Author: Carl Friedrich Bolz <[email protected]>
Branch: extradoc
Changeset: r4444:a32696e76d14
Date: 2012-08-07 08:31 +0200
http://bitbucket.org/pypy/extradoc/changeset/a32696e76d14/
Log: add the final version of the code
diff --git a/talk/dls2012/licm.pdf b/talk/dls2012/licm.pdf
index
6cfb629737169edbc7f5a6758cc35f3fa2862e07..b5095a01d7df94cc6bf06124503d77a8b740596a
GIT binary patch
[cut]
diff --git a/talk/dls2012/paper.tex b/talk/dls2012/paper.tex
--- a/talk/dls2012/paper.tex
+++ b/talk/dls2012/paper.tex
@@ -831,24 +831,36 @@
jump($L_1$, $p_{0}$, $i_8$)
\end{lstlisting}
-If all the optimizations presented above are applied, the resulting
-optimized peeled loop will consist of a single integer addition
+If all the optimizations presented above are applied, the resulting loop looks
+as in Figure~\ref{fig:opt-trace}.
+The resulting optimized peeled loop consists of a single integer addition
only. That is it will become type-specialized to the types of the
variables \lstinline{step} and \lstinline{y}, and the overhead of
using boxed values is removed.
-\revc{
-This paper presents an elegant, if simple, technique, and demonstrates that
-it's effective in small cases. The worked example is particularly helpful, and
-would be better if it were worked more thoroughly. Some of the omitted steps
-are not entirely obvious, and the paper would be improved by making the
-clearer. In particular, the final program presented on the bottom of page 5,
-first column, still has memory access, boxing, and type checks, which the paper
-then claims can be removed. There's enough space to show this.
-}
-\cfbolz{
-we have space now, can someone add the final optimized version of the loop?
-}
+
+\begin{figure}
+\begin{lstlisting}[mathescape,numbers =
right,basicstyle=\setstretch{1.05}\ttfamily\scriptsize]
+$L_0$($p_{0}$, $p_{1}$):
+# inside f: y = y.add(step)
+guard_class($p_{1}$, BoxedInteger)
+ # inside BoxedInteger.add
+ $i_{2}$ = get($p_{1}$, intval)
+ guard_class($p_{0}$, BoxedInteger)
+ # inside BoxedInteger.add__int
+ $i_{3}$ = get($p_{0}$, intval)
+ $i_{4}$ = $i_{2}+i_{3}$
+ # inside BoxedInteger.__init__
+jump($L_1$, $p_{0}$, $i_{4}$)
+
+$L_1$($p_{0}$, $i_{3}$, $i_{4}$):
+ $i_{8}$ = $i_{4}+i_{3}$
+jump($L_1$, $p_{0}$, $i_{3}$, $i_8$)
+\end{lstlisting}
+\caption{The fully optimized loop of the Example Interpreter}
+\label{fig:opt-trace}
+\end{figure}
+
\section{Benchmarks}
_______________________________________________
pypy-commit mailing list
[email protected]
http://mail.python.org/mailman/listinfo/pypy-commit
