FactorDB PRP's - Page 6

EdH · 2011-01-19, 16:46

Quote:

Originally Posted by R. Gerbicz

The term of "unproven prime" is meaningless. "probable prime" would be good.

The db calls them "Unproven (probable) primes." I left off the parenthesized portion. Thanks for the note. . .

EdH · 2011-01-21, 19:19

Even a random grab of all the 613 digit PRPs last night didn't seem quite worth it:

Code:

Found 196 files, 27 certificates added.

. . .

RichD · 2011-10-17, 21:06

I noticed there is a new 64-bit Primo for Linux here. Still in alpha testing but appears to create valid certificates.

This should help speed up the PRP processing in factorDB.

EdH · 2011-10-17, 22:37

Quote:

Originally Posted by RichD

I noticed there is a new 64-bit Primo for Linux here. Still in alpha testing but appears to create valid certificates.

This should help speed up the PRP processing in factorDB.

Well, it certified ten 300 digit candidates in just under 5 minutes. Machine info:

Code:

Fedora 15 64-bit (x86_64)
AMD Athlon(tm) 64 X2 Dual Core Processor 6000+ 3GHz

Not sure how that compares to other setups...

Perhaps I'll run a few more in the next couple days...

ET_ · 2011-10-18, 09:02

Quote:

Originally Posted by EdH

Well, it certified ten 300 digit candidates in just under 5 minutes. Machine info:

Code:

Fedora 15 64-bit (x86_64)
AMD Athlon(tm) 64 X2 Dual Core Processor 6000+ 3GHz

Not sure how that compares to other setups...

Perhaps I'll run a few more in the next couple days...

Note that the version is still in Alpha, and bugs are coming out from time to time... Apart from this, the new 64 bit version is definitely faster than the 1-cpu version.

Luigi

EdH · 2011-10-19, 01:11

I was in error as to the size of the aforementioned test candidates. All the ~300 digit ones have apparently been certified and the remainder are a bit larger. I didn't check all of the ones I tested, but the ones I checked were in the range of 400-500 digits. Those were what I reported on earlier. My mistake was that the default for the download page said 300 and I didn't think to check if they really were that size.

So far I have run a couple hundred (some at least as large as 1039) and all "seems" fine. The current size for the smallest PRP appears to be 1047 digits. Tomorrow I'll try some a little bigger and see how they run.

schickel · 2011-10-19, 04:40

Quote:

Originally Posted by EdH

I was in error as to the size of the aforementioned test candidates. All the ~300 digit ones have apparently been certified and the remainder are a bit larger. I didn't check all of the ones I tested, but the ones I checked were in the range of 400-500 digits. Those were what I reported on earlier. My mistake was that the default for the download page said 300 and I didn't think to check if they really were that size.

So far I have run a couple hundred (some at least as large as 1039) and all "seems" fine. The current size for the smallest PRP appears to be 1047 digits. Tomorrow I'll try some a little bigger and see how they run.

What we should do is pick a couple of benchmark numbers of various sizes so we have something to measure with....

I'd be willing to run some tests for my rigs.

[Edit: Actually, Marcel has some run times for various numbers posted on the Primo-for-Linux page. They all start at 1000 digits, though. Might it make sense to get some smaller ones just for completeness?]

EdH · 2011-10-19, 15:23

Quote:

Originally Posted by schickel

What we should do is pick a couple of benchmark numbers of various sizes so we have something to measure with....

I'd be willing to run some tests for my rigs.

[Edit: Actually, Marcel has some run times for various numbers posted on the Primo-for-Linux page. They all start at 1000 digits, though. Might it make sense to get some smaller ones just for completeness?]

The current floor in the db is over 1000, though. I just did the 34 candidates at 1051 digits this morning:

Code:

Candidate 1=Certified, 177s
Candidate 2=Certified, 188s
Candidate 3=Certified, 174s
Candidate 4=Certified, 171s
Candidate 5=Certified, 153s
Candidate 6=Certified, 183s
Candidate 7=Certified, 173s
Candidate 8=Certified, 190s
Candidate 9=Certified, 145s
Candidate 10=Certified, 190s
Candidate 11=Certified, 171s
Candidate 12=Certified, 185s
Candidate 13=Certified, 173s
Candidate 14=Certified, 194s
Candidate 15=Certified, 167s
Candidate 16=Certified, 194s
Candidate 17=Certified, 172s
Candidate 18=Certified, 195s
Candidate 19=Certified, 174s
Candidate 20=Certified, 167s
Candidate 21=Certified, 176s
Candidate 22=Certified, 166s
Candidate 23=Certified, 235s
Candidate 24=Certified, 252s
Candidate 25=Certified, 178s
Candidate 26=Certified, 206s
Candidate 27=Certified, 193s
Candidate 28=Certified, 213s
Candidate 29=Certified, 168s
Candidate 30=Certified, 209s
Candidate 31=Certified, 172s
Candidate 32=Certified, 148s
Candidate 33=Certified, 162s
Candidate 34=Certified, 192s

total: ~1h 43m 26s, for an average of ~3 minutes per candidate.

schickel · 2011-10-19, 18:15

Quote:

Originally Posted by EdH

The current floor in the db is over 1000, though. I just did the 34 candidates at 1051 digits this morning:
total: ~1h 43m 26s, for an average of ~3 minutes per candidate.

Too bad the Windows version is not going to see that kind of speed....

[Edit: I guess your work is why Syd has 5 cores processing certs now!]

EdH · 2011-10-20, 03:24

Quote:

Originally Posted by schickel

Too bad the Windows version is not going to see that kind of speed....

[Edit: I guess your work is why Syd has 5 cores processing certs now!]

I hadn't realized there would be a lot of work processing the certificates... I have uploaded quite a few (maybe >1000), but it hasn't seemed to dent the 69k PRPs awaiting certification. The current floor seems to be at 1047 digits.

schickel · 2011-10-20, 04:09

Quote:

Originally Posted by EdH

I hadn't realized there would be a lot of work processing the certificates...

I don't think a lot of work, as such, just that they verify slower than you can generate them (I think). I think there were some benchmarks in the thread about the verifier that WraithX wrote...

Quote:

I have uploaded quite a few (maybe >1000), but it hasn't seemed to dent the 69k PRPs awaiting certification. The current floor seems to be at 1047 digits.

It'll be a while, I think. If you check the PRP size distribution, there is a slight bulge from 1988-2013 digits where it jumps to a 3 digit count, but mostly it's one or two digits up to 3000 digits.

I don't remember the largest numbers I checked, but beyond a certain point it gets to be a diminishing return. Although I could do a whole lot better now with a new hex-core on my side....

2011-01-21, 19:19	#57
EdH "Ed Hall" Dec 2009 Adirondack Mtns 11·347 Posts	Even a random grab of all the 613 digit PRPs last night didn't seem quite worth it: Code: Found 196 files, 27 certificates added. . . .

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
A suggestion for factordb.	enzocreti	FactorDB	15	2021-06-24 07:15
Other Factordb Problems	wblipp	FactorDB	470	2021-06-13 16:58
Accessing FactorDB from Python	shortcipher	FactorDB	21	2018-12-03 17:03
Extending Factordb	carpetpool	FactorDB	6	2017-01-23 11:04
FactorDB question	Raman	Factoring	15	2010-01-28 10:24

2011-10-17, 21:06	#58
RichD Sep 2008 Kansas 2⁴·211 Posts	I noticed there is a new 64-bit Primo for Linux here. Still in alpha testing but appears to create valid certificates. This should help speed up the PRP processing in factorDB.

2011-10-19, 01:11	#61
EdH "Ed Hall" Dec 2009 Adirondack Mtns 11×347 Posts	I was in error as to the size of the aforementioned test candidates. All the ~300 digit ones have apparently been certified and the remainder are a bit larger. I didn't check all of the ones I tested, but the ones I checked were in the range of 400-500 digits. Those were what I reported on earlier. My mistake was that the default for the download page said 300 and I didn't think to check if they really were that size. So far I have run a couple hundred (some at least as large as 1039) and all "seems" fine. The current size for the smallest PRP appears to be 1047 digits. Tomorrow I'll try some a little bigger and see how they run.