Polynomial Request Thread - Page 172

swellman · 2020-05-08, 23:26

Quote:

Originally Posted by EdH

I'm playing with some alternate scripts for my machines that aren't working on the 198 team project. I might as well try to make them useful during this time.

Is there a polyselect request that I might try some things with? I won't be trying to set too many values ATM, so something where I wouldn't duplicate effort and if I don't perform well, nothing is really detracted from the overall picture, would be preferred.

Ed - here is a C182 I’m chasing now

https://www.mersenneforum.org/showpo...9&postcount=74

I’m working c5 < 1M, using msieve-GPU. Any help welcome!

EdH · 2020-05-09, 00:15

Quote:

Originally Posted by swellman

Ed - here is a C182 I’m chasing now

https://www.mersenneforum.org/showpo...9&postcount=74

I’m working c5 < 1M, using msieve-GPU. Any help welcome!

Thanks! I'll see if I can get something working here with CADO-NFS. I might look into setting up a Colab instance with msieve GPU as well. It's been a long time since i used msieve with a GPU. All my GPUs were 1.x architecture.

EdH · 2020-05-09, 14:57

I'm sure this is a poor poly, but am I on the right track?

Code:

skew: 23193644.637
c0: -27677017276391468847497030636368115901471780
c1: -10552743874205252207093977517839426343
c2: 1210164270654515308018946622990
c3: 43735071840910583726351
c4: -886925003073138
c5: 6165600
Y0: -120879887180705227225827835478045022
Y1: 1217953568360884501991
# MurphyE (Bf=4.295e+09,Bg=2.147e+09,area=2.416e+16) = 1.110e-08
# f(x) = 6165600*x^5-886925003073138*x^4+43735071840910583726351*x^3+1210164270654515308018946622990*x^2-10552743874205252207093977517839426343*x-27677017276391468847497030636368115901471780
# g(x) = 1217953568360884501991*x-120879887180705227225827835478045022

cownoise said:

Code:

Best Skew: 31354947.23598
. . .
31354947.23598      5.51074631e-14
. . .

Should I be using the Best Scores listing as a reference of what to look for?

VBCurtis · 2020-05-09, 16:01

Quote:

Originally Posted by EdH

Should I be using the Best Scores listing as a reference of what to look for?

Yep! I look to break the record for one digit tougher, or get within 5% of the current record if the first digit of my composite is 1 or 2. In this case, the C182-183 records are a few years old, so breaking them is fairly likely.

I'd guess 6.5e-14 is a good goal for this composite, and a 7 would not surprise me.

swellman · 2020-05-09, 18:11

I got a decent poly searching down low with msieve-GPU

Code:

n: 26521232090195873108384905824300492852413283081683568418163219479089273132380406501680155963531361683795706304607082425988301635509432877463621844114521741860720947862338201013214619
# norm 1.051570e-17 alpha -9.121892 e 6.462e-14 rroots 5
skew: 148988649.66
c0: -6813699213478967095358229786320166183214771200
c1: 534615773158213719542512689605226366504
c2: 9522073202739549597183040910818
c3: -51164897082315737549521
c4: -420892801747988
c5: 965712
Y0: -122390924023389854749571531023242691
Y1: 2233936328262373

For reference, the goals for e-score per msieve are as follows:

Threshold: 6.28e-14

Objective: > 7.23e-14

Best found to date: 6.732e-14

Searching 8-10M next using msieve-GPU

EdH · 2020-05-09, 21:24

Thanks Guys!

I've got another run going. I had to adjust some scripts. I couldn't get a Colab GPU session compiled. I thought I had made the necessary changes, but I need to go look through some threads to see if I missed something.

How do you derive threshold and objective?

swellman · 2020-05-09, 21:55

Quote:

Originally Posted by EdH

How do you derive threshold and objective?

It’s output by msieve at the start of poly search. They are hard coded values based on GNFS difficulty. Finding a poly with a score above threshold is “good enough” to start sieving in most circumstances.

In this case, I mentioned this composite in the thread for finding better CADO parameters in case that work went beyond C180. We now have a baseline for a C182, if the work continues. But even though the poly found is above threshold, it was found in < 60 hours by msieve-GPU. CADO can probably do much better. Heck, even msieve-GPU will likely keep finding better polys with higher search ranges for c5. So I’ll go on searching until it’s clearly counterproductive.

Think we can find a 7-handle?

EdH · 2020-05-09, 22:11

Quote:

Originally Posted by swellman

It’s output by msieve at the start of poly search. They are hard coded values based on GNFS difficulty. Finding a poly with a score above threshold is “good enough” to start sieving in most circumstances.

In this case, I mentioned this composite in the thread for finding better CADO parameters in case that work went beyond C180. We now have a baseline for a C182, if the work continues. But even though the poly found is above threshold, it was found in < 60 hours by msieve-GPU. CADO can probably do much better. Heck, even msieve-GPU will likely keep finding better polys with higher search ranges for c5. So I’ll go on searching until it’s clearly counterproductive.

Think we can find a 7-handle?

I'm playing around with both CADO-NFS and trying to get a Colab msieve-GPU session working. I don't expect much from my current work, but since my interest is piqued, I kind of think once the 198 team sieve is finished, I will try a full "farm" poly search run and see what turns up. By then I should have my "poly" scripts fine tuned. (This is where my ego says, "7? No problem!")

See what you did by putting my name on a list?

VBCurtis · 2020-05-10, 02:17

I can't let a poly search just sit there with others having all the fun.
Taking 14-15M on CADO with incr=2310, P=3M, nq=15625.

That's about 3% of a full search, just to see what turns up.

EdH · 2020-05-10, 02:54

Quote:

Originally Posted by VBCurtis

I can't let a poly search just sit there with others having all the fun.
Taking 14-15M on CADO with incr=2310, P=3M, nq=15625.

That's about 3% of a full search, just to see what turns up.

Earlier today I started a default CADO-NFS run with the minor machines here. I expect to add the major machines when they finish sieving. How does my default run affect your 14-15M area? Is it full duplication, or is there enough randomness to not worry? I left everything at the values in your modified params.c180.

VBCurtis · 2020-05-10, 04:51

I believe the default settings only go to 2-3M; I went for 14M to avoid any duplication.
That said, I used admin 14e6 and admax 15e6; after running all the workunits, it thinks it is 99.6% done and won't proceed to rootopt.
I've tried changing admax to a multiple of 2310, changing admax to far below 15e6 so that it is clearly more than 100% done, and tried setting admin/admax/adrange all to zero to try to force it to finish sizeopt, no joy.

So, I guess I won't be providing a poly for this after all.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
GIMPS wiki account request thread	ixfd64	mersennewiki	169	2018-09-21 05:43
Polynomial Discriminant is n^k for an n-1 degree polynomial	carpetpool	Miscellaneous Math	14	2017-02-18 19:46
Lost Prime Raider password request thread	cheesehead	Forum Feedback	6	2009-07-28 13:02
Polynomial	R.D. Silverman	NFSNET Discussion	13	2005-09-16 20:07
Deutscher Thread (german thread)	TauCeti	NFSNET Discussion	0	2003-12-11 22:12

2020-05-09, 21:24	#1887
EdH "Ed Hall" Dec 2009 Adirondack Mtns DDA₁₆ Posts	Thanks Guys! I've got another run going. I had to adjust some scripts. I couldn't get a Colab GPU session compiled. I thought I had made the necessary changes, but I need to go look through some threads to see if I missed something. How do you derive threshold and objective?

2020-05-10, 02:17	#1890
VBCurtis "Curtis" Feb 2005 Riverside, CA 3·1,543 Posts	I can't let a poly search just sit there with others having all the fun. Taking 14-15M on CADO with incr=2310, P=3M, nq=15625. That's about 3% of a full search, just to see what turns up.

2020-05-10, 04:51	#1892
VBCurtis "Curtis" Feb 2005 Riverside, CA 3×1,543 Posts	I believe the default settings only go to 2-3M; I went for 14M to avoid any duplication. That said, I used admin 14e6 and admax 15e6; after running all the workunits, it thinks it is 99.6% done and won't proceed to rootopt. I've tried changing admax to a multiple of 2310, changing admax to far below 15e6 so that it is clearly more than 100% done, and tried setting admin/admax/adrange all to zero to try to force it to finish sizeopt, no joy. So, I guess I won't be providing a poly for this after all.