pipstransfo.tex

   1 \documentclass[pdftex, a4paper, 11pt]{report}
   2
   3 \usepackage[utf8]{inputenc}
   4 \usepackage[T1]{fontenc}
   5 \usepackage[english]{babel}
   6 \usepackage{lmodern}
   7
   8 \usepackage{listings}
   9 \usepackage{hyperref}
  10
  11
  12 \title{PIPS~--- List of code transformations}
  13
  14
  15
  16 \begin{document}
  17
  18
  19 \chapter{List of Pips transformations}
  20
  21 \section{Memory allocation alteration}
  22
  23 \section{Loop transformations}
  24
  25 \section{Inter-procedural transformations}
  26
  27 \section{Base blocs transformations}
  28
  29 \section{Dead code removal}
  30
  31
  32
  33
  34 \begin{description}
  35 \item[loop unrolling]{
  36   is a loop transformation. Unrolling a loop by a factor of $n$ consists in the substitution of a loop
  37   body by itself, replicated $n$ times. A prelude and/or postlude are
  38   added to preserve the number of iteration.}
  39
  40 \item[inlining]{
  41   is a function transformation. Inlining a function
  42   \texttt{foo} in its caller \texttt{bar} consists in the substitution
  43   of the calls to \texttt{foo} in \texttt{bar} by the function body
  44   after replacement of the formal parameters by their effective
  45   parameters.
  46 }
  47
  48 \item[forward substitution]{
  49   is the process of replacing a reference read in an
  50   expression by the latest expression affected to it.}
  51
  52 \item[loop fusion]{
  53   is a loop transformation that replaces two loops by a single loops whose
  54   body is the concatenation of the bodies of the two initial loops.}
  55
  56 \item[loop tiling]{
  57   is a loop nest transformation that changes the loop execution order
  58   through a partitions of the iteration space into
  59   chunks, so that the iteration is performed over each chunk and in the
  60   chunks.}
  61
  62 \item[reduction detection]{
  63   is an analysis that identifies statements that perform a reduction over a
  64   variable.}
  65
  66 \item[parallelism detection]{
  67   is a common name for analysis that detect if a loop can be run in parallel.}
  68
  69 \item[parallelism extraction]{
  70   is a common name for code transformations that modifies loop nest to make it legal to
  71   run them in parallel.}
  72
  73 \item[directive generation]{
  74   is a common name for code transformations that annotate the code with directives.}
  75
  76 \item[constant propagation]{
  77   is a pass that replaces a variable by its value when this value is
  78   known at compile time.}
  79
  80 \item[instruction selection]{
  81   is the process of mapping parts of the IR to machine instructions.}
  82
  83 \item[goto elimination]{
  84   is the process of replacing \texttt{goto} instructions by a hierarchical control flow
  85   graph.}
  86 \item[outlining]{
  87   is the process of extracting part of a function body into a new function
  88   and replacing it in the initial function by a function call.}
  89
  90 \item[common subexpression elimination]{
  91   is the process of replacing similar expressions by a variable that holds
  92   the result of their evaluation.}
  93
  94 \item[loop interchange]{is a loop transformation that permutes two
  95   loops from a loop nest.}
  96
  97 \item[loop unswitching]{is a loop transformation that replaces a
  98   loop containing a test independent from the loop execution by a test
  99   containing the loop without the test in both true and false branch.}
 100
 101 \item[statement isolation]{is the process of replacing all
 102   variables referenced in a statement by newly declared variables. A
 103   prologue
 104   and an epilogue are added to copy old variable values to new variable, back
 105   and forth.}
 106
 107 \item[dead code elimination]{is the process of pruning from a
 108   function all the statements whose results are
 109   never used.}
 110
 111 \item[array linearization]{is the process of converting
 112   multidimensional array into unidimensional arrays, possibly with a
 113   conversion from array to pointer.}
 114
 115 \item[privatization]{is the process of  detecting variables that
 116   are private to a loop body, i.e.\ written first, then read.}
 117
 118 \item[loop normalization]{is a loop transformation that changes
 119   the loop initial increment value or the loop range to enforce certain values, generally~1.}
 120
 121 \item[iteration clamping]{is a loop transformation that extends
 122   the loop range but guards the loop body with the former range.}
 123
 124 \item[flatten code]{is the process of pruning a function body from
 125   declaration blocks so that all declarations are made at the top level.}
 126
 127 \item[strength reduction]{is the process of replacing an operation
 128   by an operation of lower cost.}
 129
 130 \item[split update operator]{is the process of replacing an update
 131   operator by its expanded form.}
 132
 133 \item[n address code generation]{is the process of splitting
 134   complex expression in simpler ones that take at most $n$ operands.}
 135
 136 \item[memory footprint reduction]{is the process of tiling a loop
 137   to make sure the iteration over the tile has a memory footprint bounded by
 138   a given value.}
 139
 140 \item[redundant load-store elimination]{is an inter procedural transformation
 141   that optimizes data transfers by delaying and merging them.}
 142
 143 \item[invariant code motion]{is a loop transformation that moves outside
 144   of the loop
 145   the code from its body that is independent from the iteration.}
 146
 147
 148
 149 \item[loop rerolling]{finds manually unrolled loop and replace
 150   them by their non-unrolled version.}
 151 \end{description}
 152
 153 \end{document}