What to do after my DCs are confirmed?

edorajh · 2003-12-30, 11:22

I run one DC test and will run more on machine that has "Error: Round Off" error doing first time LL. My first DC is confirmed. If I get several more DCs confirmed what would be the best way to go:

1. to start stoped first time LL test from begining
2. to start from where I stop first time LL test (21%) when it report "Error: Round Off" errors (will chance for bad result (with "Error: Round Off" errors) be the same even if I get confirmed DCs?)

GP2 · 2003-12-30, 14:36

If you get several confirmed DCs, that's a general indication that your machine is good and not error-prone.

The answer to your question is yes. If your machine is good (non error-prone), then there is much less chance of a bad result even if you get an error reported.

For instance, here are some error rates for non-reproducible round off errors:

Code:

err code	   bad	  good	bad/total
--------	   ----	  -----	------	----
00000100            274     445         0.381
00000200             83      47         0.638
00000300             40      15         0.727

However, there is a big difference in error rates for these error codes depending on whether the machine overall is error prone or not

Code:

Machines with overall error rate >= 0.333

00000100            120      26         0.822
00000200             33       5         0.868
00000300             15       0         1.000

Code:

Machines with overall error rate < 0.333

00000100            154     419         0.269
00000200             50      42         0.543
00000300             25      15         0.625

Code:

Machines with overall error rate >= 0.100

00000100            193      68         0.739
00000200             65       9         0.878
00000300             29       1         0.967

Code:

Machines with overall error rate < 0.100

00000100             81     377         0.177
00000200             18      38         0.321
00000300             11      14         0.440

Note 00000100 means 1 "non-disregardable" round off error, 00000200 means 2, and so forth.

However, from your other thread, you said you got "several" errors like this. So even if your machine successfully completes 90% of its double checks, I'd say you have at least a 40% chance of a bad result for your stopped LL test.

edorajh · 2003-12-30, 15:28

Thanks for explanation! I will restart my LL test from begining.

2003-12-30, 11:22	#1
edorajh Oct 2003 Croatia 2³×3×19 Posts	What to do after my DCs are confirmed? I run one DC test and will run more on machine that has "Error: Round Off" error doing first time LL. My first DC is confirmed. If I get several more DCs confirmed what would be the best way to go: 1. to start stoped first time LL test from begining 2. to start from where I stop first time LL test (21%) when it report "Error: Round Off" errors (will chance for bad result (with "Error: Round Off" errors) be the same even if I get confirmed DCs?)

2003-12-30, 14:36	#2
GP2 Sep 2003 5·11·47 Posts	If you get several confirmed DCs, that's a general indication that your machine is good and not error-prone. The answer to your question is yes. If your machine is good (non error-prone), then there is much less chance of a bad result even if you get an error reported. For instance, here are some error rates for non-reproducible round off errors: Code: err code bad good bad/total -------- ---- ----- ------ ---- 00000100 274 445 0.381 00000200 83 47 0.638 00000300 40 15 0.727 However, there is a big difference in error rates for these error codes depending on whether the machine overall is error prone or not Code: Machines with overall error rate >= 0.333 00000100 120 26 0.822 00000200 33 5 0.868 00000300 15 0 1.000 Code: Machines with overall error rate < 0.333 00000100 154 419 0.269 00000200 50 42 0.543 00000300 25 15 0.625 Code: Machines with overall error rate >= 0.100 00000100 193 68 0.739 00000200 65 9 0.878 00000300 29 1 0.967 Code: Machines with overall error rate < 0.100 00000100 81 377 0.177 00000200 18 38 0.321 00000300 11 14 0.440 Note 00000100 means 1 "non-disregardable" round off error, 00000200 means 2, and so forth. However, from your other thread, you said you got "several" errors like this. So even if your machine successfully completes 90% of its double checks, I'd say you have at least a 40% chance of a bad result for your stopped LL test. Last fiddled with by GP2 on 2003-12-30 at 14:37

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
New confirmed pi(10^27), pi(10^28) prime counting function records	kwalisch	Computer Science & Computational Number Theory	34	2020-10-29 10:04

2003-12-30, 15:28	#3
edorajh Oct 2003 Croatia 2³·3·19 Posts	Thanks for explanation! I will restart my LL test from begining.