Home My Page Projects Code Snippets Project Openings SML/NJ
Summary Activity Forums Tracker Lists Tasks Docs Surveys News SCM Files

SCM Repository

[smlnj] View of /sml/trunk/src/MLRISC/Doc/latex/mlrisc-compiler.tex
ViewVC logotype

View of /sml/trunk/src/MLRISC/Doc/latex/mlrisc-compiler.tex

Parent Directory Parent Directory | Revision Log Revision Log


Revision 651 - (download) (as text) (annotate)
Thu Jun 1 18:34:03 2000 UTC (19 years, 1 month ago) by monnier
File size: 2799 byte(s)
bring revisions from the vendor branch to the trunk
\section{MLRISC Based Compiler}
    A traditional compiler will typically consist of a 
    \begin{color}{#dd0000}lex/yacc\end{color} based front end, an 
    \begin{color}{#DD0000}optimization\end{color} 
    phase that is repeatedly invoked
    over some intermediate representation, and finally a 
    \begin{color}{#DD0000}back end\end{color} 
    code generation phase. The intermediate
    representation is usually at a level of detail appropriate to the
    optimization being performed, and may be far removed from the
    native instructions of the target architecture. The back end
    proceeds by translating the intermediate representation into
    instructions and registers for an abstract machine that is much
    closer to the target architecture. Retargetting is then achieved
    by mapping the registers and instructions of the abstract machine
    to registers and instructions of the target architecture.

    \br{clear=left}

    \image{MLRISC based compiler}{pictures/png/compiler-2.png}{align=right}

  An MLRISC based compiler, on the other hand, translates the
  intermediate representation into MLRISC instructions and it is the
  MLRISC instructions that get mapped onto instructions of the target
  architecture. Another possibility is to translate the front end
  abstract machine instructions instead of the intermediate
  representation.  Once MLRISC instructions have been generated,
  nearly all aspects of high quality code generation come for free. A
  long story would be cut short if MLRISC were just another abstract
  machine.

  \begin{color}{#580000} The key idea behind MLRISC is that there is no
  single MLRISC instruction set or intermediate program
  representation, \end{color} but the MLRISC intermediate representation
  is specialized to the needs of the front end source language being
  compiled. The specialization does not stop there, but the:
   \begin{itemize}
  \item  \begin{color}{#005500}target instruction set\end{color},
  \item  \begin{color}{#005500}flowgraph\end{color}, and
  \item  \begin{color}{#005500}entire optimization suite\end{color}
  \end{itemize}

  are specialized to the needs of the front end. The ability to
  consistently specialize each of these to create a back end for a
  specific language, summarizes the characteristics of MLRISC that
  distinguishes it from other retargetable backends.

  \begin{color}{#580000} It is important to emphasize that little
  optimizations performed on the MLRISC intermediate
  representation. \end{color} Most optimizations are done on a flowgraph of
  target machine instructions, to enable optimizations that take advantage
  of the characteristics of each architectural.
  The MLRISC intermediate representation is just used as a stepping 
  stone to get to the flowgraph.

root@smlnj-gforge.cs.uchicago.edu
ViewVC Help
Powered by ViewVC 1.0.0