mersenneforum.org  

Go Back   mersenneforum.org > Factoring Projects > Factoring

Reply
 
Thread Tools
Old 2018-06-25, 14:14   #1
VictordeHolland
 
VictordeHolland's Avatar
 
"Victor de Hollander"
Aug 2011
the Netherlands

2·587 Posts
Cool Factoring on a ARM board (Odroid-U2) with Cortex A9

These little ARM boards are a lot of fun .
Hint: They're not particularly fast

GMP-ECM 7.0.5-dev
http://mersenneforum.org/showthread.php?t=23070

CADO-NFS 3.0-dev (factoring RSA-100 in about 3.5h)
http://mersenneforum.org/showpost.ph...&postcount=129

and now also Msieve (SVN1022)
Quadratic sieve:
Code:
odroid@odroid-u2:~/Documents/msieve-SVN1022$ ./msieve -v 482023953171263179598463399469620779294192640218599890595570781655655552082529

Msieve v. 1.54 (SVN Unversioned directory)
Mon Jun 25 14:41:10 2018
random seeds: 93cbf29d 83c8d2b2
factoring 482023953171263179598463399469620779294192640218599890595570781655655552082529 (78 digits)
searching for 15-digit factors
commencing quadratic sieve (78-digit input)
using multiplier of 3
using generic 32kb sieve core
sieve interval: 12 blocks of size 32768
processing polynomials in batches of 17
using a sieve bound of 952381 (37824 primes)
using large prime bound of 95238100 (26 bits)
using trial factoring cutoff of 27 bits
polynomial 'A' values have 10 factors

sieving in progress (press Ctrl-C to pause)
38137 relations (19841 full + 18296 combined from 204445 partial), need 37920
38137 relations (19841 full + 18296 combined from 204445 partial), need 37920
sieving complete, commencing postprocessing
begin with 224286 relations
reduce to 54111 relations in 2 passes
attempting to read 54111 relations
recovered 54111 relations
recovered 42193 polynomials
attempting to build 38137 cycles
found 38137 cycles in 1 passes
distribution of cycle lengths:
   length 1 : 19841
   length 2 : 18296
largest cycle: 2 relations
matrix is 37824 x 38137 (5.0 MB) with weight 1145593 (30.04/col)
sparse part has weight 1145593 (30.04/col)
filtering completed in 4 passes
matrix is 26555 x 26619 (3.8 MB) with weight 893449 (33.56/col)
sparse part has weight 893449 (33.56/col)
saving the first 48 matrix rows for later
matrix includes 64 packed rows
matrix is 26507 x 26619 (2.4 MB) with weight 655532 (24.63/col)
sparse part has weight 470989 (17.69/col)
commencing Lanczos iteration
memory use: 2.4 MB
lanczos halted after 421 iterations (dim = 26505)
recovered 16 nontrivial dependencies
p1 factor: 3
p33 factor: 300704334686783305051368312538631
p45 factor: 534327685114509832669825288562536388785831453
 elapsed time 00:13:22
Next I did polysearch+sieve of RSA100 on a computer (x86_64) and copied the relations and poly to the ARM board and did the filtering/LA and SQR with msieve there:
Code:
Sun Jun 24 13:42:13 2018  
Sun Jun 24 13:42:13 2018  
Sun Jun 24 13:42:13 2018  Msieve v. 1.54 (SVN Unversioned directory)
Sun Jun 24 13:42:13 2018  random seeds: 3202ded0 6ba54a8e
Sun Jun 24 13:42:13 2018  factoring 1522605027922533360535618378132637429718068114961380688657908494580122963258952897654000350692006139 (100 digits)
Sun Jun 24 13:42:16 2018  searching for 15-digit factors
Sun Jun 24 13:42:18 2018  commencing number field sieve (100-digit input)
Sun Jun 24 13:42:18 2018  R0: -1500948764646399765833785
Sun Jun 24 13:42:18 2018  R1: 4412959449373
Sun Jun 24 13:42:18 2018  A0: 58495626606615351512386040384
Sun Jun 24 13:42:18 2018  A1: -5772374246691267911604
Sun Jun 24 13:42:18 2018  A2: -8833629323103645
Sun Jun 24 13:42:18 2018  A3: 595383944
Sun Jun 24 13:42:18 2018  A4: 300
Sun Jun 24 13:42:18 2018  skew 4593762.87, size 1.165e-13, alpha -4.779, combined = 1.277e-08 rroots = 2
Sun Jun 24 13:42:18 2018  
Sun Jun 24 13:42:18 2018  commencing relation filtering
Sun Jun 24 13:42:18 2018  estimated available RAM is 2023.1 MB
Sun Jun 24 13:42:18 2018  commencing duplicate removal, pass 1
Sun Jun 24 13:47:23 2018  found 364484 hash collisions in 4508009 relations
Sun Jun 24 13:48:46 2018  commencing duplicate removal, pass 2
Sun Jun 24 13:48:50 2018  found 255001 duplicates and 4253008 unique relations
Sun Jun 24 13:48:50 2018  memory use: 16.3 MB
Sun Jun 24 13:48:50 2018  reading ideals above 100000
Sun Jun 24 13:48:50 2018  commencing singleton removal, initial pass
Sun Jun 24 13:51:25 2018  memory use: 94.1 MB
Sun Jun 24 13:51:25 2018  reading all ideals from disk
Sun Jun 24 13:51:25 2018  memory use: 130.9 MB
Sun Jun 24 13:51:27 2018  keeping 4582269 ideals with weight <= 200, target excess is 21260
Sun Jun 24 13:51:30 2018  commencing in-memory singleton removal
Sun Jun 24 13:51:31 2018  begin with 4253008 relations and 4582269 unique ideals
Sun Jun 24 13:51:44 2018  reduce to 1675157 relations and 1462630 ideals in 13 passes
Sun Jun 24 13:51:44 2018  max relations containing the same ideal: 106
Sun Jun 24 13:51:47 2018  removing 422012 relations and 328079 ideals in 93933 cliques
Sun Jun 24 13:51:47 2018  commencing in-memory singleton removal
Sun Jun 24 13:51:48 2018  begin with 1253145 relations and 1462630 unique ideals
Sun Jun 24 13:51:53 2018  reduce to 1170153 relations and 1045619 ideals in 9 passes
Sun Jun 24 13:51:53 2018  max relations containing the same ideal: 83
Sun Jun 24 13:51:55 2018  removing 338525 relations and 244592 ideals in 93933 cliques
Sun Jun 24 13:51:55 2018  commencing in-memory singleton removal
Sun Jun 24 13:51:55 2018  begin with 831628 relations and 1045619 unique ideals
Sun Jun 24 13:51:58 2018  reduce to 756925 relations and 720072 ideals in 9 passes
Sun Jun 24 13:51:58 2018  max relations containing the same ideal: 61
Sun Jun 24 13:52:00 2018  removing 71580 relations and 59389 ideals in 12191 cliques
Sun Jun 24 13:52:00 2018  commencing in-memory singleton removal
Sun Jun 24 13:52:00 2018  begin with 685345 relations and 720072 unique ideals
Sun Jun 24 13:52:02 2018  reduce to 680615 relations and 655865 ideals in 7 passes
Sun Jun 24 13:52:02 2018  max relations containing the same ideal: 57
Sun Jun 24 13:52:02 2018  relations with 0 large ideals: 423
Sun Jun 24 13:52:02 2018  relations with 1 large ideals: 268
Sun Jun 24 13:52:02 2018  relations with 2 large ideals: 3711
Sun Jun 24 13:52:02 2018  relations with 3 large ideals: 25970
Sun Jun 24 13:52:02 2018  relations with 4 large ideals: 92954
Sun Jun 24 13:52:02 2018  relations with 5 large ideals: 186499
Sun Jun 24 13:52:02 2018  relations with 6 large ideals: 198741
Sun Jun 24 13:52:02 2018  relations with 7+ large ideals: 172049
Sun Jun 24 13:52:02 2018  commencing 2-way merge
Sun Jun 24 13:52:04 2018  reduce to 399836 relation sets and 375086 unique ideals
Sun Jun 24 13:52:04 2018  commencing full merge
Sun Jun 24 13:52:21 2018  memory use: 39.2 MB
Sun Jun 24 13:52:21 2018  found 194290 cycles, need 191286
Sun Jun 24 13:52:22 2018  weight of 191286 cycles is about 13654981 (71.39/cycle)
Sun Jun 24 13:52:22 2018  distribution of cycle lengths:
Sun Jun 24 13:52:22 2018  1 relations: 16617
Sun Jun 24 13:52:22 2018  2 relations: 17228
Sun Jun 24 13:52:22 2018  3 relations: 17888
Sun Jun 24 13:52:22 2018  4 relations: 17484
Sun Jun 24 13:52:22 2018  5 relations: 16784
Sun Jun 24 13:52:22 2018  6 relations: 15926
Sun Jun 24 13:52:22 2018  7 relations: 14859
Sun Jun 24 13:52:22 2018  8 relations: 13350
Sun Jun 24 13:52:22 2018  9 relations: 11803
Sun Jun 24 13:52:22 2018  10+ relations: 49347
Sun Jun 24 13:52:22 2018  heaviest cycle: 23 relations
Sun Jun 24 13:52:22 2018  commencing cycle optimization
Sun Jun 24 13:52:23 2018  start with 1310580 relations
Sun Jun 24 13:52:28 2018  pruned 36004 relations
Sun Jun 24 13:52:28 2018  memory use: 32.9 MB
Sun Jun 24 13:52:28 2018  distribution of cycle lengths:
Sun Jun 24 13:52:28 2018  1 relations: 16617
Sun Jun 24 13:52:28 2018  2 relations: 17555
Sun Jun 24 13:52:28 2018  3 relations: 18514
Sun Jun 24 13:52:28 2018  4 relations: 17903
Sun Jun 24 13:52:28 2018  5 relations: 17390
Sun Jun 24 13:52:28 2018  6 relations: 16354
Sun Jun 24 13:52:28 2018  7 relations: 15270
Sun Jun 24 13:52:28 2018  8 relations: 13614
Sun Jun 24 13:52:28 2018  9 relations: 11896
Sun Jun 24 13:52:28 2018  10+ relations: 46173
Sun Jun 24 13:52:28 2018  heaviest cycle: 23 relations
Sun Jun 24 13:52:29 2018  RelProcTime: 611
Sun Jun 24 13:52:29 2018  
Sun Jun 24 13:52:29 2018  commencing linear algebra
Sun Jun 24 13:52:29 2018  read 191286 cycles
Sun Jun 24 13:52:30 2018  cycles contain 662298 unique relations
Sun Jun 24 13:52:48 2018  read 662298 relations
Sun Jun 24 13:52:50 2018  using 20 quadratic characters above 4294917295
Sun Jun 24 13:53:00 2018  building initial matrix
Sun Jun 24 13:53:30 2018  memory use: 74.0 MB
Sun Jun 24 13:53:31 2018  read 191286 cycles
Sun Jun 24 13:53:31 2018  matrix is 191105 x 191286 (54.7 MB) with weight 18411203 (96.25/col)
Sun Jun 24 13:53:31 2018  sparse part has weight 12990646 (67.91/col)
Sun Jun 24 13:53:40 2018  filtering completed in 2 passes
Sun Jun 24 13:53:40 2018  matrix is 191010 x 191191 (54.7 MB) with weight 18406796 (96.27/col)
Sun Jun 24 13:53:40 2018  sparse part has weight 12989219 (67.94/col)
Sun Jun 24 13:53:46 2018  matrix starts at (0, 0)
Sun Jun 24 13:53:46 2018  matrix is 191010 x 191191 (54.7 MB) with weight 18406796 (96.27/col)
Sun Jun 24 13:53:46 2018  sparse part has weight 12989219 (67.94/col)
Sun Jun 24 13:53:46 2018  saving the first 48 matrix rows for later
Sun Jun 24 13:53:46 2018  matrix includes 64 packed rows
Sun Jun 24 13:53:46 2018  matrix is 190962 x 191191 (52.4 MB) with weight 14528553 (75.99/col)
Sun Jun 24 13:53:46 2018  sparse part has weight 12576621 (65.78/col)
Sun Jun 24 13:53:46 2018  using block size 8192 and superblock size 49152 for processor cache size 512 kB
Sun Jun 24 13:53:48 2018  commencing Lanczos iteration
Sun Jun 24 13:53:48 2018  memory use: 41.9 MB
Sun Jun 24 13:54:57 2018  linear algebra at 6.3%, ETA 0h16m
Sun Jun 24 14:11:54 2018  lanczos halted after 3021 iterations (dim = 190962)
Sun Jun 24 14:11:55 2018  recovered 34 nontrivial dependencies
Sun Jun 24 14:11:55 2018  BLanczosTime: 1166
Sun Jun 24 14:11:55 2018  
Sun Jun 24 14:11:55 2018  commencing square root phase
Sun Jun 24 14:11:55 2018  reading relations for dependency 1
Sun Jun 24 14:11:55 2018  read 95430 cycles
Sun Jun 24 14:11:56 2018  cycles contain 330468 unique relations
Sun Jun 24 14:12:06 2018  read 330468 relations
Sun Jun 24 14:12:11 2018  multiplying 330468 relations
Sun Jun 24 14:13:02 2018  multiply complete, coefficients have about 12.85 million bits
Sun Jun 24 14:13:03 2018  initial square root is modulo 24181847
Sun Jun 24 14:14:23 2018  GCD is N, no factor found
Sun Jun 24 14:14:23 2018  reading relations for dependency 2
Sun Jun 24 14:14:23 2018  read 95866 cycles
Sun Jun 24 14:14:23 2018  cycles contain 331562 unique relations
Sun Jun 24 14:14:34 2018  read 331562 relations
Sun Jun 24 14:14:38 2018  multiplying 331562 relations
Sun Jun 24 14:15:30 2018  multiply complete, coefficients have about 12.89 million bits
Sun Jun 24 14:15:31 2018  initial square root is modulo 25612739
Sun Jun 24 14:16:51 2018  GCD is 1, no factor found
Sun Jun 24 14:16:51 2018  reading relations for dependency 3
Sun Jun 24 14:16:51 2018  read 95879 cycles
Sun Jun 24 14:16:51 2018  cycles contain 330932 unique relations
Sun Jun 24 14:17:02 2018  read 330932 relations
Sun Jun 24 14:17:07 2018  multiplying 330932 relations
Sun Jun 24 14:17:58 2018  multiply complete, coefficients have about 12.87 million bits
Sun Jun 24 14:18:00 2018  initial square root is modulo 24799387
Sun Jun 24 14:19:20 2018  GCD is N, no factor found
Sun Jun 24 14:19:20 2018  reading relations for dependency 4
Sun Jun 24 14:19:20 2018  read 95523 cycles
Sun Jun 24 14:19:21 2018  cycles contain 331356 unique relations
Sun Jun 24 14:19:31 2018  read 331356 relations
Sun Jun 24 14:19:36 2018  multiplying 331356 relations
Sun Jun 24 14:20:27 2018  multiply complete, coefficients have about 12.88 million bits
Sun Jun 24 14:20:28 2018  initial square root is modulo 25307939
Sun Jun 24 14:21:47 2018  GCD is 1, no factor found
Sun Jun 24 14:21:47 2018  reading relations for dependency 5
Sun Jun 24 14:21:47 2018  read 95805 cycles
Sun Jun 24 14:21:48 2018  cycles contain 331526 unique relations
Sun Jun 24 14:21:58 2018  read 331526 relations
Sun Jun 24 14:22:03 2018  multiplying 331526 relations
Sun Jun 24 14:22:55 2018  multiply complete, coefficients have about 12.89 million bits
Sun Jun 24 14:22:56 2018  initial square root is modulo 25568393
Sun Jun 24 14:24:16 2018  sqrtTime: 741
Sun Jun 24 14:24:16 2018  p50 factor: 37975227936943673922808872755445627854565536638199
Sun Jun 24 14:24:16 2018  p50 factor: 40094690950920881030683735292761468389214899724061
 Sun Jun 24 14:24:16 2018  elapsed time 00:42:03

I could not find a pure-c version of the lasieve4 sievers (they all use ASM and/or intrinsics and requires cweb??), so If anybody knows where to get that please let me know (and some basic compile instructions if there is no makefile)!

So I went the Msieve --> las (CADO-NFS siever) --> Msieve route for a full NFS factorization

Poly search with Msieve -np 1,1000
Code:
Sun Jun 24 14:27:34 2018  
Sun Jun 24 14:27:34 2018  
Sun Jun 24 14:27:34 2018  Msieve v. 1.54 (SVN Unversioned directory)
Sun Jun 24 14:27:34 2018  random seeds: 4f90fbf4 d86a5791
Sun Jun 24 14:27:34 2018  factoring 1522605027922533360535618378132637429718068114961380688657908494580122963258952897654000350692006139 (100 digits)
Sun Jun 24 14:27:37 2018  searching for 15-digit factors
Sun Jun 24 14:27:39 2018  commencing number field sieve (100-digit input)
Sun Jun 24 14:27:39 2018  commencing number field sieve polynomial selection
Sun Jun 24 14:27:39 2018  polynomial degree: 4
Sun Jun 24 14:27:39 2018  max stage 1 norm: 1.36e+17
Sun Jun 24 14:27:39 2018  max stage 2 norm: 3.19e+15
Sun Jun 24 14:27:39 2018  min E-value: 9.14e-09
Sun Jun 24 14:27:39 2018  poly select deadline: 1286
Sun Jun 24 14:27:39 2018  time limit set to 0.36 CPU-hours
Sun Jun 24 14:27:39 2018  expecting poly E from 1.29e-08 to > 1.49e-08
Sun Jun 24 14:27:39 2018  searching leading coefficients from 1 to 1000
Sun Jun 24 14:35:08 2018  polynomial selection complete
Sun Jun 24 14:35:08 2018  R0: -1160318224499030718294235
Sun Jun 24 14:35:08 2018  R1: 378654317647
Sun Jun 24 14:35:08 2018  A0: -12084851459361393651319621216
Sun Jun 24 14:35:08 2018  A1: -1180605682597027204428
Sun Jun 24 14:35:08 2018  A2: -8614727289369279
Sun Jun 24 14:35:08 2018  A3: -2848466002
Sun Jun 24 14:35:08 2018  A4: 840
Sun Jun 24 14:35:08 2018  skew 3254900.80, size 9.828e-14, alpha -5.221, combined = 1.210e-08 rroots = 2
Sun Jun 24 14:35:08 2018  elapsed time 00:07:34

Which resulted in the msieve.poly (which I renamed to rsa100.poly)
Code:
n: 1522605027922533360535618378132637429718068114961380688657908494580122963258952897654000350692006139
skew: 3254900.80
c0 -12084851459361393651319621216
c1 -1180605682597027204428
c2 -8614727289369279
c3 -2848466002
c4 840
Y0: -1160318224499030718294235
Y1: 378654317647
Created a factorbase for the las siever with makefb (from the CADO-NFS suite), I had no idea what to put with the -maxbits (if anybody can explain if or how the -maxbits corresponds to the siever ( -I) or the lpb I'd appreciate that.

Code:
odroid@odroid-u2:~/Documents/makefb$ ./makefb -poly rsa100.poly -lim 650000 -maxbits 12 -out rsa100factorbase.fb
started las with (I looked at the CADO-NFS logs of RSA100 on how it start it):
Code:
./las -I 12 -poly rsa100.poly -q0 400000 -q1 500000 -lim0 650000 -lim1 650000 -lpb0 25 -lpb1 25 -mfb0 50 -mfb1 50 -ncurves0 11 -ncurves1 16 -fb rsa100factorbase.fb -out rsa100.q400k-500k.gz -stats-stderr
At first it looked like it failed, but somehow corrected (probably because of my stupid parameters):

Code:
madvise failed: Invalid argument
# redoing q=400277, rho=5345 because buckets are full
# Fullest level-1s bucket #44, wrote 23894/23855
# Maybe you have too many threads compared to the size of the factor bases.
# Please try less threads, or a larger -bkmult parameter (at some cost!).
# The code will now try to adapt by allocating more memory for buckets.
# Average J=1517 for 7765 special-q's, max bucket fill -bkmult 1,1s:1.1018
# Discarded 0 special-q's out of 7765 pushed
# Total cpu time 6710.98s [norm 9.25+77.0, sieving 3548.4 (1294.4 + 396.3 + 1857.7), factor 3076.4 (2740.1 + 336.3)]
# Total elapsed time 6869.65s, per special-q 0.884694s, per relation 0.0101343s
# PeakMemusage (MB) = 56 
# Total 677859 reports [0.0099s/r, 87.3r/sq]
So it produced relations (about 100/s) and after 100k special Q I got about 677,000 relations, which took almost 2h (on a single core)


After Q 400k-500k I did 500k-600k the same way (up to 1000k)

I unzipped the relations and put them together in a single file with the cat (miauw!):

Code:
odroid@odroid-u2:~/Documents/msieve-SVN1022$ cat rsa100.q400k-500k rsa100.q500k-600k rsa100.q600k-700k rsa100.q700k-800k rsa100.q800k-900k rsa100.q900k-1000k > msieve.dat

I had to open the msieve.dat file and put

Code:
N 1522605027922533360535618378132637429718068114961380688657908494580122963258952897654000350692006139
in as the first line to get msieve to recognize the relations.

Convert the .poly to a .fb file for use with Msieve (there should be a program available somewhere in Msieve or CADO-NFS, but I couldn't find it, so I did it manually).

Code:
N 1522605027922533360535618378132637429718068114961380688657908494580122963258952897654000350692006139
SKEW 3254900.80
A4 840
A3 -2848466002
A2 -8614727289369279
A1 -1180605682597027204428
A0 -12084851459361393651319621216
R1 378654317647
R0 -1160318224499030718294235
FAMAX 650000
 FRMAX 650000

and it worked!!!!
Code:
Mon Jun 25 00:20:30 2018  
Mon Jun 25 00:20:30 2018  
Mon Jun 25 00:20:30 2018  Msieve v. 1.54 (SVN Unversioned directory)
Mon Jun 25 00:20:30 2018  random seeds: 6c9d5866 1efb149a
Mon Jun 25 00:20:30 2018  factoring 1522605027922533360535618378132637429718068114961380688657908494580122963258952897654000350692006139 (100 digits)
Mon Jun 25 00:20:33 2018  searching for 15-digit factors
Mon Jun 25 00:20:35 2018  commencing number field sieve (100-digit input)
Mon Jun 25 00:20:35 2018  R0: -1160318224499030718294235
Mon Jun 25 00:20:35 2018  R1: 378654317647
Mon Jun 25 00:20:35 2018  A0: -12084851459361393651319621216
Mon Jun 25 00:20:35 2018  A1: -1180605682597027204428
Mon Jun 25 00:20:35 2018  A2: -8614727289369279
Mon Jun 25 00:20:35 2018  A3: -2848466002
Mon Jun 25 00:20:35 2018  A4: 840
Mon Jun 25 00:20:35 2018  skew 3254900.80, size 9.828e-14, alpha -5.221, combined = 1.210e-08 rroots = 2
Mon Jun 25 00:20:35 2018  
Mon Jun 25 00:20:35 2018  commencing relation filtering
Mon Jun 25 00:20:35 2018  estimated available RAM is 2023.1 MB
Mon Jun 25 00:20:35 2018  commencing duplicate removal, pass 1
Mon Jun 25 00:24:36 2018  found 542632 hash collisions in 4243088 relations
Mon Jun 25 00:25:34 2018  added 1544 free relations
Mon Jun 25 00:25:34 2018  commencing duplicate removal, pass 2
Mon Jun 25 00:25:42 2018  found 571058 duplicates and 3673574 unique relations
Mon Jun 25 00:25:42 2018  memory use: 41.3 MB
Mon Jun 25 00:25:42 2018  reading ideals above 100000
Mon Jun 25 00:25:42 2018  commencing singleton removal, initial pass
Mon Jun 25 00:27:11 2018  memory use: 94.1 MB
Mon Jun 25 00:27:11 2018  reading all ideals from disk
Mon Jun 25 00:27:11 2018  memory use: 110.1 MB
Mon Jun 25 00:27:13 2018  keeping 3374382 ideals with weight <= 200, target excess is 20465
Mon Jun 25 00:27:16 2018  commencing in-memory singleton removal
Mon Jun 25 00:27:17 2018  begin with 3673574 relations and 3374382 unique ideals
Mon Jun 25 00:27:31 2018  reduce to 2235657 relations and 1751368 ideals in 12 passes
Mon Jun 25 00:27:31 2018  max relations containing the same ideal: 147
Mon Jun 25 00:27:35 2018  removing 729096 relations and 498821 ideals in 230275 cliques
Mon Jun 25 00:27:36 2018  commencing in-memory singleton removal
Mon Jun 25 00:27:36 2018  begin with 1506561 relations and 1751368 unique ideals
Mon Jun 25 00:27:41 2018  reduce to 1356738 relations and 1083362 ideals in 8 passes
Mon Jun 25 00:27:41 2018  max relations containing the same ideal: 103
Mon Jun 25 00:27:44 2018  removing 601706 relations and 371431 ideals in 230275 cliques
Mon Jun 25 00:27:44 2018  commencing in-memory singleton removal
Mon Jun 25 00:27:46 2018  begin with 755032 relations and 1083362 unique ideals
Mon Jun 25 00:27:48 2018  reduce to 627990 relations and 565221 ideals in 8 passes
Mon Jun 25 00:27:48 2018  max relations containing the same ideal: 57
Mon Jun 25 00:27:49 2018  removing 141279 relations and 102250 ideals in 39029 cliques
Mon Jun 25 00:27:49 2018  commencing in-memory singleton removal
Mon Jun 25 00:27:49 2018  begin with 486711 relations and 565221 unique ideals
Mon Jun 25 00:27:50 2018  reduce to 456187 relations and 430334 ideals in 8 passes
Mon Jun 25 00:27:50 2018  max relations containing the same ideal: 50
Mon Jun 25 00:27:51 2018  relations with 0 large ideals: 368
Mon Jun 25 00:27:51 2018  relations with 1 large ideals: 307
Mon Jun 25 00:27:51 2018  relations with 2 large ideals: 4539
Mon Jun 25 00:27:51 2018  relations with 3 large ideals: 27250
Mon Jun 25 00:27:51 2018  relations with 4 large ideals: 83283
Mon Jun 25 00:27:51 2018  relations with 5 large ideals: 137570
Mon Jun 25 00:27:51 2018  relations with 6 large ideals: 120688
Mon Jun 25 00:27:51 2018  relations with 7+ large ideals: 82182
Mon Jun 25 00:27:51 2018  commencing 2-way merge
Mon Jun 25 00:27:52 2018  reduce to 284206 relation sets and 258353 unique ideals
Mon Jun 25 00:27:52 2018  commencing full merge
Mon Jun 25 00:28:05 2018  memory use: 28.5 MB
Mon Jun 25 00:28:05 2018  found 137264 cycles, need 132553
Mon Jun 25 00:28:05 2018  weight of 132553 cycles is about 9600137 (72.42/cycle)
Mon Jun 25 00:28:05 2018  distribution of cycle lengths:
Mon Jun 25 00:28:05 2018  1 relations: 9273
Mon Jun 25 00:28:05 2018  2 relations: 11186
Mon Jun 25 00:28:05 2018  3 relations: 12585
Mon Jun 25 00:28:05 2018  4 relations: 12807
Mon Jun 25 00:28:05 2018  5 relations: 12657
Mon Jun 25 00:28:05 2018  6 relations: 11770
Mon Jun 25 00:28:05 2018  7 relations: 11054
Mon Jun 25 00:28:05 2018  8 relations: 9737
Mon Jun 25 00:28:05 2018  9 relations: 8417
Mon Jun 25 00:28:05 2018  10+ relations: 33067
Mon Jun 25 00:28:05 2018  heaviest cycle: 22 relations
Mon Jun 25 00:28:05 2018  commencing cycle optimization
Mon Jun 25 00:28:06 2018  start with 904230 relations
Mon Jun 25 00:28:10 2018  pruned 32551 relations
Mon Jun 25 00:28:10 2018  memory use: 22.0 MB
Mon Jun 25 00:28:10 2018  distribution of cycle lengths:
Mon Jun 25 00:28:10 2018  1 relations: 9273
Mon Jun 25 00:28:10 2018  2 relations: 11434
Mon Jun 25 00:28:10 2018  3 relations: 13045
Mon Jun 25 00:28:10 2018  4 relations: 13348
Mon Jun 25 00:28:10 2018  5 relations: 13212
Mon Jun 25 00:28:10 2018  6 relations: 12248
Mon Jun 25 00:28:10 2018  7 relations: 11603
Mon Jun 25 00:28:10 2018  8 relations: 10001
Mon Jun 25 00:28:10 2018  9 relations: 8490
Mon Jun 25 00:28:10 2018  10+ relations: 29899
Mon Jun 25 00:28:10 2018  heaviest cycle: 21 relations
Mon Jun 25 00:28:10 2018  RelProcTime: 455
Mon Jun 25 00:28:10 2018  elapsed time 00:07:40
Mon Jun 25 00:30:42 2018  
Mon Jun 25 00:30:42 2018  
Mon Jun 25 00:30:42 2018  Msieve v. 1.54 (SVN Unversioned directory)
Mon Jun 25 00:30:42 2018  random seeds: e81300e3 2b61f139
Mon Jun 25 00:30:42 2018  factoring 1522605027922533360535618378132637429718068114961380688657908494580122963258952897654000350692006139 (100 digits)
Mon Jun 25 00:30:45 2018  searching for 15-digit factors
Mon Jun 25 00:30:47 2018  commencing number field sieve (100-digit input)
Mon Jun 25 00:30:47 2018  R0: -1160318224499030718294235
Mon Jun 25 00:30:47 2018  R1: 378654317647
Mon Jun 25 00:30:47 2018  A0: -12084851459361393651319621216
Mon Jun 25 00:30:47 2018  A1: -1180605682597027204428
Mon Jun 25 00:30:47 2018  A2: -8614727289369279
Mon Jun 25 00:30:47 2018  A3: -2848466002
Mon Jun 25 00:30:47 2018  A4: 840
Mon Jun 25 00:30:47 2018  skew 3254900.80, size 9.828e-14, alpha -5.221, combined = 1.210e-08 rroots = 2
Mon Jun 25 00:30:47 2018  
Mon Jun 25 00:30:47 2018  commencing linear algebra
Mon Jun 25 00:30:47 2018  read 132553 cycles
Mon Jun 25 00:30:47 2018  cycles contain 430159 unique relations
Mon Jun 25 00:30:56 2018  read 430159 relations
Mon Jun 25 00:30:57 2018  using 20 quadratic characters above 4294917295
Mon Jun 25 00:31:03 2018  building initial matrix
Mon Jun 25 00:31:22 2018  memory use: 48.8 MB
Mon Jun 25 00:31:22 2018  read 132553 cycles
Mon Jun 25 00:31:22 2018  matrix is 132373 x 132553 (38.0 MB) with weight 13013892 (98.18/col)
Mon Jun 25 00:31:22 2018  sparse part has weight 9028499 (68.11/col)
Mon Jun 25 00:31:27 2018  filtering completed in 2 passes
Mon Jun 25 00:31:27 2018  matrix is 132295 x 132475 (38.0 MB) with weight 13009752 (98.21/col)
Mon Jun 25 00:31:27 2018  sparse part has weight 9026861 (68.14/col)
Mon Jun 25 00:31:29 2018  matrix starts at (0, 0)
Mon Jun 25 00:31:29 2018  matrix is 132295 x 132475 (38.0 MB) with weight 13009752 (98.21/col)
Mon Jun 25 00:31:29 2018  sparse part has weight 9026861 (68.14/col)
Mon Jun 25 00:31:29 2018  saving the first 48 matrix rows for later
Mon Jun 25 00:31:29 2018  matrix includes 64 packed rows
Mon Jun 25 00:31:29 2018  matrix is 132247 x 132475 (36.6 MB) with weight 10287295 (77.65/col)
Mon Jun 25 00:31:29 2018  sparse part has weight 8790586 (66.36/col)
Mon Jun 25 00:31:29 2018  using block size 8192 and superblock size 49152 for processor cache size 512 kB
Mon Jun 25 00:31:30 2018  commencing Lanczos iteration
Mon Jun 25 00:31:30 2018  memory use: 28.8 MB
Mon Jun 25 00:32:16 2018  linear algebra at 9.2%, ETA 0h 7m
Mon Jun 25 00:39:47 2018  lanczos halted after 2093 iterations (dim = 132246)
Mon Jun 25 00:39:48 2018  recovered 33 nontrivial dependencies
Mon Jun 25 00:39:48 2018  BLanczosTime: 541
Mon Jun 25 00:39:48 2018  elapsed time 00:09:06
Mon Jun 25 00:40:39 2018  
Mon Jun 25 00:40:39 2018  
Mon Jun 25 00:40:39 2018  Msieve v. 1.54 (SVN Unversioned directory)
Mon Jun 25 00:40:39 2018  random seeds: be7f2f8d e1df76bb
Mon Jun 25 00:40:39 2018  factoring 1522605027922533360535618378132637429718068114961380688657908494580122963258952897654000350692006139 (100 digits)
Mon Jun 25 00:40:41 2018  searching for 15-digit factors
Mon Jun 25 00:40:44 2018  commencing number field sieve (100-digit input)
Mon Jun 25 00:40:44 2018  R0: -1160318224499030718294235
Mon Jun 25 00:40:44 2018  R1: 378654317647
Mon Jun 25 00:40:44 2018  A0: -12084851459361393651319621216
Mon Jun 25 00:40:44 2018  A1: -1180605682597027204428
Mon Jun 25 00:40:44 2018  A2: -8614727289369279
Mon Jun 25 00:40:44 2018  A3: -2848466002
Mon Jun 25 00:40:44 2018  A4: 840
Mon Jun 25 00:40:44 2018  skew 3254900.80, size 9.828e-14, alpha -5.221, combined = 1.210e-08 rroots = 2
Mon Jun 25 00:40:44 2018  
Mon Jun 25 00:40:44 2018  commencing square root phase
Mon Jun 25 00:40:44 2018  reading relations for dependency 1
Mon Jun 25 00:40:44 2018  read 66295 cycles
Mon Jun 25 00:40:44 2018  cycles contain 215178 unique relations
Mon Jun 25 00:40:49 2018  read 215178 relations
Mon Jun 25 00:40:53 2018  multiplying 215178 relations
Mon Jun 25 00:41:23 2018  multiply complete, coefficients have about 8.63 million bits
Mon Jun 25 00:41:24 2018  initial square root is modulo 91577
Mon Jun 25 00:42:12 2018  GCD is N, no factor found
Mon Jun 25 00:42:12 2018  reading relations for dependency 2
Mon Jun 25 00:42:12 2018  read 66301 cycles
Mon Jun 25 00:42:13 2018  cycles contain 214640 unique relations
Mon Jun 25 00:42:18 2018  read 214640 relations
Mon Jun 25 00:42:21 2018  multiplying 214640 relations
Mon Jun 25 00:42:52 2018  multiply complete, coefficients have about 8.61 million bits
Mon Jun 25 00:42:53 2018  initial square root is modulo 88951
Mon Jun 25 00:43:41 2018  sqrtTime: 177
Mon Jun 25 00:43:41 2018  p50 factor: 37975227936943673922808872755445627854565536638199
Mon Jun 25 00:43:41 2018  p50 factor: 40094690950920881030683735292761468389214899724061
 Mon Jun 25 00:43:41 2018  elapsed time 00:03:02
VictordeHolland is offline   Reply With Quote
Old 2018-06-25, 14:42   #2
paulunderwood
 
paulunderwood's Avatar
 
Sep 2002
Database er0rr

65668 Posts
Default

Were you using 4 cores on both the X86_64 and the ARM? What is the power usage of the 2 systems and what was each system's total time?
paulunderwood is offline   Reply With Quote
Old 2018-06-25, 15:28   #3
henryzz
Just call me Henry
 
henryzz's Avatar
 
"David"
Sep 2007
Cambridge (GMT/BST)

131448 Posts
Default

lasieve4 should compile with no asm. You should copy the generic folder and call it asm. I have had it running on a raspberry pi 2b.
It is worth bearing in mind that you would probably get a factor of several improvement in speed on a modern arm chip such as a A57.

How does msieve qs compare with c code to yafu qs with c code?
henryzz is online now   Reply With Quote
Old 2018-06-25, 17:10   #4
bsquared
 
bsquared's Avatar
 
"Ben"
Feb 2007

2·17·97 Posts
Default

Quote:
Originally Posted by henryzz View Post
How does msieve qs compare with c code to yafu qs with c code?
I'm actually not sure if it is possible to compile yafu qs without any x86 assembly or sse2 code. defining FORCE_GENERIC somewhere globally might work, but without looking I can't remember if I've guarded all assembly sections with that symbol. However with that defined the speed is pretty similar through c75 or so. After that yafu starts to get slightly faster (10 to 20%).

If anyone does manage to get yafu working on an ARM I'd like to hear about it
bsquared is offline   Reply With Quote
Old 2018-06-26, 00:53   #5
VictordeHolland
 
VictordeHolland's Avatar
 
"Victor de Hollander"
Aug 2011
the Netherlands

2·587 Posts
Default

Quote:
Originally Posted by henryzz View Post
lasieve4 should compile with no asm. You should copy the generic folder and call it asm. I have had it running on a raspberry pi 2b.
Thanks for the tip! It still took me more than 2 hours to fix all the errors (note errors, not warnings) I was getting (undefined variables, problems with header files and of course the cweb/ctangle/cweave stuff and a little editing of the makefile in the mix). I don't understand why they used the cweb stuff???



But persistence pays of (the Odroid-U2:) 10k special-Q

Code:
odroid@odroid-u2:~/Documents/ggnfs-source/bin/testing$ ./gnfs-lasieve4I12e -a -v -o testRels testPoly
gnfs-lasieve4I12e: L1_BITS=15, SVN $Revision$
FBsize 89232+0 (deg 4), 169510+0 (deg 1)
total yield: 79211, q=2310019 (0.01136 sec/rel)
648 Special q, 974 reduction iterations
reports: 294157337->26568664->24297997->5506929->5506836->5140364
Number of relations with k rational and l algebraic primes for (k,l)=:

Total yield: 79211
0/0 mpqs failures, 42287/41995 vain mpqs
milliseconds total: Sieve 45468 Sched 0 medsched 31
TD 234474 (Init 0, MPQS 47318) Sieve-Change 210096
TD side 0: init/small/medium/large/search: 0 0 0 0 0
sieve: init/small/medium/large/search: 0 0 0 23372 0
TD side 1: init/small/medium/large/search: 0 0 0 0 124656
sieve: init/small/medium/large/search: 0 0 0 21900 196
Intel Xeon E5-2650 (Sandy-bridge based Xeon) generic

Code:
./gnfs-lasieve4I12e -a -v -o testRels testPoly
gnfs-lasieve4I12e: L1_BITS=15, SVN $Revision$
FBsize 89232+0 (deg 4), 169510+0 (deg 1)
total yield: 79211, q=2310019 (0.00380 sec/rel)
648 Special q, 974 reduction iterations
reports: 294157337->26568664->24297997->5506929->5506836->5140364
Number of relations with k rational and l algebraic primes for (k,l)=:

Total yield: 79211
0/0 mpqs failures, 42287/41995 vain mpqs
milliseconds total: Sieve 15557 Sched 0 medsched 0
TD 10260 (Init 0, MPQS 0) Sieve-Change 81015
TD side 0: init/small/medium/large/search: 0 0 0 0 0
sieve: init/small/medium/large/search: 0 0 0 7780 0
TD side 1: init/small/medium/large/search: 0 0 0 0 3211
sieve: init/small/medium/large/search: 0 0 0 7777 0
Intel Xeon E5-2650 (Sandy-bridge based Xeon) asm64 (downloaded somewhere from the forum)

Code:
./gnfs-lasieve4I12e -a -v -o testRels testPoly
gnfs-lasieve4I12e (with asm64): L1_BITS=15, SVN $Revision: 399 $
FBsize 89232+0 (deg 4), 169510+0 (deg 1)
total yield: 79211, q=2310019 (0.00189 sec/rel) 
648 Special q, 974 reduction iterations
reports: 199726752->18930110->17344713->5507541->5507448->5140389
Number of relations with k rational and l algebraic primes for (k,l)=:

Total yield: 79211
0/0 mpqs failures, 42287/41995 vain mpqs
milliseconds total: Sieve 37300 Sched 0 medsched 23990
TD 53880 (Init 2820, MPQS 32270) Sieve-Change 34440
TD side 0: init/small/medium/large/search: 780 2330 960 1030 4930
sieve: init/small/medium/large/search: 2260 10750 940 4340 2440
TD side 1: init/small/medium/large/search: 320 2080 910 1090 4050
sieve: init/small/medium/large/search: 880 6210 880 4210 4390
Intel Xeon E5-2650 (Sandy-bridge based Xeon) asm64 (compiled on the machine)

Code:
./gnfs-lasieve4I12e -a -v -o testRels testPoly
gnfs-lasieve4I12e (with asm64): L1_BITS=15, SVN $Revision$
FBsize 89232+0 (deg 4), 169510+0 (deg 1)
total yield: 79211, q=2310019 (0.00178 sec/rel) 
648 Special q, 974 reduction iterations
reports: 199726752->18930110->17344713->5507541->5507448->5140389
Number of relations with k rational and l algebraic primes for (k,l)=:

Total yield: 79211
0/0 mpqs failures, 42287/41995 vain mpqs
milliseconds total: Sieve 34144 Sched 0 medsched 24430
TD 47818 (Init 2512, MPQS 27253) Sieve-Change 34580
TD side 0: init/small/medium/large/search: 770 2264 960 1034 4545
sieve: init/small/medium/large/search: 1644 9387 954 4325 2375
TD side 1: init/small/medium/large/search: 314 2164 907 1118 3643
sieve: init/small/medium/large/search: 849 5945 932 4329 3404
Quote:
It is worth bearing in mind that you would probably get a factor of several improvement in speed on a modern arm chip such as a A57.
I know, it is mostly just for the fun of it. I sometimes have no idea what I'm doing .

Quote:
How does msieve qs compare with c code to yafu qs with c code?
The YAFU code is full of ASM intrinsics and files (SSE2, SSE4.1, AVX2, AVX-512) so I'm going nowhere near that. That stuff is way above my head/skill. If it doesn't work after 'make' or editing the makefile to point to the libs I just scream for help .
VictordeHolland is offline   Reply With Quote
Old 2018-06-26, 08:36   #6
henryzz
Just call me Henry
 
henryzz's Avatar
 
"David"
Sep 2007
Cambridge (GMT/BST)

22×1,433 Posts
Default

A factor of 4 difference isn't much given the difference in power draw.
I wonder how it would perform on a more modern ARM chip.
It looks like you might get better performance by optimizing parameters differently.
henryzz is online now   Reply With Quote
Old 2018-06-26, 09:47   #7
VictordeHolland
 
VictordeHolland's Avatar
 
"Victor de Hollander"
Aug 2011
the Netherlands

2×587 Posts
Default

Quote:
Originally Posted by paulunderwood View Post
Were you using 4 cores on both the X86_64 and the ARM? What is the power usage of the 2 systems and what was each system's total time?
The results are in for RSA-100

Odroid-U2
<1 W idle (my wattmeter is not accurate enough)
7.2W sieveing with 4 threads
Code:
Tue Jun 26 03:04:52 2018 -> factmsieve.py (v0.86)
Tue Jun 26 03:04:52 2018 -> This is client 1 of 1
Tue Jun 26 03:04:52 2018 -> Running on 4 Cores with 1 hyper-thread per Core
Tue Jun 26 03:04:52 2018 -> Working with NAME = RSA100
...
Tue Jun 26 06:16:43 2018  p50 factor: 37975227936943673922808872755445627854565536638199
Tue Jun 26 06:16:43 2018  p50 factor: 40094690950920881030683735292761468389214899724061
3:11:51

(which is comparable to the 3:31:18 it took with CADO-NFS 3.0-dev)

A reference point:
Intel Core i7 3770k @3.5GHz (80+ Gold PSU)
Using the pythonscript+lasieve4+msieve
54.5W idle
88-89W - 4core/4thread sieve - 25:12
96-97.5W - 4core/8thread sieve - 21:17

Some other RSA100 times:
Dual Intel Xeon E5-2650 (16core/32thread) - YAFU (NFS) - 10:25
Dual Intel Xeon E5-2650 (16core/32thread) - lasieve4+msieve+py - 11:20
Dual Intel Xeon E5-2650 (16core/32thread) - CADO-NFS 3.0-dev - 11:56
KNL 7210 (256-thread) AVX512 SIQS - 6:08
from
http://mersenneforum.org/showpost.ph...&postcount=241
and
http://mersenneforum.org/showthread.php?t=23362

Last fiddled with by VictordeHolland on 2018-06-26 at 09:50 Reason: somehow the forum adds extra newlines to my posts...
VictordeHolland is offline   Reply With Quote
Old 2018-07-19, 13:19   #8
VictordeHolland
 
VictordeHolland's Avatar
 
"Victor de Hollander"
Aug 2011
the Netherlands

2·587 Posts
Default

I bought myself a little present:
Raspberry Pi 3B+
It was only 35 euros :)

Should have some CADO-NFS and Msieve results soon.
VictordeHolland is offline   Reply With Quote
Old 2018-07-20, 10:36   #9
VictordeHolland
 
VictordeHolland's Avatar
 
"Victor de Hollander"
Aug 2011
the Netherlands

2·587 Posts
Default

Quote:
Originally Posted by VictordeHolland View Post
I bought myself a little present:
Raspberry Pi 3B+
It was only 35 euros :)

Should have some CADO-NFS and Msieve results soon.
Factmsieve v0.86 + Msieve SVN1022 + Lasieve4 (generic)

Code:
Thu Jul 19 09:49:47 2018 -> factmsieve.py (v0.86)
Thu Jul 19 09:49:47 2018 -> This is client 1 of 1
Thu Jul 19 09:49:47 2018 -> Running on 4 Cores with 1 hyper-thread per Core
Thu Jul 19 09:49:47 2018 -> Working with NAME = RSA100
...
Thu Jul 19 13:41:11 2018  p50 factor: 37975227936943673922808872755445627854565536638199
Thu Jul 19 13:41:11 2018  p50 factor: 40094690950920881030683735292761468389214899724061
3:51:24



CADO-NFS 3.0-dev

Code:
Info:Square Root: Factors: 37975227936943673922808872755445627854565536638199 40094690950920881030683735292761468389214899724061
Info:Square Root: Total cpu/real time for sqrt: 514.38/159.773
Info:Polynomial Selection (size optimized): Aggregate statistics:
Info:Polynomial Selection (size optimized): potential collisions: 1867.92
Info:Polynomial Selection (size optimized): raw lognorm (nr/min/av/max/std): 3725/33.340/37.848/38.690/0.707
Info:Polynomial Selection (size optimized): optimized lognorm (nr/min/av/max/std): 3725/32.980/37.005/38.680/1.127
Info:Polynomial Selection (size optimized): Total time: 1634.75
Info:Polynomial Selection (root optimized): Aggregate statistics:
Info:Polynomial Selection (root optimized): Total time: 460.95
Info:Polynomial Selection (root optimized): Rootsieve time: 457.98
Info:Generate Factor Base: Total cpu/real time for makefb: 7.48/2.16876
Info:Generate Free Relations: Total cpu/real time for freerel: 205.35/53.9807
Info:Lattice Sieving: Aggregate statistics:
Info:Lattice Sieving: Total number of relations: 2362062
Info:Lattice Sieving: Average J: 1019.58 for 69667 special-q, max bucket fill -bkmult 1.0,1s:1.204870
Info:Lattice Sieving: Total time: 19829.4s
Info:Filtering - Duplicate Removal, splitting pass: Total cpu/real time for dup1: 22.71/51.1323
Info:Filtering - Duplicate Removal, splitting pass: Aggregate statistics:
Info:Filtering - Duplicate Removal, splitting pass: CPU time for dup1: 50.7s
Info:Filtering - Duplicate Removal, removal pass: Total cpu/real time for dup2: 90.65/73.286
Info:Filtering - Duplicate Removal, removal pass: Aggregate statistics:
Info:Filtering - Duplicate Removal, removal pass: CPU time for dup2: 68.4s
Info:Filtering - Singleton removal: Total cpu/real time for purge: 70.99/55.5308
Info:Filtering - Merging: Total cpu/real time for merge: 346.7/329.282
Info:Filtering - Merging: Total cpu/real time for replay: 33.93/32.8305
Info:Linear Algebra: Total cpu/real time for bwc: 18085.5/4799.38
Info:Linear Algebra: Aggregate statistics:
Info:Linear Algebra: Krylov: WCT time 2911.73, iteration CPU time 0.45, COMM 0.01, cpu-wait 0.02, comm-wait 0.0 (6000 iterations)
Info:Linear Algebra: Lingen CPU time 712.82, WCT time 189.49
Info:Linear Algebra: Mksol: WCT time 1482.32, iteration CPU time 0.47, COMM 0.01, cpu-wait 0.02, comm-wait 0.0 (3000 iterations)
Info:Quadratic Characters: Total cpu/real time for characters: 21.5/10.3837
Info:Square Root: Total cpu/real time for sqrt: 514.38/159.773
Info:HTTP server: Shutting down HTTP server
Info:Complete Factorization: Total cpu/elapsed time for entire factorization: 51928.4/16344.8
CPU time: 51928.4 Elapsed time: 16344.8

4:32:25
VictordeHolland is offline   Reply With Quote
Old 2018-07-20, 10:44   #10
henryzz
Just call me Henry
 
henryzz's Avatar
 
"David"
Sep 2007
Cambridge (GMT/BST)

131448 Posts
Default

Where is the qs/gnfs border on systems like this?

Also how does the raspberry pi 3b compare powerwise to your Odroid U2?

It looks like arm is more efficient for nfs sieving than x86 according to your power report a few posts ago.
henryzz is online now   Reply With Quote
Old 2018-07-20, 14:56   #11
VictordeHolland
 
VictordeHolland's Avatar
 
"Victor de Hollander"
Aug 2011
the Netherlands

2·587 Posts
Default

Quote:
Originally Posted by henryzz View Post
Where is the qs/gnfs border on systems like this?

Also how does the raspberry pi 3b compare powerwise to your Odroid U2?

It looks like arm is more efficient for nfs sieving than x86 according to your power report a few posts ago.
Yes, but the differences are not that shockingly big.


Quote:
Originally Posted by VictordeHolland View Post
Odroid-U2
<1 W idle (my wattmeter is not accurate enough)
7.2W sieveing with 4 threads
3:11:51
Factoring RSA100 using the Odroid-U2 works out to roughly:
11,511 seconds * 7.2W = 82,879 Joules

Quote:
A reference point:
Intel Core i7 3770k @3.5GHz (80+ Gold PSU)
Using the pythonscript+lasieve4+msieve
54.5W idle
88-89W - 4core/4thread sieve - 25:12
96-97.5W - 4core/8thread sieve - 21:17
And for a quadcore of a few generations old, then factoring RSA100 takes:
1,277 seconds * 97.5W = 124,508 Joules

Somehow the Raspberry Pi 3B+ takes longer than the ODROID-U2
(wombatman got similar timings on his Pi3B+) see:
http://mersenneforum.org/showpost.ph...&postcount=273
it is using about ~6.7W (fluctuates a lot, so eye-balling it a bit)

3:51:24
13,884 seconds * 6.7W = 93,023 Joules

The biggest obstacle with using these ARM boards for any useful factorizations is their severely limited memory (1GB for the RPi3B+ and 2GB for the ODROID-U2)
VictordeHolland is offline   Reply With Quote
Reply

Thread Tools


Similar Threads
Thread Thread Starter Forum Replies Last Post
Next-gen Odroid announcement ewmayer Mlucas 116 2020-07-23 09:24
mprime on Odroid 64bit ET_ Software 2 2017-02-24 15:42
ARM reveals Cortex A72 details VictordeHolland Hardware 0 2015-04-24 14:49
N queens on an N/N board. Peter Hackman Math 4 2008-09-24 16:34
Board colors Prime Monster Lounge 38 2003-05-23 19:14

All times are UTC. The time now is 15:54.

Tue Oct 27 15:54:39 UTC 2020 up 47 days, 13:05, 1 user, load averages: 2.49, 2.78, 2.63

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2020, Jelsoft Enterprises Ltd.

This forum has received and complied with 0 (zero) government requests for information.

Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.2 or any later version published by the Free Software Foundation.
A copy of the license is included in the FAQ.