Some CRUS stats

[rogue]

The attached script addresses the extra ")" in the stats and the missing stats for the bases that haven't been started. The latter was caused by a bad edit, i.e. I put an extra character in a function that prevented it from running. The former was due to a bad assumption regarding the format of the html.

[rogue]

I had a mistake that affected the even/odd conjectures. The attached scripts will address that.

Gary, let me know if you find anything else amiss. As the even/odd conjectures have such a low weight, it might tempt people to take them on.

[gd_barnes]

Originally Posted by rogue

I understand where you are coming from, but what you are computing and what I am computing are not the same. If I change the input base to srsieve, then the range of n for the input base is no longer the same, i.e. sieving 9519*4^n-1 from 10e4 to 11e4 is not the same as sieving 9519*2^n-1 10e4 to 11e4 from 10e4 to 11e4. Sieving 9519*4^n-1 from 10e4 to 11e4 is closer to sieving 9519*2^n-1 20e4 to 22e4. My method is to sieve base 2, then only count the even/odd n in the output based upon the conjecture.

Your method is technically the most accurate way to go. But it should result in the weight and "n per 25000" being almost exactly half of what I show above are. Previously it was far higher. I see that you posted a new script for it so hopefully that is now the case. I'll edit this post with what I find.

Edit: 2.5 of the 3 problems have been fixed. The ")" issue, "not started" issue, and base 2 odd-n issues are fixed. But the base 2 even-n issue still remains. As expected, the odd-n "n per 25000" are almost exactly half of what I posted. But the even-n "n per 25000" is very high.

In looking closer, it looks like the even-n "n per 25000" is about 10 times too high. Could there be a character truncation or manipulation issue?

Thanks for the quick fixes on all of this. We're just about there. :-)

[rogue]

Originally Posted by gd_barnes

Edit: 2.5 of the 3 problems have been fixed. The ")" issue, "not started" issue, and base 2 odd-n issues are fixed. But the base 2 even-n issue still remains. As expected, the odd-n "n per 25000" are almost exactly half of what I posted. But the even-n "n per 25000" is very high.

In looking closer, it looks like the even-n "n per 25000" is about 10 times too high. Could there be a character truncation or manipulation issue?

Thanks for the quick fixes on all of this. We're just about there. :-)

Is one of the 2 even-n conjectures wrong or are both of them wrong? I would expect 162 (Riesel) and 265 (Sierpinski) as that is what I get when I run srsieve manually and remove the n we don't need. Out of curiosity, what numbers are you getting? I presume that is after you deleted the rows from the difficulty files before running the script.

[rogue]

I found the problem associated with base 2 conjectures. Apparently gawk was appending the files created when processing previous conjectures of the same base. You need to add this line to the 22_* and 23_* scripts immediately before the close of srsieve.out:

system("rm " tempFile)
close("srsieve.out")

[gd_barnes]

Originally Posted by rogue

I found the problem associated with base 2 conjectures. Apparently gawk was appending the files created when processing previous conjectures of the same base. You need to add this line to the 22_* and 23_* scripts immediately before the close of srsieve.out:

system("rm " tempFile)
close("srsieve.out")

This does not work. Both sides of even-n are still too high. I added the above "system" line in one place in the 22_ file and two places in the 23_file. I then deleted all base 2 lines from the difficulty files on both sides. Here is what is now in the difficulty files:

Riesel:
2,2,57,4403
2 even-n,2,3,2658 <-- too high
2 odd-n,2,5,477

Sierp:
2,2,6,427
2 2nd-c,2,31,2525
2 even-n,2,2,1723 <-- too high
2 odd-n,2,4,237

With each of these changes that you are having me make, the even-n values on both sides change a little bit but they stay way too high.

[rogue]

I found the cause of this issue. The temp file passed as input to srsieve is not closed. This means that the script is unable to delete the input file and then concatenates onto it when processing the "even" conjectures. You need to add:

close(tempFile)

to the function getAllK() in 23_build_unproven_list.txt immediately before the return.

[gd_barnes]

That did it. Thanks!

[rogue]

Did I fix all of the problems or are they some that I missed?

[gd_barnes]

Originally Posted by rogue

Did I fix all of the problems or are they some that I missed?

No more problems that I can see right now.

[gd_barnes]

A little closer inspection of the "CRUS statistics and progress" page revealed a couple of problems with the Sierpinski side only of the 1st table labeled "Status summary by conjecture":

1. On the "base not 2^x" column, the total should be 1019 bases (vs. 1018).

2. On the "all bases" column, the total should be 1032 bases (vs. 1033).

Note that the column in #1 adds correctly, there is just a missing base somewhere. As for the column in #2, the addition is also incorrect. If you add down, it only totals 1031 (vs. 1033). Correcting the addition still leaves it one base short just like column #1. Then...subsequently correcting #1 should fully correct #2.

The bottom line is that we have the same number of "base not 2^x" bases on both sides (1019) but we have one more "base 2^x" bases on the Sierp side (13 vs. 12) due to the Sierp base 2 2nd conjecture.