Running Msieve

[fivemack]

I am now slightly more suspicious about the threading code in 1.33.

At least, I ran a job with -nc 1000000,1000000 -t4 using 1.33 on my quad-core of unclear reliability, which failed oddly; I've just rerun the same job with -t4 using 1.29, and it's completed without issue.

[jasonp]

Originally Posted by fivemack

I am now slightly more suspicious about the threading code in 1.33.

At least, I ran a job with -nc 1000000,1000000 -t4 using 1.33 on my quad-core of unclear reliability, which failed oddly; I've just rerun the same job with -t4 using 1.29, and it's completed without issue.

I guess either there is a threading problem in the code that only shows up for multithreaded runs on fast machines, or both your machines are slightly unstable but only for a version of msieve that has high multithreaded concurency. I try not to blame the computer because in general that means making two mistakes (including blaming the computer), but I frankly don't know where to start looking. Is it possible to underclock the memory to see if the problems persist? Failing that, can anyone audit the matrix multiply code?

[Wacky]

Well, the LA for "2,841- C245" finally finished.
But the sqrt has failed.
:(



Fri Feb 22 10:16:00 2008 Msieve v. 1.33
Fri Feb 22 10:16:00 2008 random seeds: f602689b 5099f3da
Fri Feb 22 10:16:00 2008 factoring 27312187167948796049898887842782596405655381943670578373120353781997447000328700858938482409418918448758778194857586271196003004729048486053340769161697167767956404785889687728602979284674332324623675660860406658739006337124213538943374972682241 (245 digits)
Fri Feb 22 10:16:03 2008 no P-1/P+1/ECM available, skipping
Fri Feb 22 10:16:03 2008 commencing number field sieve (245-digit input)
Fri Feb 22 10:16:03 2008 R0: -1393796574908163946345982392040522594123776
Fri Feb 22 10:16:03 2008 R1: 1
Fri Feb 22 10:16:03 2008 A0: -1
Fri Feb 22 10:16:03 2008 A1: 0
Fri Feb 22 10:16:03 2008 A2: 0
Fri Feb 22 10:16:03 2008 A3: 0
Fri Feb 22 10:16:03 2008 A4: 0
Fri Feb 22 10:16:03 2008 A5: 0
Fri Feb 22 10:16:03 2008 A6: 2
Fri Feb 22 10:16:03 2008 size score = 4.189742e-12, Murphy alpha = 2.413277, combined = 2.102509e-12
Fri Feb 22 10:16:03 2008
Fri Feb 22 10:16:03 2008 commencing linear algebra
Fri Feb 22 10:16:12 2008 read 13558374 cycles
Fri Feb 22 10:17:01 2008 matrix is 13558175 x 13558374 (3832.4 MB) with weight 1195000318 (88.14/col)
Fri Feb 22 10:17:01 2008 sparse part has weight 855491974 (63.10/col)
Fri Feb 22 10:17:02 2008 saving the first 48 matrix rows for later
Fri Feb 22 10:17:08 2008 matrix is 13558127 x 13558374 (3675.6 MB) with weight 906838153 (66.88/col)
Fri Feb 22 10:17:08 2008 sparse part has weight 827953701 (61.07/col)
Fri Feb 22 10:17:08 2008 matrix includes 64 packed rows
Fri Feb 22 10:17:08 2008 using block size 65536 for processor cache size 4096 kB
Fri Feb 22 10:18:53 2008 commencing Lanczos iteration (4 threads)
Fri Feb 22 10:18:53 2008 memory use: 3945.0 MB
Fri Feb 22 10:19:43 2008 restarting at iteration 26887 (dim = 1700224)
Thu Mar 13 21:27:55 2008 lanczos halted after 214402 iterations (dim = 13558116)
Thu Mar 13 21:28:43 2008 recovered 42 nontrivial dependencies
Thu Mar 13 21:28:47 2008 elapsed time 490:12:47
Thu Mar 13 21:41:39 2008
Thu Mar 13 21:41:39 2008
Thu Mar 13 21:41:39 2008 Msieve v. 1.33
Thu Mar 13 21:41:39 2008 random seeds: 5a79d3c3 d4b0c533
Thu Mar 13 21:41:39 2008 factoring 27312187167948796049898887842782596405655381943670578373120353781997447000328700858938482409418918448758778194857586271196003004729048486053340769161697167767956404785889687728602979284674332324623675660860406658739006337124213538943374972682241 (245 digits)
Thu Mar 13 21:41:42 2008 no P-1/P+1/ECM available, skipping
Thu Mar 13 21:41:42 2008 commencing number field sieve (245-digit input)
Thu Mar 13 21:41:42 2008 R0: -1393796574908163946345982392040522594123776
Thu Mar 13 21:41:42 2008 R1: 1
Thu Mar 13 21:41:42 2008 A0: -1
Thu Mar 13 21:41:42 2008 A1: 0
Thu Mar 13 21:41:42 2008 A2: 0
Thu Mar 13 21:41:42 2008 A3: 0
Thu Mar 13 21:41:42 2008 A4: 0
Thu Mar 13 21:41:42 2008 A5: 0
Thu Mar 13 21:41:42 2008 A6: 2
Thu Mar 13 21:41:42 2008 size score = 4.189742e-12, Murphy alpha = 2.413277, combined = 2.102509e-12
Thu Mar 13 21:41:42 2008
Thu Mar 13 21:41:42 2008 commencing square root phase
Thu Mar 13 21:41:42 2008 reading relations for dependency 1
Thu Mar 13 21:41:45 2008 read 6778074 cycles
Thu Mar 13 21:42:34 2008 cycles contain 21733920 unique relations
Thu Mar 13 21:58:59 2008 read 21733920 relations
Thu Mar 13 22:04:39 2008 multiplying 30208440 relations
Thu Mar 13 23:40:41 2008 multiply complete, coefficients have about 808.62 million bits
Thu Mar 13 23:40:54 2008 error: relation product is incorrect; if this is a debug binary, make sure that NDEBUG is not defined
Thu Mar 13 23:40:54 2008 algebraic square root failed
Thu Mar 13 23:40:54 2008 reading relations for dependency 2
Thu Mar 13 23:41:07 2008 read 6778294 cycles
Thu Mar 13 23:41:57 2008 cycles contain 21741379 unique relations
Thu Mar 13 23:56:52 2008 read 21741379 relations
Fri Mar 14 00:02:32 2008 multiplying 30220600 relations
Fri Mar 14 01:38:36 2008 multiply complete, coefficients have about 808.95 million bits
Fri Mar 14 01:38:48 2008 error: relation product is incorrect; if this is a debug binary, make sure that NDEBUG is not defined
Fri Mar 14 01:38:48 2008 algebraic square root failed
Fri Mar 14 01:38:48 2008 reading relations for dependency 3
Fri Mar 14 01:39:02 2008 read 6778113 cycles
Fri Mar 14 01:39:52 2008 cycles contain 21738263 unique relations
Fri Mar 14 01:55:02 2008 read 21738263 relations
Fri Mar 14 02:00:42 2008 multiplying 30216356 relations
Fri Mar 14 03:36:50 2008 multiply complete, coefficients have about 808.83 million bits
Fri Mar 14 03:37:02 2008 error: relation product is incorrect; if this is a debug binary, make sure that NDEBUG is not defined
Fri Mar 14 03:37:02 2008 algebraic square root failed
Fri Mar 14 03:37:02 2008 reading relations for dependency 4
Fri Mar 14 03:37:16 2008 read 6778859 cycles
Fri Mar 14 03:38:05 2008 cycles contain 21736172 unique relations
Fri Mar 14 03:53:30 2008 read 21736172 relations
Fri Mar 14 03:59:11 2008 multiplying 30215154 relations
Fri Mar 14 05:35:14 2008 multiply complete, coefficients have about 808.79 million bits
Fri Mar 14 05:35:27 2008 error: relation product is incorrect; if this is a debug binary, make sure that NDEBUG is not defined
Fri Mar 14 05:35:27 2008 algebraic square root failed

[miklin]

At me too such was with version 1.33. Use version 1.32.

[Andi47]

Just for the records: successful 131 digits postprocessing (was c131 of HP782(81)) with 1.33

Tue Mar 18 06:42:59 2008  
Tue Mar 18 06:42:59 2008  
Tue Mar 18 06:42:59 2008  Msieve v. 1.33
Tue Mar 18 06:42:59 2008  random seeds: 3d085994 57fd92d4
Tue Mar 18 06:42:59 2008  factoring 43222886169321329059036880032650709108733196405535056581042817111139574013523184223970207895448911636597349075156599534230820924243 (131 digits)
Tue Mar 18 06:43:00 2008  searching for 15-digit factors
Tue Mar 18 06:43:02 2008  commencing number field sieve (131-digit input)
Tue Mar 18 06:43:02 2008  R0: -19064967440603921797335263
Tue Mar 18 06:43:02 2008  R1:  63486192432329
Tue Mar 18 06:43:02 2008  A0:  86517693452891581716471954832800
Tue Mar 18 06:43:02 2008  A1:  826599889106007076977737118
Tue Mar 18 06:43:02 2008  A2: -10804596144270735568004
Tue Mar 18 06:43:02 2008  A3: -11580666747757990
Tue Mar 18 06:43:02 2008  A4:  191514913087
Tue Mar 18 06:43:02 2008  A5:  17160
Tue Mar 18 06:43:02 2008  size score = 1.095187e-013, Murphy alpha = -6.730026, combined = 1.032189e-012
Tue Mar 18 06:43:14 2008  restarting with 13505843 relations
Tue Mar 18 06:43:18 2008  added 3307 free relations
Tue Mar 18 06:43:18 2008  
Tue Mar 18 06:43:18 2008  commencing relation filtering
Tue Mar 18 06:43:18 2008  commencing duplicate removal, pass 1
Tue Mar 18 06:46:13 2008  found 1911814 hash collisions in 13509148 relations
Tue Mar 18 06:46:13 2008  commencing duplicate removal, pass 2
Tue Mar 18 06:46:57 2008  found 1818184 duplicates and 11690964 unique relations
Tue Mar 18 06:46:57 2008  memory use: 65.3 MB
Tue Mar 18 06:47:01 2008  ignoring smallest 367531 rational and 368278 algebraic ideals
Tue Mar 18 06:47:01 2008  filtering rational ideals above 5294233
Tue Mar 18 06:47:01 2008  filtering algebraic ideals above 5294233
Tue Mar 18 06:47:01 2008  need 1250875 more relations than ideals
Tue Mar 18 06:47:01 2008  commencing singleton removal, pass 1
Tue Mar 18 06:49:46 2008  relations with 0 large ideals: 3037
Tue Mar 18 06:49:46 2008  relations with 1 large ideals: 178629
Tue Mar 18 06:49:46 2008  relations with 2 large ideals: 1323773
Tue Mar 18 06:49:46 2008  relations with 3 large ideals: 3625483
Tue Mar 18 06:49:46 2008  relations with 4 large ideals: 4363158
Tue Mar 18 06:49:46 2008  relations with 5 large ideals: 2035624
Tue Mar 18 06:49:46 2008  relations with 6 large ideals: 156156
Tue Mar 18 06:49:46 2008  relations with 7+ large ideals: 5104
Tue Mar 18 06:49:46 2008  11690964 relations and about 9685407 large ideals
Tue Mar 18 06:49:46 2008  commencing singleton removal, pass 2
Tue Mar 18 06:52:32 2008  found 3649185 singletons
Tue Mar 18 06:52:32 2008  current dataset: 8041779 relations and about 5662564 large ideals
Tue Mar 18 06:52:32 2008  commencing singleton removal, pass 3
Tue Mar 18 06:54:54 2008  found 542090 singletons
Tue Mar 18 06:54:54 2008  current dataset: 7499689 relations and about 5107285 large ideals
Tue Mar 18 06:54:54 2008  commencing singleton removal, pass 4
Tue Mar 18 06:57:11 2008  found 92584 singletons
Tue Mar 18 06:57:11 2008  current dataset: 7407105 relations and about 5014234 large ideals
Tue Mar 18 06:57:11 2008  commencing singleton removal, final pass
Tue Mar 18 06:59:35 2008  memory use: 155.1 MB
Tue Mar 18 06:59:35 2008  commencing in-memory singleton removal
Tue Mar 18 06:59:36 2008  begin with 7407105 relations and 5310239 unique ideals
Tue Mar 18 06:59:44 2008  reduce to 6995326 relations and 4894051 ideals in 9 passes
Tue Mar 18 06:59:44 2008  max relations containing the same ideal: 56
Tue Mar 18 06:59:50 2008  removing 1453446 relations and 1053446 ideals in 400000 cliques
Tue Mar 18 06:59:51 2008  commencing in-memory singleton removal
Tue Mar 18 06:59:52 2008  begin with 5541880 relations and 4894051 unique ideals
Tue Mar 18 06:59:55 2008  reduce to 5407136 relations and 3698117 ideals in 6 passes
Tue Mar 18 06:59:55 2008  max relations containing the same ideal: 47
Tue Mar 18 06:59:59 2008  removing 1130803 relations and 730803 ideals in 400000 cliques
Tue Mar 18 07:00:00 2008  commencing in-memory singleton removal
Tue Mar 18 07:00:00 2008  begin with 4276333 relations and 3698117 unique ideals
Tue Mar 18 07:00:03 2008  reduce to 4148372 relations and 2830542 ideals in 7 passes
Tue Mar 18 07:00:03 2008  max relations containing the same ideal: 37
Tue Mar 18 07:00:06 2008  removing 254415 relations and 187461 ideals in 66954 cliques
Tue Mar 18 07:00:07 2008  commencing in-memory singleton removal
Tue Mar 18 07:00:07 2008  begin with 3893957 relations and 2830542 unique ideals
Tue Mar 18 07:00:09 2008  reduce to 3887204 relations and 2636221 ideals in 4 passes
Tue Mar 18 07:00:09 2008  max relations containing the same ideal: 36
Tue Mar 18 07:00:12 2008  dataset has 70.0% excess relations
Tue Mar 18 07:00:15 2008  ignoring smallest 333170 rational and 333987 algebraic ideals
Tue Mar 18 07:00:15 2008  filtering rational ideals above 4764809
Tue Mar 18 07:00:15 2008  filtering algebraic ideals above 4764809
Tue Mar 18 07:00:15 2008  need 1125884 more relations than ideals
Tue Mar 18 07:00:15 2008  commencing singleton removal, final pass
Tue Mar 18 07:01:35 2008  memory use: 78.6 MB
Tue Mar 18 07:01:35 2008  commencing in-memory singleton removal
Tue Mar 18 07:01:35 2008  begin with 3887204 relations and 2704807 unique ideals
Tue Mar 18 07:01:38 2008  reduce to 3886580 relations and 2704183 ideals in 6 passes
Tue Mar 18 07:01:38 2008  max relations containing the same ideal: 36
Tue Mar 18 07:01:40 2008  removing 102225 relations and 73969 ideals in 28256 cliques
Tue Mar 18 07:01:40 2008  commencing in-memory singleton removal
Tue Mar 18 07:01:41 2008  begin with 3784355 relations and 2704183 unique ideals
Tue Mar 18 07:01:42 2008  reduce to 3782930 relations and 2628779 ideals in 4 passes
Tue Mar 18 07:01:42 2008  max relations containing the same ideal: 36
Tue Mar 18 07:01:44 2008  removing 92207 relations and 63951 ideals in 28256 cliques
Tue Mar 18 07:01:44 2008  commencing in-memory singleton removal
Tue Mar 18 07:01:44 2008  begin with 3690723 relations and 2628779 unique ideals
Tue Mar 18 07:01:46 2008  reduce to 3689335 relations and 2563430 ideals in 4 passes
Tue Mar 18 07:01:46 2008  max relations containing the same ideal: 36
Tue Mar 18 07:01:47 2008  dataset has 53.0% excess relations
Tue Mar 18 07:01:49 2008  ignoring smallest 298637 rational and 299396 algebraic ideals
Tue Mar 18 07:01:49 2008  filtering rational ideals above 4235386
Tue Mar 18 07:01:49 2008  filtering algebraic ideals above 4235386
Tue Mar 18 07:01:49 2008  need 900724 more relations than ideals
Tue Mar 18 07:01:49 2008  commencing singleton removal, final pass
Tue Mar 18 07:02:53 2008  memory use: 76.6 MB
Tue Mar 18 07:02:53 2008  commencing in-memory singleton removal
Tue Mar 18 07:02:54 2008  begin with 3689335 relations and 2632514 unique ideals
Tue Mar 18 07:02:55 2008  reduce to 3688830 relations and 2632009 ideals in 4 passes
Tue Mar 18 07:02:55 2008  max relations containing the same ideal: 36
Tue Mar 18 07:02:58 2008  removing 230657 relations and 152609 ideals in 78048 cliques
Tue Mar 18 07:02:58 2008  commencing in-memory singleton removal
Tue Mar 18 07:02:58 2008  begin with 3458173 relations and 2632009 unique ideals
Tue Mar 18 07:03:00 2008  reduce to 3450957 relations and 2472060 ideals in 5 passes
Tue Mar 18 07:03:00 2008  max relations containing the same ideal: 33
Tue Mar 18 07:03:02 2008  removing 228818 relations and 150770 ideals in 78048 cliques
Tue Mar 18 07:03:02 2008  commencing in-memory singleton removal
Tue Mar 18 07:03:02 2008  begin with 3222139 relations and 2472060 unique ideals
Tue Mar 18 07:03:04 2008  reduce to 3212412 relations and 2311369 ideals in 5 passes
Tue Mar 18 07:03:04 2008  max relations containing the same ideal: 32
Tue Mar 18 07:03:06 2008  dataset has 22.5% excess relations
Tue Mar 18 07:03:09 2008  ignoring smallest 263796 rational and 264642 algebraic ideals
Tue Mar 18 07:03:09 2008  filtering rational ideals above 3705963
Tue Mar 18 07:03:09 2008  filtering algebraic ideals above 3705963
Tue Mar 18 07:03:09 2008  need 794673 more relations than ideals
Tue Mar 18 07:03:09 2008  commencing singleton removal, final pass
Tue Mar 18 07:04:05 2008  memory use: 68.1 MB
Tue Mar 18 07:04:05 2008  commencing in-memory singleton removal
Tue Mar 18 07:04:05 2008  begin with 3212412 relations and 2380911 unique ideals
Tue Mar 18 07:04:07 2008  reduce to 3211899 relations and 2380398 ideals in 4 passes
Tue Mar 18 07:04:07 2008  max relations containing the same ideal: 32
Tue Mar 18 07:04:09 2008  removing 68550 relations and 50136 ideals in 18414 cliques
Tue Mar 18 07:04:09 2008  commencing in-memory singleton removal
Tue Mar 18 07:04:09 2008  begin with 3143349 relations and 2380398 unique ideals
Tue Mar 18 07:04:11 2008  reduce to 3142656 relations and 2329566 ideals in 4 passes
Tue Mar 18 07:04:11 2008  max relations containing the same ideal: 31
Tue Mar 18 07:04:12 2008  removing 59715 relations and 41301 ideals in 18414 cliques
Tue Mar 18 07:04:12 2008  commencing in-memory singleton removal
Tue Mar 18 07:04:13 2008  begin with 3082941 relations and 2329566 unique ideals
Tue Mar 18 07:04:14 2008  reduce to 3082380 relations and 2287702 ideals in 4 passes
Tue Mar 18 07:04:14 2008  max relations containing the same ideal: 31
Tue Mar 18 07:04:15 2008  dataset has 8.0% excess relations
Tue Mar 18 07:04:15 2008  relations with 0 large ideals: 2167
Tue Mar 18 07:04:15 2008  relations with 1 large ideals: 85369
Tue Mar 18 07:04:15 2008  relations with 2 large ideals: 472717
Tue Mar 18 07:04:15 2008  relations with 3 large ideals: 978271
Tue Mar 18 07:04:15 2008  relations with 4 large ideals: 947586
Tue Mar 18 07:04:15 2008  relations with 5 large ideals: 462670
Tue Mar 18 07:04:15 2008  relations with 6 large ideals: 116505
Tue Mar 18 07:04:15 2008  relations with 7+ large ideals: 17095
Tue Mar 18 07:04:15 2008  commencing 2-way merge
Tue Mar 18 07:04:18 2008  reduce to 2301851 relation sets and 1507173 unique ideals
Tue Mar 18 07:04:18 2008  commencing full merge
Tue Mar 18 07:04:55 2008  memory use: 137.8 MB
Tue Mar 18 07:04:55 2008  found 1201254 cycles, need 935373
Tue Mar 18 07:04:56 2008  weight of 935373 cycles is about 60842890 (65.05/cycle)
Tue Mar 18 07:04:56 2008  distribution of cycle lengths:
Tue Mar 18 07:04:56 2008  1 relations: 71062
Tue Mar 18 07:04:56 2008  2 relations: 106231
Tue Mar 18 07:04:56 2008  3 relations: 123181
Tue Mar 18 07:04:56 2008  4 relations: 126953
Tue Mar 18 07:04:56 2008  5 relations: 125832
Tue Mar 18 07:04:56 2008  6 relations: 117267
Tue Mar 18 07:04:56 2008  7 relations: 104114
Tue Mar 18 07:04:56 2008  8 relations: 87702
Tue Mar 18 07:04:56 2008  9 relations: 63710
Tue Mar 18 07:04:56 2008  10+ relations: 9321
Tue Mar 18 07:04:56 2008  heaviest cycle: 12 relations
Tue Mar 18 07:04:56 2008  commencing cycle optimization
Tue Mar 18 07:04:57 2008  start with 4590689 relations
Tue Mar 18 07:05:08 2008  pruned 87279 relations
Tue Mar 18 07:05:08 2008  memory use: 120.6 MB
Tue Mar 18 07:05:08 2008  distribution of cycle lengths:
Tue Mar 18 07:05:08 2008  1 relations: 71062
Tue Mar 18 07:05:08 2008  2 relations: 107888
Tue Mar 18 07:05:08 2008  3 relations: 127250
Tue Mar 18 07:05:08 2008  4 relations: 130820
Tue Mar 18 07:05:08 2008  5 relations: 130854
Tue Mar 18 07:05:08 2008  6 relations: 120941
Tue Mar 18 07:05:08 2008  7 relations: 105640
Tue Mar 18 07:05:08 2008  8 relations: 82824
Tue Mar 18 07:05:08 2008  9 relations: 51411
Tue Mar 18 07:05:08 2008  10+ relations: 6683
Tue Mar 18 07:05:08 2008  heaviest cycle: 12 relations
Tue Mar 18 07:05:10 2008  
Tue Mar 18 07:05:10 2008  commencing linear algebra
Tue Mar 18 07:05:11 2008  read 935373 cycles
Tue Mar 18 07:05:13 2008  cycles contain 2394472 unique relations
Tue Mar 18 07:06:21 2008  read 2394472 relations
Tue Mar 18 07:06:25 2008  using 32 quadratic characters above 134215002
Tue Mar 18 07:06:52 2008  building initial matrix
Tue Mar 18 07:07:31 2008  memory use: 300.9 MB
Tue Mar 18 07:07:33 2008  read 935373 cycles
Tue Mar 18 07:07:57 2008  matrix is 934883 x 935373 (251.4 MB) with weight 90139280 (96.37/col)
Tue Mar 18 07:07:57 2008  sparse part has weight 58407509 (62.44/col)
Tue Mar 18 07:08:11 2008  filtering completed in 3 passes
Tue Mar 18 07:08:11 2008  matrix is 929498 x 929698 (250.4 MB) with weight 89759497 (96.55/col)
Tue Mar 18 07:08:11 2008  sparse part has weight 58213982 (62.62/col)
Tue Mar 18 07:08:28 2008  read 929698 cycles
Tue Mar 18 07:09:51 2008  matrix is 929498 x 929698 (250.4 MB) with weight 89759497 (96.55/col)
Tue Mar 18 07:09:54 2008  sparse part has weight 58213982 (62.62/col)
Tue Mar 18 07:09:54 2008  saving the first 48 matrix rows for later
Tue Mar 18 07:09:55 2008  matrix is 929450 x 929698 (242.0 MB) with weight 69029264 (74.25/col)
Tue Mar 18 07:09:55 2008  sparse part has weight 57857561 (62.23/col)
Tue Mar 18 07:09:55 2008  matrix includes 64 packed rows
Tue Mar 18 07:09:55 2008  using block size 65536 for processor cache size 4096 kB
Tue Mar 18 07:10:04 2008  commencing Lanczos iteration (2 threads)
Tue Mar 18 07:10:04 2008  memory use: 247.7 MB
Tue Mar 18 09:00:48 2008  lanczos halted after 14697 iterations (dim = 929448)
Tue Mar 18 09:00:50 2008  recovered 43 nontrivial dependencies
Tue Mar 18 09:00:51 2008  
Tue Mar 18 09:00:51 2008  commencing square root phase
Tue Mar 18 09:00:51 2008  reading relations for dependency 1
Tue Mar 18 09:00:52 2008  read 464618 cycles
Tue Mar 18 09:00:53 2008  cycles contain 1464214 unique relations
Tue Mar 18 09:02:04 2008  read 1464214 relations
Tue Mar 18 09:02:16 2008  multiplying 2245242 relations
Tue Mar 18 09:10:54 2008  multiply complete, coefficients have about 113.04 million bits
Tue Mar 18 09:11:00 2008  initial square root is modulo 130015399
Tue Mar 18 09:25:03 2008  prp56 factor: 32954971892100042977799569415093678837254428701547341173
Tue Mar 18 09:25:03 2008  prp76 factor: 1311574056589704078537800777069018790782839890278845124803308140882436967591
Tue Mar 18 09:25:03 2008  elapsed time 02:42:04

Just a guess: Could it be possible, that with huge postprocessing jobs (larger than mine) a variable is overflowing and thus causes these jobs to fail?

[miklin]

Originally Posted by Andi47

Just for the records: successful 131 digits postprocessing (was c131 of HP782(81)) with 1.33
Just a guess: Could it be possible, that with huge postprocessing jobs (larger than mine) a variable is overflowing and thus causes these jobs to fail?

It is necessary to read through attentively conference simply.
Since # 409, and further to take version 1.32 and to start SQRT.
I already the fourth key rsa154 consider one and too occurs.
It is assured that all will turn out.
Success.

[jasonp]

Originally Posted by miklin

Since # 409, and further to take version 1.32 and to start SQRT.
I already the fourth key rsa154 consider one and too occurs.

Using the older version has not helped Richard, and at this point I am assuming the job is so large that the FFTs fail due to accumulated roundoff error. If that is true, the only solution is to use all-integer transforms.

[Andi47]

Originally Posted by jasonp

Using the older version has not helped Richard, and at this point I am assuming the job is so large that the FFTs fail due to accumulated roundoff error. If that is true, the only solution is to use all-integer transforms.

What about using a larger FFT size for larger jobs (like Prime95 does)?
And a roundoff error check (in debug mode, and also prior to doing the "real" calculations to decide whether to switch to a larger FFT size) could give you an answer to the question if the error is accumulated roundoff error or something else - if roundoff errors tend to be not far below 0.5 it is likely, that some of the roundoff errors are in fact somewhat above 0.5.

[miklin]

Originally Posted by jasonp

Using the older version has not helped Richard, and at this point I am assuming the job is so large that the FFTs fail due to accumulated roundoff error. If that is true, the only solution is to use all-integer transforms.

Yes you are possible and the rights. To mine most easier to try hours 10 will borrow. As up to sqrt I use 1.33 version, and sqrt I do by version 1.32. You when have suggested it to try me I and itself were surprised. As it is new version on old to change. And all has earned. Why I do not know. Yes all the same the program fine works. The good fellow.

[jasonp]

Originally Posted by Andi47

What about using a larger FFT size for larger jobs (like Prime95 does)?

The FFTs here are 30x larger than the ones used in Prime95, and previous experience shows that failure due to roundoff error is pretty much inevitable for convolution products above 512 megabits in size. We've tried switching to fewer bits per convolution element and it doesn't help. I'll do something about it when time allows.

[Prime95]

Originally Posted by jasonp

The FFTs here are 30x larger than the ones used in Prime95, and previous experience shows that failure due to roundoff error is pretty much inevitable for convolution products above 512 megabits in size.

Prime95 can handle 596 million bit numbers using an FFT size of 32M (about 17.75 bits per double).