mersenneforum.org

mersenneforum.org (https://www.mersenneforum.org/index.php)
-   YAFU (https://www.mersenneforum.org/forumdisplay.php?f=96)
-   -   yafu ignoring yafu.ini (https://www.mersenneforum.org/showthread.php?t=24817)

chris2be8 2019-10-07 15:52

yafu ignoring yafu.ini
 
Hello,

I've found that yafu ignores yafu.ini even though it's called with a full path to the executable and yafu.ini is in the same directory.

Eg (from the log of my script calling it):
[code]
Mon Oct 7 15:25:54 2019 =>"/home/chris/yafu/yafu" 'factor(2387558269606450709622009990622002596238721211851386839362889832990611106360073)' -v -p -ecm_path /usr/bin/ecm -logfile ggnfs.log -threads 4
10/07/19 15:25:54 v1.34.5 @ rigel,
10/07/19 15:25:54 v1.34.5 @ rigel, ****************************
10/07/19 15:25:54 v1.34.5 @ rigel, Starting factorization of 2387558269606450709622009990622002596238721211851386839362889832990611106360073
10/07/19 15:25:54 v1.34.5 @ rigel, using pretesting plan: normal
10/07/19 15:25:54 v1.34.5 @ rigel, no tune info: using qs/gnfs crossover of 95 digits
10/07/19 15:25:54 v1.34.5 @ rigel, ****************************
[/code]

But yafu.ini in /home/chris/yafu definitely contains tune info for this system (the last line):
[code]
chris@rigel:~/yafu$ cat yafu.ini
%B1pm1=100000
%B1pp1=20000
%B1ecm=11000
%rhomax=1000
threads=4
%pretest_ratio=0.25
%ggnfs_dir=..\ggnfs-bin\Win32\
ggnfs_dir=/home/chris/lasieve4_64/
ecm_path=/usr/bin/ecm
%ecm_path=..\gmp-ecm\bin\x64\Release\ecm.exe
%ecm_path=../ecm/current/ecm
tune_info= Intel(R) Xeon(R) CPU E5-4650 0 @ 2.70GHz,LINUX64,1.73786e-05,0.200412,0.400046,0.0987873,98.8355,2699.98
tune_info=AMD Athlon(tm) II X2 240 Processor,LINUX64,1.29672e-05,0.208991,0.561878,0.0975652,95.8178,2800.02
tune_info=Intel(R) Core(TM) i5 CPU 650 @ 3.20GHz,LINUX64,1.51053e-05,0.203897,0.33923,0.101122,97.4879,3209.99
[/code]

I managed to get round this by adding -ecm_path /usr/bin/ecm to the command. But it was rather confusing. It would help if yafu called with -v said whether it was reading yafu.ini or not.

Another point was that when I ran tune() the first time it spend a few minutes running QS, then aborted because it could not find the lattice siever. I had to run [c]yafu 'tune()'[/c] in the same dir as the lattice siever to make tune work. It would have saved time if tune() checked if it can find the lattice siever *before* benchmarking QS.

Chris

bsquared 2019-10-16 14:03

Thanks for the report - not something I test out very often. I've been pretty busy lately but will try to look into it soon.

bsquared 2019-10-16 16:06

Fix checked into yafu-wip r379.

Behavior is now: If you specify -xover, that will be preferred over what is in tune_info, else if valid tune_info is found, use that xover, else use default xover (currently 95).
Oh, and I now check for sievers before starting any qs tuning, thanks for that suggestion.

chris2be8 2019-10-16 16:18

The tune info was just a side issue, the problem for me was that is was ignoring ecm_path in yafu.ini so ran ECM single threaded. And ignoring ggnfs_dir when running tune.

And a message saying whether it had found yafu.ini would have helped me work out what was going wrong.

But thanks for fixing tune to check for sievers first.

Chris

bsquared 2019-10-16 16:23

It should not be ignoring yafu.ini... let me know if it still seems to be doing that (I forgot to include a message that yafu.ini was found, sorry).

chris2be8 2019-10-17 15:44

Hello,

I'm afraid I don't know if the latest version ignores yafu.ini. The only option to download anything I can find on [url]http://sourceforge.net/projects/yafu/[/url] just gives me yafu-1.34.zip which is exactly the same as the version I'm running. It's probably my slightly old web browser causing the problem, but trying Firefox ESR didn't work either.

I'm not too worried because I can get round the problem by passing yafu parms from my script.

Chris

bsquared 2019-10-17 16:22

[QUOTE=chris2be8;528218]Hello,

I'm afraid I don't know if the latest version ignores yafu.ini. The only option to download anything I can find on [url]http://sourceforge.net/projects/yafu/[/url] just gives me yafu-1.34.zip which is exactly the same as the version I'm running. It's probably my slightly old web browser causing the problem, but trying Firefox ESR didn't work either.

I'm not too worried because I can get round the problem by passing yafu parms from my script.

Chris[/QUOTE]

Everything I've been doing lately has been in the "wip" branch of the repository and will involve a build from source to use. If you are not accustomed to doing that, and if you already have a workaround, then that's probably easier.


All times are UTC. The time now is 03:05.

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