mersenneforum.org

mersenneforum.org (https://www.mersenneforum.org/index.php)
-   LMH > 100M (https://www.mersenneforum.org/forumdisplay.php?f=46)
-   -   100M digits prefactor project. (https://www.mersenneforum.org/showthread.php?t=2558)

lpmurray 2004-06-02 04:23

i've done 332200000 - 332300000 to 50bits and need that program to strip out the factors or i could just send someone the results file its 222k. Also i'm working on 1 number in that range to large bit. i'm running it on a 2.8gig p4 and it took 100hrs to do just 67bit is that normal for factor3_2? At that rate 73 start 74 end would take a year....Since time doubles every bit depth.

jinydu 2004-06-02 05:19

[QUOTE=lpmurray]i've done 332200000 - 332300000 to 50bits and need that program to strip out the factors or i could just send someone the results file its 222k. Also i'm working on 1 number in that range to large bit. i'm running it on a 2.8gig p4 and it took 100hrs to do just 67bit is that normal for factor3_2? At that rate 73 start 74 end would take a year....Since time doubles every bit depth.[/QUOTE]

That does seem a bit high. But then again, factoring to a given bit depth takes longer for [B]smaller[/B] exponents (because there are more values of k to test, since k is smaller).

Uncwilly 2004-06-02 05:56

[QUOTE=lpmurray]i've done 332200000 - 332300000 to 50bits and need that program to strip out the factors or i could just send someone the results file its 222k.[/QUOTE]

I'll PM you an e-mail address, zip it and send to me. What bit interval do you want to do next on it? 50-?? When I strip it I will save the factors to send to Will E. and return you a worktodo.ini for factor3_2b At this expo level and with Factor3_2 it is not worth fatoring them very far (70).

ET_ 2004-06-02 09:59

[QUOTE=Uncwilly]I'll PM you an e-mail address, zip it and send to me. What bit interval do you want to do next on it? 50-?? When I strip it I will save the factors to send to Will E. and return you a worktodo.ini for factor3_2b At this expo level and with Factor3_2 it is not worth fatoring them very far (70).[/QUOTE]

Working so fast you'll soon complete the gap from 100MP to 1BP! :smile:

As jinydu said the speed of factor3_2 depends on both the bit depth and the exponent size.

I'm working on 3,321,928,171 from 71 to 72 bits and this will take less than a week on an Athlon XP 2100+

Luigi

lpmurray 2004-06-02 23:46

[QUOTE=ET_]

I'm working on 3,321,928,171 from 71 to 72 bits and this will take less than a week on an Athlon XP 2100+

Luigi[/QUOTE]
Are you running windows or linux? I'm running windows xp with a p4 2.8gig. I am running a number in the 300 million range which is 10 times smaller and it took 100hrs just to do 67 to 68 bits. by doubling the time for each bit depth it should take me 67 days to do 71 to 72 bits. I can't figure out why you are running so much faster.

Uncwilly 2004-06-02 23:58

Factors for Mersennes must be in the form 2[I]k[/I][B]p[/B]+1
[B]P[/B] is the expo.
As it increases the max that [I]k[/I] can be and still have the whole factor below a certain bit level shrinks. You reduce the max that [I]k[/I] can be by the same amount that [B]p[/B] goes up. p'=p*10 -> k'=k/10

jinydu 2004-06-03 01:26

That is, if the exponent is ten times smaller, you have ten times as many values of k to test to reach a given bit depth. This is why it has been possible to factor MM127 to over 170 bits.

However, it was mentioned in another thread ([url]http://www.mersenneforum.org/showthread.php?t=2335&page=2[/url], biwema's posts) that the amount of CPU power needed to test a particular value of k for M(p) is proportional to the amount of CPU power to LL test M(log base 2 [p]). Thus, larger exponents do take longer to test for each value of k. However, this increase grows very slowly. Each value of k in a billion digit exponent takes only 12.5% longer than each value of k in a 100 million digit exponent. Still, it puts a limit on the largest exponents that can be factored: M(M(p)), where M(p) is the largest exponent that can be LL tested.

Plesae correct me if I'm wrong.

Uncwilly 2004-06-03 06:56

[QUOTE=Uncwilly]332209991 - 332249999 out to 54 or 55 in three or four passes. 2014 expos[/QUOTE]

Duh, I feel dumb. I stepped on toes here.

I will now be doing 332300001 - 332399999, likely only to 50 bits.

I have improved my tracking system. I am thinking about setting up a page for this, with worktodo.ini's all zipped and ready for open ranges, etc. Next week, though, not now.

ET_ 2004-06-03 09:07

[QUOTE=Uncwilly]Duh, I feel dumb. I stepped on toes here.[/QUOTE]

No, you just did some double-check :rolleyes:

Waiting for your page, I will link it to mine.

Luigi

ET_ 2004-06-03 09:11

[QUOTE=lpmurray]Are you running windows or linux? I'm running windows xp with a p4 2.8gig. I am running a number in the 300 million range which is 10 times smaller and it took 100hrs just to do 67 to 68 bits. by doubling the time for each bit depth it should take me 67 days to do 71 to 72 bits. I can't figure out why you are running so much faster.[/QUOTE]

I am running on Windows XP.

Mind that the program be factor3_2 instead of factor3_1. Note also that the software has been optimized for [B]Athlons[/B] in both Cygwin and GMP environments. I still have to (buy and) set up a [B]Pentium 4[/B] and recompile an optimized version. :innocent:

Luigi

Uncwilly 2004-06-03 22:07

[QUOTE=lpmurray]Also i'm working on 1 number in that range to large bit.[/QUOTE]

Let me know what it is, I will track individual expos that are being run 'deep'.


All times are UTC. The time now is 04:05.

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2021, Jelsoft Enterprises Ltd.