Author: Carl Friedrich Bolz <[email protected]>
Branch: extradoc
Changeset: r4445:3da0148e6fdd
Date: 2012-08-07 08:37 +0200
http://bitbucket.org/pypy/extradoc/changeset/3da0148e6fdd/
Log: make all traces figures
diff --git a/talk/dls2012/paper.tex b/talk/dls2012/paper.tex
--- a/talk/dls2012/paper.tex
+++ b/talk/dls2012/paper.tex
@@ -623,8 +623,9 @@
the preamble might make the guards of the peeled loop
redundant and thus removed. Therefore one effect of combining redundant
guard removal with loop peeling is that loop-invariant guards are moved out of
the
-loop. The peeled loop of the example reduces to
+loop. The peeled loop of the example reduces to the trace in
Figure~\ref{fig:guard-trace}.
+\begin{figure}
\begin{lstlisting}[mathescape,numbers =
right,basicstyle=\setstretch{1.05}\ttfamily\scriptsize]
$L_1$($p_{0}$, $p_{5}$):
# inside f: y = y.add(step)
@@ -638,6 +639,9 @@
set($p_{9}$, intval, $i_{8}$)
jump($L_1$, $p_{0}$, $p_{9}$)
\end{lstlisting}
+\caption{Peeled loop after redundant guard removal}
+\label{fig:guard-trace}
+\end{figure}
The guard on $p_5$ on line 17 of Figure~\ref{fig:peeled-trace} can be
removed since $p_5$ is allocated on line 10 with a known class. The
@@ -673,8 +677,10 @@
form, because the argument $i_3$ is not passed along the loop arguments.
Therefore $i_3$ needs to be added to the loop arguments.
-Doing this, the trace from Figure~\ref{fig:peeled-trace} will be optimized to:
+Doing this, the trace from Figure~\ref{fig:peeled-trace} will be optimized to
+the trace in Figure~\ref{fig:cse-trace}.
+\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)
@@ -703,6 +709,9 @@
set($p_{9}$, intval, $i_{8}$)
jump($L_1$, $p_{0}$, $p_{9}$, $i_3$)
\end{lstlisting}
+\caption{Trace after common subexpression elimination}
+\label{fig:cse-trace}
+\end{figure}
After loop peeling and redundant operation removal the peeled loop
will typically no longer be in SSA form but operate on variables that are the
result
@@ -734,7 +743,7 @@
In the optimized trace $J$ is replaced by $\hat J$ and $K$ by $\hat
K$.
-\subsection{Allocation Removals}
+\subsection{Allocation Removal}
\label{sub:allocation}
RPython's allocation removal optimization~\cite{bolz_allocation_2011} makes it
@@ -805,8 +814,10 @@
.
\end{equation}
In the optimized trace $I$ is replaced by $\hat I$ and $K$ by $\hat
-K$. The trace from Figure~\ref{fig:unopt-trace} will be optimized into
+K$. The trace from Figure~\ref{fig:unopt-trace} will be optimized to
+the trace in Figure~\ref{fig:virtual-trace}.
+\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)
@@ -830,6 +841,9 @@
# inside BoxedInteger.__init__
jump($L_1$, $p_{0}$, $i_8$)
\end{lstlisting}
+\caption{Trace after allocation removal}
+\label{fig:virtual-trace}
+\end{figure}
If all the optimizations presented above are applied, the resulting loop looks
as in Figure~\ref{fig:opt-trace}.
_______________________________________________
pypy-commit mailing list
[email protected]
http://mail.python.org/mailman/listinfo/pypy-commit
