Camil Demetrescu  over 8 years ago

Commit id: 13a5181c4ef49141f0437f88116f848a064a1532

deletions | additions      

       

\section{Conclusions}  \label{se:conclusions}  In this paper, we have proposed a combination of features of extant OSR techniques that no previous solution provided simultaneously. Relevant aspects include platform independence~\cite{lameed2013modular}, generation of highly optimized continuation functions~\cite{fink2003design}, and performing deoptimization without the need for an interpreter~\cite{bebenita2010spur}. A Two  novel abstraction features  we propose is are  OSR with compensation code, which allows to extend the range of points where OSR transitions can be fired. fired, and the ability to inject OSR points at arbitrary locations.  Using this technique, these features,  we have shown how to improve the state of the art of \feval\ optimization in MATLAB virtual machines. We have also investigated the feasibility of our approach in LLVM, showing that it is efficient in practice. \ifx\noauthorea\undefined  \paragraph{Acknowledgements.}