Reserved for MF - Sequence 4788

[Batalov]

msieve143_gpu.exe -np 1200,2400
didn't find polynomials yet, but it only took 9 hrs. I'll try another interval.

[jrk]

Here's a preliminary poly found with Jason's new msieve-gpu branch.

# norm 2.523819e-15 alpha -7.877120 e 1.839e-12
skew: 2785154.52
c0: -138570565461588730821881652221421126805
c1: -356118891405782364779335898545631
c2:  157627527368055796683392529
c3:  368273140050413578367
c4: -24904624740140
c5:  949200
Y0: -1035248474311472807522765292042
Y1:  832063911771386519

Note that msieve's final_norm cutoff for this composite is 1.835e-12.

[bsquared]

My curvefits suggest > 1.95e-12 would be good.

[10metreh]

Originally Posted by bsquared

My curvefits suggest > 1.95e-12 would be good.

The previous c157 had a 2.00e-12 poly.

[bsquared]

Originally Posted by 10metreh

The previous c157 had a 2.00e-12 poly.

And it was good .

[Greebley]

Originally Posted by Greebley

I was thinking it would be neat to get 314718 to index 9000, however it is at 8902 now so that would be a long 92 indices more.
It is index 2540 for 4788.

Both those figures are wrong: 314718 is at index 8908, and 4788 is at index 2448.

Yeah the index is 8908 which is 92 indices from 9000 so I mis-typed that one.

However, the other value (2540) is the current index + 92 i.e. we could stop at index 2540 for that is when we reach 9000 for 314718. So for that one it just wasn't clear what I was trying to say.

To summarize:
My suggestion is to run the sequence 4788 to index 2540 if we can.

[10metreh]

Originally Posted by Greebley

My suggestion is to run the sequence 4788 to index 2540 if we can.

Unless we acquire the downdriver at something ridiculous like 172 digits, in which case we just have to continue.

[axn]

I'll run pol51 for 0-200M using following parameters:

pol51m0b -b c157_1 -p 7 -n 2.8e23 -a 0 -A 200000
pol51opt -b c157_1 -n 3.6e22 -N 3.6e19 -e 1.7e-12

[bsquared]

I left the range running overnight, and woke up to these:

polysel_0-2500.cand:BEGIN POLY #skewness 979863.43 norm 8.85e+20 alpha -5.12 Murphy_E 2.14e-12
polysel_0-2500.cand:BEGIN POLY #skewness 1249264.69 norm 1.13e+21 alpha -5.19 Murphy_E 2.07e-12
polysel_0-2500.cand:BEGIN POLY #skewness 1239920.67 norm 1.08e+21 alpha -5.17 Murphy_E 2.07e-12
polysel_0-2500.cand:BEGIN POLY #skewness 1064109.97 norm 9.82e+20 alpha -4.76 Murphy_E 1.97e-12
polysel_0-2500.cand:BEGIN POLY #skewness 979626.75 norm 9.01e+20 alpha -4.60 Murphy_E 1.94e-12

I'll let it continue to run... in the mean time, anyone care to test sieve the top 2 or 3? We can probably duplicate the siever and parameters from the last C157.

details:

n: 1128681916333165583281832717120127496638972860118923964844791402387721331170448754199775637678274646951935918929237315749341487244542351920306209843985409563
#skewness 979863.43 norm 8.85e+20 alpha -5.12 Murphy_E 2.14e-12
c5: 156600
c4: 5511109690356
c3: 2556240664351703353
c2: -2709038327031089573784479
c1: 772591600803924819123938074991
c0: -127641604647397600699119628378829341
Y1: 64393121809011209
Y0: -1484416691862834512911536697536
skew: 197324.88 
rlim: 45000000
alim: 45000000 
lpbr: 29 
lpba: 29 
mfbr: 58 
mfba: 58 
rlambda: 2.6 
alambda: 2.6
 
n: 1128681916333165583281832717120127496638972860118923964844791402387721331170448754199775637678274646951935918929237315749341487244542351920306209843985409563
#skewness 1249264.69 norm 1.13e+21 alpha -5.19 Murphy_E 2.07e-12
c5: 156600
c4: 5314399732356
c3: -2883058196108786471
c2: -2573465483932011297872477
c1: 2271343202342143863498206577119
c0: 23472793401947121132547837819239017
Y1: 64393121809011209
Y0: -1484416708040060932502186689770
skew: 197324.88 
rlim: 45000000
alim: 45000000 
lpbr: 29 
lpba: 29 
mfbr: 58 
mfba: 58 
rlambda: 2.6 
alambda: 2.6
 
n: 1128681916333165583281832717120127496638972860118923964844791402387721331170448754199775637678274646951935918929237315749341487244542351920306209843985409563
#skewness 1239920.67 norm 1.08e+21 alpha -5.17 Murphy_E 2.07e-12
c5: 156600
c4: 5335915006356
c3: -2297759499328129799
c2: -2787019491787729028389067
c1: 2123826836917266179771394505903
c0: 18333066845778520513050090022694295
Y1: 64393121809011209
Y0: -1484416706270666731434176688868
skew: 197324.88 
rlim: 45000000
alim: 45000000 
lpbr: 29 
lpba: 29 
mfbr: 58 
mfba: 58 
rlambda: 2.6 
alambda: 2.6

I just realized all of those (and a half dozen more 1.9x 's) have the same leading coefficient! Never seen that before...

[henryzz]

once i have finally managed to compile msieve-gpu i will have a CUDA gpu i can run on polynomial selection
it is proving difficult to get environment variables and such like correct

all you need is a nvidia series >=8
those look good polys

[bsquared]

Originally Posted by bsquared

I left the range running overnight, and woke up to these:

polysel_0-2500.cand:BEGIN POLY #skewness 979863.43 norm 8.85e+20 alpha -5.12 Murphy_E 2.14e-12
polysel_0-2500.cand:BEGIN POLY #skewness 1249264.69 norm 1.13e+21 alpha -5.19 Murphy_E 2.07e-12
polysel_0-2500.cand:BEGIN POLY #skewness 1239920.67 norm 1.08e+21 alpha -5.17 Murphy_E 2.07e-12
polysel_0-2500.cand:BEGIN POLY #skewness 1064109.97 norm 9.82e+20 alpha -4.76 Murphy_E 1.97e-12
polysel_0-2500.cand:BEGIN POLY #skewness 979626.75 norm 9.01e+20 alpha -4.60 Murphy_E 1.94e-12

I decided to do some test sieving on the top 3 as well as jrk's msieve poly. I was dissapointed by the performance of the top poly (about 1100 rels in a range of 1000 Q starting at 10M, 0.170 sec/rel), and the others were even worse in comparison. So on a whim I drug out a 1.92e-12 scoring poly with a better alpha (-6.44), and it beat the pants off of the 2.14e-12 poly. Rough curve fits of test sieving 1k blocks from 5M to 50M in steps of 5M suggest we'll need about 43MQ and about 7 million cpu seconds (based on data from a 3.16 GHz core 2) or 81 CPU days.

I think there are better polys to be found, but this is the baseline so far. Lesson to be learned: Murphy_E is a crapshoot!