In the previous log Aiming for w27 I tell how the scanning was performed. 3489660928 bytes (3.5GB) were generated by two 12-threads computers in scalar mode... Here are the results!

Let's start with the basics: the sum.

$ ./dump7ck -S log.27.p7k
opened log.27.p7k
Sum = 9007199321849850

We would expect this value:

$ echo $(( 2**((2*27)-1) + 2**(27-1) -1 ))
9007199321849855

Pretty close ! There is a very very small difference, but it's still here.

Now let's fuse all the semi-arcs...

 LOOP FOUND: len=545890867990900 arcs=8134767 at 134217705
 LOOP FOUND: len=545890867990900 arcs=8136577 at 134217724
 LOOP FOUND: len=272945433995450 arcs=4068239 at 134217710
 LOOP FOUND: len=545890867990900 arcs=8134487 at 134217718
 LOOP FOUND: len=545890867990900 arcs=8130516 at 134217723
 LOOP FOUND: len=545890867990900 arcs=8130560 at 134217725
 LOOP FOUND: len=545890867990900 arcs=8138879 at 134217726
 LOOP FOUND: len=545890867990900 arcs=8135180 at 134217632
 LOOP FOUND: len=272945433995450 arcs=4066985 at 134217666
 LOOP FOUND: len=545890867990900 arcs=8133247 at 134217720
 LOOP FOUND: len=545890867990900 arcs=8136918 at 134217648
 LOOP FOUND: len=545890867990900 arcs=8133306 at 134217711
 LOOP FOUND: len=272945433995450 arcs=4066387 at 134217721
 LOOP FOUND: len=545890867990900 arcs=8134107 at 134217659
 LOOP FOUND: len=545890867990900 arcs=8134882 at 134217719
 LOOP FOUND: len=545890867990900 arcs=8132133 at 134217727
 LOOP FOUND: len=545890867990900 arcs=8136446 at 134217695
 LOOP FOUND: len=545890867990900 arcs=8134112 at 134217709
 65 passes
789.18user 55.77system 6:08.18elapsed 229%CPU

There are 3 pairs of orbits  of length 272945433995450 and 15 pairs of orbits of length 545890867990900.

And that's it. w27 is almost well behaved but there are 5 or so states that have not been reached, probably forming a very short pair of orbits... It's not recommended since w26 behaves better.

And this heavy run took only about 6 minutes. This is because I used a pair of SSD over USB3, one for reading and the other for writing, and vice versa. The bulk of the time was taken by sort, while the fusion program was clearly I/O bound. Maybe, later, I'll try with 3 SSD because fusion reads 2 streams...