Message boards : Fermat Divisor Search : Mistake in description on project preferences page

 Subscribe SortOldest firstNewest firstHighest rated posts first
Author Message
numbermaniac
Volunteer tester

Joined: 28 Mar 14
Posts: 192
ID: 305955
Credit: 9,806,045
RAC: 9,375

Message 137818 - Posted: 13 Feb 2020 | 3:35:38 UTC

On the project preferences page (the one where you select which subprojects you want to run), the description for this project says:

k·2^n+1 for k in (1323, 2187, 3125, 3267, 3375, 19683) and 5<=k<=49

I'm not sure how k can be between 1323 and 19683, yet also be less than 49. Is one of those letters meant to be "n" instead of "k"?

Thanks :)

Eudy Silva

Joined: 26 Aug 17
Posts: 1541
ID: 918937
Credit: 341,795,816
RAC: 181,700

Message 137819 - Posted: 13 Feb 2020 | 3:47:42 UTC - in response to Message 137818.

Click the link Fermat Divisor Search LLR (PPS-DIV) right above the line you mentioned [k·2n+1 for k in (1323, 2187, 3125, 3267, 3375, 19683) and 5<=k<=49] and it'll take you to the corresponding forum thread about the search.
You'll see it's a 2 phase search:

We'll be searching for Fermat Divisors in two stages:
First, we'll search k=19683 up to n=4M (probably; there's still some talk about where exactly it stops). This will get done really quickly.

Second phase is searching 5<=k<=49 up to 9M. This part will take longer.

The search will be done for k=1323, k=2187, k=3125, k=3267, k=3375, k=19683 and for all odd values 5<=k<=49.

Edit: Also, this link shows the current status.
____________
"Accidit in puncto, quod non contingit in anno."
Something that does not occur in a year may, perchance, happen in a moment.

dannyridel
Volunteer tester

Joined: 3 Feb 19
Posts: 919
ID: 1097922
Credit: 20,097,369
RAC: 20,484

Message 137862 - Posted: 14 Feb 2020 | 5:26:46 UTC - in response to Message 137819.

Also, the choice on the prefs says that MT isn't recommended but that's not the case now--it's going to catch up with SR5 before summer vacation, and it's already 1.6M digits so i think MTing should be recommended.
____________
SHSIDElectronicsGroup@outlook.com

waiting for a TdP prime...
Proth "SoB": 44243*2^440969+1

numbermaniac
Volunteer tester

Joined: 28 Mar 14
Posts: 192
ID: 305955
Credit: 9,806,045
RAC: 9,375

Message 139899 - Posted: 26 Apr 2020 | 13:19:47 UTC - in response to Message 137862.

Also, the choice on the prefs says that MT isn't recommended but that's not the case now--it's going to catch up with SR5 before summer vacation, and it's already 1.6M digits so i think MTing should be recommended.

That is true, especially since PPS Mega is smaller than DIV, yet multi-threading is recommended for Mega while it is currently not recommended for DIV. Perhaps that should be updated to recommend multi-threading for both.
____________
1 PPSE (+2 DC) & 5 SGS primes

Reggie
Volunteer moderator
Volunteer tester
Project scientist

Joined: 10 May 14
Posts: 204
ID: 311759
Credit: 179,767,605
RAC: 53,451

Message 139902 - Posted: 26 Apr 2020 | 14:25:46 UTC - in response to Message 139899.

Perhaps that should be updated to recommend multi-threading for both.

Updated.

dannyridel
Volunteer tester

Joined: 3 Feb 19
Posts: 919
ID: 1097922
Credit: 20,097,369
RAC: 20,484

Message 139906 - Posted: 26 Apr 2020 | 15:10:55 UTC

Hmmm that was two months ago, now I don't think it can catch up with SR5 b4 July.
____________
SHSIDElectronicsGroup@outlook.com

waiting for a TdP prime...
Proth "SoB": 44243*2^440969+1

dannyridel
Volunteer tester

Joined: 3 Feb 19
Posts: 919
ID: 1097922
Credit: 20,097,369
RAC: 20,484

Message 139914 - Posted: 27 Apr 2020 | 5:37:20 UTC

Hello, just wanted to inquire whether PPS-DIV will ever get a 10% long job bonus before it reaches n=9M. SR5 doesn't get it but TRP does. Is there a threshold value? Or is it implemented manually?
____________
SHSIDElectronicsGroup@outlook.com

waiting for a TdP prime...
Proth "SoB": 44243*2^440969+1

Reggie
Volunteer moderator
Volunteer tester
Project scientist

Joined: 10 May 14
Posts: 204
ID: 311759
Credit: 179,767,605
RAC: 53,451

Message 140044 - Posted: 3 May 2020 | 2:50:04 UTC - in response to Message 139914.

Hello, just wanted to inquire whether PPS-DIV will ever get a 10% long job bonus before it reaches n=9M. SR5 doesn't get it but TRP does. Is there a threshold value? Or is it implemented manually?

It's doubtful we will give it a bonus. A 10% long job bonus is (very roughly) defined as >1 day single threaded on "standard" hardware. Our estimates don't show PPS-DIV reaching this point. Actually, SR5 may be getting this bonus soon-ish. We'll keep an eye on it to see. It is something we have to implement manually.

(Sorry for the long wait before this response; we had to discuss it a bit.)

dannyridel
Volunteer tester

Joined: 3 Feb 19
Posts: 919
ID: 1097922
Credit: 20,097,369
RAC: 20,484

Message 140045 - Posted: 3 May 2020 | 2:55:47 UTC - in response to Message 140044.

No problem, thank you!
n=9M is around 2.5million digits right?
____________
SHSIDElectronicsGroup@outlook.com

waiting for a TdP prime...
Proth "SoB": 44243*2^440969+1

Ravi Fernando
Volunteer tester
Project scientist

Joined: 21 Mar 19
Posts: 176
ID: 1108183
Credit: 10,215,312
RAC: 5,086

Message 140047 - Posted: 3 May 2020 | 4:18:32 UTC - in response to Message 140045.

It's a little over 2.7M digits.

dannyridel
Volunteer tester

Joined: 3 Feb 19
Posts: 919
ID: 1097922
Credit: 20,097,369
RAC: 20,484

Message 140048 - Posted: 3 May 2020 | 5:58:57 UTC - in response to Message 140047.

Hmmm, SR5 won't be reaching that point for a while, but we'll see. Meanwhile I'll crunch happily with MTed DIV without a bonus :D
____________
SHSIDElectronicsGroup@outlook.com

waiting for a TdP prime...
Proth "SoB": 44243*2^440969+1

Ravi Fernando
Volunteer tester
Project scientist

Joined: 21 Mar 19
Posts: 176
ID: 1108183
Credit: 10,215,312
RAC: 5,086

Message 140049 - Posted: 3 May 2020 | 6:41:14 UTC - in response to Message 140048.

True--but keep in mind that DIV tasks are faster than similarly-sized SR5 tasks. (Much like 321 vs. ESP or GCW.) It may be that DIV will never catch up to the current average SR5 runtime.

In any case, it's a manual change, so we'll cross that bridge when we get to it.

dannyridel
Volunteer tester

Joined: 3 Feb 19
Posts: 919
ID: 1097922
Credit: 20,097,369
RAC: 20,484

Message 140055 - Posted: 3 May 2020 | 10:27:17 UTC - in response to Message 140049.

Ah, yes, forgot about that really strange phenomenon.
Could you explain (here or via PM) why this happens?
____________
SHSIDElectronicsGroup@outlook.com

waiting for a TdP prime...
Proth "SoB": 44243*2^440969+1

stream
Volunteer moderator
Volunteer developer
Volunteer tester

Joined: 1 Mar 14
Posts: 920
ID: 301928
Credit: 513,298,826
RAC: 4,577

Message 140057 - Posted: 3 May 2020 | 11:02:17 UTC - in response to Message 140055.

Ah, yes, forgot about that really strange phenomenon.
Could you explain (here or via PM) why this happens?

Some operations in base 2 tests can be optimized as bitmaps.
To calculate k*2^n, you just take k and shift it left n bits, and have a complete answer in just a few CPU cycles.
For any other base, for example, k*5^n, you must calculate 5^n in standard way, and also do final multiplication.
So why non base 2 tests are significantly slower (or base 2 tests are significantly faster, comparing to generic tests).

dannyridel
Volunteer tester

Joined: 3 Feb 19
Posts: 919
ID: 1097922
Credit: 20,097,369
RAC: 20,484

Message 140058 - Posted: 3 May 2020 | 11:06:05 UTC - in response to Message 140057.

Ah, okay. Thanks for the clarification!
____________
SHSIDElectronicsGroup@outlook.com

waiting for a TdP prime...
Proth "SoB": 44243*2^440969+1

Ravi Fernando
Volunteer tester
Project scientist

Joined: 21 Mar 19
Posts: 176
ID: 1108183
Credit: 10,215,312
RAC: 5,086

Message 140070 - Posted: 3 May 2020 | 16:27:19 UTC - in response to Message 137818.

To return to the original purpose of this thread: Reggie edited the six larger k's out of the description today. No point mentioning the k's we already finished.

Eudy Silva

Joined: 26 Aug 17
Posts: 1541
ID: 918937
Credit: 341,795,816
RAC: 181,700

Message 140072 - Posted: 3 May 2020 | 16:47:33 UTC - in response to Message 140070.

Suggestion: In the Official PrimeGrid Badge List, the description of the badge could also list Fermat Divisor Search as an pertinent project.
____________
"Accidit in puncto, quod non contingit in anno."
Something that does not occur in a year may, perchance, happen in a moment.

Reggie
Volunteer moderator
Volunteer tester
Project scientist

Joined: 10 May 14
Posts: 204
ID: 311759
Credit: 179,767,605
RAC: 53,451

Message 140084 - Posted: 3 May 2020 | 23:38:54 UTC - in response to Message 140072.

Suggestion: In the Official PrimeGrid Badge List, the description of the badge could also list Fermat Divisor Search as an pertinent project.

Updated. Now, let's keep threads on topic please. This thread has gone in so many directions it's ridiculous.

Message boards : Fermat Divisor Search : Mistake in description on project preferences page