Patent Application
20070179970
It is sometimes necessary to format an electronic data set in different ways. One way to format a data set in different ways is by means of a plurality of data formatters, each of which is configured to format the data set in a different way, and each of which accesses the data set in series (i.e., when one of the data formatters is done formatting the data, another of the data formatters then accesses the data). Alternately, each of the data formatters may access the data set in parallel. Or, the data set can be replicated, and each of the data formatters can access a different replica of the data set.
When a plurality of data formatters access a data set in series, it takes longer to format the data set than if the data formatters access the data set in parallel. However, when data formatters access a data set in parallel (or access replicas of the data set in parallel), memory, storage and/or processing resources are often used inefficiently, and memory, storage and/or processing limitations can be neared or reached.
In one embodiment, a method comprises, in response to a plurality of events, 1) creating a plurality of data objects in memory, 2) relating data corresponding to ones of the events to ones of the data objects, and 3) storing the data in the memory. At least some of the data objects correspond to logical groupings of data implied by ones of the events. During or after creation of the data objects, a number of data formatters are provided access to the plurality of data objects and data in the memory.
In another embodiment, apparatus comprises computer readable code stored on computer readable media. The computer readable code includes code to, in response to a plurality of events, 1) create a plurality of data objects in memory, 2) relate data corresponding to ones of the events to ones of the data objects, and 3) store the data in the memory. At least some of the data objects correspond to logical groupings of data implied by ones of the events. The computer readable code also includes code to provide a number of data formatters access to the plurality of data objects and data in the memory.
In yet another embodiment, a method comprises, in response to an ordered sequence of events corresponding to execution of a plurality of tests on at least one device under test (DUT), creating a plurality of data objects, and relating ones of the data objects to others of the data objects in a hierarchical tree structure. At least some of the data objects correspond to logical groupings of test results implied by ones of the events. The method further comprises 1) relating data corresponding to ones of the events to ones of the data objects in the hierarchical tree structure, the data including test results, and 2) providing a number of data formatters access to the plurality of data objects and data in the hierarchical tree structure.
In still another embodiment, apparatus comprises computer readable code stored on computer readable media. The computer readable code includes code to, in response to an ordered sequence of events corresponding to execution of a plurality of tests on at least one device under test (DUT), 1) create a plurality of data objects, 2) relate ones of the data objects to others of the data objects in a hierarchical tree structure, and 3) relate data corresponding to ones of the events to ones of the data objects in the hierarchical tree structure, the data including test results. At least some of the data objects correspond to logical groupings of test results implied by ones of the events. The computer readable code also includes code to provide a number of data formatters access to the plurality of data objects and data in the hierarchical tree structure.
Other embodiments are also disclosed.
Illustrative embodiments of the invention are illustrated in the drawings, in which:
As a preliminary manner, it is noted that, in the following description, like reference numbers appearing in different drawing figures refer to like elements/features. Often, therefore, like elements/features that appear in different drawing figures will not be described in detail with respect to each of the drawing figures.
To improve the manner in which data is formatted by a plurality of data formatters,
At least some of the data objects created by the method 100 correspond to logical groupings of data that are implied by ones of the events (e.g., a wafer object, a device object, and so on). As used in this description, “implied” groupings can be those that are specifically mentioned, or those that are only inferred. In any case, it is preferable that the logical groupings correspond to real-world groupings that can be understood by an engineer or user that is tasked with formatting data (or tasked with creating a data formatter for formatting the data).
After creating one or more of the data objects, data corresponding to ones of the events is related to ones of the data objects (at step 104) and stored in the memory. During or after creation of the data objects, a number of data formatters (i.e., one or more data formatters) is provided access to the plurality of data objects and data in the memory. See step 106.
In one embodiment, the method 100 may include steps for formatting data. The steps include, for and via each of a number of active data formatters, accessing ones of the data objects and data stored in the memory (at step 108), formatting the data in accord with rules maintained by the data formatter (at step 110), and writing the formatted data to a file (at step 112). For purposes of this description, an “active” data formatter is one of a number of “available” data formatters that a user has selected to format a particular data set. If the method 100 is applied to circuit test, one or more of the data formatters may write data as a number of test records.
It is noted that the order of the method steps shown in
The method 100 shown in
In one embodiment, the computer readable code that implements the method 100 may cause the functional units (or processes) shown in
The data populator 200 1) receives a plurality of events 210, 2) creates a plurality of data objects in memory 212, 3) relates data corresponding to ones of the events to ones of the data objects, and 4) stores the data in the memory 212. At least some of the data objects created by the data populator 200 correspond to logical groupings of data that are implied by ones of the events received by the data populator 200. For purposes of this disclosure, “implied” groupings include those that are specifically “expressed”.
The number of data formatters 202, 204, 206, 208 access ones of the data objects, and then retrieve and format the data that is related to the data objects.
In one embodiment, computer readable code may further instantiate or run a notification manager 214. The notification manager 214 may receive indications of events from the data populator 200, and in response to these indications, may provide notifications of ones of the events to one or more of the data formatters 202, 204, 206, 208. The data formatters 202, 204, 206, 208 may then be configured to initiate their access of data objects (and retrieval of data) in response to the notifications. Note that, in some cases, the indications of events received by the notification manager 214 may correspond to fewer or different events than those that are received by the data populator 200. For example, in the case of circuit test, the data populator 200 may receive indications of test setup events that are not deemed sufficiently important to pass to the notification manager 214, and that are not used as a basis for creating data objects. Also, there may be events that the data populator 200 infers from the events it receives. For example, based on a change in part numbers or other indicia, the data populator 200 might infer that a new “lot” of devices is being tested, and then provide an indication of such event to the notification manager 214 (i.e., even though the data populator 200 itself might not receive a new “lot” event).
The method 100 and apparatus disclosed above can be used in many applications, one being the storing and formatting of test results such as circuit test results. In one particular application, the test results may be generated by a 93000 SOC Series tester offered by Agilent Technologies, Inc.
The 93000 SOC Series tester (hereinafter referred to as the “93000 tester”) is an SOC (System On a Chip) tester that logs test results and events to a binary data file known as an EDL (Event Data Logging) file. The events in this EDL file correspond to the execution of a plurality of tests on at least one device under test (DUT), and are stored in an ordered sequence. However, the events stored in the EDL file are not “thrown” to any other process, and are merely logged to the EDL file. In such an application, the method 100 shown in
In one embodiment, the EDL file is parsed in response to method calls made by an event manager. As shown in
The shared library 302 may take the form of compiled code, such as a data retrieval library (DRL), that executes a method or methods when called by the event manager 300.
The data objects that are created in memory 212 may be related to one another in various ways. However, in one embodiment, they are related to one another in a hierarchical tree structure. Data objects that are children of other data objects may maintain pointers to their parent data objects, but parent objects need not maintain a list of pointers to all of their children. As will be explained later in this description, these pointers from children to their parents can aid the process of deleting data objects that are no longer needed.
In the case of data objects that are based on the execution of a plurality of circuit tests, the logical groupings of data (e.g., groupings of test results) that are implied by the events may comprise one or more hardware groupings, such as groupings for lots, wafers and DUTs, and one or more test groupings, such as groupings of test results corresponding to tests and subtests.
Data may be related to data objects in various ways, including, by 1) directly storing data within the data objects, or 2) storing data in data structures that are related to the data objects (e.g., related by pointers or other means).
In an EDL file 304, data is stored as attributes of events. Thus, if the data populator 200 receives events drawn from an EDL file 304, the data populator 200 may extract data corresponding to events by extracting the data from attributes of the events. In the case of circuit test, the extracted data may comprise test results.
By way of example,
In addition to storing individual data items in the memory 212, the data populator 200 may accumulate data statistics, such as test data statistics, and then relate the data statistics to ones of the data objects (e.g., by storing the data statistics within the data objects, or by storing the data statistics in data structures that are related to the data objects). In one embodiment, the data statistics may be accumulated by the data populator 200 and then related to a data object when fully compiled. In an alternate embodiment, incomplete data statistics may be related to data objects and then updated. Similarly to data statistics, the data populator 200 may accumulate data interpretations, such as test data interpolations, and relate them to ones of the data objects.
The choice of whether to have the data populator 200 or individual data formatters 202, 204, 206, 208 compile statistics and/or interpret data may be based on the usefulness of the statistic or interpretation in different data formats. That is, if a statistic or interpretation is likely to be needed by multiple data formatters 202, 204, 206, 208, it is often best to have the statistic or interpretation compiled once by the data populator 200. On the other hand, statistics and interpretations that are format-specific may best be compiled by a particular one of the data formatters 202, 204, 206, 208.
There are many ways in which the data formatters 202, 204, 206, 208 may access the data objects and data that are created by the data populator 200. In one embodiment, the data formatters 202, 204, 206, 208 can simply monitor the data objects. However, this can require a lot of memory bandwidth, and is often not very efficient. In a preferred embodiment, the data populator 200 generates pointers to the data objects that it creates, and then passes the pointers to the notification manager 214. The notification manager 214 then distributes ones of the pointers to ones of the data formatters 202, 204, 206, 208.
It is noted that that the notification manager 214 need only receive one pointer to each data object. The notification manager 214 can then replicate the pointer, or broadcast it, to each of the data formatters 202, 204, 206, 208. Alternately, the notification manager 214 may not replicate or broadcast the pointer for/to each of the data formatters 202, 204, 206, 208, and may only replicate or broadcast the pointer for/to certain ones of the data formatters 202, 204, 206, 208 that have subscribed for a type of data that is represented by the data object.
Typically, the operation of the data populator 200 will be given priority over the operation of the data formatters 202, 204, 206, 208. To further control access to the memory 212 in which the data objects and data are stored, the data populator 200 (or other process that creates the plurality of data objects and stores data in the memory 212) may be provided with direct access to the data objects and the data. However, the number of data formatters 202, 204, 206, 208 may be provided with access to the plurality of data objects via a structured interface that coordinates/arbitrates the data formatters' access to the data objects and data.
To reduce the number of data objects that is maintained in the memory 212, the data populator 200 may monitor reservations of the data objects, and may delete data objects that are no longer reserved by any object or process that references or accesses the data objects. In one embodiment, a data object is considered reserved when a pointer is generated to reference the data object. The data populator 200 may generate one such pointer upon creation of the data object, and the notification manager 214 may generate and distribute additional copies of this pointer to the data formatters 202, 204, 206, 208. The data populator 200 may also generate a pointer to an object when a child object of the data object is created. As pointers are generated, a count of the number of pointers that reference a particular data object may be maintained (possibly within the data object itself). The data formatters 202, 204, 206, 208 and other processes can then be programmed to relinquish their pointer to a data object when they have finished accessing the data object, and the data populator 200 can delete the data object when all such pointers have been released. Due to child objects referencing their parents, a parent object cannot be deleted until all of its children are first deleted.
In light of the above teachings,
The data formatters shown in
Although the hierarchical tree structure that is created by the method 600 is not required to be stored in memory 212, it is noted that is most always will be advantageous to do so, as this significantly speeds data creation/retrieval processes.
Depending on their implementation, the data models disclosed herein (i.e., where data objects are created based on logical groupings of data, and/or where data objects are related to one another in a hierarchical tree structure) can provide a number of advantages—especially in the context of circuit test. For example, by providing a separate process to organize data in memory, and then making the data available to a plurality of data formatters, much of the overhead of having to read and organize data is removed from the data formatters; and, by creating data objects “in memory”, they can be accessed more quickly than data stored on disk. This enables the code of the data formatters to be lighter weight, and also provides a standard data model that can be leveraged by new data formatters. It also 1) enables easier revisions and patches to the data populator and individual data formatters, 2) enables parallel and faster coding of each of these functional units, and 3) reduces the likelihood of error in coding each functional unit (e.g., because each is lighter weight, and there is less code to test for each unit). Further, it enables a tester to generate data in a form that is more efficient to generate in a run-time test environment, while making test data available to data formatters (and the authors thereof) in a form that is more user-friendly and logically compartmentalized.
The data models disclosed herein, and the manner in which data is populated and removed from them, also help to balance 1) the speed with which a plurality of data formatters can format a common data set (i.e., they can each format the data of the data set in parallel), with 2) the real-world limitations on memory, storage and/or processing resources. In the realm of circuit test, and particularly when conducting parametric tests using an SOC tester, so much data is generated that resource limitations of even a well-provisioned computer system can be reached after testing as few as 10-20 chips. However, by using the data models disclosed herein, and removing data objects that are no longer needed, resource limitations can typically be avoided.
