The present invention relates generally to computer software compilers. More particularly, the present invention relates to a XLANG/s compilation method that performs static data flow analysis of program code at compile time. Even more particularly, the present invention relates to such a method that automatically determines data object create, lock and destroy points at compile time by way of such static data flow analysis.
A business may use a web service to interact with other entities via a distributed environment, such as the Internet or World Wide Web. To ensure that such an interaction is accomplished successfully, one or more protocols must be in place for carrying messages to and from participants, and specific business applications must also be in place at each participant's end. Such interactions are message driven. For example, a buyer sends a purchase order to a seller. The seller then checks its inventory to determine if it can provide the ordered items. If so, the seller sends an acknowledgement back to the buyer with a price. Finally, the buyer accepts or rejects the seller's offer (and/or possibly places another order). As evident in this example, each participant's business application reacts to the receipt of messages.
Web services use computer programs based on software code to implement message-based business transactions such as those described above. Such software code is in turn implemented in a programming language. Programming languages typically allow a programmer to declare one or more data objects, such as a variable, that hold the state of the program.
In a message-oriented language that is capable of implementing the business transactions discussed above, a data object may be a message that is received as input from some outside source. Such a message may be quite large, and therefore the storage of multiple data objects may consume a great deal of memory resources. Thus, a data object should be deleted, or “destroyed,” once the data object is no longer needed for the proper operation of the software. Conventionally, destroy points, which are code instructions to destroy one or more data objects, are manually inserted into the software code by a programmer. A programmer will usually select the particular data object to be destroyed and the location within the code where such deletion is to occur. Typically, a programmer will attempt to place the destroy point immediately after the last use of the data object. In such a manner, the data object is retained in memory for as brief a duration as possible.
As a result of being performed manually, such a conventional method of destroying data objects therefore has a potential for human error in the placement of the destroy points and the selection of data objects to be deleted. An additional shortcoming is present in the context of a program that is implementing a message-based transaction. For example, in such a program, messages may be interrupted or other errors may occur that invoke special, “compensation” software components that may resend a message or may attempt some other process to rectify, or compensate for, the error. Compensation components add complexity to the software code, thereby making the programmer's task of selecting the proper location within the code where a destroy point should be located even more difficult. In addition, program code for message-based transactions may be especially complex, with many operations being processed in parallel, with the possibility that some of the operations may not complete or may invoke a compensation component at runtime. Therefore, selecting a location for a destroy point in such a situation is difficult, as every possible logic path that may be taken during processing must be accounted for.
The above description of the shortcomings of conventional placement of destroy points applies equally to “create” and “lock” points, as well as to any other type of data object handling operations. A create point is a location within software code where a data object is created. Ideally, a create point should be placed immediately before the data object is first used, again to reduce the consumption of memory resources. A lock point is a location within software code where a set of shared objects, such as a variable, is made inaccessible to program components that are not involved with a current transaction. In this way, the value(s) assigned to the data object cannot be changed by an outside component during the transaction, which therefore avoids errors during processing. Like create points, lock points should be placed as close to the use of a data object as possible, to avoid locking the variable unnecessarily. However, in software program that implements message-based transactions, an improper placement of a lock point may, in the event of a message error or another type of error, cause the program to have faults or to even stop operating. In complex software, it becomes increasingly difficult for a programmer to keep track of all the possible logic paths and attendant compensation components to place lock points at an optimal location: as close as possible to the use of a data object without being in a location that could cause faults if an error occurs.
As should be known to those skilled in the art, a compiler converts the high-level code used by programmers into a machine-readable language that can be more quickly executed at run time. At compile time, a compiler maps out the different possible logic paths that may be taken by the program to create a flowgraph of such logic paths. The flowgraphs are used for semantic analysis of program correctness and to generate the most efficient code.
A conventional compiler lacks the functionality to analyze a program, and/or produce flowgraphs, in such a way as to be able to determine the optimal placement of create, lock and/or destroy points. In addition, a conventional compiler lacks the ability to insert such points into the program so the corresponding data objects will be created, locked and/or created at the appropriate points at run time.
Accordingly, what is needed is a method for analyzing software, such as XLANG/s, that is capable of implementing message-based transactions in such a way as to be able to determine appropriate points for data object operations. More particularly, what is needed is a method for creating a flowgraph of a program that is capable of supporting analysis of the program's handling of data objects. More particularly, what is needed is a method of compiling computer code that involves generating a flowgraph and performing static data flow analysis on such a flowgraph to determine appropriate create, lock and destroy points for data objects in the program. Even more particularly, what is needed is a compiler that performs such an analysis of a flowgraph associated with such a program and determines whether data object errors will occur at runtime.
In light of the foregoing limitations and drawbacks, a XLANG/s compilation method is provided that uses data flow analysis of a program's flow graph to determine lifetimes of its data objects, such as variables. According to the method, a flowgraph is created according to abstract computer instructions, wherein the flowgraph has a plurality of basic blocks and a data object, and wherein the abstract instructions are translated from a parse tree formed from computer code. A depth-first order is assigned to the basic blocks and a dominance relationship between the basic blocks is determined. A determination is made as to whether any loops are present within the flowgraph and, if so, the loops are identified. A determination is also made as to the usage of the data object, and according to the usage, identified loops, dominance relationships and depth-first order of the plurality of basic blocks, a creation point, destruction point and lock point for the data object is determined. Instructions are inserted into the computer code to create the data object at the creation point, to destroy the data object at the destruction point and to lock the data object at the lock point.
The foregoing summary, as well as the following detailed description of preferred embodiments, is better understood when read in conjunction with the appended drawings. For the purpose of illustrating the invention, there is shown in the drawings exemplary embodiments of the invention; however, the invention is not limited to the specific methods and instrumentalities disclosed. In the drawings:
Overview
A XLANG/s compilation method is provided that parses XLANG/s code into one or more flowgraphs representative of the logical operation of the code. The method identifies a data object such as, for example, a variable in the flowgraph and tracks the operations that are performed on such data object. The method also places a creation point for such an object at an optimal location within XLANG/s code as late as possible prior to the object's first use. In addition, the method places a destruction point for such data object as early as possible after such object's last use. The method further places a lock point for a shared data object at an optimal location: at a location that avoids deadlocked code, while being as close prior to the first use of such object as possible.
Exemplary Computing Environment
The invention is operational with numerous other general purpose or special purpose computing system environments or configurations. Examples of well known computing systems, environments, and/or configurations that may be suitable for use with the invention include, but are not limited to, personal computers, server computers, hand-held or laptop devices, multiprocessor systems, microprocessor-based systems, set top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, distributed computing environments that include any of the above systems or devices, and the like.
The invention may be described in the general context of computer-executable instructions, such as program modules, being executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform particular tasks or implement particular abstract data types. The invention may also be practiced in distributed computing environments where tasks are performed by remote processing devices that are linked through a communications network or other data transmission medium. In a distributed computing environment, program modules and other data may be located in both local and remote computer storage media including memory storage devices.
With reference to
Computer 110 typically includes a variety of computer readable media. Computer readable media can be any available media that can be accessed by computer 110 and includes both volatile and non-volatile media, removable and non-removable media. By way of example, and not limitation, computer readable media may comprise computer storage media and communication media. Computer storage media includes both volatile and non-volatile, removable and non-removable media implemented in any method or technology for storage of information such as computer readable instructions, data structures, program modules or other data. Computer storage media includes, but is not limited to, RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital versatile disks (DVD) or other optical disk storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other medium which can be used to store the desired information and which can accessed by computer 110. Communication media typically embodies computer readable instructions, data structures, program modules or other data in a modulated data signal such as a carrier wave or other transport mechanism and includes any information delivery media. The term “modulated data signal” means a signal that has one or more of its characteristics set or changed in such a manner as to encode information in the signal. By way of example, and not limitation, communication media includes wired media such as a wired network or direct-wired connection, and wireless media such as acoustic, RF, infrared and other wireless media. Combinations of any of the above should also be included within the scope of computer readable media.
The system memory 130 includes computer storage media in the form of volatile and/or non-volatile memory such as ROM 131 and RAM 132. A basic input/output system 133 (BIOS), containing the basic routines that help to transfer information between elements within computer 110, such as during start-up, is typically stored in ROM 131. RAM 132 typically contains data and/or program modules that are immediately accessible to and/or presently being operated on by processing unit 120. By way of example, and not limitation,
The computer 110 may also include other removable/non-removable, volatile/non-volatile computer storage media. By way of example only,
The drives and their associated computer storage media, discussed above and illustrated in
The computer 110 may operate in a networked environment using logical connections to one or more remote computers, such as a remote computer 180. The remote computer 180 may be a personal computer, a server, a router, a network PC, a peer device or other common network node, and typically includes many or all of the elements described above relative to the computer 110, although only a memory storage device 181 has been illustrated in
When used in a LAN networking environment, the computer 110 is connected to the LAN 171 through a network interface or adapter 170. When used in a WAN networking environment, the computer 110 typically includes a modem 172 or other means for establishing communications over the WAN 173, such as the Internet. The modem 172, which may be internal or external, may be connected to the system bus 121 via the user input interface 160, or other appropriate mechanism. In a networked environment, program modules depicted relative to the computer 110, or portions thereof, may be stored in the remote memory storage device. By way of example, and not limitation,
Exemplary Distributed Computing Frameworks or Architectures
Various distributed computing frameworks have been and are being developed in light of the convergence of personal computing and the Internet. Individuals and business users alike are provided with a seamlessly interoperable and web-enabled interface for applications and computing devices, making computing activities increasingly web browser or network-oriented.
For example, MICROSOFT®'s .NET platform includes servers, building-block services, such as web-based data storage, and downloadable device software. Generally speaking, the .NET platform provides (1) the ability to make the entire range of computing devices work together and to have user information automatically updated and synchronized on all of them, (2) increased interactive capability for web sites, enabled by greater use of XML rather than HTML, (3) online services that feature customized access and delivery of products and services to the user from a central starting point for the management of various applications, such as e-mail, for example, or software, such as Office NET, (4) centralized data storage, which will increase efficiency and ease of access to information, as well as synchronization of information among users and devices, (5) the ability to integrate various communications media, such as e-mail, faxes, and telephones, (6) for developers, the ability to create reusable modules, thereby increasing productivity and reducing the number of programming errors, and (7) many other cross-platform integration features as well.
While exemplary embodiments herein are described in connection with software residing on a computing device, one or more portions of the invention may also be implemented via an operating system, API, or a “middle man” object between a coprocessor and requesting object, such that services may be performed by, supported in, or accessed via all of NET's languages and services, and in other distributed computing frameworks as well.
Exemplary Embodiments
Introduction to Web Services
Businesses are more frequently interacting via distributed environments, such as the Internet or World Wide Web. For example, a consumer may want to know the prices of rental cars for an upcoming trip. The consumer may request the prices through an intermediate business on the Internet, such as a travel website. The intermediate business, upon receiving the consumer's request, sends price quote requests to a variety of rental car businesses. After responses are received from the rental car businesses, the intermediate business then sends the responses to the consumer. The consumer may then reserve a car and pay for the reservation via the intermediate business. The business processes discussed above are implemented using messages. For example, the consumer's request to the intermediate business is implemented as an electronic message to the intermediate business that contains the particulars of the proposed car rental, for example: rental dates and times, type of car, additional insurance desired, consumer's name and address, credit card information and/or the like. The intermediate business' price quote request is another message, as are the responses from the rental car businesses—both to the intermediate business from the rental car businesses and from the intermediate business to the consumer—and the reservation ultimately sent by the consumer.
A computer language designed to automate business processes and the messages used to carry out such processes is XLANG/s, which is described in more detail, below. As may be appreciated, it is important that the processes that are implemented using XLANG/s execute properly, and without errors. In addition, such processes should be robust enough to compensate for external errors, such as communications link failures and the like. Providing a method of programming and compiling XLANG/s code to accomplish such processes in a robust and error-free manner is an application of an embodiment of the present invention.
XLANG/s Introduction
XLANG/s is a language that describes the logical sequencing of business processes, as well as the implementation of the business process by using various technology components or services. XLANG/s is described in more detail than is disclosed herein in a document titled “XLANG/s Language Specification,” Version 0.55, Copyright© Microsoft 1999–2000, and a document titled “XLANG Web Services For Business Process Design, Satish Thatte, Copyright© Microsoft Corporation 2001, both of which are hereby incorporated by reference in their entirety. The XLANG language is expressed in Extensible Markup Language (XML). As noted above, business processes may be implemented using software code. The XLANG/s language is a modem, domain specific, special purpose language used to describe business processes and protocols. XLANG/s is also a declarative language, which means that it defines an explicit instruction set that describes and implements steps in a business process, the relationship between those steps, as well as their semantics and interactions. XLANG/s code is not just descriptive; it is also designed to be executable. Because of the declarative nature of XLANG/s and its specific semantics, the resulting executable code is deterministic; that is, the behavior of the running business process is well defined by the semantics of the collection of XLANG/s instructions. Therefore, by examining XLANG/s code one is able to determine the business process that is carried out by such code. The definition of such a business process in executable form is an “orchestration.”
XLANG/s is compatible with many Internet standards. XLANG/s is designed to use XML, XSLT (http://www.w3.org/TR/xslt), XPATH (http://www.w3.org/TR/xpath), XSD (XML Schema Definition) and WSDL (Web Services Description Language) as supported standards and has embedded support for working with NET based objects and messages. WSDL is described in a document titled “Web Services Description Language (WSDL) 1.1,” W3C Note January 2001, by Microsoft and IBM Research, Copyright© 2000 Ariba, International Business Machines Corporation, Microsoft, which is hereby incorporated by reference in its entirety. The XLANG/s language is syntactically similar to C#, thus a C# specification may be referenced as an aid to understanding the exact syntax. The semantics embodied in XLANG/s are a reflection of those defined in a document entitled “Business Process Execution Language for Web Services,” Version 1.1, dated Mar. 31, 2003, published by Microsoft, IBM and BEA for the definition of Business Process semantics, which is also hereby incorporated by reference in its entirety. The Business Process Execution Language for Web Services specification is commonly referred to as the “BPEL4WS” specification. As may be appreciated, therefore, the use of XLANG/s is most advantageous when applied to a business process.
XLANG/s defines a rich set of high-level constructs used to define a business process. XLANG/s statements generally fall into one of two categories: simple statements that act on their own, such as receive or send, and complex statements that contain or group simple statements and/or other complex statements. XLANG/s also supports low-level data types such as String and Int32 (integer), for example. High-level data types are also defined such as, for example, messages, ports (locations to which messages are sent and received), correlations and service links. The data types are used to rigorously define the semantics associated with the business process and are complemented by process control statements such as “while,” and “scope.”
As noted above, a XLANG/s service communicates with the outside world by sending and/or receiving messages. The message type is the structural definition of such a message. Messages are acted upon by operations (e.g., receive, response), and an operation may be either a single asynchronous message or a request-response pair of messages. Operations may be either incoming or outgoing. For example, a seller may offer a service/product that begins an interaction by accepting a purchase order (from a potential buyer) via an input message. The seller may then return an acknowledgement to the buyer if the order can be fulfilled. The seller may send additional messages to the buyer (e.g., shipping notices, invoices). Typically, these input and output operations occur in accordance with a defined sequence, referred to as a service process. The seller's service remembers the state of each purchase order interaction separately from other similar interactions. This is particularly advantageous in situations in which the buyer may be conducting many simultaneous purchase processes with the same seller. Also, each instance of a service process may perform activities in the background (e.g., update inventory, update account balance) without the stimulus of an input operation.
A service process may represent an interaction utilizing several operations. As such the interaction has a well-defined beginning and end. This interaction is referred to as an instance of the service. An instance can be started in either of two ways. A service can be explicitly instantiated using some implementation-specific functionality or a service can be implicitly instantiated with an operation in its behavior that is meant to be an instantiation operation. A service instance terminates when the process that defines its behavior terminates.
Services are instantiated to act in accordance with the history of an extended interaction. Messages sent to such services are delivered not only to the correct destination port, but to the correct instance of the service that defines the port. A port is an end point where messages are sent and received by a service. The infrastructure hosting the service supports this routing, thus avoiding burdening every service implementation with the need to implement a custom mechanism for instance routing.
Turning now to
For example, consider a typical supply chain situation in which a buyer sends a purchase order to a seller. The buyer sends the message from, for example, first computer 220 to the seller's second computer 230 by way of the network 210 and communications links 222, 233. Assume the buyer and seller have a stable business relationship and are statically configured to send documents related to the purchasing interaction to the URLs associated with the relevant ports. When the seller returns an acknowledgement for the order, the acknowledgement is routed to the correct service instance at the buyer's end at computer 220 or, optionally, another computer such as additional computers 224, 226. One way to implement such a routing is to carry an embedded token (e.g., cookie) in the order message that is copied into the acknowledgement for correlation. The token may be in the message envelope in a header or in the business document (purchase order) itself. The structure and position of the tokens in each message can be expressed declaratively in the service description. This declarative information allows XLANG/s compliant infrastructure to use tokens to provide instance routing automatically.
During its lifetime, a service instance may typically hold one or more conversations with other service instances representing other participants involved in the interaction. Conversations may use a sophisticated transport infrastructure that correlates the messages involved in a conversation and routes them to the correct service instance. In many cases correlated conversations may involve more than two parties or may use lightweight transport infrastructure with correlation tokens embedded directly in the business documents being exchanged. XLANG/s addresses correlation scenarios by providing a very general mechanism to specify correlated groups of operations within a service instance. A set of correlation tokens can be defined as a set of properties shared by all messages in the correlated group. Such a set of properties is called a correlation set.
Description of Embodiments of the Present Invention
Features such as data types, tokens, correlation sets and the like are implemented in XLANG/s code using variables to store values necessary to carry out such features. As is the case with other high-level software languages, XLANG/s is compiled by a compiler prior to run time. Details relating to compilations processes should be known to those of skill in the art and are therefore not described in detail herein. An embodiment of the present invention is carried out as part of an XLANG/s code compilation process, after some steps of the compilation process have completed. For example, it is assumed herein that a programmer has written a textual description of a business process, wherein the textual description conforms to the grammar and specification of the business process language such as, for example, XLANG/s. It is also assumed herein that the textual program is read by a parser, which ensures that the input is well-formed, meaning that there are no textual or syntax errors in the textual representation, and then the parser translates the textual representation into a canonical high-level parse tree representation. Finally, it is assumed herein that the high-level parse tree is analyzed to produce a sequence of low-level abstract instructions. These instructions, when executed in the proper sequence, will perform the actions that the business process writer expressed in the textual representation.
As noted above, the output of the aforementioned steps of the compilation process is a sequence of low-level abstract instructions. The instructions are then used as input for an embodiment of the present invention. As may be appreciated, additional steps or processes may take place prior to initiating an embodiment of the present invention and, likewise, some or all of the above steps may be consolidated. Any such situation is equally compatible with an embodiment of the present invention. An example of the previously-discussed process that results in input for an embodiment is discussed below in connection with
First, a textual description of a business process is provided. As noted above, the description preferably conforms to the grammar and specification of the business process language being utilized such as, for example, XLANG/s. It will be appreciated that the description describes the aspects of the business process that will ultimately be carried out by the program that results from the complete compilation process according to an embodiment of the present invention, An exemplary fragment of a textual description of a business process may be, as follows:
As may be appreciated, the above fragment is merely exemplary, as a textual description of a business process may describe any aspect of such business process that is capable of being carried out by business software. This particular fragment describes a business process that involves a request to withdraw funds from an account. In particular, and with reference to the above textual description, a message is received on port myPort. The message contains an instruction to withdraw an amount from an account, which is labeled myAccount. If the balance after withdrawal is less than zero, then a result message is assigned the value “overdrawn” to indicate that the account had insufficient funds. Otherwise, the result message is assigned the value “okay.” Finally, the result message is sent back to the sender of the original message as a response.
Second, and as noted above, a parser reads the above disclosed textual representation and builds a parse tree that captures the form and structure contained in the text. Turning now to
The parse tree 300 comprises blocks that represent the logical structure of the textual description. Block 305, for example, represents the receive(myPort, msgAmount) statement, which then leads to block 303, which in turn represents the balance=myAccount.Withdraw(msgAmount) statement. As can be seen in
Block 315 represents the “if” portion of the if (balance <0) statement. Block 317, along with blocks 319–323, represents the logic involved with determining whether the balance of myAccount is less than zero, according to the above if (balance <0) statement. Blocks 325 and 333 represent the logic of generating a message according to the result of the determination made in connection with block 317. For example, if the balance determined in block 317 was less than zero, a message is generated according to block 325, and interior blocks 327–331. As may be appreciated, blocks 325–331 represent the msgResult=“overdrawn” statement of the above textual description. If the balance determined in block 317 was greater than zero, a message is generated according to block 333, and interior blocks 335–339. As may be appreciated, blocks 333–339 represent the msgResult=“okay” statement. Finally, block 341 represents the send(myPort, msgResult) statement, which describes sending a message generated in either blocks 325–331 or blocks 333–339 in accordance with the balance determination of blocks 317–323.
The parse tree is then translated into a sequence of abstract instructions that, when executed, implement the business process.
As may be appreciated, the instructions listed above are representative of a list of instructions, as an actual list of instructions may take any form, and any such form is equally consistent with an embodiment of the present invention. In addition, the numbering and parenthetical explanations above, while consistent with an embodiment of the present invention, are not required to be present in abstract instructions that are used in an embodiment of the present invention. As discussed above, the abstract instructions—in whatever form they are presented—are then used as input by an embodiment of the present invention.
Referring now to
Building the Flowgraph
In one embodiment of the present invention, a flowgraph is constructed by partitioning the instruction list into basic blocks, and then connecting each basic block to its immediate successor(s) and predecessor(s). A basic block is a series of consecutive programming statements, or instructions, in which flow of control enters at the first statement and leaves at the last statement without stopping and without any branching possibilities except at the end of the block. Additional references with respect to basic blocks may be found in Compilers: Principles, Techniques, and Tools, by Aho, Sethi and Ullman, which is herein incorporated by reference in its entirety. One exemplary algorithm that is compatible with an embodiment of the present invention is Algorithm 9.1 of Compilers: Principles, Techniques, and Tools, by Aho, Sethi and Ullman. Thus, an embodiment of the present invention applies, for example, algorithm 9.1 to abstract instructions 1–7, above. In the above example, therefore, the basic blocks are {1, 2, 3}, {4, 5}, {6} and {7}.
Finding Successors and Predecessors
Once the set of abstract instructions, or “instruction stream,” has been partitioned into basic blocks, the blocks are connected bi-directionally to their successors and predecessors. The successors of a first basic block are simply any blocks whose first instructions can be reached by the last instruction in the first basic block. In the above example, therefore, the last instruction in the basic block {1, 2, 3} is the JumpIfFalse instruction. This instruction is defined to perform a jump to another instruction if the condition is false; otherwise, the next sequential instruction will be processed. Thus, the successors of {1, 2, 3} are {4, 5} and {6}. Predecessors, as the name implies, are the opposite of successors. As a result, in the present example the predecessor of {4, 5} is {1, 2, 3}.
Turning now to
Flowgraphs for Transactions
A transaction in XLANG/s encapsulates a sub-process that exhibits transactional characteristics. Thus, a process contained within a XLANG/s transaction either completes in its entirety, or fails in its entirety. As may be appreciated, planning for the success or failure of a particular XLANG/s transaction allows a programmer to create a business application that is robust enough to compensate for errors. A transaction may be, for example, atomic or long-running.
An atomic XLANG/s transaction is a software transaction in which no action takes place and no program state is changed unless the transaction successfully completes. It is atomic in the sense that it completes successfully or it does nothing at all, which is to say that it does not “commit.” A long-running XLANG/s transaction is, as the name implies, a software transaction that takes place over a period of time that is significant for computing purposes. At any point during the processing of such a transaction, an exception may occur. When an exception occurs in, or is “thrown from,” a long-running transaction, the compiler cannot be certain of which, if any, of the instructions within the transaction have executed successfully.
Regardless of its type, each transaction comprises a body process, zero or more exception handler processes and an optional compensation process. As may be appreciated, an atomic transaction does not have an exception handler; however, an atomic transaction may be retry-able. An exception handler process is a sub-process that executes whenever any instruction in the transaction body throws an exception of the kind for which the handler is designed, or whenever the transaction aborts. A compensation process is a sub-process that might execute, depending on the configuration of the program, after a transaction commits in order to “undo” the work that was performed in the transaction body. As may be appreciated, therefore, in one embodiment transactions can nest, and if an inner transaction commits, but an outer transaction aborts, the outer transaction can invoke the inner transaction's compensation handler.
Discussed below in connection with
Long Running Transaction
The simplest form of XLANG/s transaction is a long-running transaction with no exception handlers and no compensation. A textual description of such a transaction may, for example, consist of the following high-level code:
This, in turn, produces the following abstract instruction stream:
Again, and as noted above, this abstract instruction stream consists of numbering and bracketed comments, which are inserted herein for ease of discussion. Thus, such numbering and comments, while consistent with an embodiment of the present invention, are not required. The special abstract instruction BeginTransaction has as its successors the transaction body process, Pbody and, if any were present in the present example, all exception handler processes. The special abstract instruction CommitTransaction has as its successors the next sequential instruction and, if one were present in the present example, the transaction's compensation process.
Because the above two special abstract instructions generally interrupt the straight-line execution of the program by causing a jump or branch to an instruction in another area of the code, such abstract instructions mark the end of a basic block. Thus, the above fragment has three basic blocks: {1, 2}, {3, 4} and {5}. Turning now to
Atomic Transaction
The present example illustrates how an embodiment may treat an atomic transaction differently than a long-running transaction. As may be appreciated, the textual description below is similar to that of the long-running transaction discussed above in connection with
This, in turn, produces the following abstract instruction stream, which is also to that of the long-running transaction discussed above in connection with
In the present atomic transaction example, the CommitTransaction comprises slightly different instructions than that was described above in connection with
As may be appreciated, accounting for the implied loop, as represented by dashed arrow A, is very important when determining destroy points and the like with respect to data objects that are affected by the loop. For example, if the last lexical use of a data object ν occurs in the transaction body Pbody, the data object ν cannot be destroyed until the transaction commits. If data object ν was destroyed in Pbody and if the transaction aborted and retried, the value of the program object ν would be lost during the retry. Accordingly, the presence of the implicit loop is accounted for during the determination of destruction points as will be discussed below in connection with
Long-Running Transaction with Exception Handler
The present example illustrates the handling of an XLANG/s exception handler as implemented in a long-running transaction. As discussed above, atomic transactions do not have exception handlers. A textual description of such a transaction may, for example, consist of the following high-level code:
Turning directly to
The flowgraph presumes that an exception can occur immediately after the transaction body is entered; thus, both the body and the exception handler can logically follow the BeginTransaction abstract instruction. The unlabeled, solid arrow following block 601 indicates normal program flow, while dashed arrow A indicates the logical path that is taken in the event of an exception. Block 607 contains the special abstract instruction Resume, which is defined to jump to the next statement after the transaction. Thus, upon completing the processing of the exception handler represented by block 607, the logic path indicated by dashed arrow B is taken to Pend at block 605. As may be appreciated, both the body of block 603 and the exception handler of block 607 flow to the same successor, Pend at block 605.
Long Running Transaction with Compensation
The present example illustrates the handling of a XLANG/s compensation handler as implemented in a long-running transaction. As discussed above, compensation handlers are XLANG/s code modules that attempt to undo, or “compensate” for, a transaction that does not complete successfully due to an error or the like. Such a compensation handler may perform any task desired by a programmer such as, for example, performing a function that is the opposite of the failed transaction, performing a function that nullifies the effect of the failed transaction, or the like. A textual description of such a transaction may, for example, consist of the following high-level code:
Turning now to
Dashed arrow A represents the logical path taken in the event of a program state for which compensation is desired. As can be seen in
Atomic Transaction with Compensation
The flowgraph of an atomic transaction with compensation, as may be appreciated, may be described as a combination of the atomic transaction discussed above in connection with
Long Running Transaction with Exception Handler and Compensation
The flowgraph of a long-running transaction with exception handler and compensation, as may be appreciated, may be described as a combination of the long-running transaction with exception handler discussed above in connection with
Preparing the Flowgraph
As may be appreciated, once a flowgraph such as, for example, one or more of the flowgraphs as described above in connection with
As may also be appreciated, an embodiment of the present invention may create flowgraphs, such as those discussed above in connection with
Depth-First Ordering
According to an embodiment, a depth-first search is performed on the flow graph and each block contained within the flowgraph is assigned a corresponding depth-first number. As may be appreciated, any algorithm for finding a depth-first ordering of a flowgraph is equally consistent with an embodiment of the present invention. For example, one such algorithm is algorithm 10.14, as described in Compilers: Principles, Techniques, and Tools.
It will be appreciated that the nature of a depth-first search ensures that the last basic block in the program, which corresponds to the end of the program, receives the lowest number, while the first basic block in the program, which corresponds to the start of the program, receives the highest number. Referring now to
Finding Dominators
The dominance relationships of all blocks 705–720 are determined bi-directionally using, for example, algorithm 10.16, as described in Compilers: Principles, Techniques, and Tools. As noted above, any algorithm for finding dominators of basic blocks is equally consistent with an embodiment of the present invention.
Determining whether a block pre-dominates another block involves starting at the top—representing the beginning of the control flow—of flowgraph 700 and determining if all logic paths leading to a particular block must have first passed through a previous block. For example, in flowgraph 700, the pre-dominators of block 720 are 720 and 705. It will be appreciated that a block dominates itself. It will also be appreciated that blocks 710 and 715 do not pre-dominate block 720 because control flow does not need to pass through either block 710 or 715 prior to reaching block 720. For example, control flow may pass from block 705 to block 710 and then to 720, bypassing block 715. Likewise, control flow may pass from block 705 to block 715 and then to 720, bypassing block 710.
Determining whether a block post-dominates another block involves starting at the bottom—representing the end of the control flow—of flowgraph 700 and determining whether all paths leading from a particular block must ultimately pass through another block. For example, in flowgraph 700 the post-dominators of block 705 are blocks 705 and 720.
Finding Loops
Referring now to
Analyzing Data Object Usage
Thus, at the completion of step 420 of
As discussed above in connection with FIGS. 4 and 6A–F, the flowgraphs produced by an embodiment of the present invention are used by a XLANG/s compiler to perform several tasks. For example, the calculation and insertion of create, destroy and lock points are some of the tasks that may be performed by an embodiment of the present invention, and are now discussed in detail below in connection with
Computing the Creation Point for an Object
Turning now to
At step 910, a compiler according to an embodiment of the present invention computes the intersection I of Bcreate's pre-dominators and Bnext's pre-dominators. At step 915, a determination is made as to whether the set I contains Bcreate. If so, at step 920 the creation point of the data object of interest remains Bcreate. If not, at step 925 a new creation point Bcreate is chosen from the set I, such that the depth-first number of the new creation point is the smallest number that is greater than the depth-first number of the current creation point. As a result, therefore, the creation point of the data object occurs before such data object is used so as to avoid an error that would result if the object was referred to and had no value. Likewise, the creation point is created as close to the first use of the object as possible so as to avoid unnecessarily burdening memory resources. As may be appreciated, if the new creation point is in a different loop than the previous creation point, then successively higher- and higher-numbered blocks from the set I are chosen until all inner loops are bypassed.
At step 930, a determination is made as to whether another use of the data object within the set Suse is present and, if so, the method 900 returns to step 910 for such next use. Once all uses of the data object are accounted for, the method 900 proceeds to step 935. At step 935, an instruction that performs the data objects creation and initialization is inserted within the basic block Bcreate, as determined in steps 905–930, prior to any intra-block uses.
Computing the Destruction Point for an Object
Turning now to
At step 1010, a compiler according to an embodiment of the present invention computes the intersection I of Bdestroy's post-dominators and Bprev's post-dominators. At step 1015, a determination is made as to whether the set I contains Bdestroy. If so, at step 1020 the destruction point of the data object of interest remains Bdestroy. If not, at step 1025 a new destruction point Bdestroy is chosen from the set I, such that the depth-first number of the new destruction point is the largest number that is less than the depth-first number of the current destruction point. As a result, therefore, the destruction point of the data object occurs after such data object is last used so as to avoid unnecessarily burdening memory resources. As may be appreciated, if the new destruction point is in a different loop than the previous destruction point, then successively lower- and lower-numbered blocks from the set I are chosen until all inner loops are bypassed.
At step 1030, a determination is made as to whether another use of the data object within the set Suse is present and, if so, the method 1000 returns to step 1010 for such next use. Once all uses of the data object are accounted for, the method 1000 proceeds to step 1035. At step 1035, an instruction that performs the data object's destruction and cleanup is inserted within the basic block Bdestroy, as determined in steps 1005-1030, after any intra-block uses.
Computing Lock Points for Shared Objects
As may be appreciated, the computation of lock points appears similar to that of creation points, because it is desirable to place the locks as late as possible prior to the first use of a shared data object. However, locks for shared objects cannot be placed individually. The following exemplary XLANG/s code provides an example of the complications that arise when computing lock points.
In the above exemplary code, in each parallel task, the objects a and b are accessed in opposite orders. An example of improper lock placement is provided in bold text in the following code example.
If the above code is able to execute without interruption or error, the above lock placement may be acceptable. However, if the first task proceeds far enough to acquire a's lock, and then is interrupted, a deadlocked program may result. If, for example, after the above interruption, the second task proceeds to acquire b's lock, it will then become blocked because it cannot acquire a's lock. If the first task resumes after the interruption, it will not be able to proceed because it now cannot acquire b's lock. The process is therefore deadlocked.
An embodiment of the present works around this problem by considering the use-points of all the shared data objects within a synchronized scope, and placing locks for all such objects prior to the first use of any of the objects. Deadlocks are avoided by acquiring locks in a consistent order across all synchronized scopes. As may be appreciated, any lock order may be used if such order is consistent.
Using the method of
As may be appreciated, the above lock placement eliminates the deadlock problem as discussed above. As will be seen in the discussion of
Turning now to
At step 1140, the sets Swrite and Sread are added to V. At step 1145, a determination is made as to whether another data object is within the set L and, if so, the method 1100 returns to step 1115 for the next data object. Otherwise, the method 1100 proceeds to step 1150 on
Turning now to
At step 1155, a compiler according to an embodiment of the present invention computes the intersection I of Block'S pre-dominators and Bnext's pre-dominators. At step 1160, a determination is made as to whether the set I contains Block. If so, at step 1170 the lock point of the data object of interest remains Block. If not, at step 1165 a new lock point Bcreate is chosen from the set I, such that the depth-first number of the new lock point is the smallest number that is greater than the depth-first number of the current lock point. As a result, the lock point is created as close to the first use of the object as possible so as to avoid early locking of resources, while at the same time being placed in such a way as to avoid deadlocking the transaction. As may be appreciated, if the new lock point is in a different loop than the previous lock point, successively higher- and higher-numbered blocks are chosen from the set I until all inner loops are bypassed.
At step 1175, a determination is made as to whether another lexically subsequent block, Bnext is present within the set V is present and, if so, the method 1100 returns to step 1155 for such next block. Once all uses of the data object are accounted for, the method 1100 proceeds to step 1180. At step 1180, an instruction that locks—as either a read or write block, as determined above in connection with FIG. 11A—each shared object is placed in a deterministic order, prior to any intra-block uses.
Thus, a method and system for compiling XLANG/s code involving static dataflow analysis has been provided. While the present invention has been described in connection with the exemplary embodiments of the various figures, it is to be understood that other similar embodiments may be used or modifications and additions may be made to the described embodiment for performing the same function of the present invention without deviating therefrom. For example, one skilled in the art will recognize that the present invention as described in the present application may apply to any configuration of business method software in any type of application environment. Therefore, the present invention should not be limited to any single embodiment, but rather should be construed in breadth and scope in accordance with the appended claims.
Number | Name | Date | Kind |
---|---|---|---|
6016398 | Radigan | Jan 2000 | A |
20040117772 | Brand et al. | Jun 2004 | A1 |
Number | Date | Country | |
---|---|---|---|
20050050527 A1 | Mar 2005 | US |