Try to improve the run of a factor-1.4 paragraph

results.tex

@@ -54,6 +54,7 @@ This makes it possible to write
 a Haswell implementation that is significantly faster than the others.
 
 \subheading{The cost of completeness.}
+
 Another question we might be able to answer is if the factor-$1.4$ penalty
 claimed in~\cite{RCB16}---for complete formulas vs.\ incomplete formulas---%
 % On versus as 'vs.': https://english.stackexchange.com/a/5395
@@ -63,12 +64,38 @@ for optimized implementations.
 
 In~\cite{BCLN16}, Bos, Costello, Longa, and Naehrig present performance results
 for scalar multiplication on a prime-order Weierstraß curve over $\mathds{F}_{2^{256}-189}$
-using parameter $a=-3$. The implementation uses non-complete formulas for addition
-and doubling and they report $278\,000$ for variable-base scalar multiplication
-on Intel Sandy Bridge. 
-The curve is very similar to Curve13318, the software in~\cite{BCLN16} is
-seriously optimized and also claimed to run in constant time, so these 
+using parameter $a=-3$.
+The curve is very similar to Curve13318 and
+the implementation uses non-complete formulas for addition
+and doubling.
+The authors report $278\,000$ cycles for variable-base scalar multiplication on Intel Sandy Bridge.
+The software in~\cite{BCLN16} is seriously optimized, and claimed to
+run in constant time,
+so these 
 $278\,000$ cycles are reasonably comparable to our $389\,546$ cycles with
-complete formulas. 
-In other words, this comparison confirms the factor-$1.4$ performance-penalty
+complete formulas.
+In other words, this comparison affirms the factor-$1.4$ performance-penalty
 claim from~\cite{RCB16}.
+
+% -Aside from comparing the Renes-Costello-Batina formulas to Curve25519's formulas,
+% -we can also look briefly at other prime-order curve implementations,
+% -in order to see how the formulas fare against others.
+%
+% [...]
+% 
+% -In their paper, Renes, Costello, and Batina reported the overhead to be $1.38\times$.
+% -On the other hand, \cite{BCLN16},
+% -implement a similar $a=-3$ curve over $\mathds{F}_{2^{256}-189}$
+% -(which they call ``w-256-mers'');
+% -they chose to use incomplete formulas for their implementation,
+% -as using complete formulas would incur a performance cost of a factor 2.\footnote{%
+% -A relevant note to this estimate is that it was constructed \emph{before} \cite{RCB16} was published.
+% -I.e.\ it is based on the formulas from Bosma and Lenstra~(\cite{BL95}),
+% -\emph{not} those from Renes, Costello, and Batina~(\cite{RCB16}).
+% -}
+% -Their variable-basepoint scalar-multiplication runs in $278\unit{kcc}$
+% -on the Sandy Bridge microarchitecture.
+% -Comparing that measurement to ours, suggests that the complete formulas add---%
+% -relative to their incomplete formulas based on conditional masking---%
+% -an overhead of about $40\%$,
+% -which strongly affirms the overhead measured by Renes, Costello, and Batina.