Factoring database

[chris2be8]

Is anyone running yafu.pl to factor small composites in factordb? I'm running similar scripts (changed to do ECM and poly selection with my GPU, the rest on other systems), but I'm slowly losing ground as new composites are entered faster than I can factor them (I'm only using a couple of systems too slow for more serious work).

Chris

[Antonio]

Originally Posted by chris2be8

Is anyone running yafu.pl to factor small composites in factordb? I'm running similar scripts (changed to do ECM and poly selection with my GPU, the rest on other systems), but I'm slowly losing ground as new composites are entered faster than I can factor them (I'm only using a couple of systems too slow for more serious work).

Chris

I'm running my Perl script when my machine is idle, I've cleared >35000 composites since April 16th. Not run it since July 1st though.

[Antonio]

Originally Posted by Antonio

I'm running my Perl script when my machine is idle, I've cleared >35000 composites since April 16th. Not run it since July 1st though.

OK, cleared 2600 small (C70 to C91) composites overnight.

[chris2be8]

Thanks.

Chris

[Antonio]

Originally Posted by chris2be8

Thanks.

Chris

No problem, cleared another 2100 today between other jobs.
Will leave it running overnight now I've finished working on my machine for the day

[R. Gerbicz]

5 years old post:

Originally Posted by R. Gerbicz

Now there is more than 19000 numbers in the db for that 10 Miller-Rabin tests fails (with gmp-5.0.1), the previous count was approx. 1500. See http://factordb.com/prooffailed.php

Now a new world record (in this gmp version): for n=5376825678759197295101416291 I have found that 17 Miller Rabin tests fails!

A new record with some luck for
n=1396981702787004809899378463251
21 Miller-Rabin tests fails (on gmp-6.1.1, but it should be the same on not very-very old gmp versions also). More large orders at: http://factordb.com/prooffailed.php .

[paulunderwood]

Originally Posted by R. Gerbicz

5 years old post:


A new record with some luck for
n=1396981702787004809899378463251
21 Miller-Rabin tests fails (on gmp-6.1.1, but it should be the same on not very-very old gmp versions also). More large orders at: http://factordb.com/prooffailed.php .

All the more reason for GMP to include a Lucas sequence type test?

[chris2be8]

That explains why the following code from factMsieve.pl said a lot of small numbers in factordb are prime when they are really composite.

if (eval('use GMP::Mpz;1;')) {
  *probab_prime_p = sub {
    GMP::Mpz::probab_prime_p("$_[0]", $_[1]);  #convert Math::BigInt to scalar
  };
}

I had to comment out that check to get my version of it to factor them. Now I'm relying on ECPP which seems to work OK.

Chris

[ATH]

Originally Posted by R. Gerbicz

A new record with some luck for
n=1396981702787004809899378463251
21 Miller-Rabin tests fails (on gmp-6.1.1, but it should be the same on not very-very old gmp versions also). More large orders at: http://factordb.com/prooffailed.php .

Also 21 tests on MPIR 2.7.2 and GMP 6.1.2

[ATH]

Testing for fun which bases n=1396981702787004809899378463251 fails for:

Between bases 2 and 1000: 265 of them fails. Base 2 actually does not fail, so GMP does not use base 2 apparantly.

These bases<1000 shows n as PRP:

3 9 16 17 19 20 22 25 27 31 41 43 46 48 51 53 56 57 58 60 61 66 70 75 77 81 83 93 101 103 104 113 118 123 129 130 138 139 143 144 153 157 159 161 163 168 171 174 180 183 188 193 196 198 203 210 225 227 229 231 235 239 243 245 249 254 256 262 268 271 272 277 279 281 289 292 296 299 302 303 304 309 312 316 317 320 323 334 335 339 340 347 352 354 361 362 364 365 369 370 373 374 377 380 382 387 388 390 395 398 400 407 413 414 417 418 421 425 428 429 431 432 433 440 446 455 459 461 463 466 471 475 477 483 484 485 487 489 491 496 499 500 502 504 509 513 522 523 527 535 540 547 548 549 550 557 564 568 577 579 588 589 594 596 601 605 609 620 622 625 630 656 658 674 675 676 681 682 685 686 687 688 693 697 698 705 710 712 716 717 719 727 729 731 734 735 736 745 747 751 757 762 767 768 775 778 779 781 782 786 787 797 802 804 809 813 816 817 820 821 831 837 843 845 848 851 859 860 867 872 874 876 886 888 889 890 895 896 897 898 901 902 906 909 912 914 917 920 927 928 936 938 941 946 947 948 951 952 960 961 969 976 979 983 986

2-10000: 2559 bases fails

2-100000: 25017 bases fails

[henryzz]

Originally Posted by ATH

Testing for fun which bases n=1396981702787004809899378463251 fails for:

Between bases 2 and 1000: 265 of them fails. Base 2 actually does not fail, so GMP does not use base 2 apparantly.

These bases<1000 shows n as PRP:

3 9 16 17 19 20 22 25 27 31 41 43 46 48 51 53 56 57 58 60 61 66 70 75 77 81 83 93 101 103 104 113 118 123 129 130 138 139 143 144 153 157 159 161 163 168 171 174 180 183 188 193 196 198 203 210 225 227 229 231 235 239 243 245 249 254 256 262 268 271 272 277 279 281 289 292 296 299 302 303 304 309 312 316 317 320 323 334 335 339 340 347 352 354 361 362 364 365 369 370 373 374 377 380 382 387 388 390 395 398 400 407 413 414 417 418 421 425 428 429 431 432 433 440 446 455 459 461 463 466 471 475 477 483 484 485 487 489 491 496 499 500 502 504 509 513 522 523 527 535 540 547 548 549 550 557 564 568 577 579 588 589 594 596 601 605 609 620 622 625 630 656 658 674 675 676 681 682 685 686 687 688 693 697 698 705 710 712 716 717 719 727 729 731 734 735 736 745 747 751 757 762 767 768 775 778 779 781 782 786 787 797 802 804 809 813 816 817 820 821 831 837 843 845 848 851 859 860 867 872 874 876 886 888 889 890 895 896 897 898 901 902 906 909 912 914 917 920 927 928 936 938 941 946 947 948 951 952 960 961 969 976 979 983 986

2-10000: 2559 bases fails

2-100000: 25017 bases fails

How many do average numbers fail?