M.W. Hall, J.M. Mellor-Crummey, A. Carle, and R. Rodriguez. FIAT: A Framework for Interprocedural Analysis and Transformation. In The Sixth Anunual Workshop on Parallel Languages and Compilers, August 1993.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....schemas, compact names were devised for EITH constructs (shown at the bottom of Fig. 3) The problem of representing programs has been studied extensively in the context of software engineering [1] For example, in traditional interprocedural analysis, a program is abstracted by a call graph [5]. Compilers use a variety of intermediate representations of programs, such as abstract syntax trees (AST) control flow graphs, def use chains, etc. For more complex manipulations of programs, the program dependence graph (PDG) 3] and the dependence flow graph (DPG) 12] have been used. Finally, ....

M. Hall, J. M. M. Cmmmey, A. Carle, and R. G. Rodriguez, "FIAT: A framework for interprocedural analysis and transformations," 6th Workshop on Languages and Compilers for Parallel Computing, 1993.

....to optimize the mapping and execution of the program. It is imperative that good performance be achievable with modest effort and the highest levels of performance be available with reasonable tuning effort. 3. 1 Automatic techniques These techniques exploit aggressive interprocedural analysis [48, 49, 47, 18, 30, 29, 11, 1], profile data, and run time statistics to optimize program implementation choices are essential to the programmability of the machine and accessibility of high performance. Aggressive compiler analysis has been essential to high performance computing based on vector, shared memory, and ....

Hall, M. W., Mellor-Crummey, J. M., Clarle, A., and Rodr' iguez, R. G. FIAT: A framework for interprocedural analysis and transformation. In Proceedings of the Sixth Workshop for Languages and Compilers for Parallel Machines (August 1993), pp. 522--545.

....with a node that is connected (through a tunneling value connection) to the right dependence region. 9. RELATED WORK A number of analysis frameworks have been developed for making intra and interprocedural analyses easier to write and reason about, including Sharlit [26] SPARE [27] FIAT [17], McCAT [19] System Z [33] PAG [2] the k tuple dataflow analysis framework [22] and Dwyer and Clarke s system [15] However, none of these systems address integrating transformations with analyses, nor automatically combining analyses profitably. Nelson and Oppen [23] describe how under ....

M.W. Hall, J.M. Mellor-Crummey, A. Carle, and R. Rodriguez. Fiat: A framework for interprocedural analysis and transformation. In The Sixth Anunual Workshop on Parallel Languages and Compilers, August 1993.

....tools designed to help scientists to use languages like HPF. It is designed to support Fortran D, a precursor and contributor to HPF that contains a subset of features found in HPF as well as support for capabilities not currently in HPF. The D System contains an interprocedural analysis framework [9], the Fortran D compiler [11] and tools for performing automatic data decomposition [2] data race detection, static performance estimation [3, 13] and performance profiling. All these components will be integrated in the D editor, the core of the D system. The goal of the D editor is to help ....

M. W. Hall, J. Mellor-Crummey, A. Carle, and R. Rodriguez. FIAT: A framework for interprocedural analysis and transformation. In Proceedings of the Sixth Workshop on Languages and Compilers for Parallel Computing, Portland, OR, August 1993.

....answers the question of whether there are any exposed uses in the region, the top down phase determines at the end of a region whether there are any exposed uses in the rest of the execution. We implemented our proposed liveness analysis in an interprocedural region based analysis framework[51][55] in SUIF[106] We found our algorithm to be both precise and efficient across a suite of realworld applications, despite the fact that potentially exponential algorithms are used in manipulating the array sections. Because array liveness analysis was deemed to be too expensive[104] we also ....

M. W. Hall, J. Mellor-Crummey, A. Carle, and R. Rodriguez. "FIAT: A framework for interprocedural analysis and transformation." In Proceedings of the Sixth Workshop on Languages and Compilers for Parallel Computing, Portland, OR, August 1993. Springer-Verlag.

....the equivalent of a CFG node with a list of analysis specific objects to apply to a node. The result is code that iterates over these objects to perform the appropriate analysis. Other tools that automate the process of implementing a data flow analysis, such as Sharlit [22] SPARE [6] FIAT [8], System Z [27] PAG [1] and Dwyer and Clarke system [5] can expedite the construction of new data flow analyses. They provide a higher level of abstraction than used in our system. The McCAT [2] and PROLANGS [15, 20] pointer analysis systems are implemented in C and thus, do not use the ....

Mary W. Hall, John M. Mellor-Crummey, Alan Carle, and Rene G. Rodriguez. Fiat: A framework for interprocedural analysis and transformations. In 6th Workshop on Languages and Compilers for Parallel Computing, 1993.

....possible paths, namely E R and E R . Of course, these two criteria must be taken into account when chosing or implementing an interprocedural analysis framework. 1. 3 Interprocedural Analysis Frameworks Many interprocedural analysis frameworks have been defined (some examples can be found in [64, 21, 46, 79]) They differ in several ways: The degree of summarization and the accuracy of the translation across procedure calls are two important issues, and directly depend on the specific semantic problem and the representation of its solutions. But the representation of the program structure on which ....

....on the specific semantic problem and the representation of its solutions. But the representation of the program structure on which the information is conveyed and the ability to perform or not flow sensitive or context sensitive analyses are also of importance. These last two problems are related [48, 46], as shown in this section. Interprocedural analyses can be performed on the program call graph, where nodes represent individual procedures, and edges represent call sites 4 . Each edge is labeled with the actual parameters associated to the corresponding call site. 4 Mathematically, this is ....

[Article contains additional citation context not shown here]

M. Hall, J. Mellor-Crummey, A. Carle, and R. Rodr'iguez. FIAT: A framework for interprocedural analysis and transformation. In Sixth International Workshop on Languages and Compilers for Parallel Computing, August 1993.

....Compiler systems do not, however, provide high level analyzer generator capabilities. Recent work has exploited the inherent generality of data flow frameworks and attempted to explore some of the issues in supporting a flexible, general approach for constructing data flow analyzers. FIAT [33] is a framework for rapid prototyping of interprocedural analyses and transformations; it provide interfaces for describing a data flow analysis problem as a data flow framework and provides a general iterative solver with which to construct analyzers. Sharlit [78] is a tool for generating ....

Hall, M.W., Mellor-Crummey, J., Carle, A., and Rodriguez, R. Fiat: A framework for interprocedural analysis and transformation. In Proceedings of the Sixth Annual Workshop on Languages and Compilers for Parallel Computing, Portland, OR, August 1993.

....which will allow efficient data flow analysis, while maintaining sufficient precision to allow useful transformations and to ensure safety and correctness of transformations. 2. 1 Program Representation In traditional interprocedural analysis, program is abstracted by a call graph [17, 18]. In a call graph G = V; E) V is the set of procedures and directed edge e = i; j) e 2 E) represents a call site in which procedure i invokes procedure j. The limitation of call graph is that no information is available about control flow relationships between various call sites within a ....

....as a new program abstraction. This abstraction records any loop(s) enclosing a procedure call. Again, this abstraction does not allow to look for redundant communication preprocessing calls or communication calls in adjacent procedures. Framework for Interprocedural Analysis and Transforms (FIAT) [18] has recently been proposed as a general environment for interprocedural analysis. This is based upon Call Graph program abstraction and is targeted more towards flow insensitive interprocedural analysis. Our implementation uses several facilities available from FIAT as part of the Fortran D ....

Mary Hall, John M Mellor Crummey, Alan Carle, and Rene G Rodriguez. FIAT: A framework for interprocedural analysis and transformations. In Proceedings of the 6th Workshop on Languages and Compilers for Parallel Computing, pages 522--545. Springer-Verlag, August 1993.

....rely on some control flow information. How much information should be attached to the call graph node for each procedure In prior work, this has ranged from a relatively complete representation [Mye81] to one highly tailored for Kill and Use [Cal88] Concurrently with other researchers at Rice [HMBCR93] we have decided to take a middle road of a lightweight controlflow graph for each procedure, with annotations to the control flow nodes as needed for each particular problem. The core procedure summary control flow graph (G PS CF ) consists of five kinds of nodes: ffl a unique entry node, ....

....Callahan s difficulties Callahan with using the PSG to solve the Live problem. FIAT FIAT is an acronym of Framework for Interprocedural Analysis and Transformation, an interprocedural analysis system descended from the IR n program compiler and available in the ParaScope programming environment [HMBCR93, Hal90]. Its main data structure, the annotated call graph is fundamentally a variation of Myers s supergraph [Mye81] However, FIAT supports a much richer set of annotations than Myers proposes, and provides tools for interprocedural analyses and transformations. Our scheme for symbolic and ....

Mary W. Hall, John M. Mellor-Brummey, Alan Carle, and Ren'e G. Rodr ' iguez. Fiat: A framework for interprocedural analysis and transformations. In Banerjee et al. [BGNP93], pages 522--545.

....method does not impose any restrictions on the shapes of call graph and CFGs of the procedures. Several other projects have performed interprocedural optimizations for parallelism and for dealing with memory hierarchies. FIAT has been proposed as a general framework for interprocedural analysis [15], but largely targets flow insensitive problems. Compiler optimizations for improving I O accesses have been addressed by at least two projects. The PASSION compiler (based upon Syracuse F90D system) performs loop transformations for improving locality in out of core applications [29] Similar ....

Mary Hall, John Mellor Crummey, Alan Carle, and Rene G Rodriguez. FIAT: A framework for interprocedural analysis and transformations. In Proceedings of the 6th Workshop on Languages and Compilers for Parallel Computing, pages 522--545. Springer-Verlag, August 1993.

....gotos are forbidden because they can be eliminated by well known algorithms [3, 2] at the cost of some code duplication in the rare cases where the control graph is not reducible [1] Procedure calls are not allowed. The problem of interprocedural parallelization has been widely studied [41, 18, 38, 10] in the non recursive case, and we believe such methods can be readily included in our framework. The only data structures are scalar types and arrays of scalars. Array subscripts are unrestricted. Since our framework is based on the polytope model, our main interest is to abstract values and ....

M.W. Hall, J.M. Mellor-Crummey, A. Carle, and R. Rodriguez. Fiat: A framework for interprocedural analysis and transformation. In Proceedings of the Sixth Annual Workshop on Languages and Compilers for Parallel Computing. Springer-Verlag, January 1994. updated version for book.

....Compiler systems do not, however, provide high level analyzer generator capabilities. Recent work has exploited the inherent generality of data flow frameworks and attempted to explore some of the issues in supporting a flexible, general approach for constructing data flow analyzers. FIAT [14] is a framework for rapid prototyping of interprocedural analyses and transformations; it provide interfaces for describing a data flow analysis problem as a data flow framework and provides a general iterative solver with which to construct analyzers. Sharlit [21] is a tool for generating ....

M.W. Hall, J. Mellor-Crummey, A. Carle, and R. Rodriguez. Fiat: A framework for interprocedural analysis and transformation. In Proceedings of the Sixth Annual Workshop on Languages and Compilers for Parallel Computing, Portland, OR, August 1993.

....of the classes of function arguments. In contrast, our analysis [41] is flow sensitive with respect to both the classes of function arguments and the classes of member variables of data structures. Cloning and specialization have been studied in the context of FORTRAN by Cooper [17] and Hall [24, 25]. In the context of object oriented programs, customization [7] by Chambers and Ungar and later specialization [18] by Dean, Chambers and Grove have addressed the problem of selecting versions of functions to duplicate based on the classes of the functions arguments. Their decisions have been ....

Mary W. Hall, John M. Mellor-Crummey, Alan Clarle, and Ren'e G. Rodr ' iguez. FIAT: A framework for interprocedural analysis and transformation. In Proceedings of the Sixth Workshop for Languages and Compilers for Parallel Machines, pages 522--545, August 1993.

....summary information is recomputed from scratch, after which their recompilation tests are used to determine what procedures need to be recompiled as a result of changes in the interprocedural information. The FIAT system improves on this by computing interprocedural summary information lazily [Hall et al. 93] Using our dependency framework, however, the computation and update of the interprocedural summary information could be made incremental, with the framework only invalidating and recomputing those summary sets that are out of date. A impl B impl decl of structure S decl of proc P size of S ....

M.W. Hall, J.M. Mellor-Crummey, A. Carle, and R. Rodriguez. FIAT: A Framework for Interprocedural Analysis and Transformation. In The Sixth Anunual Workshop on Parallel Languages and Compilers, August 1993.

.... analysis techniques merge information into a flow independent summary form, which saves time or space but loses the precision needed to perform techniques such as array privatization ( 20] More recent interprocedural analysis techniques, such as atom image ( 15] and those developed for FIAT ([10]) can efficiently propagate precise array subscript information over procedure boundaries. However, demand driven partial inline expansion is still useful as a supplement to such techniques to aid code scheduling ( 18] and automatic parallelization ( 11] and inliner technology can be reused ....

Mary W. Hall, John M. Mellor-Cummey, Alan Carle, and Ren'e G. Rodr'iguez. Fiat: A framework for interprocedural analysis and transformation. In Proceedings of the Sixth Workshop on Languages and Compilers for Parallel Computers, vol. 2, February 1995.

....summary information is recomputed from scratch, after which their recompilation tests are used to determine what procedures need to be recompiled as a result of changes in the interprocedural information. The FIAT system improves on this by computing interprocedural summary information lazily [Hall et al. 93] Using our dependency framework, however, the computation and update of the interprocedural summary information could be made incremental, with the framework only invalidating and recomputing those summary sets that are out of date. A impl B impl decl of structure S decl of proc P size of S ....

M.W. Hall, J.M. Mellor-Crummey, A. Carle, and R. Rodriguez. FIAT: A Framework for Interprocedural Analysis and Transformation. In The Sixth Anunual Workshop on Parallel Languages and Compilers, August 1993.

.... side effects of array accesses using regular section descriptors are detailed in [13] Program transformations that need to be performed for optimizing interprocedural redistributions are presented in [12] These techniques have been implemented in the Fortran D compiler and in the FIAT system [10] at Rice University. However, these works assume static distributions and do not deal with the problem of determining dynamic data decompositions in the presence of procedure calls. Palermo et al. have recently presented an interprocedural data flow analysis framework for array redistribution ....

M. Hall, J. Crummey, A. Carle, and R. Rodriguez. FIAT : A framework for interprocedural analysis and transformations. In Proceedings of Languages and Compilers for Parallel Computing, August 1994.

....however, many efficiency issues and tradeoffs involved in interprocedural analysis. Due to space limitation, we do not discuss those issues here, instead, we give an example of interprocedural symbolic analysis in our framework and refer the interested reader to the literature [Cooper et al. 1992; Hall et al. 1993; Havlak 1994; Havlak and Kennedy 1991; Irigoin 1993; Triolet et al. 1986] Parallelization of a toy Fortran code using interprocedural symbolic analysis of Parafrase 2 is demonstrated in Figure 23. The loop of procedure test contains calls to procedures f and ndx. The dependences introduced by ....

Hall, M. W., Mellor-Crummey, J., Carle, A., and Rodr' iguez, R. G. 1993. FIAT: A framework for interprocedural analysis and transformations. In Proceedings of the 6th Annual Workshop on Languages and Compilers for Parallel Computing. Lecture Notes in Computer Science, vol. 768. Springer-Verlag, Berlin, 522--545.

....can be incorporated in the Locust system. Data pattern analysis becomes more challenging when computation inside a parallel loop involves calls to subroutines that might access shared data. A sizeable amount of research has been performed in the area of inter procedural analysis [26] 45] [46] in the recent years with encouraging results that further vindicates the feasibility of compiler based shared memory systems. Once it is known which tasks access what data items, further analysis is needed to split the data items into non overlapping data items. Once data items have been defined ....

M W Hall, J Mellor-Crummey, A Carle, and R Rodriguez. Fiat: a framework for interprocedural analysis and transformation. In Proceedings of the 6th Workshop on Languages and Compilers for Parallel Computing, August 1993.

....The flow information now precisely models the fact that only true is passed as the second argument to z:w:w. 4 Analysis Frameworks We conclude by comparing BANE with other program analysis frameworks. There have been many such frameworks in the past; see for example [ATGL96,AM95,Ass96,CDG96,DC96,HMCCR93,TH92,Ven89,YH93] Most frameworks are based on standard dataflow analysis, as first proposed by Cocke [Coc70] and developed by Kildall [Kil73] and Kam and Ullman [KU76] while others are based on more general forms of abstract interpretation [Ven89,YH93] In previous frameworks the user ....

.... are based on more general forms of abstract interpretation [Ven89,YH93] In previous frameworks the user specifies a lattice and a set of transfer functions, either in a specialized language [AM95] in a Yacc like system [TH92] or as a module conforming to a certain interface [ATGL96,CDG96,DC96,HMCCR93] The framework traverses a program representation (usually a control flow graph) either forwards or backwards, calling user defined transfer functions until the analysis reaches a fixed point. A fundamental distinction between BANE and these frameworks is the interface with a client analysis. In ....

M. Hall, J. Mellor-Crummey, A. Carle, and R. Rodr'iguez. FIAT: A Framework for Interprocedural Analysis and Transformation. In U. Banerjee, D. Gelernter, A. Nicolau, and D. Padua, editors, Proceedings of the 6th International Workshop on Parallel Languages and Compilers, pages 522--545, Portland, Oregon, August 1993. Springer-Verlag.

....help reduce development time and improve correctness by capturing these common features in a single module [13] In an interprocedural setting, a framework is even more important because of the complexity of collecting and managing information about all the procedures in a program. We use Fiat [6], a tool which encapsulates the common features of interprocedural analysis, in combination with the Stanford SUIF compiler to constitute our interprocedural parallelization system. The Fiat system has been described previously, but we have extended the system to obtain precise flow sensitive ....

M. W. Hall, J. Mellor-Crummey, A. Carle, and R. Rodriguez. FIAT: A framework for interprocedural analysis and transformation. In Proceedings of the Sixth Workshop on Languages and Compilers for Parallel Computing, Portland, OR, August 1993.

....representation, which will allow efficient data flow analysis, while maintaining sufficient precision to allow useful transformations and to ensure safety and correctness of transformations. 2. 1 Program Representation In traditional interprocedural analysis, program is abstracted by a call graph [18, 19]. In a call graph G = V; E) V is the set of procedures and directed edge e = i; j) e 2 E) represents a call site in which procedure i invokes procedure j. The limitation of call graph is that no information is available about control flow relationships between various call sites within a ....

....as a new program abstraction. This abstraction records any loop(s) enclosing a procedure call. Again, this abstraction does not allow to look for redundant communication preprocessing calls or communication in adjacent procedure. Framework for Interprocedural Analysis and Transforms (FIAT) [19] has recently been proposed as a general environment for interprocedural analysis. This is based up Call Graph program abstraction and is targeted more towards flow insensitive interprocedural analysis. Our implementation uses several facilities available from FIAT as part of the Fortran D ....

Mary Hall, John M Mellor Crummey, Alan Carle, and Rene G Rodriguez. FIAT: A framework for interprocedural analysis and transformations. In Proceedings of the 6th Workshop on Languages and Compilers for Parallel Computing, pages 522--545. Springer-Verlag, August 1993.

....and array references. The input language is C and Fortran through the f2c translator. The compiler outputs machine code for different architectures. The compiler initial design and the current distribution of SUIF does not incorporate interprocedural analysis. Such an addition is being worked on [12, 11]. Another problem is that the consistency of analyses is not automatically inforced (for instance invalidation after code transformations) The originality of Pips with respect to these other prototypes is the conjunction of: ffl Interprocedurality and coherency automatic management in a demand ....

Mary W. Hall, John M. Mellor-Crummey, Alan Carle, and Ren'e G. Rodr'iguez. FIAT : A framework for interprocedural analysis and transformation. In Sixth International Workshop on Languages and Compilers for Parallel Computing, August 1993.

....analysis is used to construct the call graph and solve a number of data flow problems. Transformation techniques are described to increase the availability of this information through linkage optimization including cloning. However, this work does not address clone minimization. Cooper and Hall [19, 21, 13, 14, 20, 22] present comprehensive interprocedural compilation techniques and cloning for FORTRAN. This work is general over forward data flow problems, and presents mechanisms for preserving information across clones and minimizing their number. However, concrete types are not a forward data flow problem. ....

Mary W. Hall, John M. Mellor-Crummey, Alan Clarle, and Ren'e G. Rodr ' iguez. FIAT: A framework for interprocedural analysis and transformation. In Proceedings of the Sixth Workshop for Languages and Compilers for Parallel Machines, pages 522--545, August 1993.

....much longer time to converge. Several ideas in the design of our representation are similar to the ideas used in Callahan s Program Summary Graph [4] and Interprocedural Flow Graph used by Soffa et al. 16] FIAT has been introduced as a general framework for performing interprocedural analysis [13], but is more targeted towards flow insensitive problems. Interval based approach for solving interprocedural data flow equations has been investigated in [2] Recompilation in a compiler performing interprocedural analysis has been investigated in [3] Program Foo Procedure P(x,y) a = 1 ....

Mary Hall, John M Mellor Crummey, Alan Carle, and Rene G Rodriguez. FIAT: A framework for interprocedural analysis and transformations. In Proceedings of the 6th Workshop on Languages and Compilers for Parallel Computing, pages 522--545. Springer-Verlag, August 1993.

....tools designed to help scientists to use languages like HPF. The D System supports Fortran D, a precursor and contributor to HPF that contains a subset of features found in HPF as well as support for capabilities not currently in HPF. The system contains FIAT, an interprocedural analysis framework [15], the Fortran D compiler [17] and tools for performing automatic data decomposition [3] data race detection, static performance estimation [4, 20] and performance profiling. All these components will be integrated in the D Editor, the core of the D system. The goal of the D Editor is to help ....

.... as a part of ParaScope, a programming environment that pioneered research on interprocedural optimization in an efficient compilation system [8, 9] Its interprocedural compilation framework has been encapsulated into a system called FIAT (Framework for Interprocedural Analysis and Transformation) [15] and used to solve a number of interprocedural problems in both ParaScope and the Stanford SUIF compiler. A key component of the ParaScope environment is the ParaScope Editor (PED) an intelligent interactive editor for sharedmemory machines [13, 21, 22] The ParaScope Editor provides the results ....

[Article contains additional citation context not shown here]

M. W. Hall, J. Mellor-Crummey, A. Carle, and R. Rodriguez. FIAT: A framework for interprocedural analysis and transformation. In Proceedings of the Sixth Workshopon Languagesand Compilers for Parallel Computing, Portland, OR, August 1993.

....of the flexibility to optimize the mapping and execution of the program. It is imperative that good performance be achievable with modest effort and the highest levels of performance be available with reasonable tuning effort. Automatic techniques which exploit aggressive interprocedural analysis [35, 36, 37, 38, 39, 40, 41, 42], profile data, and run time statistics to optimize program implementation choices are essential to the programmability of the machine and accessibility of high performance. Aggressive compiler analysis has been essential to high performance computing based on vector, shared memory, and ....

M. W. Hall, J. M. Mellor-Crummey, A. Clarle, and R. G. Rodr ' iguez, "FIAT: A framework for interprocedural analysis and transformation," in Proceedings of the Sixth Workshop for Languages and Compilers for Parallel Machines, pp. 522--545, August 1993.

....representation, which will allow efficient data flow analysis, while maintaining sufficient precision to allow useful transformations and to ensure safety and correctness of transformations. 2. 1 Program Representation In traditional interprocedural analysis, program is abstracted by a call graph [19, 20]. In a call graph G = V; E) V is the set of procedures and directed edge e = i; j) e 2 E) represents a call site in which procedure i invokes procedure j. The limitation of call graph is that no information is available about control flow relationships between various call sites within a ....

....a new program abstraction. This abstraction records any loop(s) enclosing a procedure call. Again, this abstraction does not allow to look for redundant communication preprocessing calls or communication calls in adjacent procedures. Framework for Interprocedural Analysis and Transforms (FIAT) [20] has recently been proposed as a general environment for interprocedural analysis. This is based upon Call Graph program abstraction and is targeted more towards flow insensitive interprocedural analysis. Our implementation uses several facilities available from FIAT as part of the Fortran D ....

Mary Hall, John M Mellor Crummey, Alan Carle, and Rene G Rodriguez. FIAT: A framework for interprocedural analysis and transformations. In Proceedings of the 6th Workshop on Languages and Compilers for Parallel Computing, pages 522--545. Springer-Verlag, August 1993.

.... To manage the software engineering complexity of building interprocedural ver sions of the suite of analyses required for parallelization, we utilize a common frame work, analogous to a traditional data flow analysis framework [70] The common framework, a significant extension of the FIAT [52] system, facilitates implementa tion of interprocedural data flow analyses by providing parameterized templates for solving data flow problems. Each analysis problem is implemented by instantiating the templates with functions to compute solutions to data flow equations. The pa rameterized ....

....analyses by providing parameterized templates for solving data flow problems. Each analysis problem is implemented by instantiating the templates with functions to compute solutions to data flow equations. The pa rameterized template for flow insensitive analysis has been previously described [52]. The flow sensitive, context sensitive interprocedural data flow analysis framework is the topic of this section. 2.2.1 Overview Precise and efficient flow and context sensitive interprocedural analysis is difficult because information flows into a procedure both from its callers (representing ....

[Article contains additional citation context not shown here]

M. W. Hall, J. Mellor-Crummey, A. Carle, and R. Rodriguez. FIAT: A frame- work for interprocedural analysis and transformation. In Proceedings of the Sixth August 1993.

....Compiler systems do not, however, provide high level analyzer generator capabilities. Recent work has exploited the inherent generality of data flow frameworks and attempted to explore some of the issues in supporting a flexible, general approach for constructing data flow analyzers. FIAT [HMCCR93] is a framework for rapid prototyping of interprocedural analyses and transformations; it provide interfaces for describing a data flow analysis problem as a data flow framework and provides a general iterative solver with which to construct analyzers. Sharlit [TH92] is a tool for generating ....

M.W. Hall, J. Mellor-Crummey, A. Carle, and R. Rodriguez. Fiat: A framework for interprocedural analysis and transformation. In Proceedings of the Sixth Annual Workshop on Languages and Compilers for Parallel Computing, Portland, OR, August 1993.

....compilation process in ParaScope preserves separate compilation of procedures to a large extent. An interprocedural compilation framework, called Fiat, provides an implementation platform for the interprocedural analyses in ParaScope, including those implemented as part of the Fortran D compiler [20]. Fiat collects information from other tools in the environment so that a procedure only needs to be examined once during compilation. Additional passes over the code can be added if necessary, but should be avoided since experience has shown that examination of source code dominates analysis ....

....so that a procedure only needs to be examined once during compilation. Additional passes over the code can be added if necessary, but should be avoided since experience has shown that examination of source code dominates analysis time. The compilation system uses the following 3 phase approach [8, 16, 20]: 1. Local Analysis. When a procedure is edited, ParaScope calculates and stores summary information concerning all local interprocedural effects for the procedure. This information includes details on call sites, formal parameters, scalar and array section uses and definitions, local constants, ....

M. W. Hall, J. Mellor-Crummey, A. Carle, and R. Rodriguez. FIAT: A framework for interprocedural analysis and transformation. In Proceedings of the Sixth Workshop on Languages and Compilers for Parallel Computing, Portland, OR, August 1993.

....results we have gathered with this system. 2 Interprocedural Framework Interprocedural parallelization depends upon the solution of a large number of interprocedural data flow analysis problems. These problems share many commonalities. We have encapsulated these common features in a tool, Fiat [1], which we have combined with the Stanford SUIF compiler to constitute our interprocedural parallelization system. Fiat is an interprocedural framework, analogous to traditional data flow analysis frameworks [5] A framework is even more important for interprocedural optimization because of the ....

M. W. Hall, J. Mellor-Crummey, A. Carle, and R. Rodriguez. FIAT: A framework for interprocedural analysis and transformation. In Proceedings of the Sixth Workshop on Languages and Compilers for Parallel Computing, Portland, OR, August 1993. 10

....an interprocedural framework to manage the complexity of the implementation and allow code reuse. The region based analysis and selective procedure cloning techniques are encapsulated in a common interprocedural framework as part of Fiat, a tool for de8 veloping interprocedural analysis systems [11]. Fiat facilitates adding new interprocedural analyses by providing parameterized templates to drive flow sensitive analysis and cloning; each analysis problem is implemented by instantiating the templates with functions to compute solutions to data flow equations. For the interprocedural ....

M. W. Hall, J. Mellor-Crummey, A. Carle, and R. Rodriguez. FIAT: A framework for interprocedural analysis and transformation. In Proceedings of the Sixth Workshop on Languages and Compilers for Parallel Computing, Portland, OR, August 1993.

....scheduling 1 . Ongoing research projects using SUIF include global data and computation decomposition for both shared and distributed address space machines [2] communication optimizations for distributed address space machines [1] array privatization [7] interprocedural parallelization [5], efficient pointer analysis, and optimization of Verilog simulations. SUIF has also been used for courses on compiler optimizations at Stanford. This paper only presents the base infrastructure of SUIF and does not include our latest research. However, as our work progresses, we will continue to ....

M. W. Hall, J. Mellor-Crummey, A. Carle, and R. Rodriguez. FIAT: A Framework for Interprocedural Analysis and Transformation. In Proceedings of the Sixth Workshop on Languages and Compilers for Parallel Computing, Aug. 1993.

....have an interprocedural framework to manage the complexity of the implementation and allow code reuse. The region based analysis and selective procedure cloning techniques are encapsulated in a common interprocedural framework as part of Fiat, a tool for developing interprocedural analysis systems [11]. Fiat facilitates adding new interprocedural analyses by providing parameterized templates to drive flow sensitive analysis and cloning; each analysis problem is implemented by instantiating the templates with functions to compute solutions to data flow equations. For the interprocedural ....

M. W. Hall, J. Mellor-Crummey, A. Carle, and R. Rodriguez. FIAT: A framework for interprocedural analysis and transformation. In Proceedings of the Sixth Workshop on Languages and Compilers for Parallel Computing, Portland, OR, August 1993.

....results we have gathered with this system. 2 Interprocedural Framework Interprocedural parallelization depends upon the solution of a large number of interprocedural data flow analysis problems. These problems share many commonalities. We have encapsulated these common features in a tool, Fiat [1], which we have combined with the Stanford SUIF compiler to constitute our interprocedural parallelization system. Fiat is an interprocedural framework, analogous to traditional data flow analysis frameworks [4] A framework is even more important for interprocedural optimization because of the ....

M. W. Hall, J. Mellor-Crummey, A. Carle, and R. Rodriguez. FIAT: A framework for interprocedural analysis and transformation. In Proceedings of the Sixth Workshop on Languages and Compilers for Parallel Computing, Portland, OR, August 1993.

No context found.

M.W. Hall, J.M. Mellor-Crummey, A. Carle, and R. Rodriguez. FIAT: A Framework for Interprocedural Analysis and Transformation. In The Sixth Anunual Workshop on Parallel Languages and Compilers, August 1993.

No context found.

M. Hall, J. Mellor-Crummey, A. Carle, and R. Rodriguez. Fiat: a framework for interprocedural analysis and transformation. In Proceedings of the Sixth Annual Workshop on Compilers for Parallel Processing, Aug 1993.

No context found.

Mary W. Hall, John M. Mellor-Crummey, Alan Carle, and Rene G. Rodriguez. Fiat: A framework for interprocedural analysis and transformations. In Workshop on Languages and Compilers for Parallel Computing (LCPC), 1993.

No context found.

M. W. Hall and et al. Fiat: A framework for interprocedural analysis and transformation. In Proceedings of the 6th International Workshop on Languages and Compilers for Parallel Computing, Lecture Notes in Computer Science, pages 522-- 545, Portland, Oregon, August 12--14, 1993. Springer-Verlag.

No context found.

MaryW. Hall, JohnM. Mellor-Crummey, Alan Carle, and ReneG. Rodriguez. FIAT: A Framework for Interprocedural Analysis and Transformation. In 6th Workshop on Parallel Languages and Compilers, pages 522--545, 1994.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC