3,436+ done
 

Hi Garo!

You can remove 3,436+.C194 from the list. I've just completed it's factorization using SNFS today :smile: :

[CODE]Factorization completed after 1359.83 seconds, at Wed Apr 13 14:46:16 2005
Original number had 194 digits:
18997425358698322763156002312739527870163513890018498798374568915773691007512954920003162204762433805813898404689542855333209192749957246007989412071919528458412143440543357623150838837246326401
Probable prime factor 1 has 72 digits:
338094490749880892504429319841868531846553260084269152434648779275477217
Probable prime factor 2 has 122 digits:
56189692167307270376356872786519195521886966899523011573488080476665147644811313856523986484373216836371571570505580046753

[/CODE]
--
Cheers,
Jes

[QUOTE=JHansen]Hi Garo!

You can remove 3,436+.C194 from the list. I've just completed it's factorization using SNFS today :smile: :

[CODE]Factorization completed after 1359.83 seconds, at Wed Apr 13 14:46:16 2005
Original number had 194 digits:
18997425358698322763156002312739527870163513890018498798374568915773691007512954920003162204762433805813898404689542855333209192749957246007989412071919528458412143440543357623150838837246326401
Probable prime factor 1 has 72 digits:
338094490749880892504429319841868531846553260084269152434648779275477217
Probable prime factor 2 has 122 digits:
56189692167307270376356872786519195521886966899523011573488080476665147644811313856523986484373216836371571570505580046753

[/CODE]
--
Cheers,
Jes[/QUOTE]

Nice.

What parameters did you use? Factor base size; bound on large primes, etc?
How big was the matrix?

Bob

Thanks.

I used 3X^5+1 and X-3^87 as polynomials. Factor bases were 14.5M each with a LPB of 29 bit. Franke's lattice sieve was used to gather relations and the CWI tools were used for post-processing.

I don't know how long it took to gather the relations, as I'm only using spare cycles at our server at the math. dept.

Matrix size was 2637142 x 2642146, total weight 163113210, requirering about 8.5 days of CPU time an a single Intel Xeon 2.4GHz CPU.

--
Cheers,
Jes

[QUOTE=JHansen]Thanks.

I used 3X^5+1 and X-3^87 as polynomials. Factor bases were 14.5M each with a LPB of 29 bit. Franke's lattice sieve was used to gather relations and the CWI tools were used for post-processing.

I don't know how long it took to gather the relations, as I'm only using spare cycles at our server at the math. dept.

Matrix size was 2637142 x 2642146, total weight 163113210, requirering about 8.5 days of CPU time an a single Intel Xeon 2.4GHz CPU.

--
Cheers,
Jes[/QUOTE]


Interesting. I am doing the LA for 2,1294L now. I used a factor base
bound of 18.8M (1.2 million primes for each polynomial), and a large prime bound of 600M. My final matrix was 4.3 million rows with a weight of
about 145 million and will take 184 hours on my 3.4GHz P IV. (about 60% done)

I used about 380,000 special q's, each with a 6K x 6K sieve region. This
yielded 65 million relations, of which 1/3 were duplicates. This was my own
lattice siever. I use CWI's filter and square root code, but my own solver.
Sieving took just over a month using two Pentiums at 1.5 GHz and another
part time (40 to 50%) at 2.6 GHz. But of course, this number is smaller
than yours.

I suspect that a sextic polynomial would have been slightly better for you:

x^6 + 9, with root 3^73

I am currently sieving 2,737+. What are your plans?

Bob

I have begun small test runs for 3,437-.C151.

I was thinking about using very similar parameters, but any advice from you would be greatly appreciated. :bow:

--
Cheers,
Jes

[QUOTE=JHansen]I have begun small test runs for 3,437-.C151.

I was thinking about using very similar parameters, but any advice from you would be greatly appreciated. :bow:

--
Cheers,
Jes[/QUOTE]

It is *theoretically* possible to take advantage of the factor 3^23 - 1,
but to do so requires using a (reciprocal) 9th degree polynomial... This
isn't useful for numbers of this size.

I would use x^6 - 3 with root 3^73. An optimal factor base would have
a bound around 24 million.

[QUOTE=R.D. Silverman]It is *theoretically* possible to take advantage of the factor 3^23 - 1,
but to do so requires using a (reciprocal) 9th degree polynomial... This
isn't useful for numbers of this size.

I would use x^6 - 3 with root 3^73. An optimal factor base would have
a bound around 24 million.[/QUOTE]

Thanks, I appreciate your advice a lot. Those parameters was also what I had in mind. I will probably use a somewhat smaller factor base in order to keep the matrix size down.

--
Cheers,
Jes

[QUOTE=R.D. Silverman]I used about 380,000 special q's, each with a 6K x 6K sieve region. This
yielded 65 million relations, of which 1/3 were duplicates. This was my own
lattice siever. I use CWI's filter and square root code, but my own solver.[/QUOTE]

Bob, if your sievers produce relations in CWI format, I would like to try them out in order to compare them with Franke's sievers. His sievers all sieve over a half-square (4K X 2K for SNFS ~150 digits or GNFS ~110 digits, 8K X 4K for SNFS ~180 GNFS ~130-140 and 16K X 8K for larger projects).

Also I would like to try and see how your Lanczos code compares to the CWI code I currently use. Sean Irvine claims that Frankes Lanczos code is faster than the CWI code, so maybe yours is too. :smile:

If I'm allowed to use your code, please send me a mail at [email]Jes dot H@math.ku.dk [remove spaces][/email]

--
Cheers,
Jes

[QUOTE=JHansen]Bob, if your sievers produce relations in CWI format, I would like to try them out in order to compare them with Franke's sievers. His sievers all sieve over a half-square (4K X 2K for SNFS ~150 digits or GNFS ~110 digits, 8K X 4K for SNFS ~180 GNFS ~130-140 and 16K X 8K for larger projects).

Also I would like to try and see how your Lanczos code compares to the CWI code I currently use. Sean Irvine claims that Frankes Lanczos code is faster than the CWI code, so maybe yours is too. :smile:

If I'm allowed to use your code, please send me a mail at [email]Jes dot H@math.ku.dk [remove spaces][/email]

--
Cheers,
Jes[/QUOTE]

I can send it, but it is Windows only right now; My Lanczos code is about
the same speed as the CWI code.

I am working on porting it to Unix (A Sparc)

[QUOTE=R.D. Silverman]I can send it, but it is Windows only right now; My Lanczos code is about
the same speed as the CWI code.

I am working on porting it to Unix (A Sparc)[/QUOTE]

That would be great, please send them! :w00t:

--
Cheers,
Jes

[QUOTE=JHansen]That would be great, please send them! :w00t:

--
Cheers,
Jes[/QUOTE]

I am stupid today.....

Send them after I finish the Unix port? Or send the Windows version?///

My Lanczos solver is clumsy; it requires extra code to convert to/from the
CWI matrix storage format and I no longer have the source to do this;
It got left behind when RSA Labs laid me off.

:question:

[QUOTE=R.D. Silverman]Send them after I finish the Unix port? Or send the Windows version?///[/QUOTE]

I was looking for the windows version. :rolleyes:

--
Cheers,
Jes

[QUOTE=JHansen]I was looking for the windows version. :rolleyes:

--
Cheers,
Jes[/QUOTE]

Hi,

I sent the code and the binaries. Do you have any feedback? Any problems? :question: :question:

[QUOTE=R.D. Silverman]I sent the code and the binaries. Do you have any feedback? Any problems? :question: :question:[/QUOTE]
I only recieved the code. There was no binaries attacted in the mail you sent me.

I had a friend help me try to compile the code, but we had no luck. We used Metroweks 7.0 to compile, but got these errors:

[CODE]Error : illegal implicit conversion from 'unsigned int[2]' to
'int *'
asmmpp.c line 300 divrem_asm(r,a[i],b,x);

Error : illegal implicit conversion from 'unsigned int[150]' to
'int *'
asmmpp.c line 489 mpcopy(input,junk1);

Error : illegal implicit conversion from 'int *' to
'unsigned int *'
asmmpp.c line 781 square(a,junkc);

Error : illegal implicit conversion from 'int *' to
'unsigned int *'
asmmpp.c line 824 square(result,temp1);

Error : illegal implicit conversion from 'int *' to
'unsigned int *'
asmmpp.c line 829 square(result,temp1);

Error : illegal implicit conversion from 'int[150]' to
'unsigned int *'
asmmpp.c line 961 square(temp1,temp);

Error : illegal implicit conversion from 'int[150]' to
'unsigned int *'
asmmpp.c line 1424 square(xi,temp);

Error : illegal implicit conversion from 'int[150]' to
'unsigned int *'
asmmpp.c line 1427 square(x2i,temp);

Error : illegal implicit conversion from 'int[150]' to
'unsigned int *'
asmmpp.c line 1430 square(x2i,temp);

Error : illegal implicit conversion from 'int[2]' to
'unsigned int *'
int64.c line 131 mul_single_precI(a[1],b,x1);

Error : illegal implicit conversion from 'int[2]' to
'unsigned int *'
int64.c line 141 mul_single_precI(a[2],b,x21);

Error : illegal implicit conversion from 'int[2]' to
'unsigned int *'
int64.c line 317 mul_single_prec_asm(a[1],b,x1);

Error : illegal implicit conversion from 'int[2]' to
'unsigned int *'
int64.c line 327 mul_single_prec_asm(a[2],b,x21);

Error : illegal implicit conversion from 'int[2]' to
'unsigned int *'
int64.c line 361 mul_single_prec_asm(a[1],a[1],x1);

Error : illegal implicit conversion from 'int[2]' to
'unsigned int *'
int64.c line 372 mul_single_prec_asm(a[1],a[2],x21);

Error : illegal implicit conversion from 'int[2]' to
'unsigned int *'
int64.c line 373 mul_single_prec_asm(a[2],a[2],x22);

+ lots and lots of more of the same....
[/CODE]

My friend also said that there was little hope of compiling the code, since a lot of it is non-standard (i.e. non-unix, I guess :rolleyes: ) code.

Could you re-send the binary and the sample inputs, please :question:

--
Best regards,
Jes Hansen

[QUOTE=JHansen]I only recieved the code. There was no binaries attacted in the mail you sent me.

I had a friend help me try to compile the code, but we had no luck. We used Metroweks 7.0 to compile, but got these errors:

[CODE]Error : illegal implicit conversion from 'unsigned int[2]' to
'int *'
asmmpp.c line 300 divrem_asm(r,a[i],b,x);

Error : illegal implicit conversion from 'unsigned int[150]' to
'int *'
asmmpp.c line 489 mpcopy(input,junk1);

Error : illegal implicit conversion from 'int *' to
'unsigned int *'
asmmpp.c line 781 square(a,junkc);

Error : illegal implicit conversion from 'int *' to
'unsigned int *'
asmmpp.c line 824 square(result,temp1);

Error : illegal implicit conversion from 'int *' to
'unsigned int *'
asmmpp.c line 829 square(result,temp1);

Error : illegal implicit conversion from 'int[150]' to
'unsigned int *'
asmmpp.c line 961 square(temp1,temp);

Error : illegal implicit conversion from 'int[150]' to
'unsigned int *'
asmmpp.c line 1424 square(xi,temp);

Error : illegal implicit conversion from 'int[150]' to
'unsigned int *'
asmmpp.c line 1427 square(x2i,temp);

Error : illegal implicit conversion from 'int[150]' to
'unsigned int *'
asmmpp.c line 1430 square(x2i,temp);

Error : illegal implicit conversion from 'int[2]' to
'unsigned int *'
int64.c line 131 mul_single_precI(a[1],b,x1);

Error : illegal implicit conversion from 'int[2]' to
'unsigned int *'
int64.c line 141 mul_single_precI(a[2],b,x21);

Error : illegal implicit conversion from 'int[2]' to
'unsigned int *'
int64.c line 317 mul_single_prec_asm(a[1],b,x1);

Error : illegal implicit conversion from 'int[2]' to
'unsigned int *'
int64.c line 327 mul_single_prec_asm(a[2],b,x21);

Error : illegal implicit conversion from 'int[2]' to
'unsigned int *'
int64.c line 361 mul_single_prec_asm(a[1],a[1],x1);

Error : illegal implicit conversion from 'int[2]' to
'unsigned int *'
int64.c line 372 mul_single_prec_asm(a[1],a[2],x21);

Error : illegal implicit conversion from 'int[2]' to
'unsigned int *'
int64.c line 373 mul_single_prec_asm(a[2],a[2],x22);

+ lots and lots of more of the same....
[/CODE]

My friend also said that there was little hope of compiling the code, since a lot of it is non-standard (i.e. non-unix, I guess :rolleyes: ) code.

Could you re-send the binary and the sample inputs, please :question:

--
Best regards,
Jes Hansen[/QUOTE]


The lines given above do compile under the Sparc C compiler. I believe
Solaris qualifies as Unix. C is not C++. It is not supposed to be strongly
typed. There is no reason why one should not be able to pass an
signed int to an unsigned int and vice-versa.

I already sent the binaries. The mail did not bounce. I just sent them again.

[QUOTE=R.D. Silverman]I already sent the binaries. The mail did not bounce. I just sent them again.[/QUOTE]

You first mail contained the following files: 163.set, lrange.dat, 737.set, nfsl.dat, callqs.c, mpdim.h, clockticks.c, functions.h, setup.dsp, alphad.h, lattice.dsp, alphag.h, basic.h, 64aux.c, newset.c, int64.c, basic2.c, basic.c, newmpp.c, lpqs.c, asmmpp.c, latticev22.c.

No binaries as far as I can see. :question:

I haven't recieved your second mail. Could you please try again or try jesh AT mail DOT dk ? :question: :unsure:

--
Best regards,
Jes Hansen

Here is the factorisation of 3,424+ c156 with SNFS:

[CODE]
Probable prime factor 1 has 58 digits:
1143472587826871098584068464034179620956699328397935002321
Probable prime factor 2 has 99 digits:
732033695730396234463835677672412999987388556190269939346190255227442809340817557330640494411685249
[/CODE]

Alex

Factors of 3,485+
 

Probable prime factor 1 has 79 digits:
2261326016862042466642129611025196517199337771814425126851286138924310369069241
Probable prime factor 2 has 74 digits:
27541676396773727812323951318988846019722180150493778108865109045202198511

Not an ECM miss. :smile:

Alex

PS: by SNFS.