deletions | additions       diff --git a/case-study.tex b/case-study.tex  index 4ded052..c2840ad 100644  --- a/case-study.tex  +++ b/case-study.tex 
...
\caption{\label{tab:feval} Speedup comparison for \feval\ optimization.}   \end{table}     Unfortunately, we are unable to compute direct performance metrics for the solution by Lameed and Hendren since its source code has not been released. Numbers in their paper~\cite{lameed2013feval} show that for these benchmarks the speed-up of the  OSR-based approach yields is equal on average to  a fraction $30.1\%$ percentage  of the speed-up from hand-coded calls on average equal to $30.1\%$ (ranging calls, ranging  from $9.2\%$ to $73.9\%$); $73.9\%$;  for the JIT-based approach this fraction the percentage  grows to $84.7\%$ (ranging $84.7\%$, ranging  from $75.7\%$ to $96.5\%$). $96.5\%$.  Our optimization  technique [...] yields speed-ups that are very close to the upper bound from by-hand optimization. In particular, in the worst case we observe a $94.1\%$ percentage when the optimized code is JIT-compiled, which becomes $97.5\%$ when a cached version is available ({\tt odeRK4} benchmark). Compared to their OSR-based approach, more type-specialized code enabled by the compensation entry block is the key driver for better performance.   We believe our approach can generate slightly better code than their JIT-based approach