly registered as the minimum first difference boundary.

[0014] Preferably, the semantic correspondence of the common functional propositions is versioned and registered, the currently verified correspondence is saved as the confirmed version, and the newly formed composite candidate correspondence is saved as the pending version; Extract the verification trajectory from the input event interval where there is no irreversible logical difference and the key outputs on both sides are consistent, and verify the input support inclusion relationship, output influence sequence relationship and state transition direction consistency of the version to be determined in turn. When all three relationships remain consistent in the continuous verification trajectory, the pending version is merged into the confirmed version and the effective input event identifier is recorded; When the minimum first difference boundary of subsequent irreversible logical differences contains the updated common functional proposition, revert to the previous confirmed version, place the relevant composite candidate correspondence and its subordinate observations into the isolation set, and stop using the state change record containing the minimum first difference boundary to update the semantic correspondence.

[0015] Preferably, for each boundary candidate proposition in the minimum first difference boundary, a boundary fingerprint is extracted, consisting of the set of incoming edge preceding propositions, the state transition relationship where the divergence occurs, the difference polarity sequence, the type of the input event to which it belongs, and the set of the affected key comparison items. Retrieve other input event intervals with the same input support set and the same preceding stable functional state from the state change record, and align the other input event intervals with the input event intervals corresponding to the boundary fingerprint according to the causal completion mark; Compare the consistent states of the preceding propositions of the incoming edges, the divergence positions of the state transitions, and the sequence of differences in polarity in the aligned intervals one by one; Boundary candidate propositions that maintain the same state transition divergence position during multiple alignments are registered as persistent first difference propositions, and boundary candidate propositions that only share the affected key comparison terms but have different state transition divergence positions are registered as independent first difference branches.

[0016] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. By acquiring state change records, key output records, and observable logic records of two heterogeneous logic circuits driven by the same input event, and converting different physical nodes into common functional propositions to establish semantic correspondence, states with the same functional meaning under different logic encoding methods and logic organization methods can be uniformly compared. By establishing local causal cones based on input events and determining causal completion markers based on the completion state of the preceding proposition, state transition conditions, and stable holding intervals, the key comparison process occurs after both sides' logic states have reached the corresponding completion stage, reducing misjudgments caused by differences in heterogeneous propagation timing. Furthermore, by verifying the difference propagation process, normal propagation differences are distinguished from irreversible logical differences based on whether the difference is recovered, whether it changes subsequent state transitions, and whether it affects the key output path. Forward propagation verification and backward dependency pruning are used to determine the minimum first difference boundary, enabling the diagnostic results to correspond to the earliest point of logical deviation. The above technical means establish a closed loop of functional unification, stage matching, and difference localization around the problem of the inability to directly compare heterogeneous logic states, transforming the comparison process of dual heterogeneous logic circuits from result judgment to state diagnosis based on logical causal relationships.

[0017] 2. By establishing common functional propositions and semantic correspondences, functionally consistent states in different logical structures can be transformed into unified comparison objects, eliminating the need for fixed node names or completely identical internal state representations in judgments between two heterogeneous logic circuits. Combining local causal cones formed by input event indices and causal completion markers ensures that the comparison process only executes on corresponding states that have completed relevant logical propagation, reducing transient inconsistencies caused by different propagation paths within the two heterogeneous logic circuits. When logical differences are detected, the source, direction, recovery location, and associated preceding propositions are recorded through difference propagation descriptions. Furthermore, dependency pruning determines the minimum first-difference boundary, enabling diagnostic results to trace back from critical output anomalies to the functional proposition region that initially caused the irreversible change, reducing interference from irrelevant logical nodes in the diagnostic process.

[0018] 3. By associating composite function descriptors, event genealogies, and boundary fingerprints, the logical state evolution relationship can be maintained under complex feedback paths and overlapping events, avoiding confusion in comparison results caused by state dependencies between input events. Versioning and registering confirmed semantic correspondences, and updating these correspondences based on verification trajectories without irreversible logical differences, ensures that the functional mapping between heterogeneous logics remains consistent throughout the state change process. The construction of difference dependency subgraphs and minimum first-difference boundaries further limits the logical scope requiring further analysis, allowing subsequent diagnosis to focus on anomaly-related regions and reducing redundant analysis of complete logical paths. These technical processes enable the critical comparison and diagnostic process to adapt to different logical implementations, different state propagation rhythms, and complex state transition relationships. Attached Figure Description

[0019] Figure 1 The overall flowchart for key comparison diagnosis of dual heterogeneous logic circuits; Figure 2 This is a diagram for determining irreversibility. Figure 3 A consistency graph of composite migration paths; Figure 4 To restore the boundary determination map; Figure 5 This is a trend chart of confirmed version numbers for pending releases. Detailed Implementation

[0020] In one embodiment, please refer to Figure 1 The key comparison diagnostic method based on dual heterogeneous logic circuits includes acquiring state change records, key output records, and observable intermediate logic records formed by the first and second heterogeneous logic circuits under the same input event sequence. Based on the input support set, output influence set, state transition conditions, and stability interval of each logic quantity, the physical nodes on both sides are converted into common functional propositions and semantic correspondences are established. Using the input event identifier as a unified index, local causal cones pointing to each key comparison item are constructed along the logical dependency relationship, and causal completion markers are determined for the common functional propositions. When the corresponding common functional propositions on both sides reach the causal completion state, key comparisons are performed. The comparison differences are then jointly judged for subsequent recovery, propagation direction, and consistency of the preceding propositions to obtain propagation beat differences or irreversible logical differences. For irreversible logical differences, forward propagation verification and backward dependency pruning are alternately performed to determine the minimum first difference boundary.

[0021] In this embodiment, the first heterogeneous logic circuit and the second heterogeneous logic circuit represent two logic execution entities that implement the same target logic function using different logic decomposition relationships, different state encoding methods, or different condition execution orders. The two logic execution entities can be formed by different program logic, different programmable logic descriptions, or different state machine descriptions. The diagnostic process does not require the two sides to have the same physical node names and the same number of intermediate states. Instead, it requires the two sides to receive input events that can be numbered according to unified event rules and output at least one key comparison item with the same functional meaning. The key comparison item is a logic result that can characterize the target logic function completion state, permitted state, prohibited state, switching state, or protected state. The diagnostic program obtains the logic execution trajectory of the two sides by bypass reading or execution log reading without changing the original input event content and logic execution order.

[0022] Specifically, state change records consist of basic record items consisting of logical quantity identifier, state before change, state after change, triggering event identifier, previous state reference, logical time position, and stable end position. Key output records consist of basic record items consisting of key comparison item identifier, output value, corresponding input event identifier, output generation position, and subsequent holding interval. Observable intermediate logic records include condition judgment results, state transition results, branch selection results, latch results, release results, and intermediate proposition results formed by multiple conditions. The diagnostic program organizes record items triggered by the same input event into event trajectory segments and records state propagation that is not completed within the current event trajectory segment as pending updates, so that logic propagation across events is not mistakenly truncated as an independent execution process.

[0023] In a preferred embodiment, the input event sequence consists of discrete trigger quantities formed by external input changes, internal state feedback changes, or logic execution environment. Each input event is configured with a globally unique event identifier. If multiple input changes are read simultaneously by both circuits at the same logic time position, the multiple input changes are merged into a composite input event and the original order of each component is saved. If the two circuits use different sampling times for the input changes, the local receiving position is established at the moment when the input value enters its respective target logic calculation domain. Then, the cross-side event correspondence is established through the input value content, the reference of the preceding event, and the subsequent state response. This allows the comparison index to be established on the causal source of the input rather than on a fixed clock sequence number.

[0024] In this embodiment, when extracting the input support set for each observable logic quantity, the diagnostic program identifies necessary input conditions, exclusionary input conditions, preceding state conditions, and feedback constraints from the conditional expressions or execution trajectories that generate the current logic quantity change. Necessary input conditions represent conditions that cannot form the current change when missing, exclusionary input conditions represent conditions that prevent the current change when present, preceding state conditions represent existing states that the current change depends on, and feedback constraints represent conditions that are returned from downstream states and participate in the current logic judgment. The output influence set is traced downstream from the current logic quantity along the actual execution dependency edge, recording the key comparison items that can be reached, the types of state transitions traversed, and the relative positions of each key comparison item in the propagation order.

[0025] Preferably, when establishing functional descriptors for the logical quantities on both sides, the candidate semantic correspondence score is calculated using the following formula: ; in, The scores represent the candidate semantic correspondences, where 'a' represents the candidate logic quantity in the first heterogeneous logic circuit, and 'b' represents the candidate logic quantity in the second heterogeneous logic circuit. and These represent the normalized input support sets, and These represent the sets of influence of the normalized output, and These represent the state transition condition sequences, and Let represent the stable interval sequence, J represent the intersection, union, and ratio of sets, C represent the consistency of the transition condition order, and R represent the overlap of stable intervals. , , and The weights are non-negative and their sum is 1.

[0026] When determining specific values, if candidate logic quantity a and candidate logic quantity b have the same necessary input conditions and exclusionary input conditions, then If the two sets of input conditions only share some of the input conditions, the value is determined by the ratio of the number of elements in the intersection to the number of elements in the union. If both sets of input conditions can propagate to the same key comparison term and the propagation order is consistent, then... and If the values ​​are close to 1, and both remain stable with their intervals overlapping under the corresponding input event, then... Take 1, for example, take 1. , , and When the values ​​are 0.3, 0.3, 0.25 and 0.15 respectively, only candidate logic quantities whose input sources, output destinations and transition orders are similar will obtain higher corresponding scores, thus avoiding the establishment of incorrect semantic relationships simply because the output values ​​are the same by chance.

[0027] Furthermore, the diagnostic procedure does not directly determine semantic correspondence based on a single score. Instead, it performs forward consistency checks from input to output and reverse consistency checks from key output to input on candidate pairs. The forward consistency check starts from the normalized input conditions and verifies whether the candidate logic quantities on both sides produce state changes in the same direction under the same input event type and the same preceding functional state. The reverse consistency check starts from the same key comparison item and verifies whether the candidate logic quantities on both sides are located on the actual dependency path that can explain the changes in the key comparison item. Only candidate pairs that pass both checks simultaneously and do not have one-to-many conflicts are registered as direct common functional propositions. Candidate logic quantities that fail the checks enter the compound proposition processing flow.

[0028] In this embodiment, local causal cones are established according to key comparison items. The diagnostic program traverses the actual logical dependency edges in reverse from the output node corresponding to a key comparison item, and includes input events, conditional propositions, state propositions, and intermediate transition nodes that can affect the key comparison item into the same local causal cone. A partial order hierarchy is formed according to the dependency direction. Propositions at lower partial order levels must be completed before propositions at higher partial order levels. If a logical branch only enters the actual execution path of a key comparison item when a specific condition is met, then the logical branch is included in the active causal cone only in the event trajectory segment where the specific condition is met. The unselected candidate branches are retained as inactive dependencies and do not participate in the current completion status determination.

[0029] Preferably, the causal completion flag of the common functional proposition p in the heterogeneous logic circuit q is determined according to the following formula: ; in, For causal completion markers, This is an instruction operation that takes a value of 1 when the condition is true and a value of 0 when the condition is false. This indicates whether all necessary preconditions of the common functional proposition p have acquired stable values. This indicates whether all necessary transition branches related to the common functional proposition p have been selected. This represents the number of pending upstream updates that point to a common functional proposition p and belong to the same input event spectrum.

[0030] When the necessary prerequisites for a common functional proposition are stable and all necessary migration branches have been completed, but there is still one pending upstream update, and Take 1 for all If the value is not 0, the causal completion flag remains 0, and the diagnostic procedure does not perform the corresponding comparison. This continues until the pending upstream update arrives and the state merging is complete. When the causal completion flag becomes 0, the causal completion flag becomes 1. For example, after the current input event, the first heterogeneous logic circuit forms a stable proposition value through two intermediate transitions, and the second heterogeneous logic circuit forms a proposition value with the same function through four intermediate transitions. The diagnostic program waits for the causal completion flags on both sides to become 1 before comparing them. Therefore, it will not identify the intermediate state of the second heterogeneous logic circuit that has not yet completed propagation as a real anomaly.

[0031] In one embodiment, when a common functional proposition in the first heterogeneous logic circuit reaches a causal completion state first, the diagnostic program freezes its stable proposition value, preceding proposition version, associated input event identifier, and completion partial order level. The frozen value is saved as the current comparison candidate value instead of stopping the first heterogeneous logic circuit from continuing execution. If the first heterogeneous logic circuit subsequently receives a new input event and generates a new version proposition value, the new version proposition value is saved separately according to the new event spectrum. When the corresponding common functional proposition in the second heterogeneous logic circuit reaches the same input event identifier and partial order level as the frozen value, the diagnostic program unfreezes the corresponding frozen record and performs a key comparison to avoid cross-comparing proposition values ​​with the same name generated by different input events.

[0032] In this embodiment, the key comparison results include at least four categories: consistency, value inconsistency, completion status inconsistency, and event spectrum inconsistency. Consistency indicates that the corresponding common functional propositions on both sides have the same proposition value under the same input event spectrum and the same functional completion level. Value inconsistency indicates that both sides have completed but the proposition values ​​are different. Completion status inconsistency indicates that one side has completed while the other side has not yet completed within the allowed causal propagation range. Event spectrum inconsistency indicates that the currently available proposition values ​​on both sides come from different input events. Completion status inconsistency and event spectrum inconsistency will not be directly registered as irreversible logical differences, but will continue to enter the pending propagation tracking. Only when a comparable state cannot be formed after the corresponding causal branch is closed will it be transformed into an abnormal candidate.

[0033] Specifically, after detecting a value inconsistency, the diagnostic program records the polarity of the difference, the duration of the event interval, the source input event, the subsequent recovery position, the set of downstream affected propositions, and the consistency state of the preceding propositions. The polarity of the difference is used to describe the direction of the proposition value deviation of the first heterogeneous logic circuit relative to the second heterogeneous logic circuit. The duration of the event interval consists of the position where the difference first appears until both sides re-enter the same functional state or the diagnostic cutoff position. The set of downstream affected propositions consists of propositions that carry the current difference source mark and have undergone state deviation. The consistency state of the preceding propositions is used to determine whether the current difference is inherited from a known difference upstream.

[0034] Preferably, the following formula is used when calculating the irreversible determination quantity for the difference d: ; in, This indicates an irreversible decision quantity. This indicates whether the difference has propagated to the key comparison terms; it is 1 if the difference has propagated to any key comparison term, and 0 otherwise. This indicates whether the difference changes the subsequent state transition branch; a value of 1 indicates a change, and a value of 0 indicates otherwise. This indicates whether the difference crosses the boundary of the current input event and continues to be retained; a value of 1 indicates a crossover, and a value of 0 indicates otherwise. This indicates whether there exists a recovery boundary that satisfies the recovery condition; the value is 1 if it exists, and 0 otherwise. , , and They are respectively represented as , , and The corresponding non-negative weights sum to 1.

[0035] In one example of a value, , , and If we take values ​​of 0.35, 0.3, 0.2, and 0.15 respectively, and a difference does not propagate to the key comparison item, does not change the migration branch, does not cross the input event boundary, and a recovery boundary exists, then... If a difference is set to 0, it is recorded as a propagation beat difference. If a difference has changed subsequent migration branches and crossed the input event boundary, even if the final critical output has not yet been reached, It may also obtain a value greater than that of a simple transient offset. The diagnostic program retains it as an irreversible logical difference candidate and continues to trace its subsequent propagation results along the active causal cone.

[0036] Please refer to Figure 2 , Figure 2 The horizontal axis represents the Chinese segment position of the difference record in the difference propagation queue, corresponding to the front segment, the middle front segment, the middle back segment, and the back segment, respectively. The vertical axis represents the difference retention determination value, ranging from 0 to 1. The curve represents the comprehensive determination result of whether each difference propagates to the key comparison item, whether it changes the subsequent migration branch, whether it continues to be retained across the current input event boundary, and whether there is a recovery boundary. The dashed line and the dotted line represent the review reference threshold and the tracking reference line, respectively.

[0037] from Figure 2 As can be seen, the difference retention determination value is significantly high in multiple positions, with the highest value reaching 1.000. This indicates that these differences simultaneously possess characteristics such as reaching key comparison items, changing migration branches, or being retained across events, and cannot be directly eliminated as ordinary propagation beat differences. Figure 2There are also several low-value points, with the lowest value being about 0.350. This indicates that although there is a local inconsistency in the corresponding differences, the scope of its propagation, branching influence, or recovery relationship is insufficient to prove that it has formed an irreversible logical difference.

[0038] The overall curve shows an alternating distribution of high and low points, indicating that the difference between real anomalies and normal heterogeneous propagation in dual heterogeneous logic circuits cannot be distinguished by comparing a single output value alone. Instead, it requires a joint judgment based on the direction of propagation of the difference, the duration of the event, and the recovery boundary. Figure 2 To provide input for constructing the difference-dependent subgraph: records above the tracking reference line should continue to be analyzed along the active causal cone, while records near the verification reference line should be further confirmed in conjunction with the recovery boundary determination results.

[0039] Furthermore, the forward propagation verification starts from the earliest common functional proposition where inconsistency occurs, and copies the difference source marker to the downstream dependency edge actually driven by the common functional proposition. If the downstream proposition has a different proposition value, branch selection, or completion position than the indifferent trajectory after receiving the difference source marker, the downstream proposition is added to the set of affected propositions. If the downstream proposition has the same proposition value but its predecessor version, branch source, or pending update set is different, the difference source marker is retained and propagation continues. If the downstream proposition becomes consistent again in state value, migration branch, and pending update set, the recovery boundary determination process is entered.

[0040] Backward dependency pruning ends with the affected key comparison terms and the set of affected propositions. It removes dependency branches from the local causal cone that do not lead to any affected proposition. Then, it removes nodes from the remaining dependency branches that only carry consistent preceding propositions and do not participate in the selection of difference branches. For converging propositions with multiple preceding branches, it retains the preceding branches that can explain the current difference polarity and the control branches that have an exclusion relationship with them. Then, it re-performs forward propagation verification on the pruned dependencies. If the new propagation results show that some candidate first difference propositions fail to cover the original affected key comparison terms, it restores the corresponding dependency branches until the set of candidate first difference propositions and their coverage remain stable.

[0041] In a preferred embodiment, establishing semantic correspondence includes segmenting the state change records on both sides according to the arrival order of input events, extracting an input support set for each logical quantity consisting of necessary input conditions, exclusionary input conditions, and preceding states, forming an output influence sequence based on the key outputs that can be reached after the logical quantity changes and the order of propagation, encoding the input support set, output influence sequence, state transition conditions, and stability intervals into functional descriptors, generating candidate pairs between the functional descriptors on both sides, and performing forward consistency checks from input to output and reverse consistency checks from key output to input respectively, retaining candidate pairs that pass both checks and do not have one-to-many conflicts with other candidate pairs that pass both checks, and marking logical quantities that fail the checks as non-directly corresponding propositions.

[0042] In this embodiment, the input support set is represented by functional condition identifiers instead of original variable names. The diagnostic program merges the original input conditions that can be determined to be equivalent through design logic, program branches, or execution trajectories into the same functional condition identifier. The negation conditions are uniformly converted into condition identifiers and polarity combinations, and the preceding states are converted into state function identifiers and state version combinations. The output impact sequence is encoded according to the key comparison items that can be reached, the functional transition categories passed through, and the order of arrival. This enables the first heterogeneous logic circuit implemented with multi-level condition judgments and the second heterogeneous logic circuit implemented with combined condition judgments to be compared at the functional condition level.

[0043] In a preferred embodiment, forward consistency verification uses a set of verification events to drive the execution records of both sides of the logic one by one. The set of verification events includes events that make the candidate proposition true, events that make the candidate proposition remain true, and events that make the candidate proposition revoked. If the candidate logic quantities on both sides form the same functional change direction for the three types of events in the corresponding preceding functional state, then forward consistency verification is passed. Reverse consistency verification extracts the actual dependency path from the key comparison items that the candidate logic quantities can reach. If the candidate logic quantities on both sides are in the same functional position and there is no other candidate logic quantity that can explain the corresponding key output with less dependency deviation, then reverse consistency verification is passed.

[0044] For candidate pairs that simultaneously satisfy two-way verification but generate one-to-many conflicts, the diagnostic program compares the independent change capabilities of multiple candidate logic quantities under different input events. If a single logic quantity on the first side corresponds to multiple continuously changing logic quantities on the second side, and the multiple logic quantities on the second side jointly complete the functional migration represented by the single logic quantity on the first side, then instead of arbitrarily selecting one logic quantity to establish a direct correspondence, the relevant logic quantities are marked as non-direct correspondence propositions and enter the construction of a composite migration path. If multiple candidate logic quantities participate in different key comparison items or have state changes that can be triggered independently, then they are kept as multiple undecided candidate correspondence items to avoid erroneously merging different functional propositions.

[0045] In a preferred embodiment, for non-directly corresponding propositions, multiple logical quantities with continuous dependencies are selected from the state change records on the same side, based on the input support set and key output influence range that can be normalized to the same functional meaning on both sides. A composite migration path is synthesized according to the migration order of the multiple logical quantities. The starting preconditions, ending stable state, path exclusion conditions, and crossed state change intervals of the composite migration path are encoded as composite functional descriptors. On the other side, a single logical quantity or another composite migration path is searched as a composite candidate corresponding item. Event order maintenance verification, state reachability verification, and key output boundary verification are performed on the composite candidate corresponding items. Composite candidate corresponding items that simultaneously meet the three verifications and whose internal intermediate states do not participate in other key comparison items are registered as common functional propositions. The logical quantities inside the path are limited to the subordinate observations of the common functional proposition.

[0046] Preferably, a composite migration path The path consistency value is determined by the following formula: ; in, Indicates path consistency quantity. This represents the composite migration path in the first heterogeneous logic circuit. This represents a single logic path or a compound transition path in the second heterogeneous logic circuit. E represents the event order preservation value, which is 1 when the corresponding order of the input events and the transition order within the path are consistent; otherwise, it is the ratio of the number of transition pairs with the preserved order to the total number of transition pairs. V represents the state reachability consistency value, which is 1 when both sides can reach the corresponding terminal stable state from the corresponding starting pre-state; otherwise, it is the ratio of the number of transition pairs with the preserved order to the total number of transition pairs. Y represents the critical output boundary consistency value, which is 1 when both paths have the same critical output influence boundary. , and They are respectively represented as , and The corresponding non-negative weights sum to 1.

[0047] For example, the first heterogeneous logic circuit completes the establishment of the permitted state with a single state transition, and the second heterogeneous logic circuit completes the same function by sequentially going through three internal logic quantities: condition confirmation, state temporary storage, and permission release. If both sides enter the permitted state from the prohibited state after receiving the same input event, and the order of the three internal logic quantities on the second side remains consistent in each verification trajectory, and the internal logic quantities do not individually affect other key comparison items, then E, V, and Y are all set to 1. The composite transition path is registered as a common functional proposition corresponding to the single state transition on the first side. Subsequent key comparisons are executed for the stable state of the composite path termination, and the three logic quantities inside the path are retained as subordinate observations used during positioning.

[0048] Please refer to Figure 3 , Figure 3 The horizontal axis represents the Chinese segmentation position of the composite candidate path in the path normalization process, corresponding to the front segment, the middle front segment, the middle back segment, and the back segment in sequence. The vertical axis represents the consistency quantity of the composite migration path, with a value ranging from 0 to 1. The curve represents the comprehensive consistency between the composite migration path in the first heterogeneous logic circuit and the single logic path or another composite migration path in the second heterogeneous logic circuit in terms of event order preservation, state reachability, and key output boundaries.

[0049] from Figure 3 As can be seen, the curve starts at a low point, with a minimum value of approximately 0.392, indicating that although the composite candidate path may have some input support or similar termination states, the preservation of event order or key output boundaries is still insufficient. Subsequently, the curve rises rapidly and remains at a high level, reaching a maximum value of 1.000, indicating that subsequent composite paths have strong consistency in starting preconditions, termination stable states, internal migration order, and the scope of influence of key outputs, and can be registered as candidates for common functional propositions.

[0050] Event order preservation verification is completed by comparing the trigger event references and preceding migration references of each logical quantity within the composite migration path. If different events can be inter-inserted within the path, the diagnostic program constructs corresponding event spectrum versions respectively, and does not regard the path formed by inter-inserted events and the path formed by non-inter-inserted events as the same composite functional proposition. State reachability verification is completed by performing reachability search from the initial preconditions along the actual migration edges. Key output boundary verification verifies whether there are independent dependency edges in the logical quantities within the path pointing to other key comparison items. If there are independent dependency edges, the corresponding internal logical quantities cannot be treated as subordinate observations only.

[0051] In a preferred embodiment, constructing a local causal cone and determining causal completion markers includes configuring a consistent event identifier across both sides for each input event, incorporating state transition nodes triggered by the same input event and belonging to the same key comparison item into the corresponding local causal cone, generating a partial order hierarchy for nodes within the cone based on the logical dependency direction, and defining the completion conditions of the corresponding common functional proposition as the preceding proposition having obtained a stable value, all necessary transition branches having been selected, and the key comparison item having no pending upstream updates. The completion positions of the common functional propositions on both sides reaching each partial order hierarchy are recorded respectively. When one side reaches the completion condition first, its stable proposition value is frozen, and the records from the other side continue to be received until the corresponding proposition on the other side reaches the same partial order hierarchy, after which the freeze is lifted and the key comparison is performed. If subsequent input events enter the same local causal cone, the propagation branches of the previous input event that have not yet been completed are separated and marked from the subsequent input events.

[0052] Furthermore, the separation labeling is accomplished through event genealogy. The diagnostic procedure establishes event genealogy identifiers for each input event entering the same local causal cone, and associates the state transition node with the earliest unclosed event genealogy based on the necessary input support set of the state transition node. When a state transition node depends only on the input conditions and preceding states in one event genealogy, a single genealogy reference is directly recorded. When a state transition node depends on multiple unclosed event genealogies, a composite genealogy reference is generated and the contribution relationship of each genealogy to the current transition conditions is saved, so that the state change formed by multiple consecutive input events can be traced to the corresponding input source.

[0053] When the feedback path causes the same state transition node to be driven by multiple event families, the diagnostic program retains the preceding proposition versions of each event family and generates a composite version containing all unclosed families at the confluence node. The composite version records the generation event, stable state, and whether it has been covered by subsequent feedback for each preceding proposition. If an event family only changes the intermediate state but does not participate in the necessary conditions of the confluence node, the propagation reference of that family is closed at the confluence node. If an event family changes the branch selection or stable result of the confluence node, the reference of that family continues to be passed to the downstream proposition.

[0054] Specifically, the set of pending upstream updates is maintained according to the event spectrum. Each pending update item records at least the source proposition, the target proposition, the event spectrum to which it belongs, the expected partial order level, and the current arrival state. When all necessary branches constituting the current common functional proposition are completed and there are no pending upstream updates belonging to the same event spectrum, the corresponding event spectrum is closed. When performing a key comparison, stable proposition values ​​with the same event spectrum identifier and the same preceding proposition version are selected from both sides. The state change records of other unclosed event spectra are retained in the local causal cone, do not participate in the current comparison, and are not deleted.

[0055] In one application example, input event A triggers the first heterogeneous logic circuit to enter the state confirmation process. Before the key comparison item is formed in this process, input event B arrives and changes an exclusion condition. The second heterogeneous logic circuit may first temporarily store the intermediate state formed by event A before processing event B. The first heterogeneous logic circuit may process event A and event B simultaneously in the same conditional expression. The diagnostic program establishes event spectrums for event A and event B respectively, and compares the composite spectrum references at the final convergence of propositions on both sides. The key comparison is only performed after both sides have completed the necessary transitions corresponding to event A and event B, avoiding the registration of normal intermediate differences formed by different event processing orders as the first difference.

[0056] In a preferred embodiment, jointly determining and identifying the minimum first difference boundary includes starting with the earliest common functional proposition where inconsistency occurs, recording the difference polarity, duration of events, subsequent recovery position, downstream affected proposition set, and consistency state of preceding propositions to form a difference propagation description. Downstream propositions that inherit the difference polarity or are deviated from the migration triggered by the difference are marked layer by layer along the local causal cone. When the difference disappears and both sides re-enter the same functional state, a recovery boundary is generated. Differences that do not cross the key comparison item and are eliminated within the recovery boundary are classified as propagation beat differences. Differences that cross the key comparison item, change subsequent migration branches, or do not have a recovery boundary are classified as irreversible logical differences. For irreversible logical differences, preceding branches that do not have a dependency path with the affected propositions are deleted in reverse. Forward marking and reverse deletion are repeated on the remaining branches until the first difference proposition set no longer shrinks.

[0057] In this embodiment, a propagation marker carrying the difference polarity, the source first difference proposition, and the identifier of the input event to which it belongs is established for each inconsistent common functional proposition. The propagation marker is copied along each downstream branch of the local causal cone, and the branch selection change caused by the propagation marker is recorded at the state transition node. If the same difference source reaches the converging proposition through multiple parallel branches, the converging proposition saves all source branches and their proposition value combinations. If different difference sources form the same output value at the converging proposition, the diagnostic program still retains each propagation marker according to the source first difference proposition, the transition branch, and the pending update set to avoid the same output value masking the internal causal differences.

[0058] Preferably, the recovery decision factor for the candidate proposition r at the recovery boundary is expressed by the following formula: ; in, Indicates the recovery determination quantity. and These represent the current functional states of the candidate propositions for restoring the boundary on both sides, respectively. and These represent the sets of transition branch identifiers used when reaching the candidate proposition on both sides. and These represent the two unresolved upstream update sets on either side. This indicates the number of unresolved downstream branches that still carry the same primary difference proposition as the candidate proposition, only when all four indicator conditions are met simultaneously. Take 1 and register the restored boundary.

[0059] For example, if both sides arrive at a logical truth value at a certain converging proposition, but the first heterogeneous logic circuit arrives through a normally permitted branch, while the second heterogeneous logic circuit arrives through an abnormal bypass branch, then... Established Not valid. If the value is set to 0, the propagation marker continues to be retained. If the current functional states, migration branches, and pending update sets on both sides are consistent, but there is still one pending branch carrying the same initial difference source, then... If the value is not 0, the recovery boundary will not be registered. Only when the differences are completely closed in terms of state, path and residual propagation relationship will the corresponding position be used as the recovery boundary.

[0060] Please refer to Figure 4 , Figure 4 The horizontal axis represents the Chinese segment position of the candidate proposition for the recovery boundary in the difference propagation link, corresponding to the front segment, the middle front segment, the middle rear segment, and the rear segment in sequence. The vertical axis represents the recovery closure result, with a value range of 0 to 1. The curve indicates whether both sides simultaneously satisfy four conditions at the candidate confluence proposition: consistent current functional state, consistent arrival migration branch set, consistent pending upstream update set, and no longer carrying the same source first difference proposition. The dashed line and the dotted line represent the unclosed boundary and the closure reference line, respectively.

[0061] from Figure 4 As can be seen, most candidate positions are in the low position, with an average of about 0.176, indicating that even if the two heterogeneous logic circuits are the same again at some intermediate proposition values, they do not necessarily form an effective recovery boundary. Figure 4 Only a few peaks reach 1.000, indicating that the corresponding candidate propositions have closed in terms of state, path source, pending updates, and downstream propagation relationships, and can be registered as positions where differences disappear. Low-level segments indicate that there are still cases where branches have different sources, pending updates are not cleared, or subsequent branches continue to carry the same primary difference source.

[0062] Figure 4 It is emphasized that boundary restoration judgment is not simply a matter of determining whether the proposition values ​​on both sides are equal, but rather a simultaneous check of the arrival paths and residual propagation relationships behind the state values. For Figure 4 For candidate positions that do not reach the closed reference line, the diagnostic procedure should continue to retain the propagation markers and trace them backward to avoid misjudging cases where the surface values ​​are the same but the paths are still different as recovery. A few peak positions provide a termination basis for difference-dependent subgraph clipping and minimum first-difference boundary convergence.

[0063] In a preferred embodiment, propagation beat differences are kept as independent records instead of being directly deleted. The records include the difference between the completion positions on both sides, the partial order levels through which the propagation has passed, the recovery boundary position, and the corresponding input event type. When the same input event type is repeated in the future, the diagnostic program calls the historical propagation beat difference records as a reference to be verified, but does not replace the causal completion determination with a fixed delay value. If the new propagation trajectory is still recovered under the same functional state and the same migration branch, it continues to be registered as a propagation beat difference. If the new propagation trajectory crosses the original recovery boundary or changes the migration branch, it re-enters the irreversible logic difference determination.

[0064] In a preferred embodiment, after the set of first difference propositions no longer shrinks, the common functional propositions with comparative differences in the remaining dependency branches and their dependency edges are used to form a differential dependency subgraph. Common functional propositions that are inconsistent themselves and whose corresponding predecessor propositions are all consistent are selected from the differential dependency subgraph as boundary candidate propositions. The set of affected key comparison items that can be reached by each boundary candidate proposition is calculated, as well as the path interception set from the entry proposition without inconsistent predecessors in the differential dependency subgraph to the affected key comparison items. Boundary candidate propositions whose set of affected key comparison items is completely covered by the upstream boundary candidate propositions and do not have an exclusive path interception relationship are deleted. The retained boundary candidate propositions are used to form proposition anti-chains that do not have upstream and downstream inclusion relationships according to logical dependency relationships. The proposition anti-chains and their entry edge predecessor propositions are jointly registered as the minimum first difference boundary.

[0065] Specifically, the difference dependency subgraph only retains the dependency edges that have been executed in the actual event trajectory. Unselected conditional branches do not enter the current difference dependency subgraph. All corresponding antecedent propositions of the boundary candidate proposition remain consistent, indicating that the corresponding antecedent propositions on both sides have the same functional state under the same event spectrum and the same completion level. If an antecedent proposition itself is inconsistent, the current proposition is only registered as a difference propagation node and not as a boundary candidate proposition. If a proposition does not have any observable antecedent proposition, its input event conditions and previous stable functional states are used as incoming edge antecedent conditions to participate in the boundary judgment.

[0066] The path interception set is formed by enumerating the acyclic dependency paths from the entry proposition to the affected key comparison items in the difference dependency subgraph. If there are feedback edges in the local causal cone, the feedback is first expanded into directed edges at different logical time positions according to the event spectrum and proposition version, so that the difference dependency subgraph corresponding to the current event trajectory remains acyclic. The path interception set of a certain boundary candidate proposition consists of the path identifiers that pass through the candidate proposition. If all key comparison items and all paths that can be reached by the downstream candidate proposition have been covered by the upstream candidate proposition, and the downstream candidate proposition does not have a path intercepted only by itself, then the downstream candidate proposition is deleted.

[0067] The formation of proposition anti-links is determined by dependency reachability. If candidate proposition A can reach candidate proposition B along the difference dependency subgraph, then A and B cannot be the same level of minimum first difference propositions. The diagnostic program retains candidate propositions that can explain more exclusive paths and whose preconditions are consistent. If candidate proposition A and candidate proposition B belong to parallel dependency branches and are mutually unreachable, then both are retained in the proposition anti-link. The minimum first difference boundary formed can represent a single point first difference, multiple parallel first differences, or branch first differences triggered by common input conditions.

[0068] In one application example, a key permitted output of the first heterogeneous logic circuit and the second heterogeneous logic circuit becomes inconsistent. The difference-dependency subgraph contains conditional proposition A, state proposition B, and output proposition C. A is a prerequisite proposition of B, and B is a prerequisite proposition of C. If both sides of A are consistent, while both sides of B are inconsistent for the first time, then B is registered as a boundary candidate proposition. Although C is inconsistent, its prerequisite B is already inconsistent, so C is only an affected node. If there is another state proposition D that is parallel to B and has the same prerequisite condition, it is also inconsistent for the first time, and D affects another key comparison item through an independent path, then B and D together form a proposition anti-chain.

[0069] In a preferred embodiment, the semantic correspondence of common functional propositions is versioned and registered. The currently verified correspondence is saved as the confirmed version, and the newly formed composite candidate correspondence is saved as the pending version. The verification trajectory is extracted from the input event interval where there is no irreversible logical difference and the key outputs on both sides are consistent. The consistency of the input support inclusion relationship, the output influence sequence relationship, and the state transition direction of the pending version is checked in turn. When all three relationships are consistent in the continuous verification trajectory, the pending version is merged into the confirmed version and the effective input event identifier is recorded. When the minimum first difference boundary of the subsequent irreversible logical difference contains the updated common functional proposition, the previous confirmed version is restored, the relevant composite candidate correspondence and its subordinate observations are placed into the isolation set, and the use of the state change record containing the minimum first difference boundary to update the semantic correspondence is stopped.

[0070] In this embodiment, each semantically corresponding version stores at least a version identifier, an effective input event identifier, a corresponding set of logical quantities, a function descriptor, a verification trajectory reference, a reference to the previous confirmed version, and an isolation state. The pending version does not directly replace the confirmed version, but is calculated in parallel with the confirmed version when performing key comparisons. If the pending version and the confirmed version reach different comparison conclusions for the same input event, the confirmed version maintains the current diagnostic benchmark, and the input event trajectory that causes the divergence is added to the verification set of the pending version. The version is only merged when the pending version maintains consistent input support, output influence, and migration direction under multiple different preceding functional states and different input event types.

[0071] Preferably, the number of confirmed cases for the pending version v is calculated according to the following formula: ; Where m represents the number of trajectories used to verify the undetermined versions. This represents the state coverage weight of the j-th verification trajectory. A larger weight is assigned when a new preceding function state or a new input event type is covered. This indicates whether the undetermined version simultaneously satisfies the input support inclusion relationship, the output influence sequence relationship, and the consistency of the state transition direction in the j-th verification trajectory. A value of 1 is assigned if all three relationships are satisfied, and 0 otherwise. This indicates the degree of consistency between the pending version and the verified trajectory.

[0072] For example, a pending version is tested in four verification tracks, where two tracks cover the same preceding state, and the other two tracks cover a new preceding state and a new input event type, respectively. This allows the four tracks to be verified. If we take 1, 1, 2, and 2 in sequence, and the undetermined version satisfies all relationships in the first three trajectories but shows inconsistent migration directions in the fourth trajectory, then... The ratio is 4 to 6. The pending version remains pending and the reason for the inconsistency is recorded. If the minimum first difference boundary of the subsequent irreversible logical difference contains the common functional proposition involved in the pending version, then the update using the relevant trajectory is stopped and the version is returned to the previous confirmed version.

[0073] Please refer to Figure 5 , Figure 5 The horizontal axis represents the Chinese segmentation position of the pending semantic version verification trajectory during the confirmation process, corresponding to the front segment, the middle front segment, the middle rear segment, and the rear segment in sequence. The vertical axis represents the version confirmation quantity, with a value ranging from 0 to 1. The curve represents the cumulative degree to which the pending version satisfies the consistency of input support inclusion relationship, output influence sequence relationship, and state transition direction in the continuous verification trajectory. The dotted line and dashed line represent the incorporation reference line and the review reference line, respectively.

[0074] from Figure 5 As can be seen, the confirmation count remained at 1.000 from the beginning to the middle and later stages, indicating that the pending version simultaneously met all three relationship checks in the early verification trajectory, and still had an evidentiary basis for merging into the confirmed version after the coverage weights were accumulated. The curve dropped to about 0.667 in the later stage, indicating that the subsequent verification trajectory showed inconsistent migration directions or insufficient coverage relationships, causing the cumulative confirmation count to be lower than the merging reference line but still higher than the review reference line.

[0075] Figure 5The versioning process for semantic correspondences demonstrates that this invention does not accept newly formed composite candidate correspondences all at once, but rather confirms them step by step based on verification trajectories without irreversible logical differences. When the curve remains above the merging reference line, the pending version can be merged into the confirmed version; when the curve declines or the subsequent minimum first difference boundary contains the common functional propositions involved in the pending version, the absorption of relevant trajectories should be stopped and the previous confirmed version should be returned to avoid abnormal states being mistakenly learned as new normal correspondences.

[0076] The isolation set is used to store logical quantities, composite migration paths, and verification trajectories that may be contaminated by real anomalies. Objects entering the isolation set can still be read as evidence for anomaly localization, but they do not participate in the learning of new semantic correspondences and version confirmation. Only when the corresponding logical region in the subsequent input event trajectory continuously satisfies the input support, output influence, and migration direction constraints of the confirmed version, and there is no propagation marker carrying the original first difference source, will the relevant object be removed from the isolation set and reprocessed as a pending candidate, thereby preventing the fault state from being absorbed into a new normal semantic relationship.

[0077] In a preferred embodiment, a boundary fingerprint is extracted for each boundary candidate proposition in the minimum first difference boundary. This fingerprint consists of the set of incoming edge preceding propositions, the state transition relationship where a divergence occurs, the difference polarity sequence, the type of input event to which it belongs, and the set of affected key comparison terms. Other input event intervals with the same input support set and the same preceding stable functional state are retrieved from the state change record. Based on the causal completion marker, the other input event intervals are partially aligned with the input event intervals corresponding to the boundary fingerprint. The consistent state of incoming edge preceding propositions, the state transition divergence position, and the difference polarity sequence in the aligned intervals are compared one by one. Boundary candidate propositions that maintain the same state transition divergence position in multiple alignments are registered as persistent first difference propositions. Boundary candidate propositions that only share affected key comparison terms but have different state transition divergence positions are registered as independent first difference branches.

[0078] Specifically, the set of preceding propositions in the boundary fingerprint is encoded according to the functional proposition identifier, proposition polarity, and proposition version; the state transition relationship is encoded according to the functional state before the transition, the trigger condition category, the selected branch, and the functional state after the transition; the difference polarity sequence records the relationship between the proposition values ​​on both sides according to the local causal cone partial order hierarchy; the type of the input event is encoded according to the combination of input conditions and the category of its preceding events; the set of affected key comparison items is encoded according to the key comparison item identifier and the first affected level; the same boundary fingerprint does not require the two trajectories to have the same physical time length, but requires the functional causal order to remain consistent.

[0079] Partial order alignment uses the causal completion position of common functional propositions as anchor points to map event trajectory segments with the same input support set and the same preceding stable functional state to a unified partial order level. If one side of the trajectory contains additional intermediate logic, the additional intermediate logic is assigned to the local path interval between adjacent common functional propositions without changing the sequential relationship between common functional propositions. If a trajectory lacks the necessary preceding proposition required for the boundary fingerprint or has not reached the corresponding key comparison item, it is not considered a complete alignment sample and is only retained as a local verification record.

[0080] In multiple alignments, if the same boundary candidate proposition continues to show the same state transition divergence position under the same preceding functional state and the same input event type, even if the subsequent affected key comparison item changes due to different input conditions, the boundary candidate proposition will be registered as a persistent first difference proposition. If different event trajectories all lead to the same key comparison item anomaly, but their earliest transition divergence is located in proposition branches that are not dependent on each other, they will be registered as independent first difference branches. The diagnostic record saves the correspondence between each independent first difference branch and the input event conditions, and does not merge multiple different first difference sources into a single abnormal position.

[0081] In a preferred embodiment, the diagnostic program uses a graph data structure to store common functional propositions, logical dependency edges, event genealogies, difference propagation markers, and semantic correspondence versions. Each common functional proposition node stores references to the physical logical quantities on both sides, the current proposition value, the stable interval, the causal completion marker, and the version to which it belongs. Each logical dependency edge stores the source proposition, the target proposition, the triggering condition, the exclusion condition, the event genealogy to which it belongs, and the activity state. When an event trajectory arrives, only the local causal cone affected by the current input event is updated, avoiding repeated traversal of all logical relationships.

[0082] In this embodiment, if there are missing state change records, the diagnostic program marks the corresponding interval as an incomplete trajectory based on the previous stable state, the next stable state, and the observable dependent edges. It does not establish new semantic correspondences for missing propositions in the incomplete trajectory, nor does it directly register the missing position as the first difference proposition. If the causal completion status of the corresponding common functional propositions on both sides can be determined based on the complete preceding records, the key comparison can still be performed. If it is impossible to determine whether the pending upstream update has been cleared due to missing records on either side, the corresponding comparison will remain in a pending state and will continue to be processed when the complete trajectory is restored later.

[0083] If the same input event generates different numbers of state change records on both sides, the diagnostic program performs semantic merging based on the input support set, state transition direction, and key output influence range, without supplementing according to the number of records. If multiple consecutive state changes on one side constitute a composite functional proposition, the final stable state of the composite transition path is used as the comparison value. If there are internal maintenance state changes on one side that are unrelated to the target function, they are marked as inactive dependent observations. They are only included in the current local causal cone when the maintenance state change alters the actual dependency path of the key comparison item.

[0084] In a complete running example, two heterogeneous logic circuits receive the same permitted input event. The first heterogeneous logic circuit sequentially forms a conditional confirmation proposition, a permitted state proposition, and a critical permitted output. The second heterogeneous logic circuit forms a conditional combination proposition, a temporary proposition, a release proposition, and a critical permitted output. The diagnostic program establishes a common functional proposition between the permitted state proposition on the first side and the composite migration path composed of the temporary proposition and the release proposition on the second side based on the input support set and the output influence set. Then, it constructs a local causal cone based on the event identifier and compares the proposition values ​​after both common functional propositions on both sides have reached the causal completion state.

[0085] If the temporary proposition in the second heterogeneous logic circuit is completed later than the permitted state proposition in the first heterogeneous logic circuit, but both sides enter the same permitted state after the subsequent release proposition arrives, and the transition branch and the set of pending updates are consistent, then the difference closes within the recovery boundary and is registered as a propagation beat difference. If the release proposition in the second heterogeneous logic circuit fails to hold due to an error in interpreting the exclusion condition, and the difference continues to propagate to the critical permitted output and changes the subsequent state transition branch, then it is registered as an irreversible logic difference. The diagnostic procedure prunes in reverse along the difference dependency subgraph, retaining the exclusion condition proposition, the release proposition, and the affected output path.

[0086] When the exclusion condition proposition remains consistent on both sides while the release proposition first shows inconsistency, the release proposition becomes a boundary candidate proposition. If it can capture all the difference paths from the entry proposition to the key permitted output, and there is no downstream candidate proposition with an exclusive path, then the release proposition and its preceding entry edge proposition form the minimum first difference boundary. When the trajectories of the same input support set and the preceding permitted state show the same release migration divergence again through partial order alignment, the release proposition is registered as a persistent first difference proposition.

[0087] Through the above implementation, the physical nodes, composite migration paths, and key outputs in the dual heterogeneous logic circuit are uniformly organized into functional propositions with clear input support, state transitions, and output influence relationships. Key comparisons are performed based on the input event spectrum and causal completion state. Differences are classified based on propagation results, recovery boundaries, and prior consistent states. Irreversible logical differences converge to the minimum first difference boundary through difference dependency subgraphs, path interception sets, and proposition anti-chains. Semantic correspondences maintain update boundaries through confirmed versions, pending versions, and isolation sets. Boundary fingerprints distinguish between persistent first difference propositions and independent first difference branches through partial order alignment.

Claims

1. A key comparison diagnostic method based on dual heterogeneous logic circuits, characterized in that, include: Acquire the state change records, key output records, and observable intermediate logic records formed by the first heterogeneous logic circuit and the second heterogeneous logic circuit under the same input event sequence; Based on the input support set, output influence set, state transition conditions, and stability interval of each logical quantity, the physical nodes on both sides are converted into common functional propositions and semantic correspondences are established. Using input event identifiers as a unified index, local causal cones pointing to each key comparison item are constructed along logical dependencies, and causal completion markers are determined for common functional propositions; The critical comparison is performed when both sides of the corresponding common functional proposition have reached the causal completion state. The differences in comparison are then jointly assessed by subsequent recovery, propagation direction, and consistency of the preceding propositions to obtain differences in propagation rhythm or irreversible logical differences. For the irreversible logical differences, forward propagation verification and backward dependency pruning are performed alternately to determine the minimum first difference boundary.

2. The key comparison diagnostic method based on dual heterogeneous logic circuits according to claim 1, characterized in that, Establishing the semantic correspondence includes: dividing the state change records on both sides according to the arrival order of the input events, extracting the input support set consisting of necessary input conditions, exclusionary input conditions and previous states for each logical quantity, and forming an output influence sequence based on the key outputs that can be reached after the logical quantity changes and the order of propagation. The input support set, the output influence sequence, the state transition condition, and the stability interval are encoded as functional descriptors; Candidate pairs are generated between the two function descriptors, and forward consistency checks from input to output and reverse consistency checks from critical output to input are performed respectively. Retain candidate pairs that pass both-way checks and do not have one-to-many conflicts with other candidate pairs that pass both-way checks, and mark logic quantities that fail checks as non-directly corresponding propositions.

3. The key comparison diagnostic method based on dual heterogeneous logic circuits according to claim 1, characterized in that, Constructing the local causal cone and determining the causal completion marker includes: configuring a consistent event identifier across both sides for each input event, and incorporating state transition nodes triggered by the same input event and belonging to the same key comparison item into the corresponding local causal cone; A partial order hierarchy is generated for nodes within the cone based on the logical dependency direction, and the conditions for the completion of the corresponding common functional proposition are defined as: the preceding proposition has obtained a stable value, all necessary migration branches have been selected, and there are no pending upstream updates for the key comparison item. Record the completion positions of the common functional propositions on both sides at each partial order level; When one side meets the completion condition first, its stable proposition value is frozen, and the other side continues to receive records until the corresponding proposition on the other side reaches the same partial order level, at which point the freeze is lifted and the key comparison is performed. If subsequent input events enter the same local causal cone, the propagation branch of the previous input event that has not yet been completed is separated from the subsequent input event and marked.

4. The key comparison diagnostic method based on dual heterogeneous logic circuits according to claim 1, characterized in that, The joint determination and determination of the minimum first difference boundary includes: starting from the earliest common functional proposition inconsistency, recording the difference polarity, the duration of the event interval, the subsequent recovery position, the set of downstream affected propositions, and the consistency status of the preceding propositions to form a difference propagation description; Downstream propositions that inherit differential polarity or are triggered by differential migration deviation are marked layer by layer along the local causal cone, and a recovery boundary is generated when the differential disappears and both sides re-enter the same functional state. Differences that do not cross the key comparison item and are eliminated within the recovery boundary are classified as propagation beat differences, while differences that cross the key comparison item, change subsequent migration branches, or do not exist at the recovery boundary are classified as irreversible logic differences. In response to the aforementioned irreversible logical differences, reverse deletion is performed on the preceding branches of the affected propositions that do not have dependent paths, and forward marking and reverse deletion are repeated on the remaining branches until the set of the first difference propositions no longer shrinks.

5. The key comparison diagnostic method based on dual heterogeneous logic circuits according to claim 2, characterized in that, For the aforementioned non-directly corresponding propositions, based on the input support set and key output influence range that can be unified to the same functional meaning on both sides, multiple logical quantities with continuous dependencies are selected from the state change records on the same side, and a composite migration path is synthesized according to the migration order of the multiple logical quantities. The starting preconditions, ending stable state, exclusion conditions within the path, and the state change intervals crossed by the composite migration path are encoded as composite function descriptors. On the other side, a single logical quantity or another composite migration path is searched as a composite candidate corresponding item, and event order maintenance verification, state reachability verification and critical output boundary verification are performed on the composite candidate corresponding item. Composite candidate items that simultaneously satisfy all three checks and whose internal intermediate states do not participate in other key comparison items are registered as common functional propositions, and the internal logic quantities of the path are limited to subordinate observations of the common functional propositions.

6. The key comparison diagnostic method based on dual heterogeneous logic circuits according to claim 3, characterized in that, The separation markers include: establishing event lineage identifiers for each input event entering the same local causal cone, and associating the state transition node with the earliest-arrived event lineage that has not yet closed, based on the necessary input support set of the state transition node; When the feedback path causes the same state transition node to be driven by multiple event families, the preceding proposition versions of each event family are retained, and a composite version containing all unclosed families is generated at the merging node. Maintain a set of unresolved upstream updates based on the propagation position of each event lineage in the partial order hierarchy. When all necessary branches constituting the current common functional proposition have been completed and there are no unresolved upstream updates belonging to the same event lineage, close the corresponding event lineage. When performing the key comparison, stable proposition values ​​with the same event spectrum identifier and the same preceding proposition version are selected from both sides, and the state change records of other unclosed event spectrums are retained in the local causal cone.

7. The key comparison diagnostic method based on dual heterogeneous logic circuits according to claim 4, characterized in that, The generation of the recovery boundary includes: establishing a propagation marker for each inconsistent common functional proposition, carrying the difference polarity, the source first difference proposition and the identifier of the input event to which it belongs; copying the propagation marker along each downstream branch of the local causal cone; and recording the branch selection change caused by the propagation marker at the state transition node. When multiple downstream branches obtain the same proposition value again at the confluence proposition, check the current state of the confluence propositions on both sides, the selected migration branches, and the set of pending upstream updates respectively; The converging proposition is registered as the recovery boundary only when the current state, the selected migration branch, and the set of pending upstream updates are all consistent, and there are no more pending downstream branches carrying the source first difference proposition in the subsequent partial order level; For converging propositions with the same proposition value but different migration branches or pending upstream update sets, the propagation marker is retained and propagated to subsequent key comparison terms.

8. The key comparison diagnostic method based on dual heterogeneous logic circuits according to claim 4, characterized in that, After the set of first difference propositions no longer shrinks, the common functional propositions with the comparison differences in the remaining dependency branches and their dependency edges are used to form a difference dependency subgraph. From the difference-dependency subgraph, select common functional propositions that are inconsistent with themselves and whose corresponding predecessors are consistent as boundary candidate propositions, and calculate the set of affected key comparison items that can be reached by each boundary candidate proposition, as well as the set of path interception from the entry proposition with no inconsistent predecessors in the difference-dependency subgraph to the affected key comparison items. Delete the boundary candidate propositions whose affected key comparison item set is completely covered by upstream boundary candidate propositions and do not have exclusive path interception relationships; The retained boundary candidate propositions are used to form proposition antichains that do not have upstream or downstream inclusion relationships according to their logical dependencies, and the proposition antichains and their incoming edge preceding propositions are jointly registered as the minimum first difference boundary.

9. The key comparison diagnostic method based on dual heterogeneous logic circuits according to claim 8, characterized in that, The semantic correspondence of the common functional propositions is versioned and registered. The currently verified correspondence is saved as the confirmed version, and the newly formed composite candidate correspondence is saved as the pending version. Extract the verification trajectory from the input event interval where there is no irreversible logical difference and the key outputs on both sides are consistent, and verify the input support inclusion relationship, output influence sequence relationship and state transition direction consistency of the version to be determined in turn. When all three relationships remain consistent in the continuous verification trajectory, the pending version is merged into the confirmed version and the effective input event identifier is recorded; When the minimum first difference boundary of subsequent irreversible logical differences contains the updated common functional proposition, revert to the previous confirmed version, place the relevant composite candidate correspondence and its subordinate observations into the isolation set, and stop using the state change record containing the minimum first difference boundary to update the semantic correspondence.

10. The key comparison diagnostic method based on dual heterogeneous logic circuits according to claim 8, characterized in that, For each boundary candidate proposition in the minimum first difference boundary, extract the boundary fingerprint composed of the set of incoming edge preceding propositions, the state transition relationship where the divergence occurs, the difference polarity sequence, the type of the input event to which it belongs, and the set of the affected key comparison items. Retrieve other input event intervals with the same input support set and the same preceding stable functional state from the state change record, and align the other input event intervals with the input event intervals corresponding to the boundary fingerprint according to the causal completion mark; Compare the consistent states of the preceding propositions of the incoming edges, the divergence positions of the state transitions, and the sequence of differences in polarity in the aligned intervals one by one; Boundary candidate propositions that maintain the same state transition divergence position during multiple alignments are registered as persistent first difference propositions, and boundary candidate propositions that only share the affected key comparison terms but have different state transition divergence positions are registered as independent first difference branches.

Images