mersenneforum.org (https://www.mersenneforum.org/index.php)
-   Conjectures 'R Us (https://www.mersenneforum.org/forumdisplay.php?f=81)

 mdettweiler 2009-01-06 20:14

This is a thread where people can post LLR/PRP/etc. residuals that, though not abnormal, have a "wacky" look to them. I'll start us off with one I saw in a PRPnet client a few minutes ago:

74612*6^122915+1 is composite LLR64=[B]1390f22f5bddbbdd[/B]. (t=532.50s)

Rarely do you ever see such a repeating pattern--and, incidentally, "b" becomes a "d" if you take its mirror image, which makes this residual catch the eye more so than other "wacky" ones. :smile:

 henryzz 2009-01-06 20:17

[quote=mdettweiler;157243]This is a thread where people can post LLR/PRP/etc. residuals that, though not abnormal, have a "wacky" look to them. I'll start us off with one I saw in a PRPnet client a few minutes ago:

74612*6^122915+1 is composite LLR64=[B]1390f22f5bddbbdd[/B]. (t=532.50s)

Rarely do you ever see such a repeating pattern--and, incidentally, "b" becomes a "d" if you take its mirror image, which makes this residual catch the eye more so than other "wacky" ones. :smile:[/quote]
now your job should be to persuade ironbits to write a program to search through all the llrnet results for interesting residues:smile:

 mdettweiler 2009-01-06 20:24

[quote=henryzz;157244]now your job should be to persuade ironbits to write a program to search through all the llrnet results for interesting residues:smile:[/quote]
LOL--actually, I imagine it would be somewhat difficult to write such a program, since it's hard for a computer to determine what looks "interesting" to the human eye. Though, yes, it might be an interesting thing to try...and, hey, in that case, why limit yourself to LLRnet results? Gary's got a whole collection of results from the NPLB 1st, 2nd, 3rd, and 4th drives (which includes both manual and LLRnet results), not to mention all the CRUS results; all those base 3 efforts have probably produced millions of residuals by now. :smile:

 Flatlander 2009-01-06 20:56

I can't believe you've got me trawling through an old lresults file. :redface:

Who can find the longest word? (Or who is going to write a program to find them?)

 Mini-Geek 2009-01-06 21:17

Another thing to look for when you look for 'interesting' residuals: related words coded in the hex.
[URL]http://centricle.com/tools/ascii-hex/[/URL] or [URL]http://www.dolcevie.com/js/converter.html[/URL]
e.g. prime is 7072696d65, NPLB is 4e504c42, CRUS is 43525553, !prime (not prime, which is true for any number with this in the residual because it has a non-zero residual :wink:) is 217072696d65, 74612 (Max's example's k, this portion of residue would be interesting only if found in this specific residue, of course...or you could look for these actual numbers instead of the residue of their ASCII equivalent, or you could convert it to hex, 12374 and look for that) is 3734363132, etc. etc. you get the idea. :smile:
[quote=Flatlander;157251]I can't believe you've got me trawling through an old lresults file. :redface:

Who can find the longest word? (Or who is going to write a program to find them?)[/quote]
[URL]http://nedbatchelder.com/text/hexwords.html[/URL]
Look for some of those in residues.
Edit: I have many lresults files in a folder and subfolders and I'm trying to get them all into one big file so that I can look for interesting residuals more easily, especially with looking for hexwords and hex-encoded words as described above. I know I can use the windows command line "copy" app to copy several files into one, (by specifying many source files and one destination file) but the problem is that, as you might expect, they're named very similarly (I have four names total in case you're curious, lresults.txt, lresults1.txt, lresults2.txt, and lresults13.txt) and I have a lot of them, so I can't just dump them all into a folder together to find it easily. Any ideas how I can do this more easily than manually renaming each one to something different? BTW I can get them all in a list together by searching in the folder for anything with lresults in the file name, so if something needs to be done from that, it can be done.
:blahblah: In short, how can I combine lresults files?

 MooooMoo 2009-01-06 22:35

313*2^714[b]999[/b]-1 is not prime. LLR Res64: 931897[b]CCCCC[/b]047D5 Time : 847.984 sec.

First you've got the 999 in the exponent, then you've got the 5 C's in a row in the residue :surprised

 MrOzzy 2009-01-07 08:06

[COLOR=black]A few nice ones:[/COLOR]

[COLOR=red]222[/COLOR]*117^[COLOR=red]22[/COLOR]989-1 is not prime. RES64: 657164CA8828099D. OLD64: 0169[COLOR=red]22222[/COLOR]B9C412A Time: 200.045 sec.
33[COLOR=red]6[/COLOR]8*117^2[COLOR=red]6[/COLOR]0[COLOR=red]66[/COLOR]-1 is not prime. RES64: 92[COLOR=red]66666[/COLOR]BCD0D45B9
1[COLOR=red]8[/COLOR]16*117^1931-1 is not prime. RES64: FECB3[COLOR=red]88888[/COLOR]AC5DE9
3128*117^20527-1 is not prime. RES64: 09C35A7DA6[COLOR=red]BBBBB[/COLOR]1
3480*117^21566-1 is not prime. RES64: 85C2A[COLOR=red]99999[/COLOR]B6BB8A. OLD64: 9147F[COLOR=red]CCCC[/COLOR]D24329B Time: 375.869 sec

 henryzz 2009-01-07 08:37

[quote=mdettweiler;157246]LOL--actually, I imagine it would be somewhat difficult to write such a program, since it's hard for a computer to determine what looks "interesting" to the human eye. Though, yes, it might be an interesting thing to try...and, hey, in that case, why limit yourself to LLRnet results? Gary's got a whole collection of results from the NPLB 1st, 2nd, 3rd, and 4th drives (which includes both manual and LLRnet results), not to mention all the CRUS results; all those base 3 efforts have probably produced millions of residuals by now. :smile:[/quote]
i forgot gary had all of the residues for everything
surely looking for groups of digits wouldnt be too difficult

 kar_bon 2009-01-07 09:25

[QUOTE=Mini-Geek;157255]I know I can use the windows command line "copy" app to copy several files into one, (by specifying many source files and one destination file) but the problem is that, as you might expect, they're named very similarly (I have four names total in case you're curious, lresults.txt, lresults1.txt, lresults2.txt, and lresults13.txt) and I have a lot of them, so I can't just dump them all into a folder together to find it easily. Any ideas how I can do this more easily than manually renaming each one to something different? BTW I can get them all in a list together by searching in the folder for anything with lresults in the file name, so if something needs to be done from that, it can be done.
:blahblah: In short, how can I combine lresults files?[/QUOTE]

the command is correct:

try: help copy
and you got an overview of the parameters.

try
copy lresults.txt+lresults1.txt+lresults2.txt+lresults13.txt all.txt
and all 4 result-files will be copy to one named "all.txt"

an easier way:
copy *.txt all.txt
will copy all *.txt-file in the directory to 'all.txt'

done!

 kar_bon 2009-01-07 09:31

perhaps other goals to search for:

- find the lowest (\$000000...) or highest (\$FFFFFF...) value of a residue

- make a graph of the distribution of residues
so take 1 million residues, sort them by value and count their rate (perhaps too low for only 1M)
question: is this equally distributed? or some patterns?

- convert the residues in decimal and find primes :-)

-...

 gd_barnes 2009-01-07 12:24

Oh, you guys are just NUTS!!

Karsten, I about fell over laughing at your last suggestion. Converting residuals to decimal and finding primes. Now THAT is funny!!

I'm quite impressed by Max's first one here but I'm sorry, I can't get too impressed with only 5 repeating digits in a residual. You'll have to make it 6 or more repeating digits...7 would be better.

Sorry...no time to sift through my 100's of thousands or millions of residuals right now. I may do so at some point.

BTW, we don't collect residuals on any tests for n<25K on base 3. That would be stupid! lol The primes themselves are huge enough.

Here's a question for the higher-math types: Can you reverse engineer a residual? That is: I'll post a residual and people can tell me what test(s) it applies to. I assume that it could apply to an infinite # of tests but the percentage of total tests would be very small. I'm sure if you kept the k-value, base, and n-value relatively small when doing the reverse engineering that it would only apply to just a few possible tests.

That would be an interesting question to pose in some sort of "puzzles" forum.

Gary

All times are UTC. The time now is 04:14.