Yafu

[James Heinrich]

Originally Posted by bsquared

-logfile NUL -session NUL

Perfect, thank you. Works as desired.

[Random Poster]

Originally Posted by bsquared

I'll get a new version out soon with this (and several other things) fixed.

Here's another thing to fix: some source files try to include ecm.h, but there's no such file in yafu-1.28.4-src.zip.

[bsquared]

Originally Posted by Random Poster

Here's another thing to fix: some source files try to include ecm.h, but there's no such file in yafu-1.28.4-src.zip.

That's part of the gmp-ecm package, not yafu. I suppose I should use <ecm.h> instead of "ecm.h".

[Random Poster]

Originally Posted by bsquared

That's part of the gmp-ecm package, not yafu. I suppose I should use <ecm.h> instead of "ecm.h".

Do you mean it's mandatory to have the gmp-ecm sources available to compile yafu? In that case you should update the readme (which quite clearly states that gmp-ecm is optional) and probably also clean up the makefile and sources (no point to keep HAVE_GMP and HAVE_GMP_ECM around).

[bsquared]

Originally Posted by Random Poster

Do you mean it's mandatory to have the gmp-ecm sources available to compile yafu?

Errr (*checks quick*), yes, I guess it is. That used to not be the case; until recently I had my own implementation as backup to gmp-ecm.

Originally Posted by Random Poster

In that case you should update the readme (which quite clearly states that gmp-ecm is optional) and probably also clean up the makefile and sources (no point to keep HAVE_GMP and HAVE_GMP_ECM around).

Will do, although "clean up source" has been at the top of my todo list for years now, and, well, it's still a mess .

Thanks for the feedback.

[bsquared]

Originally Posted by yoyo

Can I get a static linked version for Linux in the next release?
I'm sending yafu automatic to registered Boinc user and most of them do not have libgmp.so installed.

thanks,
yoyo

Yes, I'll try to get you one.

By the way, it's nice to see this tool getting so much use via your new BOINC project. I joined up and started contributing cycles just to see what BOINC is like - I've never used it before till now.

[yoyo]

I saw you joining and some others fro the forum.
I'm not a math expert. But I like the simple usage of yafu which runs first rho, ecm, pp1, ppm to some extent to find small factors and than runs siqs for the rest. I must not make this by hand. Some Perl scripting and cooperation with Syd and everything was running in Boinc. 230 user have joined until now.
yoyo

[bsquared]

 
v 1.28.5 
+ no longer print SKEW line to nfs.fb files 
+ add issquare and ispow to function list 
+ cleanup of siqs tdiv code: now requires SSE2 
+ added perfect power checks to siqs and nfs 
+ on NFS restart, check any rels found against min_rels instead of always proceeding to filtering 
+ support for large nfs jobs on windows (file size limitations removed) 
+ fix for the "skipping blank line" infinite loop sometimes encountered during -batchfile runs 
+ fixed estimation when allocating memory in SoE wrapper - now gives better estimates and saves memory
+ fixed bug introduced in 1.28.4 where rho() doesn't detect PRP factors correctly (thanks wblipp!) 
+ if gnfs-lasieve binaries are not detected prior to starting an nfs job that requires sieving, siqs is started instead of aborting (and a warning message is printed to the screen and logfile) 
+ added a check for a valid siqs savefile and restart siqs inside of factor()

Hopefully that covers most things reported/requested recently.

@yoyo
I statically linked the linux executables in the download... let me know if people are still having issues with them. This new version should also work better as the BOINC project gets to larger composite values... sooner or later yafu will want to run NFS and find that the user doesn't have gnfs-lasieve binaries. This new version will revert to siqs in that case and not abort (as 1.28.4 and previous versions did). I don't know how complex a process it would be to also distribute gnfs-lasieve binaries in the project...

[LaurV]

Originally Posted by bsquared

 
+ add issquare and ispow to function list

Hopefully that covers most things reported/requested recently.

Thanks a billion! Not that I have an acute need of them right now (still use them occasionally), but that was something I reported, and now you see.... I feel like an important guy!

Still missing that loops and screws... aaa.. sorry.. ifs...

[offtopic]
By the way, after reading few posts about pari/gp on this forum, I started playing with it, and I want to tell you that yafu is between 1.5 and about 40 times faster (depending what I am doing). That's a good point for yafu! But hey, in pari/gp at least I can implement some loops... Pari also helped me to understand better a lot of things that were somehow unclear or unknown to me till now (math related). Because I have this bad habit to read the manual first, every time I get to learn a new language, to see what the new toy can do. All that functions with strange names on different kind of groups and fields puzzled me to death, so I went back to the math books...
[/offtopic]

Question: If you remember my C185, it has already over 1.6 gigs of nsf.dat and all the other stuff. Anything to know if I want to switch to the new version of yafu now? Or I should better wait for the sieving to finish first? Can the new version "take over" on the fly? Just worried a bit.

[bsquared]

Originally Posted by LaurV

Thanks a billion! Not that I have an acute need of them right now (still use them occasionally), but that was something I reported, and now you see.... I feel like an important guy!

Still missing that loops and screws... aaa.. sorry.. ifs...

I may have mentioned it before, but you are free to edit the source code of yafu - not to make it *parse* if's and loops, but to just directly code your if's and loops in C. It really shouldn't be too much harder than coding something in pari/gp. PM me if you want and I can point you to a good place to start.

Originally Posted by LaurV

[offtopic]
By the way, after reading few posts about pari/gp on this forum, I started playing with it, and I want to tell you that yafu is between 1.5 and about 40 times faster (depending what I am doing). That's a good point for yafu! ...[/offtopic]

Originally Posted by LaurV

Question: If you remember my C185, it has already over 1.6 gigs of nsf.dat and all the other stuff. Anything to know if I want to switch to the new version of yafu now? Or I should better wait for the sieving to finish first? Can the new version "take over" on the fly? Just worried a bit.

There is nothing you'd gain from switching, so I'd just continue on the way you are. That said, a switch *should* be seamless - "ctrl-C" in one version and "nfs(<the same number>) -R" in the new version should work just fine. As always if you're nervous, backup the data first.

[yoyo]

Thanks for the yafu update. I deployed the version into the Boinc system.

Regarding gnfs:
On which composite size yafu will use gnfs?
Where can I find gnfs binaries?
How to configure yafu with gnfs?

yoyo