New GGNFS problem

[10metreh]

I've tried both ways and it crashed both times. Perhaps there is a bug in the code?

Here is all in my log that is relevant (I started this last night, but it did not want to resume this morning):

 
[11/28 18:48:44] GGNFS-0.77.1-20060722-pentium4 : makefb
[11/28 18:48:50] name: 
[11/28 18:48:50] n=1230055008086286534595627201713118974133964495903086531402839064679655229427876182135719467916304601 (100 digits)
[11/28 18:48:50] c0: 4404605767354605876157364
[11/28 18:48:50] c1: -1198874897864449346549
[11/28 18:48:50] c2: 36988141245696361
[11/28 18:48:50] c3: 7321780823831
[11/28 18:48:50] c4: -140496168
[11/28 18:48:50] c5: 1260
[11/28 18:48:50] RFBsize: 135072 (upto 1799999)
[11/28 18:48:50] AFBsize: 134534 (upto 1799999)
[11/28 18:48:50] maxNumLargeRatPrimes: 3
[11/28 18:48:50] maxLargeRatPrime: 67108864
[11/28 18:48:50] maxNumLargeAlgPrimes: 3
[11/28 18:48:50] maxLargeAlgPrime: 67108864
->               minimum number of FF's: 302030
->               minimum number of FF's: 302030
->               minimum number of FF's: 302030
 LatSieveTime: 8956
[11/29 10:49:05] GGNFS-0.77.1-20060722-pentium4 : procrels
->               minimum number of FF's: 302030

VIE (very important edit): I have downloaded another version of procrels and it works! There was a bug! I will have to redo 2 1/2 hours worth of sieving, though.

[schickel]

Originally Posted by 10metreh

VIE (very important edit): I have downloaded another version of procrels and it works! There was a bug! I will have to redo 2 1/2 hours worth of sieving, though.

Hmmm, very likely the version you had was compiled for a different architecture. That'll usually result in that "quiet" type of crash (no error reported).

Originally Posted by 10metreh

LatSieveTime: 8956

If you're not comfortable with 2-1/2 hours per block, you can always reduce the "qstep" value to 1/2 that or so.....

Frank

[henryzz]

Originally Posted by schickel

Hmmm, very likely the version you had was compiled for a different architecture.

the link i gave him for binaries was an recent svn version for p4
he had said previously that he was running on a 1.7 Ghz p4

i have been using the exact same svn version complied for core 2
maybe that p4 build was compiled for prescott and his isnt a prescott?????

@10metreh
it seems like you pc does siieving at 1/5th of the speed of my pc

[10metreh]

Perhaps. The lattice siever, polyselect and makefb all worked fine.

[ET_]

Originally Posted by henryzz

i am just doing the linear algebra on a 100 digit gnfs for my aliquot sequence
q went up to 1.4M and that produced 3.6M relations

it will save the relations, at the 100 digit level, every 100k q values
it will also do a postprocessing run to see if you have sieved enough

Wouldn't it be quicker using msieve QS on C100?

Luigi

[henryzz]

Originally Posted by ET_

Wouldn't it be quicker using msieve QS on C100?

Luigi

maybe he has never completed a gnfs before so it is worth him starting small
also as the difference isnt much you could use something like RDS's arguments and say what can you learn from QS
there is a lot more to learn from doing a gnfs factorization that a QS

[ET_]

Originally Posted by henryzz

maybe he has never completed a gnfs before so it is worth him starting small
also as the difference isnt much you could use something like RDS's arguments and say what can you learn from QS
there is a lot more to learn from doing a gnfs factorization that a QS

I see what you mean...
I started my first GNFS with a C123 and now I can say that I've chosen a big candidate to start with.

Luigi

[henryzz]

Originally Posted by ET_

I see what you mean...
I started my first GNFS with a C123 and now I can say that I've chosen a big candidate to start with.

Luigi

what hardware was that with
my current record is C121 although that was with my old 3800+
i could do lots better with my newish Q6600 but i am waiting for the new version of ggnfs before i break that record

[10metreh]

On my PC msieve QS is SOOOO slow.

[henryzz]

Originally Posted by 10metreh

On my PC msieve QS is SOOOO slow.

QS seems slow to start off with but seems to speed up
the amount a QS factorization is done seems to me to be based on how many partial rels need to be found
unfortunately that number is different for different length numbers
just recently i have looked in my msieve.log for how many partial rels were needed for other factoizations and have been able to make a good guess at how many partial rels are needed for a specific factorization

can someone more knowledgable than me confirm that the rate of producing partial rels with msieve is linear throughout a factorization

[jasonp]

Yes, partial relations are expected to accumulate at an approximately linear rate, but you do not know in advance what that rate will be with great accuracy, or how many partial relations will be 'enough'. There is code to determine when 'enough' has accumulated, so that you will not sieve more than you have to, but it's tricky to find out how long it will take to get there.

What matters with QS is not the number of partial relations collected but how many cycles (full relations) they form. Cycles initially form incredibly slowly but the rate accelerates as sieving progresses. At the 100-digit level I think you need about 1.5-2 million relations, and on a modern PC it takes 11-12 hours. Individual numbers can be better or worse than that, sometimes by a significant factor. If waiting 12 hours for a factor sounds like too much to bear, that's a different problem.