Newer milestone thread - Page 81

lycorn · 2013-11-28, 08:22

@LaurV,

I´ve been doing that type of work for a while now, and still am, so tell me what exponent range you are actually factoring, so we don´t step on each other toes. Or are you only doing unreserved assignments, and reserving your work as you go?

LaurV · 2013-11-28, 10:38

Quote:

Originally Posted by lycorn

@LaurV,

I´ve been doing that type of work for a while now, and still am, so tell me what exponent range you are actually factoring, so we don´t step on each other toes. Or are you only doing unreserved assignments, and reserving your work as you go?

Whoops... I took everything done to 60, except the first 3 which are under 2000. Oliver's build of mfaktc.18 only accepts exponents over 2k. So, there were 417 exponents when I checked in yesterday, and about 30 of them were done yesterday (auto report, misfit). For today, ~~I have to go home and see...~~ [edit: wrong, I can not access my computer but still can see this gimps page, copy/paste in a text editor which shows the line numbers, only 320, so about 100 expos were done already, so about 40 hours left to finish all of them... ]

The point is that I checked Chris' Visualization page and saw activity in the last month, but nothing in the last week. I assumed no one is working there. The active assignment page of PrimeNet which I still can access, is not relevant, because many people (all??) have ECM-only assignments for this range (it is not possible to get other type of assignment, except P-1, which is already too high to be profitable).

So, sorry if I stepped on your toes. For me there was about 60 hours of work totally, so exactly what I was needing to "give a short break" to my cards. I wanted to do this for long time, although I am sure it is futile work: according with the qty of ECM done there, there should be no factor under 120-150 bits (40-50 digits?) or so. But is still fun...

lycorn · 2013-11-28, 10:46

OK, no big deal.
If you started by the 40K range, it´s perfect, as I have not done anything (yet) in that range. For the other ranges, I have only taken unreserved exponents.
But it´s odd that you have not noticed activity in the range, as I have been reporting work daily (10 - 15 exponents on average).
Just confirm you are going to do them all, and I´ll stop what I´m doing and move to other ranges/bit levels. After all, those small exponents are in better hands if done by a GTX580

.
I´ll probably take some 2.5M to 63 bits and move some CPU cores to ECM.

LaurV · 2013-11-28, 10:56

I am definitively *not* going to take them over 61 bits. Not this time

So, 61 to infinity are yours

I am just very bored by the "60" column in Chris' tables (can't see over the "72" because of it!), and wanna get rid of it! I hope someone will find factors, SNFS, ECM, whatever, for the remaining 3. Otherwise I may try doing them with Pari.

I didn't "start with 40k", but split the list in 8 parts - the number of instances of Mfaktc running, this version consumes very little GPU and you need to start lots of instances, to get a reasonable occupancy of the card. This is because the sieving can only be done with very few primes, otherwise if you sieve with primes higher than 2*p+1 (when p=3 mod 4) or 6*p+1 (when p=1 mod 4), you risk to eliminate (lose) factors, so the CPU-GPU handshake is not very productive.

So the range is split in 8 (by MISFIT) and each subrange worked independently.

(I added links to the previous post)

LaurV · 2013-11-28, 11:07

Quote:

Originally Posted by lycorn

Just confirm you are going to do them all

Sure, just to be clear, I am going to raise all of them to 61, except the first 3, which are beyond (under) the limit of the Mfaktc version I have.

lycorn · 2013-11-28, 11:08

Quote:

Originally Posted by LaurV

I am just very bored by the "60" column in Chris' tables (can't see over the "72" because of it!), and wanna get rid of it! I hope someone will find factors, SNFS, ECM, whatever, for the remaining 3.

Great minds think alike!

So that´s it. I´ll move somewhere else. It´s actually a good thing you´ve decided to give it a go. That 40K range, in particular, would have killed me

.

petrw1 · 2013-11-28, 15:37

Quote:

Originally Posted by LaurV

Sure, just to be clear, I am going to raise all of them to 61, except the first 3, which are beyond (under) the limit of the Mfaktc version I have.

I've been chugging , admittedly slowly in that range for several years. It is the best work for my ancient PC. It does the best up to 62. As long as you first checked what is assigned I am fine. All my current work is in the 100K range going to 61 bits.

petrw1 · 2013-11-28, 16:03

Quote:

Originally Posted by LaurV

The point is that I checked Chris' Visualization page and saw activity in the last month, but nothing in the last week. I assumed no one is working there. The active assignment page of PrimeNet which I still can access, is not relevant, because many people (all??) have ECM-only assignments for this range (it is not possible to get other type of assignment, except P-1, which is already too high to be profitable).
.

I have been reporting a meagre 1-3 per day all year along with Lycorn's 10 or more
http://www.mersenne.info/trial_facto...r_delta_7/3/0/

I think this is the assignments page you want.

http://www.mersenne.org/assignments/...et+Assignments

But again I have no problem if you do them. There is lots to go around for all of us. I never would have finished this alone.

LaurV · 2013-11-28, 16:50

Ok, everything I doooo, I do it for youuuu...

I found 21 reserved for you, in your list.. Because you gave me so nice list, I went through my worktodo and eliminated everything which is marked with an "x" in the following code snippet, this means they are yours, and you can continue to work them. I see you will be done in 8-10 days, which is not much.

For those which are not marked with an "x"... well, I didn't find them. As I grabbed all the "sixties", reserved or not, that would mean you were already too late, they are most probably done and reported, and you can eliminate them from 60 to 61 (better check first if they were really reported [edit: yep, they were]). You are anyhow welcome to do 61 to 62, 63.. infinity, if you want...

Code:

From my side, I will finish the remaining of them (put the third card too, there are ~~281~~ 273 of them yet to go) and move to something else, because this activity is not very "rewarding", the credit is too low, much less than a card can give if I work it on the LL front or doing DC/P-1 with it, and additionally, there will be (most probably) no factor to be found, due to the lots of ECM done there. Additionally, it does not help GIMPS at all (to find new primes, which is its main goal) because this range is long-ago settled.

My quest was only against that "60" column in Chris' tables, if he should split the table when I asked him last year

then now I wouldn't waste resources, and do some useful work at the LL front

petrw1 · 2013-11-28, 19:46

Thx...enjoy

petrw1 · 2013-11-28, 20:29

Oh and for the record if you and Lycorn hadn't helped I could have been done by 2021

2013-11-28, 08:22	#881
lycorn "GIMFS" Sep 2002 Oeiras, Portugal 2·11·67 Posts	@LaurV, I´ve been doing that type of work for a while now, and still am, so tell me what exponent range you are actually factoring, so we don´t step on each other toes. Or are you only doing unreserved assignments, and reserving your work as you go? Last fiddled with by lycorn on 2013-11-28 at 08:23

2013-11-28, 10:56	#884
LaurV Romulan Interpreter Jun 2011 Thailand 10010111000000₂ Posts	I am definitively not going to take them over 61 bits. Not this time So, 61 to infinity are yours I am just very bored by the "60" column in Chris' tables (can't see over the "72" because of it!), and wanna get rid of it! I hope someone will find factors, SNFS, ECM, whatever, for the remaining 3. Otherwise I may try doing them with Pari. I didn't "start with 40k", but split the list in 8 parts - the number of instances of Mfaktc running, this version consumes very little GPU and you need to start lots of instances, to get a reasonable occupancy of the card. This is because the sieving can only be done with very few primes, otherwise if you sieve with primes higher than 2p+1 (when p=3 mod 4) or 6p+1 (when p=1 mod 4), you risk to eliminate (lose) factors, so the CPU-GPU handshake is not very productive. So the range is split in 8 (by MISFIT) and each subrange worked independently. (I added links to the previous post) Last fiddled with by LaurV on 2013-11-28 at 11:04

2013-11-28, 16:50	#889
LaurV Romulan Interpreter Jun 2011 Thailand 2⁶·151 Posts	Ok, everything I doooo, I do it for youuuu... I found 21 reserved for you, in your list.. Because you gave me so nice list, I went through my worktodo and eliminated everything which is marked with an "x" in the following code snippet, this means they are yours, and you can continue to work them. I see you will be done in 8-10 days, which is not much. For those which are not marked with an "x"... well, I didn't find them. As I grabbed all the "sixties", reserved or not, that would mean you were already too late, they are most probably done and reported, and you can eliminate them from 60 to 61 (better check first if they were really reported [edit: yep, they were]). You are anyhow welcome to do 61 to 62, 63.. infinity, if you want... Code: 106213 110863 x 111919 116483 117259 118967 132439 134437 x 135977 x 136987 x 138727 139273 139397 x 140417 x 143093 x 143687 x 145879 146021 159979 x 160091 x 160093 x From my side, I will finish the remaining of them (put the third card too, there are ~~281~~ 273 of them yet to go) and move to something else, because this activity is not very "rewarding", the credit is too low, much less than a card can give if I work it on the LL front or doing DC/P-1 with it, and additionally, there will be (most probably) no factor to be found, due to the lots of ECM done there. Additionally, it does not help GIMPS at all (to find new primes, which is its main goal) because this range is long-ago settled. My quest was only against that "60" column in Chris' tables, if he should split the table when I asked him last year then now I wouldn't waste resources, and do some useful work at the LL front Last fiddled with by LaurV on 2013-11-28 at 17:17

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Newer X64 build needed	Googulator	Msieve	73	2020-08-30 07:47
Performance of cuda-ecm on newer hardware?	fivemack	GMP-ECM	14	2015-02-12 20:10
Cause this don't belong in the milestone thread	bcp19	Data	30	2012-09-08 15:09
Newer msieves are slow on Core i7	mklasson	Msieve	9	2009-02-18 12:58
Use of large memory pages possible with newer linux kernels	Dresdenboy	Software	3	2003-12-08 14:47

2013-11-28, 10:46	#883
lycorn "GIMFS" Sep 2002 Oeiras, Portugal 2×11×67 Posts	OK, no big deal. If you started by the 40K range, it´s perfect, as I have not done anything (yet) in that range. For the other ranges, I have only taken unreserved exponents. But it´s odd that you have not noticed activity in the range, as I have been reporting work daily (10 - 15 exponents on average). Just confirm you are going to do them all, and I´ll stop what I´m doing and move to other ranges/bit levels. After all, those small exponents are in better hands if done by a GTX580 . I´ll probably take some 2.5M to 63 bits and move some CPU cores to ECM.

2013-11-28, 19:46	#890
petrw1 1976 Toyota Corona years forever! "Wayne" Nov 2006 Saskatchewan, Canada 1001001010111₂ Posts	Thx...enjoy

2013-11-28, 20:29	#891
petrw1 1976 Toyota Corona years forever! "Wayne" Nov 2006 Saskatchewan, Canada 1001001010111₂ Posts	Oh and for the record if you and Lycorn hadn't helped I could have been done by 2021