Database for k-b-b's: - Page 3

3.14159 · 2010-08-30, 17:39

Update: The bug is persisting for k = 410 and 411. This is going to be a major issue if this persists.

I should either switch to LLR, or go to the original Pform.

Let's see if this is because I keep using the same filename. Nope. Error persists for three consecutive values. Computer error or program error.

The error happens at the 4000s and 5000s range.

Admittedly, I have left it doing PFGW work for 3-5 days on end.

The above should be irrelevant. It's either a program error or computer error.

kar_bon · 2010-08-30, 17:45

It's your turn to count them on this page!

science_man_88 · 2010-08-30, 17:52

Quote:

Originally Posted by kar_bon

It's your turn to count them on this page!

well I used find and found 1023 comma in the lists so I know it's greater than 1024. if I did my math correct 15 list each of which is missing a , at the end to help me if they were in place it would have 1023+15 comma and 1023+15+1 = 1039 k values.

3.14159 · 2010-08-30, 17:56

Let me try Multisieve, to see whether or not the problem is that NewPGen isn't supposed to be used for k-b-b's.

Confirmed: No errors with Multisieve. *massive sigh of relief*

Disproven: The error persists.

It boils down to:

1. Program error.
2. Computer has been damaged.

I'll use the -a1 flag for now. I'm guessing that is the n-1 test flag.

mdettweiler · 2010-08-30, 18:10

Quote:

Originally Posted by 3.14159

Let me try Multisieve, to see whether or not the problem is that NewPGen isn't supposed to be used for k-b-b's.

Confirmed: No errors with Multisieve. *massive sigh of relief*

Hardware errors often don't trigger any issues with sieving programs; they're a lot less hard on the CPU. Also, if a factor is missed, nobody will know that there wasn't supposed to be a factor there; you'll just have one more PRP test in your output file that you'll have to do. With primality testing, though, if the test is inaccurate, you could miss a prime.

Quote:

Disproven: The error persists.

It boils down to:

1. Program error.
2. Computer has been damaged.

I'll use the -a1 flag for now. I'm guessing that is the n-1 test flag.

-a1 still does a regular PRP test, but it used the next higher FFT length. That makes it more resistant to roundoff errors, but also a bit slower. (The slowness becomes more noticeable as tests get bigger.)

Regarding this particular error, I'm not sure what you mean; are you saying that it didn't show up when you retested 3460*308^308+1 using the latest PFGW, or did? Also, whether you get errors on other numbers as well does not give you any further information on whether the errors are hardware or software.

3.14159 · 2010-08-30, 18:13

Quote:

Originally Posted by Max

Regarding this particular error, I'm not sure what you mean; are you saying that it didn't show up when you retested 3460*308^308+1 using the latest PFGW, or did? Also, whether you get errors on other numbers as well does not give you any further information on whether the errors are hardware or software.

I'm unsure whether or not I damaged the CPU or anything related. Checking the computer showed some signs of the fans being a bit too hot (Approx. body temp.)

CRGreathouse · 2010-08-30, 18:13

Quote:

Originally Posted by mdettweiler

-a1 still does a regular PRP test, but it used the next higher FFT length. That makes it more resistant to roundoff errors, but also a bit slower. (The slowness becomes more noticeable as tests get bigger.)

Do you have any numbers on this? I've heard that it's slower but I don't even know by what order of magnitude (0.1%, 1%, 10%, 100%?).

3.14159 · 2010-08-30, 18:25

Also: Karsten: Here is the k-b-b, 60 to 250 file. Can you please update? Thank you.
kbbfile.txt

I'm up to b = 450.

CRGreathouse · 2010-08-30, 19:03

Quote:

Originally Posted by 3.14159

I'm unsure whether or not I damaged the CPU or anything related. Checking the computer showed some signs of the fans being a bit too hot (Approx. body temp.)

That's not a bad temperature at all, but the CPU could be damaged for other reasons. Alternately, it could be memory rather than CPU, if the data was large enough to spill out of the L3 cache.

3.14159 · 2010-08-30, 19:06

Quote:

Originally Posted by CRGreathouse

That's not a bad temperature at all, but the CPU could be damaged for other reasons. Alternately, it could be memory rather than CPU, if the data was large enough to spill out of the L3 cache.

Well, it's been working error free, since b = 440 without the a1 flag.

It happens to do this about every 100 values, as far as I know.

3.14159 · 2010-08-30, 19:08

I'm guessing it might do it again when b = 500 to 510.

2010-08-30, 17:39	#23
3.14159 May 2010 Prime hunting commission. 2⁴·3·5·7 Posts	Update: The bug is persisting for k = 410 and 411. This is going to be a major issue if this persists. I should either switch to LLR, or go to the original Pform. Let's see if this is because I keep using the same filename. Nope. Error persists for three consecutive values. Computer error or program error. The error happens at the 4000s and 5000s range. Admittedly, I have left it doing PFGW work for 3-5 days on end. The above should be irrelevant. It's either a program error or computer error. Last fiddled with by 3.14159 on 2010-08-30 at 17:54

2010-08-30, 17:56	#26
3.14159 May 2010 Prime hunting commission. 2⁴×3×5×7 Posts	Let me try Multisieve, to see whether or not the problem is that NewPGen isn't supposed to be used for k-b-b's. Confirmed: No errors with Multisieve. massive sigh of relief Disproven: The error persists. It boils down to: 1. Program error. 2. Computer has been damaged. I'll use the -a1 flag for now. I'm guessing that is the n-1 test flag. Last fiddled with by 3.14159 on 2010-08-30 at 18:05

2010-08-30, 18:25	#30
3.14159 May 2010 Prime hunting commission. 2⁴×3×5×7 Posts	Also: Karsten: Here is the k-b-b, 60 to 250 file. Can you please update? Thank you. kbbfile.txt I'm up to b = 450. Last fiddled with by 3.14159 on 2010-08-30 at 18:48

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
CPU Performance Database	JustinGC	Hardware	3	2012-06-22 03:11
Database question	Xyzzy	Forum Feedback	0	2012-04-01 15:37
entire database done up to 2^58	tha	Lone Mersenne Hunters	11	2010-10-17 08:27
NPLB Database	IronBits	No Prime Left Behind	177	2009-10-10 09:00
database.zip	HiddenWarrior	Data	1	2004-03-29 03:53

2010-08-30, 17:45	#24
kar_bon Mar 2006 Germany 101101011000₂ Posts	It's your turn to count them on this page!

2010-08-30, 19:08	#33
3.14159 May 2010 Prime hunting commission. 2⁴·3·5·7 Posts	I'm guessing it might do it again when b = 500 to 510.