Choosing the best polynomial - Page 9

firejuggler · 2013-07-09, 14:19

Your parameter are too stringent, enlarge your search space.

lorgix · 2013-07-09, 14:40

Quote:

Originally Posted by firejuggler

Your parameter are too stringent, enlarge your search space.

That was gonna be the next step. I understand there are a handful of parameters that I should get to know in order to get the best polys (normmax etc). I haven't really looked into this yet, and I don't know how to adjust them.

Quote:

Originally Posted by debrouxl

I often see this error. You need to have a copy / symlink of one of the files suitable for the GPU in the current working directory. But I don't currently have access to the computer with a GPU, so that I can check and post the file name...

Turns out the msieve readme was pretty clear on the point that the PTX-file should be in the working directory... Don't know how I missed that.

VBCurtis · 2013-07-12, 08:37

I have been leaving stage 1 norm at default, and setting stage 2 norm tight enough (via sample trial-and-error) to give me 150-250 hits per day. I *think* I divide the default stage2 norm by 20, rounding up slightly to a looser setting, for the C155-163 ranges I have been searching recently. I have stopped using -t option for this size, because stage 1 produces vastly more output than stage 2 can use, so it was slowing screen response without adding to the quantity of good stage 2 hits. If I were to run -np1 and -nps separately, I'd use -t 3 or 4 for this size.

For the -npr step, default score gives a bunch of polys you don't care about, but I like to have output from even 100 nps hits, so I pick an e-value a tenth or two higher than default- but I believe this saves no time.

-Curtis

lorgix · 2013-07-12, 09:11

Quote:

Originally Posted by VBCurtis

I have been leaving stage 1 norm at default, and setting stage 2 norm tight enough (via sample trial-and-error) to give me 150-250 hits per day. I *think* I divide the default stage2 norm by 20, rounding up slightly to a looser setting, for the C155-163 ranges I have been searching recently. I have stopped using -t option for this size, because stage 1 produces vastly more output than stage 2 can use, so it was slowing screen response without adding to the quantity of good stage 2 hits. If I were to run -np1 and -nps separately, I'd use -t 3 or 4 for this size.

For the -npr step, default score gives a bunch of polys you don't care about, but I like to have output from even 100 nps hits, so I pick an e-value a tenth or two higher than default- but I believe this saves no time.

-Curtis

I unfortunately don't know how to tell msieve to a certain value for a certain parameter. And -h doesn't tell me anything about -npr and -nps.

So the speed-up from a GPU should not be used to be more picky in stg1, but to be more thorough, is that right? You'll then get to be more picky in stg2, correct?

firejuggler · 2013-07-12, 14:14

http://mersenneforum.org/showpost.ph...10&postcount=1
http://mersenneforum.org/showpost.ph...8&postcount=22
those 2 post tell you how. note that it is for a RSA896 number.
For a C155, stage 1 can be run @ 6e22 (lower high bound, I usually ue tighter one, 35e21 max), and stage 2 npr @ 4 to 6e21

lorgix · 2013-07-12, 17:12

Quote:

Originally Posted by firejuggler

http://mersenneforum.org/showpost.ph...10&postcount=1
http://mersenneforum.org/showpost.ph...8&postcount=22
those 2 post tell you how. note that it is for a RSA896 number.
For a C155, stage 1 can be run @ 6e22 (lower high bound, I usually ue tighter one, 35e21 max), and stage 2 npr @ 4 to 6e21

Thanks, exactly what I needed.

The only problem now is that I can not get anything to show up in msieve.dat.m (that is where the stage1 results should go, right?).
I've tried msieve -v -np1 "stage1_norm=5e33 X,Y" with various norms and X,Y.

This is in Windows.

firejuggler · 2013-07-12, 17:36

can you try this :
a worktodo.ini with the number
3787388118933512063249259173231597770049112435945032094640277541632460519045009491164974276597271170346757143570242916437315466248541421375702840232794402591
then run
msieve151_gpu -v -np1 "stage1_norm=35e21 760000,761000" .

You should obtain a msieve.dat.m file with 111 lines.

would that work, run ( gpu version or not after np1 is not important)
msieve151_gpu -v nps "stage2_norm=1e100"
That should convert your hit into a .ms file
and finally, run
msieve151_gpu -v -npr "stage2_norm=6e21"

Should everything work, you should obtain this poly

Code:

R0: -1378574698621825449505480543177
R1: 214052541470369
A0: -175040928317314648773325949692456090200
A1: -220312295807467225271477377169614
A2: 49283065065740083251556069
A3: 25532961976751157908
A4: -3884826310244
A5: 760656
skew 3526805.93, size 2.601e-015, alpha -7.664, combined = 1.978e-012 rroots = 3

after deleting the *.fb file,should you run
msieve151_gpu -v -npr "stage2_norm=1e21"
the selected poly would be

Code:

R0: -1378574698730728104819194297831
R1: 214052541470369
A0: 121948457727252329836835394297740640
A1: -187676207057323965464416078462
A2: 3277088097400819010600521
A3: 35407735760975632884
A4: -5819805862724
A5: 760656
skew 973828.53, size 2.336e-015, alpha -6.283, combined = 1.825e-012 rroots = 1

lorgix · 2013-07-12, 18:41

Quote:

Originally Posted by firejuggler

can you try this :
a worktodo.ini with the number
3787388118933512063249259173231597770049112435945032094640277541632460519045009491164974276597271170346757143570242916437315466248541421375702840232794402591
then run
msieve151_gpu -v -np1 "stage1_norm=35e21 760000,761000" .
You should obtain a msieve.dat.m file with 111 lines.
.
.
.

Perfect, now we're cooking.

This helped me figure out that I was using the wrong binary. Long story short: I used a 1.50 with CUDA. That one wouldn't accept those commands. I'm now using 1.51.

msieve151_gpu -v -nps "stage2_norm=1e100"
Worked the way you described, when I realized I had to have the number in worktodo.ini...

-npr starts a new poly select, for some reason.

firejuggler · 2013-07-12, 19:04

what you should get should be

Code:

msieve151_gpu -v -npr "stage2_norm=6e21"


Msieve v. 1.51 (SVN 845)
Fri Jul 12 21:00:20 2013
random seeds: e55f6188 9af5f33e
factoring 3787388118933512063249259173231597770049112435945032094640277541632460
51904500949116497427659727117034675714357024291643731546624854142137570284023279
4402591 (157 digits)
searching for 15-digit factors
commencing number field sieve (157-digit input)
commencing number field sieve polynomial selection
polynomial degree: 5
max stage 1 norm: 1.31e+024
max stage 2 norm: 6.00e+021
min E-value: 1.52e-012
poly select deadline: 1079999
save 2.110925e-015 -6.2647 2764805.42 1.531848e-012 rroots 1
save 2.268584e-015 -6.9547 4176520.15 1.602621e-012 rroots 3
save 2.281509e-015 -7.3900 5544078.81 1.601885e-012 rroots 3
save 2.079984e-015 -6.3564 3112765.76 1.522669e-012 rroots 1
save 2.104211e-015 -6.7130 4107479.43 1.526810e-012 rroots 3
save 2.129228e-015 -6.7822 3890376.34 1.547907e-012 rroots 1
save 2.147478e-015 -7.0735 5120532.84 1.542681e-012 rroots 3
save 2.234085e-015 -6.9063 4094901.57 1.590176e-012 rroots 1
save 2.310882e-015 -6.5885 3067502.06 1.627121e-012 rroots 3
save 2.400789e-015 -6.6774 2943570.62 1.669219e-012 rroots 3
save 2.147813e-015 -5.9398 1972530.13 1.552128e-012 rroots 3
(more save line, check msieve.dat.p file)
(then at the end of the, get a poly)
polynomial selection complete
R0: -1378574698621825449505480543177
R1: 214052541470369
A0: -175040928317314648773325949692456090200
A1: -220312295807467225271477377169614
A2: 49283065065740083251556069
A3: 25532961976751157908
A4: -3884826310244
A5: 760656
skew 3526805.93, size 2.601e-015, alpha -7.664, combined = 1.978e-012 rroots = 3

lorgix · 2013-07-12, 19:12

Oh, I thought that was a new poly select...
It did say commencing number field sieve polynomial selection.
But I feel like I should have figured that out.

The next step worked like you described.

Thanks for the walk-through!

firejuggler · 2013-07-12, 19:17

No problem.
If you work with msieve 1.50, the second stage is run with -np2, and doesn't allow customisation. The poly score, however is slitghtly over-estimated, by 10%.

2013-07-12, 08:37	#91
VBCurtis "Curtis" Feb 2005 Riverside, CA 4,861 Posts	I have been leaving stage 1 norm at default, and setting stage 2 norm tight enough (via sample trial-and-error) to give me 150-250 hits per day. I think I divide the default stage2 norm by 20, rounding up slightly to a looser setting, for the C155-163 ranges I have been searching recently. I have stopped using -t option for this size, because stage 1 produces vastly more output than stage 2 can use, so it was slowing screen response without adding to the quantity of good stage 2 hits. If I were to run -np1 and -nps separately, I'd use -t 3 or 4 for this size. For the -npr step, default score gives a bunch of polys you don't care about, but I like to have output from even 100 nps hits, so I pick an e-value a tenth or two higher than default- but I believe this saves no time. -Curtis Last fiddled with by VBCurtis on 2013-07-12 at 08:38

2013-07-12, 14:14	#93
firejuggler Apr 2010 Over the rainbow 2³×5²×13 Posts	http://mersenneforum.org/showpost.ph...10&postcount=1 http://mersenneforum.org/showpost.ph...8&postcount=22 those 2 post tell you how. note that it is for a RSA896 number. For a C155, stage 1 can be run @ 6e22 (lower high bound, I usually ue tighter one, 35e21 max), and stage 2 npr @ 4 to 6e21 Last fiddled with by firejuggler on 2013-07-12 at 14:17

2013-07-12, 17:36	#95
firejuggler Apr 2010 Over the rainbow 2³·5²·13 Posts	can you try this : a worktodo.ini with the number 3787388118933512063249259173231597770049112435945032094640277541632460519045009491164974276597271170346757143570242916437315466248541421375702840232794402591 then run msieve151_gpu -v -np1 "stage1_norm=35e21 760000,761000" . You should obtain a msieve.dat.m file with 111 lines. would that work, run ( gpu version or not after np1 is not important) msieve151_gpu -v nps "stage2_norm=1e100" That should convert your hit into a .ms file and finally, run msieve151_gpu -v -npr "stage2_norm=6e21" Should everything work, you should obtain this poly Code: R0: -1378574698621825449505480543177 R1: 214052541470369 A0: -175040928317314648773325949692456090200 A1: -220312295807467225271477377169614 A2: 49283065065740083251556069 A3: 25532961976751157908 A4: -3884826310244 A5: 760656 skew 3526805.93, size 2.601e-015, alpha -7.664, combined = 1.978e-012 rroots = 3 after deleting the .fb file,should you run msieve151_gpu -v -npr "stage2_norm=1e21" the selected poly would be Code: R0: -1378574698730728104819194297831 R1: 214052541470369 A0: 121948457727252329836835394297740640 A1: -187676207057323965464416078462 A2: 3277088097400819010600521 A3: 35407735760975632884 A4: -5819805862724 A5: 760656 skew 973828.53, size 2.336e-015, alpha -6.283, combined = 1.825e-012 rroots = 1 Last fiddled with by firejuggler on 2013-07-12 at 17:49*

2013-07-12, 19:17	#99
firejuggler Apr 2010 Over the rainbow 2³×5²×13 Posts	No problem. If you work with msieve 1.50, the second stage is run with -np2, and doesn't allow customisation. The poly score, however is slitghtly over-estimated, by 10%. Last fiddled with by firejuggler on 2013-07-12 at 19:20

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Polynomial Discriminant is n^k for an n-1 degree polynomial	carpetpool	Miscellaneous Math	14	2017-02-18 19:46
Help choosing motherboard please.	Flatlander	GPU Computing	4	2011-01-26 08:15
Choosing the best CPU for sieving	siew	Factoring	14	2010-02-27 10:07
MPQS: choosing a good polynomial	ThiloHarich	Factoring	4	2006-09-05 07:51
Choosing amount of memory	azhad	Software	2	2004-10-16 16:41

2013-07-09, 14:19	#89
firejuggler Apr 2010 Over the rainbow 2600₁₀ Posts	Your parameter are too stringent, enlarge your search space.

2013-07-12, 19:12	#98
lorgix Sep 2010 Scandinavia 1147₈ Posts	Oh, I thought that was a new poly select... It did say commencing number field sieve polynomial selection. But I feel like I should have figured that out. The next step worked like you described. Thanks for the walk-through!