Perhaps especially we should re-work the last two lines "...it is implemented as described in the QR Decomposition essay <https://code.jsoftware.com/wiki/Essays/QR_Decomposition>. The method was chosen not because it is particularly efficient or good numerically but because it is algorithmically interesting." I'm thinking of what Roger mentioned in his email of Sep 8th on this chain.
On Sun, Sep 8, 2019 at 11:36 PM chris burke <[email protected]> wrote: > I hope someone can update > https://code.jsoftware.com/wiki/Essays/JKT_Benchmark . > > On Sun, Sep 8, 2019 at 1:14 PM Henry Rich <[email protected]> wrote: > > > For years Joey Tuttle has used the performance of (%. y) as a simple > > measure of the performance of J releases. It has improved pretty > steadily. > > > > This surprised me, because I know where the gaussian-elimination code > > is, and that code doesn't do much except C loops that do add and > > subtract. I didn't see how improvements to the JE would have any effect > > on (%. y). Not wanting to mess up a good story, I kept my doubts to > > myself. > > > > Finally I understand. The code I had been looking at is used only for > > rational matrices. The code for general (%. y) is a lovely little > > recursive QR decomposition that does lots of memory allocation and calls > > to internal J arithmetic functions. In fact, it's a pretty good > > exercise of the computational side of the system. It's quite short but > > uses general utilities so that it works on all fixed-precision numeric > > types. Bravo to Roger. He builded better than he knew: it relies on +/ > > . * for most of the numeric work, and that code has now been polished so > > well that it is the most efficient code in the system. > > > > Long live Joey's benchmark! > > > > Henry Rich > > > > --- > > This email has been checked for viruses by AVG. > > https://www.avg.com > > > > ---------------------------------------------------------------------- > > For information about J forums see http://www.jsoftware.com/forums.htm > > > ---------------------------------------------------------------------- > For information about J forums see http://www.jsoftware.com/forums.htm > -- Devon McCormick, CFA Quantitative Consultant ---------------------------------------------------------------------- For information about J forums see http://www.jsoftware.com/forums.htm
