Certificate : For all indents and purposes - Page 7

Jayder · 2016-11-01, 07:37

I recently built a dual-socket machine and I was considering trying to make a dent in the certificates under 3000dd, but the lack of support is a real downer.

schickel · 2016-11-14, 15:07

Quote:

Originally Posted by RichD

If you are using the latest Primo v.4.2.1, factordb doesn’t recognize them. The verifier supposedly would check them but it appears the parser, or gatekeeper, won’t allow them to be passed to the verifier. I don’t know of a site that archives the older Primo v.4.1.1.

Quote:

Originally Posted by Jayder

That would be the case. Thank you for telling me. The latest I have is 4.0.1. Anybody have anything later?

Was there a change between 4.0.1 and 4.1.1 that the verifier didn't like? I had been using 4.0.0 a13 back in 2012 for certs.

ETA: also, what are the chances that the PRPs I've been clearing might have beaten yours in? I've been working on the backlog for a couple of weeks ~1000-1500 at a time (from the bottom; ~10,000 cleared so far).

Jayder · 2016-11-15, 02:24

Quote:

Originally Posted by schickel

Was there a change between 4.0.1 and 4.1.1 that the verifier didn't like? I had been using 4.0.0 a13 back in 2012 for certs.

ETA: also, what are the chances that the PRPs I've been clearing might have beaten yours in? I've been working on the backlog for a couple of weeks ~1000-1500 at a time (from the bottom; ~10,000 cleared so far).

Take a look at the changelog. I was using 4.2.1. 4.2.0 introduced a new certificate format. 4.1.1 is the most recent version that still works, but I only had 4.0.1 available to me.

This is undoubtedly the issue.

I was thinking that I would email Markus and see if he can take another look at the parser.

schickel · 2016-11-15, 03:41

Quote:

Originally Posted by Jayder

Take a look at the changelog. I was using 4.2.1. 4.2.0 introduced a new certificate format. 4.1.1 is the most recent version that still works, but I only had 4.0.1 available to me.

This is undoubtedly the issue.

I was thinking that I would email Markus and see if he can take another look at the parser.

so then I guess my question is why not use 4.0.1, or do you mean you updated to 4.2.1 and don't have 4.0.1?

Jayder · 2016-11-15, 07:03

The most recent version I had was 4.0.1, but a kind soul gave me a download for 4.1.1, so this conversation is a bit moot.

I would have used it, though, if that was all that was available to me. But wouldn't you agree that it is nice to have the most recent version of a program? (Maybe we won't include Windows here.) I don't know how much the new discriminants help, but there were quite a few added after 4.0.1, not to mention a few minor features and bug fixes. Is there a reason why you're so curious?

Jayder · 2016-11-24, 08:18

Good news, folks! Markus has informed me that he has fixed the bug which caused format 4 certificates to be rejected. I have uploaded the format 4 certificates that I have and can confirm that it works splendidly. Thank you, Markus!

In a very small test, I have found that version 4.2.1 is at least 10% faster than 4.1.1. Well worth upgrading.

I am going to do some certificate testing above 2000 digits. Where is everybody working right now?

henryzz · 2016-11-24, 23:11

Quote:

Originally Posted by Jayder

Good news, folks! Markus has informed me that he has fixed the bug which caused format 4 certificates to be rejected. I have uploaded the format 4 certificates that I have and can confirm that it works splendidly. Thank you, Markus!

In a very small test, I have found that version 4.2.1 is at least 10% faster than 4.1.1. Well worth upgrading.

I am going to do some certificate testing above 2000 digits. Where is everybody working right now?

Gonna try and make a dent in the 985 digit prps.

Jayder · 2016-11-27, 08:18

An important note about Primo: Disregard my previous comment(s) about 4.2.1. I have performed a test on 195 2000dd PRPs, comparing Primo 4.2.1 and 4.1.1 with 48 concurrent tasks.

When comparing wall-clock time, I have found that 4.2.1 took 5% longer than 4.1.1 to complete the batch. Phase 1 is faster with 4.2.1, taking only 92% of the time that 4.1.1 required. The problem lies with Phase 2, which took 43% longer than 4.1.1.

When comparing process time, I have found that 4.2.1 completed the batch in 90% of the time needed by 4.1.1. Phase 1 is faster with 4.2.1, taking only 87% of the time that 4.1.1 required. Phase 2 took 8% longer than 4.1.1.

For 4.2.1, the big difference in Phase 2 time when comparing wall-clock time and process time is explained by there being polynomials that are harder to factor (or so I believe). This leaves a few threads running for a long time (sometimes for half the duration of Phase 2) on a few harder polynomials, while the rest of the threads sit idle.

Somebody please correct me if I'm wrong, but I believe the takeaway is this: If you run multiple instances of Primo and are able to max out your processor(s), use 4.2.1 over 4.1.1. If you are running only one instance and wall-clock time matters more, use 4.1.1. If you are a crazy person, use 4.2.1 for Phase 1 and 4.1.1 for Phase 2.

Data for those interested (all three of you) can be found here:

EDIT: I forgot to mention that 4.2.1 produces smaller certificates. The 195 certificates produced by 4.2.1 take up 60.1MiB of space, while 4.1.1's 195 certificates take up 90.7MiB.

Jayder · 2016-11-27, 08:24

I am currently working on 990-993. I was a bit dumb and didn't check the distribution of PRPs. Seeing how clumped it is from 985 to 1005, I've moved my machine to that range. I should have 990-993 done in 2-3 days. Once my current range is complete, I'll keep working upwards.

If I am treading in anyone's area, please let me know and I will adjust accordingly. A PM will get my attention the quickest.

schickel · 2016-11-27, 08:57

Quote:

Originally Posted by Jayder

If you are running only one instance and wall-clock time matters more, use 4.1.1. If you are a crazy person, use 4.2.1 for Phase 1 and 4.1.1 for Phase 2.

Unfortunately, according to Marcel that won't work:

Quote:

Due to numerous changes, it is impossible to resume with a 4.2.x version a certification started with a version prior to 4.2.0.

I would assume that if they're not forward compatible they're also not back compatible.

Thanks for the timing info, it'll be interesting to see how high I can push certs with the new version (biggest system I've got is a hex-core AMD).

Quote:

Originally Posted by Jayder

I am currently working on 990-993. I was a bit dumb and didn't check the distribution of PRPs. Seeing how clumped it is from 985 to 1005, I've moved my machine to that range. I should have 990-993 done in 2-3 days. Once my current range is complete, I'll keep working upwards.

If I am treading in anyone's area, please let me know and I will adjust accordingly. A PM will get my attention the quickest.

I think that you, I, and henryzz are the only ones doing certs right now. Working up from where you started should be fine. I like to keep the small stuff cleared out if I can, so I usually run the new cruft before I grab a chunk of work.

schickel · 2016-11-29, 14:57

Nice!

2016-11-27, 08:18	#74
Jayder Dec 2012 2×139 Posts	Primo An important note about Primo: Disregard my previous comment(s) about 4.2.1. I have performed a test on 195 2000dd PRPs, comparing Primo 4.2.1 and 4.1.1 with 48 concurrent tasks. When comparing wall-clock time, I have found that 4.2.1 took 5% longer than 4.1.1 to complete the batch. Phase 1 is faster with 4.2.1, taking only 92% of the time that 4.1.1 required. The problem lies with Phase 2, which took 43% longer than 4.1.1. When comparing process time, I have found that 4.2.1 completed the batch in 90% of the time needed by 4.1.1. Phase 1 is faster with 4.2.1, taking only 87% of the time that 4.1.1 required. Phase 2 took 8% longer than 4.1.1. For 4.2.1, the big difference in Phase 2 time when comparing wall-clock time and process time is explained by there being polynomials that are harder to factor (or so I believe). This leaves a few threads running for a long time (sometimes for half the duration of Phase 2) on a few harder polynomials, while the rest of the threads sit idle. Somebody please correct me if I'm wrong, but I believe the takeaway is this: If you run multiple instances of Primo and are able to max out your processor(s), use 4.2.1 over 4.1.1. If you are running only one instance and wall-clock time matters more, use 4.1.1. If you are a crazy person, use 4.2.1 for Phase 1 and 4.1.1 for Phase 2. Data for those interested (all three of you) can be found here: EDIT: I forgot to mention that 4.2.1 produces smaller certificates. The 195 certificates produced by 4.2.1 take up 60.1MiB of space, while 4.1.1's 195 certificates take up 90.7MiB. Last fiddled with by Jayder on 2016-11-27 at 08:34 Reason: Formatting.

2016-11-29, 14:57	#77
schickel "Frank <^>" Dec 2004 CDP Janesville 84A₁₆ Posts	Nice! Attached Thumbnails

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Fixup Old Primo Certificate?	wblipp	FactorDB	1	2012-05-28 03:16
Invalid certificate?	IvanP	FactorDB	3	2012-05-11 12:17
Could Moore's law be purposely used for marketing purposes?	jasong	Science & Technology	10	2007-01-19 19:04
certificate of appreciation	Unregistered	Information & Answers	13	2004-04-28 06:24

2016-11-01, 07:37	#67
Jayder Dec 2012 2×139 Posts	I recently built a dual-socket machine and I was considering trying to make a dent in the certificates under 3000dd, but the lack of support is a real downer.

2016-11-15, 07:03	#71
Jayder Dec 2012 2×139 Posts	The most recent version I had was 4.0.1, but a kind soul gave me a download for 4.1.1, so this conversation is a bit moot. I would have used it, though, if that was all that was available to me. But wouldn't you agree that it is nice to have the most recent version of a program? (Maybe we won't include Windows here.) I don't know how much the new discriminants help, but there were quite a few added after 4.0.1, not to mention a few minor features and bug fixes. Is there a reason why you're so curious?

2016-11-24, 08:18	#72
Jayder Dec 2012 100010110₂ Posts	Good news, folks! Markus has informed me that he has fixed the bug which caused format 4 certificates to be rejected. I have uploaded the format 4 certificates that I have and can confirm that it works splendidly. Thank you, Markus! In a very small test, I have found that version 4.2.1 is at least 10% faster than 4.1.1. Well worth upgrading. I am going to do some certificate testing above 2000 digits. Where is everybody working right now?

2016-11-27, 08:24	#75
Jayder Dec 2012 2·139 Posts	I am currently working on 990-993. I was a bit dumb and didn't check the distribution of PRPs. Seeing how clumped it is from 985 to 1005, I've moved my machine to that range. I should have 990-993 done in 2-3 days. Once my current range is complete, I'll keep working upwards. If I am treading in anyone's area, please let me know and I will adjust accordingly. A PM will get my attention the quickest.