deletions | additions       diff --git a/osr-llvm.tex b/osr-llvm.tex  index 933b734..57d86c8 100644  --- a/osr-llvm.tex  +++ b/osr-llvm.tex 
...
\fi  %\subsection{Example}  In this section we discuss how to implement the OSR approach of \mysection\ref{se:overview} in LLVM. Our discussion is based on a simple running  example that illustrates a profile-driven optimization scenario. scenario\footnote{The accompanying artifact allows the interested reader to repeat the process described in this section.}.  We start from a simple base function ({\tt isord}) that checks whether an array of numbers is ordered according to some criterion specified by a comparator (see \myfigure\ref{fi:isord-example}). Our goal is to instrument {\tt isord} so that, whenever the number of loop iterations exceeds a certain threshold, control is dynamically diverted to a faster version generated on the fly by inlining the comparator. The IR code shown in this section has been generated with \clang\ and instrumented with \osrkit, a library we prototyped to help VM builders implement OSR in LLVM\footnote{Virtual register names and labels in the LLVM-produced IR code shown in this paper have been refactored to make the code more readable.}. \osrkit\ provides a number of useful abstractions that include open and resolved OSR instrumentation of IR base functions without breaking the SSA form, liveness analysis, generation of OSR continuation functions, and mapping of LLVM values between different versions of a program along with compensation code generation. 
...
\paragraph{OSR Instrumentation in IR.}  To defer the compilation of the continuation function until the comparator is known at run time, we used \osrkit\ to instrument {\tt isord} with an open OSR point at the beginning of the loop body, as shown in \myfigure\ref{fig:isordfrom}. Portions added to the original code by OSR istrumentation are highlighted in grey.  %The figure illustrates how the original {\tt isord} code is instrumented by \tinyvm, highlighting in grey the added portions.   A new basic block is placed at the beginning of the loop body, which increments a hotness counter {\tt p.osr} and jumps to an OSR-firing block if the counter reaches the threshold (1000 iterations in this example). The OSR block contains a tail call to the target generation stub, which receives as parameters the four live variables at the OSR point ({\tt v}, {\tt n}, {\tt i}, {\tt c}). Notice that maintaining the SSA form requires adjusting $\phi$-nodes. The stub (see \myfigure[...]) calls a code generator that: 1) builds an optimized version of {\tt isord} by inlining the comparator (which is known when the OSR is fired), and 2) uses it to create the continuation function {\tt isordto} shown in \myfigure\ref{fig:isordascto}. The stub terminates with a tail call to {\tt isordto}. To generate the continuation function from the optimized version created by the inliner, \osrkit\ replaced the function entry point, removed dead code, replaced live variables with the function parameters, and fixed $\phi$-nodes accordingly. Additions resulting from the IR instrumentation are in grey, while removals are struck-through\footnote{The accompanying artifact allows the interested reader to repeat the instrumentation process described here.}. struck-through.  \ifdefined\noauthorea  \begin{figure}[t]