View Single Post
Old 2013-07-03, 12:23   #9
WraithX
 
WraithX's Avatar
 
Mar 2006

23·59 Posts
Default

I'm starting to do some test sieving on the polynomials I have found for my gnfs C210. I was wondering several different things about this step:

1) Should I try to find the best set of parameters (rlim,alim,lpbr,lpba,mfbr,mfba) using one polynomial, and then use those parameters to compare all the different polynomials I have? Or should I pick one set of parameters, use those to compare my different polynomials, and then try to optimize the parameters based on the winning polynomial?

For a gnfs C210:
2) Should I try sieving on the rational or algebraic side? Or does this depend on the winning polynomial?
3) Should I use 2LP or 3LP for the different sides? Or does this also depend on the winning polynomial?
4) When using 2LP that means: mfbr = 2*lpbr and/or mfba = 2*lpba and rlambda/alambda ~= 2.6 (or 2.7), right?
5) When using 3LP that means: mfbr = 3*lpbr and/or mfba = 3*lpba and rlambda/alambda ~= 3.6 (or 3.7), right?
I ask 4 and 5 because I sometimes see people quote params of 33-bit lpbr/lpba with 64-bit mfbr/mfba or a 33-bit lpbr/lpba with 96-bit mfbr/mfba, instead of 33&66 or 33&99.

Right now I'm testing these variables:
rlim = {200e6, 350e6, 500e6}
alim = {200e6, 350e6, 500e6}
lpbr = {32, 33}
lpba = {32, 33}
mfbr = 64
mfba = 96
at Q=100e6 and Q=500e6.
6) Are these reasonable test points, or should I try some other variations too?

7) Is 16e the right siever to use for a gnfs C210?
8) Does (total yield)*(sec/rel) = total time to find those relations?
9) How many raw relations should I be looking to collect for this gnfs C210?
10) What is a good range of Q to test to find these relations? Q = {50e6 to 500e6}? Or Q = {100e6 to 1000e6}? Or...?

Also, down below I am including the results of my test sieving. Do these look like reasonable outputs?

I did "-f 100000000 -c 1000" and "-f 500000000 -c 1000" for each test to hopefully get a good approximation of how sieving will go over these big ranges.

I'm sure there is probably more I should be asking, but these are the big questions I have for right now. I'd appreciate any insights or advice anyone can give on this big project I'm about to undertake.

Of the 36 tests suggested above, with results down below, it looks like I probably want to go with t29 or t32. Does this sound right from these results?
Code:
*** Test *** rlim  alim  lpbr lpba mfbr mfba

*** t01 *** 200e6 200e6 32 32 64 96 
total yield: 586, q=100001029 (2.49877 sec/rel) 
total yield: 434, q=500001001 (3.45742 sec/rel) 

*** t02 *** 350e6 200e6 32 32 64 96 
total yield: 640, q=100001029 (2.51675 sec/rel) 
total yield: 468, q=500001001 (3.50973 sec/rel) 

*** t03 *** 500e6 200e6 32 32 64 96 
total yield: 665, q=100001029 (2.61559 sec/rel) 
total yield: 480, q=500001001 (3.67786 sec/rel) 

*** t04 *** 200e6 350e6 32 32 64 96 
total yield: 586, q=100001029 (2.49873 sec/rel) 
total yield: 457, q=500001001 (3.63930 sec/rel) 

*** t05 *** 350e6 350e6 32 32 64 96 
total yield: 640, q=100001029 (2.51599 sec/rel) 
total yield: 492, q=500001001 (3.65579 sec/rel) 

*** t06 *** 500e6 350e6 32 32 64 96 
total yield: 665, q=100001029 (2.61557 sec/rel) 
total yield: 505, q=500001001 (3.81100 sec/rel)

*** t07 *** 200e6 500e6 32 32 64 96 
total yield: 586, q=100001029 (2.50044 sec/rel) 
total yield: 469, q=500001001 (3.81422 sec/rel) 

*** t08 *** 350e6 500e6 32 32 64 96 
total yield: 640, q=100001029 (2.51385 sec/rel) 
total yield: 505, q=500001001 (3.82250 sec/rel) 

*** t09 *** 500e6 500e6 32 32 64 96 
total yield: 665, q=100001029 (2.61408 sec/rel) 
total yield: 519, q=500001001 (3.96526 sec/rel)

*** t10 *** 200e6 200e6 32 33 64 96 
total yield: 897, q=100001029 (1.63303 sec/rel) 
total yield: 703, q=500001001 (2.14675 sec/rel) 

*** t11 *** 350e6 200e6 32 33 64 96 
total yield: 993, q=100001029 (1.62162 sec/rel) 
total yield: 748, q=500001001 (2.19623 sec/rel) 

*** t12 *** 500e6 200e6 32 33 64 96 
total yield: 1029, q=100001029 (1.69091 sec/rel) 
total yield: 771, q=500001001 (2.29173 sec/rel) 

*** t13 *** 200e6 350e6 32 33 64 96 
total yield: 897, q=100001029 (1.63289 sec/rel) 
total yield: 761, q=500001001 (2.18615 sec/rel) 

*** t14 *** 350e6 350e6 32 33 64 96 
total yield: 993, q=100001029 (1.62174 sec/rel) 
total yield: 808, q=500001001 (2.22832 sec/rel) 

*** t15 *** 500e6 350e6 32 33 64 96 
total yield: 1029, q=100001029 (1.69071 sec/rel) 
total yield: 833, q=500001001 (2.31128 sec/rel) 

*** t16 *** 200e6 500e6 32 33 64 96 
total yield: 897, q=100001029 (1.63459 sec/rel) 
total yield: 784, q=500001001 (2.28254 sec/rel) 

*** t17 *** 350e6 500e6 32 33 64 96 
total yield: 993, q=100001029 (1.62053 sec/rel) 
total yield: 836, q=500001001 (2.31623 sec/rel) 

*** t18 *** 500e6 500e6 32 33 64 96 
total yield: 1029, q=100001029 (1.68960 sec/rel) 
total yield: 862, q=500001001 (2.39455 sec/rel) 

*** t19 *** 200e6 200e6 33 32 64 96 
total yield: 783, q=100001029 (1.89119 sec/rel) 
total yield: 589, q=500001001 (2.57094 sec/rel) 

*** t20 *** 350e6 200e6 33 32 64 96 
total yield: 861, q=100001029 (1.89129 sec/rel) 
total yield: 644, q=500001001 (2.57473 sec/rel) 

*** t21 *** 500e6 200e6 33 32 64 96 
total yield: 900, q=100001029 (1.95379 sec/rel) 
total yield: 663, q=500001001 (2.68922 sec/rel) 

*** t22 *** 200e6 350e6 33 32 64 96 
total yield: 783, q=100001029 (1.89126 sec/rel) 
total yield: 621, q=500001001 (2.70439 sec/rel) 

*** t23 *** 350e6 350e6 33 32 64 96 
total yield: 861, q=100001029 (1.89104 sec/rel) 
total yield: 680, q=500001001 (2.67079 sec/rel) 

*** t24 *** 500e6 350e6 33 32 64 96 
total yield: 900, q=100001029 (1.95429 sec/rel) 
total yield: 702, q=500001001 (2.76947 sec/rel) 

*** t25 *** 200e6 500e6 33 32 64 96 
total yield: 783, q=100001029 (1.89111 sec/rel) 
total yield: 638, q=500001001 (2.83263 sec/rel) 

*** t26 *** 350e6 500e6 33 32 64 96 
total yield: 861, q=100001029 (1.89246 sec/rel) 
total yield: 698, q=500001001 (2.79481 sec/rel) 

*** t27 *** 500e6 500e6 33 32 64 96 
total yield: 900, q=100001029 (1.95448 sec/rel) 
total yield: 721, q=500001001 (2.88571 sec/rel) 

*** t28 *** 200e6 200e6 33 33 64 96 
total yield: 1183, q=100001029 (1.25192 sec/rel) 
total yield: 945, q=500001001 (1.60321 sec/rel) 

*** t29 *** 350e6 200e6 33 33 64 96 
total yield: 1310, q=100001029 (1.24225 sec/rel) 
total yield: 1023, q=500001001 (1.62118 sec/rel) 

*** t30 *** 500e6 200e6 33 33 64 96 
total yield: 1372, q=100001029 (1.28140 sec/rel) 
total yield: 1056, q=500001001 (1.68797 sec/rel) 

*** t31 *** 200e6 350e6 33 33 64 96 
total yield: 1183, q=100001029 (1.25142 sec/rel) 
total yield: 1017, q=500001001 (1.65128 sec/rel) 

*** t32 *** 350e6 350e6 33 33 64 96 
total yield: 1310, q=100001029 (1.24304 sec/rel) 
total yield: 1102, q=500001001 (1.64791 sec/rel) 

*** t33 *** 500e6 350e6 33 33 64 96 
total yield: 1372, q=100001029 (1.28214 sec/rel) 
total yield: 1141, q=500001001 (1.70360 sec/rel) 

*** t34 *** 200e6 500e6 33 33 64 96 
total yield: 1183, q=100001029 (1.25128 sec/rel) 
total yield: 1048, q=500001001 (1.72428 sec/rel) 

*** t35 *** 350e6 500e6 33 33 64 96 
total yield: 1310, q=100001029 (1.24330 sec/rel) 
total yield: 1138, q=500001001 (1.71471 sec/rel) 

*** t36 *** 500e6 500e6 33 33 64 96 
total yield: 1372, q=100001029 (1.28211 sec/rel) 
total yield: 1178, q=500001001 (1.76656 sec/rel)
WraithX is offline   Reply With Quote