YAFU 2.0

[James Heinrich]

Originally Posted by thyrex

But I see "unsupported 16-bit application" in header of dialog error window. And size of .exe is around 125Kb only. File with 2.07 version has size above 2Mb

Maybe your download failed. The file available at https://github.com/bbuhrow/yafu/blob...e/yafu-x64.exe is (currently) 2,582,528 bytes.

[bur]

When I try to build yafu under Ubuntu with make NFS=1 USE_SSE41=1 USE_AVX=1 it quits with:

gcc -g -m64 -DUSE_SSE2 -DUSE_AVX2 -DUSE_SSE41  -mavx2  -DUSE_SSE41 -msse4.1 -DUSE_NFS -O2 -fomit-frame-pointer -Wall  -I. -Iinclude -Itop/aprcl -Itop/cmdParser -Itop/  -I../ysieve -I../ytools -I/usr/local/include -I../msieve/zlib  top/driver.o top/test.o factor/tune.o factor/autofactor.o top/cmdParser/cmdOptions.o top/cmdParser/calc.o -o yafu -lysiqs  -lyecm  -lynfs -L. -L../ysieve -L../ytools -L/usr/local/include -L../msieve/ -lmsieve -lecm /usr/lib/x86_64-linux-gnu/libgmp.a -lytools -lysieve -lpthread -lm -ldl
/usr/bin/ld: ./libysiqs.a(SIQS.o): in function `siqs_static_init':
/home/bur/Math/yafu/factor/qs/SIQS.c:2309: undefined reference to `nextRoots_32k_avx2_intrin'
/usr/bin/ld: ../msieve//libmsieve.a(savefile.o): in function `savefile_open':
savefile.c:(.text+0x142): undefined reference to `gzopen64'
/usr/bin/ld: savefile.c:(.text+0x266): undefined reference to `gzopen64'
/usr/bin/ld: savefile.c:(.text+0x2a1): undefined reference to `gzopen64'
/usr/bin/ld: ../msieve//libmsieve.a(savefile.o): in function `savefile_close':
savefile.c:(.text+0x341): undefined reference to `gzclose'
/usr/bin/ld: ../msieve//libmsieve.a(savefile.o): in function `savefile_write_line':
savefile.c:(.text+0x461): undefined reference to `gzputs'
/usr/bin/ld: ../msieve//libmsieve.a(savefile.o): in function `savefile_flush':
savefile.c:(.text+0x4e4): undefined reference to `gzputs'
/usr/bin/ld: ../msieve//libmsieve.a(savefile.o): in function `savefile_eof':
savefile.c:(.text+0x371): undefined reference to `gzeof'
/usr/bin/ld: ../msieve//libmsieve.a(savefile.o): in function `savefile_read_line':
savefile.c:(.text+0x3f0): undefined reference to `gzgets'
/usr/bin/ld: ../msieve//libmsieve.a(savefile.o): in function `savefile_rewind':
savefile.c:(.text+0x521): undefined reference to `gzrewind'
collect2: error: ld returned 1 exit status
make: *** [Makefile:368: yafu] Error 1

Omitting the USE_AVX2=1 flag, then it builds successfully.

[BudgieJane]

I've got another of those "it won't factor it" jobbies in Yafu 2.08. For some reason it won't go into QS/SIQS.

06/20/22 16:15:39,
06/20/22 16:15:39, ****************************
06/20/22 16:15:39, Starting factorization of 7036874796802706717792076127030118075734377553821445738998743574873144601450291676951142421607390370359561
06/20/22 16:15:39, using pretesting plan: deep
06/20/22 16:15:39, using tune info for qs/gnfs crossover
06/20/22 16:15:39, ****************************
06/20/22 16:15:39, div: found prime factor = 137
06/20/22 16:15:39, div: found prime factor = 433
06/20/22 16:15:39, rho: x^2 + 3, starting 1000 iterations on C102
06/20/22 16:15:39, rho: x^2 + 2, starting 1000 iterations on C102
06/20/22 16:15:39, rho: x^2 + 1, starting 1000 iterations on C102
06/20/22 16:15:39, pm1: starting B1 = 150K, B2 = gmp-ecm default on C102
06/20/22 16:15:39, current ECM pretesting depth: 0.00
06/20/22 16:15:39, scheduled 30 curves at B1=2000 toward target pretesting depth of 34.00
06/20/22 16:15:39, ecm: commencing 32 curves using AVX-ECM method on 118623671158657249840563647393505134366149888805337835488254472697242875228844619560545884621085119441, B1=2k, B2=200k
06/20/22 16:15:39, ecm: finished 32 curves using AVX-ECM method on C102 input, B1=2k, B2=200k
06/20/22 16:15:39, current ECM pretesting depth: 15.19
06/20/22 16:15:39, scheduled 74 curves at B1=11000 toward target pretesting depth of 34.00
06/20/22 16:15:41, Finished 74 curves using GMP-ECM method on C102 input, B1=11k, B2=gmp-ecm default
06/20/22 16:15:41, current ECM pretesting depth: 20.24
06/20/22 16:15:41, scheduled 214 curves at B1=50000 toward target pretesting depth of 34.00
06/20/22 16:15:52, Finished 214 curves using GMP-ECM method on C102 input, B1=50k, B2=gmp-ecm default
06/20/22 16:15:52, pm1: starting B1 = 3750K, B2 = gmp-ecm default on C102
06/20/22 16:15:53, current ECM pretesting depth: 25.33
06/20/22 16:15:53, scheduled 430 curves at B1=250000 toward target pretesting depth of 34.00
06/20/22 16:17:25, Finished 430 curves using GMP-ECM method on C102 input, B1=250k, B2=gmp-ecm default
06/20/22 16:17:25, pm1: starting B1 = 15M, B2 = gmp-ecm default on C102
06/20/22 16:17:28, current ECM pretesting depth: 30.45
06/20/22 16:17:28, scheduled 643 curves at B1=1000000 toward target pretesting depth of 34.00
06/20/22 16:28:35, Finished 643 curves using GMP-ECM method on C102 input, B1=1M, B2=gmp-ecm default
06/20/22 16:28:35, final ECM pretested depth: 34.01
06/20/22 16:28:35, c102 cofactor = 118623671158657249840563647393505134366149888805337835488254472697242875228844619560545884621085119441
06/20/22 16:28:35, Total factoring time = 776.2737 seconds

I thought this was fixed.

[thyrex]

The next stage for this number must to be GNFS

[VBCurtis]

No, a C102 can easily be solved by QS. It's not the ideal method, but shouldn't be more than 2-3x slower than GNFS.

[BudgieJane]

Originally Posted by thyrex

The next stage for this number must to be GNFS

According to tune_info the crossover point is 112:

tune_info=11th Gen Intel(R) Core(TM) i5-11300H @ 3.10GHz,WIN64,8.83142e-06,0.197336,12.9645,0.0695818,111.146,42

[James Heinrich]

Why does YAFU (observed in both 1.34 and 2.07) like to do a bit more ECM when you've told it that the target amount of ECM has already been done?

Starting factorization of 2251614550234055854784964846511840271248417541830029305669098830420329426567781418202670939759934333698412447
using pretesting plan: deep
using specified qs/gnfs crossover of 100 digits
using specified qs/snfs crossover of 75 digits
input indicated to have been pretested to t36.33
****************************
current ECM pretesting depth: 36.33
scheduled 379 curves at B1=32765 toward target pretesting depth of 36.33
rho: x^2 + 3, starting 1000 iterations on C109
rho: x^2 + 2, starting 1000 iterations on C109
rho: x^2 + 1, starting 1000 iterations on C109
current ECM pretesting depth: 36.33
scheduled 382 curves at B1=3000000 toward target pretesting depth of 36.33
Finished 382 curves using GMP-ECM method on C109 input, B1=3M, B2=gmp-ecm default
final ECM pretested depth: 37.14

[Tray2]

Yafu (the executable from Github), when run on "101^101+2", first finds the special form, then claims it can't find it.

nfs: searching for brent special forms...
nfs: input divides 101^101 + 2
nfs: searching for homogeneous cunningham special forms...
nfs: searching for XYYXF special forms...
nfs: searching for direct special forms...
nfs: snfs form detection took 0.365021 seconds
nfs: couldn't find special form

And indeed, if I tell it to skip ECM ("-work 60" in command line), it proceeds to GNFS (instead of SNFS).

[Tray2]

With threads=3, YAFU crashes on both 2^256+1 and 2^512+1 somewhere in ECM.

With threads=1, factorization of 2^256+1 succeeds. I have no patience to wait for 2^512+1, but probably

OS: Windows 7, AVX2-capable hardware, YAFU executable from Github.

[BudgieJane]

Originally Posted by James Heinrich

Why does YAFU (observed in both 1.34 and 2.07) like to do a bit more ECM when you've told it that the target amount of ECM has already been done?

When you tell it you've done a depth of 36.33 your input is interpreted as 36.3300000...
When yafu says

scheduled 379 curves at B1=32765 toward target pretesting depth of 36.33

could that 36.33 really represent a number rounded down to 36.33, e.g. 36.334...?

What happens if you tell yafu that input has been pretested to t36.34?

[bsquared]

Yeah, it is a significant digits related issue, printed here with more precision enabled:

07/13/22 10:24:25, input indicated to have been pretested to t36.33
07/13/22 10:24:25, ****************************
07/13/22 10:24:25, current ECM pretesting depth: 36.330000000000
07/13/22 10:24:25, scheduled 379 curves at B1=0 toward target pretesting depth of 36.330000000000
07/13/22 10:24:25, rho: x^2 + 3, starting 1000 iterations on C109
07/13/22 10:24:25, rho: x^2 + 2, starting 1000 iterations on C109
07/13/22 10:24:25, rho: x^2 + 1, starting 1000 iterations on C109
07/13/22 10:24:25, current ECM pretesting depth: 36.331171089718
07/13/22 10:24:25, scheduled 382 curves at B1=3000000 toward target pretesting depth of 36.333333333333

So it really wants to get to 36.3333333333 and you've only indicated you've performed up to 36.33

Putting in -work 36.34 causes it to skip ecm as now the threshold is exceeded. I hope that's acceptable as a work around?