YAFU Version 2.10

[James Heinrich]

Originally Posted by bsquared

There is a option -snfs_xover (I think?) which you should probably set higher... to maybe 100 digits. That way siqs will run instead of these tiny snfs jobs; might be more reliable.

Oh, that works for me. It was unset in yafu.ini and defaults to 75 (and this was a c76), I've set it to 100 and it's happily SIQS'ing now.

[James Heinrich]

Originally Posted by James Heinrich

...on screen it was just looping:

unrecognized character in input: -1
=== Starting work on batchfile expression ===
factor( ♥)
=============================================
unrecognized character in input: -1
=== Starting work on batchfile expression ===
factor( ♥)

I'm not sure exactly how things changed because it I think it was working previously, but this is a permanent feature now on the latest release (Sep-25) whenever the batchfile gets empty. I've had to revert to the Sep-22 version from before the trailing-\n fix and that's working ok for now.
Is there the possibility that the trailing-newline check code you added could get confused between \n and \r\n lineends? My batchfile input file has \n only throughout the file.

[EdH]

I'm having problems getting the current pull to build at all on a machine with a GPU. I've tried with and without Msieve being built with CUDA=1.

My errors are such as this:

/usr/bin/ld: ../msieve//libmsieve.a(cuda_xface.o): in function `gpu_init':
cuda_xface.c:(.text+0x28e): undefined reference to `cuInit'
/usr/bin/ld: cuda_xface.c:(.text+0x29e): undefined reference to `cuDeviceGetCount'
/usr/bin/ld: cuda_xface.c:(.text+0x2d6): undefined reference to `cuDeviceGet'
/usr/bin/ld: cuda_xface.c:(.text+0x2f3): undefined reference to `cuDeviceGetName'
/usr/bin/ld: cuda_xface.c:(.text+0x30c): undefined reference to `cuDeviceComputeCapability'
/usr/bin/ld: cuda_xface.c:(.text+0x320): undefined reference to `cuDeviceGetProperties'
/usr/bin/ld: cuda_xface.c:(.text+0x394): undefined reference to `cuDeviceTotalMem_v2'
/usr/bin/ld: cuda_xface.c:(.text+0x3ae): undefined reference to `cuDeviceGetAttribute'
/usr/bin/ld: cuda_xface.c:(.text+0x3c8): undefined reference to `cuDeviceGetAttribute'
/usr/bin/ld: cuda_xface.c:(.text+0x3e1): undefined reference to `cuDeviceGetAttribute'
/usr/bin/ld: ../msieve//libmsieve.a(cuda_xface.o): in function `gpu_launch_init':
cuda_xface.c:(.text+0x52a): undefined reference to `cuModuleGetFunction'
/usr/bin/ld: cuda_xface.c:(.text+0x541): undefined reference to `cuFuncGetAttribute'
/usr/bin/ld: cuda_xface.c:(.text+0x5bb): undefined reference to `cuParamSetSize'
/usr/bin/ld: ../msieve//libmsieve.a(cuda_xface.o): in function `gpu_launch_set':
cuda_xface.c:(.text+0x6b0): undefined reference to `cuParamSetv'
/usr/bin/ld: cuda_xface.c:(.text+0x700): undefined reference to `cuParamSetv'
/usr/bin/ld: cuda_xface.c:(.text+0x72b): undefined reference to `cuParamSeti'
/usr/bin/ld: cuda_xface.c:(.text+0x753): undefined reference to `cuParamSeti'
/usr/bin/ld: cuda_xface.c:(.text+0x787): undefined reference to `cuParamSetv'
collect2: error: ld returned 1 exit status
make: *** [Makefile:371: yafu] Error 1

My CUDA version is 10.1 and I'm adding no switches other than NFS=1 to the make command. If this machine didn't use GPU at all with YAFU, it would be fine, at least for now. I haven't looked to see if I can tell it not try.

[James Heinrich]

I'm getting unexpected values in factor.json for runtime.pp1, always the same ridiculous value. As far as I can tell no P+1 is actually being done.
As a random example (but all similar numbers are giving the exact same output for pp1)
factor(15351101497487209665634388498612121497158603986979177288509551776957312914211)

Starting factorization of 15351101497487209665634388498612121497158603986979177288509551776957312914211
using pretesting plan: deep
using specified qs/gnfs crossover of 100 digits
using specified qs/snfs crossover of 100 digits
****************************
rho: x^2 + 3, starting 1000 iterations on C77
rho: x^2 + 2, starting 1000 iterations on C77
rho: x^2 + 1, starting 1000 iterations on C77
pm1: starting B1 = 150K, B2 = gmp-ecm default on C77
current ECM pretesting depth: 0.000000
scheduled 30 curves at B1=2000 toward target pretesting depth of 25.666667
Finished 30 curves using GMP-ECM method on C77 input, B1=2k, B2=gmp-ecm default
current ECM pretesting depth: 15.177725
scheduled 74 curves at B1=11000 toward target pretesting depth of 25.666667
Finished 74 curves using GMP-ECM method on C77 input, B1=11k, B2=gmp-ecm default
current ECM pretesting depth: 20.242996
scheduled 214 curves at B1=50000 toward target pretesting depth of 25.666667
Finished 214 curves using GMP-ECM method on C77 input, B1=50k, B2=gmp-ecm default
pm1: starting B1 = 3750K, B2 = gmp-ecm default on C77
current ECM pretesting depth: 25.332854
scheduled 29 curves at B1=250000 toward target pretesting depth of 25.666667
Finished 29 curves using GMP-ECM method on C77 input, B1=250k, B2=gmp-ecm default
final ECM pretested depth: 25.670063
scheduler: switching to sieve method
starting SIQS on c77: 15351101497487209665634388498612121497158603986979177288509551776957312914211
random seed: 3340054979937811927
==== sieve params ====
n = 78 digits, 258 bits
factor base: 33744 primes (max prime = 852833)
single large prime cutoff: 76754970 (90 * pmax)
double large prime range from 727324125889 to 62941398164503
DLP MFB = 1.75
allocating 6 large prime slices of factor base
buckets hold 2048 elements
large prime hashtables have 589824 bytes
using AVX2 enabled 32k sieve core
sieve interval: 6 blocks of size 32768
polynomial A has ~ 10 factors
using multiplier of 19
using multiplier of 19
using Q2(x) polynomials for kN mod 8 = 1
using SPV correction of 20 bits, starting at offset 31
trial factoring cutoff at 83 bits
==== sieving started ( 4 threads) ====
trial division touched 5443914 sieve locations out of 27371765760
total reports = 5443914, total surviving reports = 2076476
total blocks sieved = 835320, avg surviving reports per block = 2.49
dlp-ecm: 0 failures, 99911 attempts, 1145720 outside range, 669060 prp, 99044 useful
34344 relations found: 14037 full + 20307 from 246792 partial, using 69610 polys (277 A polys)
on average, sieving found 3.75 rels/poly and 18386.73 rels/sec
trial division touched 5443914 sieve locations out of 27371765760
==== post processing stage (msieve-1.38) ====
QS elapsed time = 14.1857 seconds.
begin singleton removal with 260829 relations
reduce to 61510 relations in 7 passes
recovered 61510 relations
recovered 40609 polynomials
attempting to build 34344 cycles
found 34344 cycles from 61510 relations in 4 passes
distribution of cycle lengths:
   length 1 : 14037
   length 2 : 11295
   length 3 : 5470
   length 4 : 2238
   length 5 : 868
   length 6 : 284
   length 7 : 105
   length 9+: 47
largest cycle: 11 relations
matrix is 33744 x 34344 (5.8 MB) with weight 1251414 (36.44/col)
sparse part has weight 1251414 (36.44/col)
filtering completed in 4 passes
matrix is 26474 x 26538 (4.6 MB) with weight 1003074 (37.80/col)
sparse part has weight 1003074 (37.80/col)
saving the first 48 matrix rows for later
matrix is 26426 x 26538 (3.8 MB) with weight 834096 (31.43/col)
sparse part has weight 739982 (27.88/col)
matrix includes 64 packed rows
using block size 10615 for processor cache size 6144 kB
commencing Lanczos iteration
memory use: 3.4 MB
lanczos halted after 419 iterations (dim = 26426)
recovered 17 nontrivial dependencies
prp27 = 180520589726984033523662249
prp50 = 85037953403010308691806417803623254338384481538539
Lanczos elapsed time = 1.3860 seconds.
Sqrt elapsed time = 0.0190 seconds.
SIQS elapsed time = 15.5857 seconds.

Total factoring time = 24.5904 seconds

In all cases it shows in factor.json

"runtime" : {"total":24.5904, "ecm":2.9725, "pm1":0.4850, "pp1":52927428516644887548067236846111591126652782165671984789388868382343165426534958353833061423718620048672828088701042409249340559637837076884380581258410371962304204496086523131765645556786332323282944.0000, "siqs":15.5857}

Not sure what that number of seconds means (it's 2^611 × 937 × 6646819000919 if that matters )

[kelvinjb]

To EdH

I change below in yafu2 Makefile seems fix it

from
ifeq ($(NFS),1)
CFLAGS += -DUSE_NFS
LIBS += -lmsieve
endif

to
ifeq ($(NFS),1)
CFLAGS += -DUSE_NFS
LIBS += -lmsieve -L/usr/local/cuda/lib64 -lcudart
INC += -I/usr/local/cuda/include/
LIBS += -L/usr/local/cuda/lib64 -lcuda
endif


Note: Below is for cuda installation on my Debian.. Pls check yours
/usr/local/cuda/lib64
/usr/local/cuda/include

[Stargate38]

Why does YAFU crash when trying to run SNFS poly select on 4195451890^16+1? Here's the output:

C:\Numbers\yafu-master>yafu "snfs(4195451890^16+1,(4195451890^16+1)/5435823450651429032129)" -np
Applying tune_info entry for WIN64 - Intel(R) Core(TM) i5-9300H CPU @ 2.40GHz


YAFU Version 2.10
Built with Microsoft Visual Studio 1931
Using GMP-ECM 7.0.4, Powered by MPIR 3.0.0
Detected Intel(R) Core(TM) i5-9300H CPU @ 2.40GHz
Detected L1 = 32768 bytes, L2 = 8388608 bytes, CL = 64 bytes
Using 1 random witness for Rabin-Miller PRP checks
Cached 664579 primes; max prime is 9999991
Parsed yafu.ini from C:\Numbers\yafu-master

===============================================================
======= Welcome to YAFU (Yet Another Factoring Utility) =======
=======             bbuhrow@gmail.com                   =======
=======     Type help at any time, or quit to quit      =======
===============================================================

>> nfs: checking for job file - job file found, testing for matching input
input from file = 29740272518858589450854274542251246345384637261381756818007424552883637946081745099049541938176119278535535553407111
input to yafu = 9214248596771890050888607398326355640430009796427922634882506077083800083753822640222841638769980323080148134556984701851008017351895689610000000000000001
nfs: number in job file does not match input
nfs: checking for poly file - no poly file found
nfs: commencing nfs on c154: 9214248596771890050888607398326355640430009796427922634882506077083800083753822640222841638769980323080148134556984701851008017351895689610000000000000001
nfs: searching for brent special forms...
nfs: checking a*b^x +/- c for 32 <= x <= 1478
nfs: checking a*b^x +/- c for 32 <= x <= 933
nfs: checking a*b^x +/- c for 32 <= x <= 637
nfs: checking a*b^x +/- c for 32 <= x <= 572
<snip>
nfs: checking x^19 +/- 1
nfs: checking x^18 +/- 1
nfs: checking x^17 +/- 1
nfs: checking x^16 +/- 1
nfs: input divides 4195451890^16 + 1
nfs: snfs form detection took 0.084631 seconds
gen: rank 1 terms:
gen: base exponent multiplier: 16
gen: multiplying by 4195451890^16 + 1 = 9214248596771890050888607398326355640430009796427922634882506077083800083753822640222841638769980323080148134556984701851008017351895689610000000000000001
gen: ========================================================
gen: considering the following polynomials:
gen: ========================================================

nfs: checking degree 4 poly
n: 1695097105419723406237626702076589959247894735221076213344744607660664326796107448808495795401150214473873473769448589776360204123969
# 4195451890^16+1, difficulty: 153.96, anorm: 3.60e+28, rnorm: 3.59e+45
# size = 5.663e-16, alpha = 2.635, combined = 8.544e-10, rroots = 0
type: snfs
size: 153
skew: 1.0000
c4: 1
c0: 1
Y1: -1
Y0: 309823946257815911188689860364098410000
nfs: checking degree 5 poly
n: 1695097105419723406237626702076589959247894735221076213344744607660664326796107448808495795401150214473873473769448589776360204123969
# 4195451890^16+1, difficulty: 192.46, anorm: 2.29e+53, rnorm: 2.54e+41
# size = 5.938e-18, alpha = -0.721, combined = 5.947e-14, rroots = 1
type: snfs
size: 192
skew: 49914071.4387
c5: 1
c0: 309823946257815911188689860364098410000
Y1: -1
Y0: 309823946257815911188689860364098410000
nfs: checking degree 5 poly
n: 1695097105419723406237626702076589959247894735221076213344744607660664326796107448808495795401150214473873473769448589776360204123969
# 4195451890^16+1, difficulty: 153.96, anorm: 2.70e+40, rnorm: 7.84e+36
# size = 7.610e-12, alpha = 1.405, combined = 2.890e-10, rroots = 1
type: snfs
size: 153
skew: 0.0119
c5: 4195451890
c0: 1
Y1: -1
Y0: 73847574559558567882586269000
nfs: checking degree 5 poly
Error: M=686000
C:\Numbers\yafu-master>

Factor.log:

11/24/22 12:04:21, nfs: commencing nfs on c154: 9214248596771890050888607398326355640430009796427922634882506077083800083753822640222841638769980323080148134556984701851008017351895689610000000000000001
11/24/22 12:04:21, nfs: input divides 4195451890^16 + 1
11/24/22 12:04:21, nfs: using supplied cofactor: 1695097105419723406237626702076589959247894735221076213344744607660664326796107448808495795401150214473873473769448589776360204123969
11/24/22 12:04:21, nfs: commencing snfs on c133: 1695097105419723406237626702076589959247894735221076213344744607660664326796107448808495795401150214473873473769448589776360204123969

Exactly what does "Error: M=686000" mean?

[storm5510]

v2.x seems to have an OS dependency in Windows 10 which is not present with a clean install. A dialog appears but does not specify what it is. It refuses to load.

[chalsall]

Originally Posted by storm5510

v2.x seems to have an OS dependency in Windows 10 which is not present with a clean install. A dialog appears but does not specify what it is. It refuses to load.

Hmmm... That might be a currently undocumented critical path error. Which some might consider fairly serious.

Please correct me if I am wrong on this, but is Windows 10 not EOL?

I am often wrong. And am welcoming of correction.

[kruoli]

The EOL of "Windows 10" (excluding LTSC versions et al.) is 2025, October 14th, when using the latest (or even "some") "release" (to be determined by M$).

[chalsall]

Originally Posted by kruoli

The EOL of "Windows 10" (excluding LTSC versions et al.) is 2025, October 14th, when using the latest (or even "some") "release" (to be determined by M$).

LOL... I actually greatly appreciate being corrected.

One cannot speak the truth if one doesn't know the current state of things.

Being corrected is part of the process...

[storm5510]

Originally Posted by chalsall

Hmmm... That might be a currently undocumented critical path error. Which some might consider fairly serious.

Please correct me if I am wrong on this, but is Windows 10 not EOL?

I am often wrong. And am welcoming of correction.

It is Windows 10 v21H1. My previous setup had run since early 2019 and was acting strange. It started out as v1809, I think, then patched as time passed. In the interim, I used their media creation tool to get a clean ISO of v21H1. I reloaded it about two weeks ago.

Off-topic: IMHO, stay away from v22H2. I tried it for just a few days. For one, the ISO will not burn onto a single-layer DVD. It is 74 MB too large. I used a USB. The context menu, right-click in any location in the File Manager, is missing an option under "New." "Folder" is gone. Only "Shortcut" remains. This was the proverbial turd-in-the-punchbowl for me! I always figured M$ would try to restrict what a person could do as time passed. It seems I was correct. My hardware it too long-in-the-tooth for W11. I have no problems with that at all!