Bases 251-500 reservations/statuses/primes

[gd_barnes]

That's cool Max. I had thought of doing a few k=6 conjectures also previously.

I was asking for parallel testing. These are brand new bases so there was nothing to parallel with. I'm sure they are correct but I just wanted to point out the difference between new testing and parallel testing.

It's best to test new and changed software with "test data" not "production data". Running new bases with newly changed software means you have "tested" it in "production", which is not a good idea.

If you'd like to help me out here, how about running the previous version of the script against these bases also and see if you get the same results? Then we'll have a true parallel test.

[mdettweiler]

Originally Posted by gd_barnes

That's cool Max. I had thought of doing a few k=6 conjectures also previously.

I was asking for parallel testing. These are brand new bases so there was nothing to parallel with. I'm sure they are correct but I just wanted to point out the difference between new testing and parallel testing.

It's best to test new and changed software with "test data" not "production data". Running new bases with newly changed software means you have "tested" it in "production", which is not a good idea.

If you'd like to help me out here, how about running the previous version of the script against these bases also and see if you get the same results? Then we'll have a true parallel test.

Sure. I don't have the version 4.2 script; could you send that one to me? (Unless you meant for me to test it with 4.1? I'd have to dig up an older version of PFGW, but I could do it if you'd like.)

[gd_barnes]

Originally Posted by mdettweiler

Sure. I don't have the version 4.2 script; could you send that one to me? (Unless you meant for me to test it with 4.1? I'd have to dig up an older version of PFGW, but I could do it if you'd like.)

Yeah, use 4.1 if you can find PFGW 3.2.3 or earlier. That would be a good parallel test. It should still write out the appropriate GFNs. It wouldn't have the compPRP and PRP files but I seriously doubt that those will apply for such small conjectured bases.

[mdettweiler]

Originally Posted by gd_barnes

Yeah, use 4.1 if you can find PFGW 3.2.3 or earlier. That would be a good parallel test. It should still write out the appropriate GFNs. It wouldn't have the compPRP and PRP files but I seriously doubt that those will apply for such small conjectured bases.

Hmm...I'm having a hard time locating a copy of 3.2.3 (I think the binaries may have been taken offline), though I do have a really old copy sitting around (the latest of the "old" versions, i.e. pre-3.0). The speed differences shouldn't be an issue for numbers this small; should the script be compatible with that version? (As I recall, no changes were made in the area of scripts until 3.2.5 or so, but I could be wrong.)

[gd_barnes]

Originally Posted by mdettweiler

Hmm...I'm having a hard time locating a copy of 3.2.3 (I think the binaries may have been taken offline), though I do have a really old copy sitting around (the latest of the "old" versions, i.e. pre-3.0). The speed differences shouldn't be an issue for numbers this small; should the script be compatible with that version? (As I recall, no changes were made in the area of scripts until 3.2.5 or so, but I could be wrong.)

As far as I know, that should work. The first good release that the new script changes worked with is 3.2.7. The last for version 4.1 of the script is 3.2.3.

[mdettweiler]

Originally Posted by gd_barnes

As far as I know, that should work. The first good release that the new script changes worked with is 3.2.7. The last for version 4.1 of the script is 3.2.3.

Okay, here's what I got:

Sierp. base 279: same output in the files, but I'm not positive if it actually proved the PRP for 2*279^4+1. Here's the console output:

PFGW Version 1.2.0 for Windows [FFT v23.8]
Factoring numbers to 30% of normal.
Script File
2*279^1+1 trivially factors as: 13*43
2*279^2+1 trivially factors as: 11*14153
2*279^3+1 trivially factors as: 113*384383
trial factoring to 19660
Switching to Exponentiating using GMP
2*279^4+1 is 3-PRP! (0.0000s+0.0081s)
4*279^1+1 trivially factors prime!: 1117

2*279^4+1 shows up in pl_prime.txt, however, so if the script did prove it, it just didn't say so on the screen.

Sierp. base 349: same as before

Sierp. base 384: similar to what happened with S279. 4*384^21+1 was shown as 3-PRP on screen, and was listed in pl_prime.txt, but I didn't see proof of a proof (no pun intended) on screen.

Sierp. base 454: same as before

Sierp. base 489: similar to S279 and S384. 4*489^5+1 was shown as 3-PRP on screen and listed in pl_prime.txt, but not shown on screen as N-1'ed or factored prime.

Note that, in all the cases where the 4.1 script neglected to log a proof to the screen, the 4.3 script quite clearly showed that it proved the primes in one way or another after having initially found them 3-PRP.

At any rate, the results seem to be resoundingly good for 4.3: the only potential problems I turned up were in 4.1, and were for issues which I believe were the primary issues addressed in 4.3, which handled them exactly as it should.

[gd_barnes]

Originally Posted by mdettweiler

Okay, here's what I got:

Sierp. base 279: same output in the files, but I'm not positive if it actually proved the PRP for 2*279^4+1. Here's the console output:

PFGW Version 1.2.0 for Windows [FFT v23.8]
Factoring numbers to 30% of normal.
Script File
2*279^1+1 trivially factors as: 13*43
2*279^2+1 trivially factors as: 11*14153
2*279^3+1 trivially factors as: 113*384383
trial factoring to 19660
Switching to Exponentiating using GMP
2*279^4+1 is 3-PRP! (0.0000s+0.0081s)
4*279^1+1 trivially factors prime!: 1117

2*279^4+1 shows up in pl_prime.txt, however, so if the script did prove it, it just didn't say so on the screen.

Sierp. base 349: same as before

Sierp. base 384: similar to what happened with S279. 4*384^21+1 was shown as 3-PRP on screen, and was listed in pl_prime.txt, but I didn't see proof of a proof (no pun intended) on screen.

Sierp. base 454: same as before

Sierp. base 489: similar to S279 and S384. 4*489^5+1 was shown as 3-PRP on screen and listed in pl_prime.txt, but not shown on screen as N-1'ed or factored prime.

Note that, in all the cases where the 4.1 script neglected to log a proof to the screen, the 4.3 script quite clearly showed that it proved the primes in one way or another after having initially found them 3-PRP.

At any rate, the results seem to be resoundingly good for 4.3: the only potential problems I turned up were in 4.1, and were for issues which I believe were the primary issues addressed in 4.3, which handled them exactly as it should.

There were no issues in 4.1. 4.3 only resolved extremely rare issues in 4.2. At the time of 4.1, the scripting language had no way to prove PRPs so 4.1 did all that it could and assume that PRPs were prime and left it up to the user to prove them.

There is nothing extraordinary about 4.3 by any means. The changes that you're observing with the proof of the PRPs started with 4.2 and PFGW 3.2.7. 4.2 has been around for several weeks now.

All that 4.3 did is resolve issues in extremely rare situations where a PRP came through the PRIMEM/PRIMEP primality proof as STILL PRP and on even more rare situations where it came through as a composite PRP yet really wasn't composite and factoring and or combined testing with PRIMEC could find it to be prime.

I'm certain that none of the very rare situations that 4.3 resolved cropped up in your testing here (hence why nothing extraordinary). I had to actually change the max factoring to a lower limit while testing 4.3 just to get some of the situations to hit such rare conditions in base 3.

The main thing that I wanted to verify is that the GFN, MOB, primes, and k's remaining files are all the same. Since there are no MOB or k's remaining for these bases, more parallel testing is needed.

BTW, screen output, the pfgw.log file, and the pfgw-prime.log file are useless when running the script. (Sometimes the pfgw.out file can be used for debugging certain situations.) Only the pl_ prefix files are applicable. If you come across a MOB, you'll see extra primes coming through those pfgw.log and pfgw-prime.log files, which don't pertain to the conjecture, hence they should just be deleted when you are through.

I have to admit, it's fun being the programmer doing the testing and debugging for a change. :-)


Gary

[mdettweiler]

Ah, okay, that makes sense. I figured the screen output was somewhat irrelevant, but nonetheless thought it worth mentioning, just in case.

BTW, for future parallel testing, is it worth doing that by comparing 4.1/4.3 as I did here, or would it really need to be 4.2/4.3 in order to be useful?

[gd_barnes]

Originally Posted by mdettweiler

Ah, okay, that makes sense. I figured the screen output was somewhat irrelevant, but nonetheless thought it worth mentioning, just in case.

BTW, for future parallel testing, is it worth doing that by comparing 4.1/4.3 as I did here, or would it really need to be 4.2/4.3 in order to be useful?

4.1 vs. 4.3 is just as relevant.

Please keep in mind to run already tested bases first, else you're "testing" in "production". For example, take both sides of bases 512 or 1024 and run them in 4.1 and then again in 4.3 to see if the files match.

After several previously tested bases come back as in sync, then you can run some new bases. The point of parallel testing is to verify that changes to existing software have not adversely affected what the prior software has already come up with. Once that is done, the new software can be rolled out on new data or as is the case here, new bases.

In other words, I would classify this as a late stage beta test.


Gary

[MyDogBuster]

Reserving Riesel base 288 as new to n=25K

[MyDogBuster]

Riesel Base 288
Conjectured k = 613
Covering Set = 5, 17, 53
Trivial Factors k == 1 mod 7(7) and k == 1 mod 41(41)

Found Primes: 505k's File attached

Remaining k's: 6k's Tested to n=25K
16*288^n-1 <---------- Proven composite by partial algebraic factors
18*288^n-1
339*288^n-1
392*288^n-1
441*288^n-1 <---------- Proven composite by partial algebraic factors
509*288^n-1

Trivial Factor Eliminations: 99k's

MOB Eliminations:
576

Base Released