Programming puzzles: processing lists!

[DavidM]

(06-15-2017 09:12 PM)pier4r Wrote:  Nice, the performance is the one of lrot7?

Short answer: yes.

There's not as much time savings as I had originally hoped with the LRLL/LRLLD commands. Probably means that the bulk of the time is spent in the copying. That's not a big surprise, but when you consider that on a 50g the copying is done by a routine that runs at ARM level (instead of Saturn), I was hoping that the initial split point computation was a bigger part of the overall time spent. Here's how the timings look for the new LROT/LRLL/LRLLD:

\begin{array}{|lrrrrc|}
\hline
\textbf{Command} & \textbf{50} & \textbf{100} & \textbf{500} & \textbf{1000} & \textbf{Cycles} \\
\hline
\textbf{LRLL} & 0.0246 & 0.0281 & 0.0586 & 0.0959 & 100 \\
\textbf{1 LROT} & 0.0274 & 0.0308 & 0.0602 & 0.0957 & 100 \\
\hline
\textbf{LRLLD} & 0.0241 & 0.0275 & 0.0541 & 0.0867 & 100 \\
\textbf{-1 LROT} & 0.0278 & 0.0309 & 0.0567 & 0.0876 & 100 \\
\hline
\end{array}