mersenneforum.org > YAFU YAFU-1.34
 Register FAQ Search Today's Posts Mark Forums Read

2013-06-23, 10:21   #111
Andi_HB

Mar 2007
Germany

26410 Posts

Quote:
 Originally Posted by henryzz Has something been fixed between 1.34.2 and 1.34.5?
Changes since 1.34.2:

v 1.34.5
+ (non-source) re-link x64 binary with ecm-6.3
+ allow brent special forms to be detected during factor() runs when the input is
partially factored

v 1.34.4
+ chose gnfs over snfs if appropriate during nfs poly selection
+ new parameter -gnfs to force use of gnfs over snfs
+ 64 bit asm base-2 fermat prp test for use in siqs DLP

v 1.34.3
+ add some documentation to the med_sieve_32k_sse4.1.c
+ move compiler definitions specific to smallmpqs.c into that file where they can be seen
+ (non-source) re-link all binaries with new gmp and gmp-ecm versions

v 1.34.2
+ fixed bug with cunningham/hcunningham algebraic reduction poly generation

2013-06-23, 20:09   #112
jasonp
Tribal Bullet

Oct 2004

354310 Posts

Quote:
 Originally Posted by danaj Ah ... prime = 0, so root2 is 0 and that loop will never exit.
When that happens in my code, it's usually a buffer overflow. Good luck tracking it down when it's a stack buffer, valgrind cannot catch those.

2013-06-23, 22:25   #113
henryzz
Just call me Henry

"David"
Sep 2007
Cambridge (GMT/BST)

2·17·173 Posts

Quote:
 Originally Posted by danaj I didn't have any problem with svn build 316 (1.34.5) with the c31. On the other hand, I get what looks like similar behavior on #103's c41. It seems to hang for quite a while (I'm not patient enough to see if it will finish after hours). gdb says it's in factor/qs/med_sieve_32k.c line 169: SIEVE_1X with i = 701, med_B = 704, and doesn't seem to proceed. Ah ... prime = 0, so root2 is 0 and that loop will never exit. fb->prime[700] = 11161 fb->prime[701] = 0 fb->prime[702] = 11173 clearly something has gone wrong. valgrind doesn't indicate any issues at runtime.
I also get a hang with that c41. I get it at a different relations count than EdH.

If it is any help tracking the bug, An older version of yafu(1.32) behaves differently. smallmpqs is not automatically selected on small numbers after calling with siqs in that version. siqs fails after sieving very slowly(input too small). smallmpqs(what we are using in 1.34.2) works fine so the bug for the c31 has been introduced past that point.

I also tried the 32-bit windows version of 1.34.2. That works fine on the c31 but still hangs on the c41.

My usual platform is the win64 binary. I just downloaded 1.32.5 and that fixes the c31. It must be a rogue binary.
Still worth attempting to fix the c41 though.

 2013-06-23, 23:00 #114 danaj   "Dana Jacobsen" Feb 2011 Bangkok, TH 22×227 Posts Using gcc --fsplit-stack or -fstack-protector --param ssp-buffer-size=1 doesn't help either. It looks like the factor_base medB and B are both ok. All the preceeding loops in factor/qs/med_sieve_32k.c have this code: Code:  // invalid root (part of poly->a) if (prime == 0) continue; but the last one does not. It looks like _SSE2_SMALL_PRIME_SIEVE_14b will set some fb->prime values to 0. Inserting those lines at line 170 after CHECK_1X_DONE makes this successfully run for me, though I am not familiar enough with all the guts to know if this is completely correct.
 2013-06-25, 14:11 #115 bsquared     "Ben" Feb 2007 22×3×293 Posts Thanks for your help Dana. I'll take a closer look, but likely that's the fix.
 2013-09-22, 08:29 #116 Antonio     "Antonio Key" Sep 2011 UK 10238 Posts Recently installed v1.34.5 (Windows 7 64bit) from sourceforge. Is there a way to stop this version outputting thousands of lines of numbers to screen during poly select, without losing the normal progress messages?
 2013-09-22, 10:28 #117 swellman     Jun 2012 2·1,543 Posts There is some discussion of the issue here, though it may be a bit dated now.
2013-09-22, 11:03   #118
Antonio

"Antonio Key"
Sep 2011
UK

32·59 Posts

Quote:
 Originally Posted by swellman There is some discussion of the issue here, though it may be a bit dated now.
Thanks for the link, as an interested user rather than a software guru it will be a matter of waiting for the clever guys to graciously provide the Windows executable then.

 2015-11-05, 15:24 #119 rmd   Nov 2015 2 Posts Yafu binaries or sources for Linux Hello, I could not find binaries or sources (as to compile myself) for YAFU 1.34 (or newer) for Linux x86 or x64, where can I download from? Thank you
 2015-11-05, 16:24 #120 Dubslow Basketry That Evening!     "Bunslow the Bold" Jun 2011 40

 Similar Threads Thread Thread Starter Forum Replies Last Post bsquared YAFU 1276 2019-01-12 04:46 EdH YAFU 8 2018-03-14 17:22 storflyt32 YAFU 2 2015-06-29 05:19 bsquared YAFU 12 2012-11-08 04:12 bsquared YAFU 21 2012-09-04 19:44

All times are UTC. The time now is 16:38.

Fri Jul 30 16:38:20 UTC 2021 up 7 days, 11:07, 0 users, load averages: 1.81, 1.84, 1.78