Fast Breeding (guru management)

[jyb]

Originally Posted by pinhodecarlos

From Jon Polynomials for homogeneous Cunningham numbers, rest on the attached file.

These polynomials didn't contain the parameters to use when I sent them to Carlos. SNFS difficulties range from 237 to 252 and judging by existing jobs at those levels, they should all be able to use the same values. The attached file contains the same polynomials as above, but with these parameter choices added, so they should all be ready to go right into the queue.

Example:

# 8-3,263
n: 1074220397608063744092456701155722449290119666365802488437623526859160975424802319394559862204759874721680339424141057973064241372305556832686404359799262209821344487053
# 8^263-3^263, difficulty: 238.89, skewness: 1.18, alpha: 0.00
skew: 1.178
c6: 3
c0: -8
Y1: -984770902183611232881
Y0: 5444517870735015415413993718908291383296
rlim: 134217727
alim: 134217727
lpbr: 31
lpba: 31
mfbr: 62
mfba: 62
rlambda: 2.6
alambda: 2.6

# 5+3,344
n: 50947357611569563899665041936661858420133932467666010404141837622987159151159644790341742059904557155078085087274009047956506693686019417370450122236059032711789321363607489
# 5^344+3^344, difficulty: 240.45, skewness: 0.84, alpha: 0.00
skew: 0.843
c6: 25
c0: 9
Y1: -1570042899082081611640534563
Y0: 6938893903907228377647697925567626953125
rlim: 134217727
alim: 134217727
lpbr: 31
lpba: 31
mfbr: 62
mfba: 62
rlambda: 2.6
alambda: 2.6

[pinhodecarlos]

Let's please queue all HCN composites. It will keep the clients busy for two months.

[swellman]

Sorry for this cross post but the following list of candidates for NFS@Home was posted in the xyyxf sub-forum:

C200_123_79 * queued 29/6
C204_126_71 * queued 4/7
C223_139_48 * queued 29/6
C224_139_50 * queued 4/7
C235_141_49 * queued 30/6
C207_143_47 * queued 30/6
C235_143_48 * queued 30/6
C185_145_43 * queued 29/6
C207_145_48 * queued 30/6
C205_148_80 * queued 29/6

Not sure if it was noticed.

[fivemack]

I've queued five of the HCN numbers with difficulty above 250.

[pinhodecarlos]

Originally Posted by fivemack

I've queued five of the HCN numbers with difficulty above 250.

Thank you. Better reserve them here http://www.chiark.greenend.org.uk/uc...mack/homcun.pl.

[fivemack]

I'm going to try to build some automation for parameter choice and range selection for XYYXF numbers

[LaurV]

Originally Posted by pinhodecarlos

Thank you. Better reserve them here http://www.chiark.greenend.org.uk/uc...mack/homcun.pl.

I don't see anything with difficulty over 250 in that list (?!)

[jyb]

Originally Posted by LaurV

I don't see anything with difficulty over 250 in that list (?!)

There are actually a small number of them. If you click on the "SNFS Difficulty (digits)" link, the table will by sorted by that and you can see the most difficult at the bottom.

However, that only partly answers the question. That web page calculates the SNFS difficulty by looking at the raw base and exponent values and doing the obvious thing. It doesn't take into account any increases in difficulty which might arise from having to actually generate practical polynomials for these numbers.

For example, 7+2,298 shows up with a difficulty of 251.8, because 298 * log₁₀7 = 251.8. However, in making a polynomial for SNFS at this approximate difficulty, one would use degree 6, which requires massaging sufficient to cause the difficulty to actually be 300*log₁₀7 + log₁₀4, which is 254.13.

After all such massaging is taken into account, there turn out to be 8 numbers with difficulty over 250. If you look at the combined .poly file I attached, you can see the most difficult numbers with something like this:

jon% grep difficulty HCN_polys.txt | sed -E 's/# ([^,]*),.*difficulty: ([^,]*),.*/\1: \2/g' | sort -n -k 2 -r
7^298+3^298: 254.48
7^298+2^298: 254.13
5^358+4^358: 252.83
5^359+3^359: 252.11
5^359+2^359: 251.93
8^277-3^277: 250.16
7^296+5^296: 250.15
7^296+4^296: 250.15
3^523-2^523: 249.53
12^226+5^226: 246.25
...

[swellman]

Here's a few more xyyx candidates for 14e:

C183_131_61 QUEUED 30/6
C208_132_61 QUEUED 30/6
C201_126_73 QUEUED 30/6

I've started preprocessing the low SNFS xyyx numbers with ECM prior to suggesting them to NFS@Home. Thank you.

[fivemack]

candidates.txt has lines of the form C123_45_67

../results2.txt is the file http://www.primefan.ru/xyyxf/results2.txt

rename the (I am a stoat; the file is attached in a new thread elsewhere) as automate.py, run automate.py and then something like

for u in *.initial_snfs; do ../gnfs-lasieve4I14e -a $u -f 130000000 -c 10000 2> $u.Axt & ../gnfs-lasieve4I14e -r $u -f 130000000 -c 10000 2> $u.Rxt & done

I am sure my python is hopelessly unidiomatic, probably make a new thread elsewhere to complain about that -- this message is already getting off-topic for this thread.

(I am testing it on #608 and #614 so don't bother rerunning on those)

[jyb]

Originally Posted by fivemack

rename the attached file as automate.py, run automate.py and then something like

for u in *.initial_snfs; do ../gnfs-lasieve4I14e -a $u -f 130000000 -c 10000 2> $u.Axt & ../gnfs-lasieve4I14e -r $u -f 130000000 -c 10000 2> $u.Rxt & done

Attached file?