Cattell, R. G. G., 1978. Formalization and automatic derivation of code generators, Carnegie -Mellon University, Computer Science Department.  Home/Search   Document Not in Database   Summary   Related Articles   Check

.... addressed by the product of two registers) Figure 4: Data flow graph and instruction patterns # ref # # ref # ref # ref # Figure 5: Covering data flow trees with instruction patterns Early approaches for covering trees were also based on heuristics (like Cattell s maximum munching method [10]) More recently, Aho et al. proposed an dynamic programming algorithm for generating an optimum cover [2] This approach, however, mainly aims at code generation for homogenous register architectures. Heterogenous register architectures can be handled with tree parsing [5] The iburg tool set ....

R.G.G. Cattell. Formalization and automatic derivation of code generators. Technical report, PhD thesis, Carnegie-Mellon University, Pittsburgh, 1978.

....optimal coverings is NP complete even for these instruction sets. Optimal polynomial algorithms were described for restricted cases, like that of dataflow trees [2] Exploitation of complex instruction sets was a major goal in the production quality compiler compiler project (PQCC) at CMU. Cattell [9] proposed the heuristic maximum munching method (MMM) to generate good coverings. With this technique, the largest instruction matching a section of the dataflow graph is selected and matching then continues for the remaining parts of the graph. The result of this technique can be seen in fig. 6 ....

....code generation. 5.1 Behavioural models Behavioural models (instruction set models) of processors have been used in compiler construction for many years. They are the basis for many of the well known pattern matching methods for code generation, such as the methods of Glanville [20] and Cattell [9]. Behavioural models provide a high abstraction of the underlying hardware. However, they do have problems with capturing the effects of pipelines, busy functional units and multiple assignments coded into one instruction word. 5.2 Structural models Due to the problems with instruction set ....

R.G.G. Cattell. "Formalization and Automatic Derivation of Code Generators", PhD thesis, Carnegie-Mellon University, Pittsburgh, 1978.

....the language being compiled, producing a syntax analyzer and compiler proper for the language, and then accepting a machine description and producing a code generator for the machine. See for compiler compilers [McKeeman, Horning and Wortman 1970] for code generator generators [Fraser 1977] and [Cattell 1978], for language description languages [Cleaveland and Uzgalis 1977] and for machine description languages [Bell and Newell 1971] 23 Many small interpreters for SCHEME subsets are presented in [Steele and Sussman 1978b] The reference description of SCHEME is [Sussman and Steele 1978a] Doyle ....

R. G. G. Cattell, "Formalization and Automatic Derivation of Code Generators," Carnegie-Mellon University, Computer Science Department, April 1978.

.... the BURS tables is described in [32] The only serious application of heuristic search techniques to code generation has been the PQCC (Production Quality Compiler Compiler) Project [34] The construction of the code generator and the code generator generator in PQCC are reported by Cattell in [4, 5, 6]. Cattell uses a means ends analysis to determine an optimal code match. This involves selecting a set of instruction templates that are semantically close to a given pattern in the input expression tree. The heuristic semantic closeness means that either the root operators of the pattern and a ....

R. G. G. Cattell. Formalization and Automatic Derivation of Code Generators. UMI Research Press, Ann Arbor, Michigan, 1982.

....especially for architectures that are not naturally stack oriented. We are considering a reimplementation of the code generator using continuation passing style[17] and Orbit[18] 5.1. Pattern matching in code generation Many modern code generators are written in the form of tree pattern matchers[21, 22]. By writing the lambda calculus tree data structure as an ML datatype (with constructors) the tree pattern matching can be directly specified as an ML function. There is one pattern for each constructor, e.g. APP(f,a) gen(f) gen(a) machine.apply( but there are also cases that match ....

R. G. G. Cattell, "Formalization and automatic derivation of code generators," Ph.D. Thesis, Carnegie-Mellon University, Pittsburgh, PA, April 1978.

....selectors from a specification of the target machine requires a formal specification language for target machine semantics. The ISP language for specifying machine instruction sets[10] has been used to automatically generate instruction selectors for intermediate representations based on trees[11] and peephole optimizers for more general graphs[12, 13] These target code generators rely on heuristics to match patterns (derived from the semantics of target machine instructions) to portions of the intermediate representation tree (or graph) The heuristic used on trees leads to a sequence of ....

R. G. G. Cattell, "Formalization and automatic derivation of code generators," Ph.D. Thesis, Carnegie-Mellon University, Pittsburgh, PA, April 1978.

No context found.

Cattell, R. G. G., 1978. Formalization and automatic derivation of code generators, Carnegie -Mellon University, Computer Science Department.

No context found.

R.G.G. Cattell, "Formalization and automatic derivation of code generators ", Ph.D. thesis, Carnegie-Mellon Univ., Pittsburgh (U.S.A.), 1978.

No context found.

R.G.G. Cattell, "Formalization and automatic derivation of code generators ", Ph.D. thesis, Carnegie-Mellon Univ., Pittsburgh (U.S.A.), 1978.

No context found.

R. G. G. Cattell, "Formalization and automatic derivation of code generators," Ph.D. Thesis, Carnegie-Mellon University, Pittsburgh, PA, April 1978.

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