Yafu

[bsquared]

Originally Posted by yoyo

A chance to get an updated version of yafu soon where nfs runs instead of siqs for larger composites?
In the meantime we have factored nearly everything until C96 in factordb and I would now deploy also the nfs binaries in the Boinc project to use nfs for larger composites with yafu.
yoyo

Have you tried the pre-compiled windows binaries I provide? They were compiled with mingw-64 and shouldn't have this problem. I assume you have been distributing binaries you have compiled yourself using visual studio, is that right?

p.s., Wow, there has been a lot of progress with this project! Enough that I'm wondering where it makes sense to stop... i.e. when will this indiscriminate factoring start annoying folks who run aliquot sequences (or something similar) and don't want "their" work trampled on. Any comments from aliquoters?

[yoyo]

I use your precompiled binaries. Also the log above was from your precompiled binary.
yoyo

[bsquared]

The Win32 pre-compiled binaries? That would make sense. The x64 precompiled binaries should print a header that looks like this:

09/13/11 16:12:51 v1.28.6 @ KEE, System/Build Info:
Using GMP-ECM 6.3, Powered by GMP 5.0.1
detected Intel(R) Xeon(R) CPU           X5687  @ 3.60GHz
detected L1 = 32768 bytes, L2 = 12582912 bytes, CL = 64 bytes
measured cpu frequency ~= 3586.218520

For systems that support it, the x64 binary is quite a bit faster than the Win32 one for siqs work. It also shouldn't have the nfs switchover problem.

[bsquared]

Originally Posted by yoyo

A chance to get an updated version of yafu soon where nfs runs instead of siqs for larger composites?
In the meantime we have factored nearly everything until C96 in factordb and I would now deploy also the nfs binaries in the Boinc project to use nfs for larger composites with yafu.
yoyo

I've checked some of the logs from the BOINC project on my computer and found that I'm getting *both* the 32 bit and 64 bit versions of yafu from BOINC. The 32 bit versions, as expected (because of the Win32 bug), are not attempting to switch to nfs (one is currently crunching on a c97, which will take maybe twice as long as it could take using 64 bit siqs). The 64 bit versions *are* trying to switch (again on a C97), but are not finding the gnfs binaries and are instead reverting to siqs.

Any ideas why I would be getting 32 bit versions of yafu downloaded to my 64 bit system by BOINC?

[RichD]

Originally Posted by bsquared

p.s., Wow, there has been a lot of progress with this project! Enough that I'm wondering where it makes sense to stop... i.e. when will this indiscriminate factoring start annoying folks who run aliquot sequences (or something similar) and don't want "their" work trampled on. Any comments from aliquoters?

Though I am no longer an active participant in aliquot sequences, I still follow all the progress. (Mostly because it is beyond my resources.)

I believe bchaffin has a script that runs against all open, unreserved sequences below 1Mil. The last status update I saw, and I can't find the post ATM, they were around 110 digits (C110). So most sequences of interest are above 110 at this time.

I was looking into a process to *ping* sequences in the range of 1M-4M since they are full of C71-C100 just to keep the queue busy. If there is no work then interest will soon die.

But I held off implementing it for a few days because of another project.

Everybody, please let me know if I am stepping on anyone's toes.

RichD.

[bsquared]

Originally Posted by RichD

I believe bchaffin has a script that runs against all open, unreserved sequences below 1Mil. The last status update I saw, and I can't find the post ATM, they were around 110 digits (C110). So most sequences of interest are above 110 at this time.

I didn't realize sequences were all up that high. I guess there is plenty of room for yoyo's project to operate in.

[yoyo]

Originally Posted by bsquared

I've checked some of the logs from the BOINC project on my computer and found that I'm getting *both* the 32 bit and 64 bit versions of yafu from BOINC. The 32 bit versions, as expected (because of the Win32 bug), are not attempting to switch to nfs (one is currently crunching on a c97, which will take maybe twice as long as it could take using 64 bit siqs). The 64 bit versions *are* trying to switch (again on a C97), but are not finding the gnfs binaries and are instead reverting to siqs.

Any ideas why I would be getting 32 bit versions of yafu downloaded to my 64 bit system by BOINC?

Is there a way to fix the win32 bug to get nfs running also there? I have only a win32 system to test it.
Usually Boinc tells the server that it is running on a win32 or win64 system and than gets the related application. Can you tell me the hostID which gets the win32 version instead the win64 version?
yoyo

[yoyo]

The Boinc client tries to be intelligent and tries to find out which app 32bit/64bit is faster on your computer. I disabled this on the server. Now you should get only 64 bit binaries.
yoyo

[schickel]

Originally Posted by bsquared

p.s., Wow, there has been a lot of progress with this project! Enough that I'm wondering where it makes sense to stop... i.e. when will this indiscriminate factoring start annoying folks who run aliquot sequences (or something similar) and don't want "their" work trampled on. Any comments from aliquoters?

Speaking only for myself, I don't see a problem with this, mainly because my sequences are way above the range that would be factored by the DB. I do 99% of my factoring offline, since my composites are usually >125 digits or so.

Originally Posted by RichD

Though I am no longer an active participant in aliquot sequences, I still follow all the progress. (Mostly because it is beyond my resources.)

What kind of number would you consider within your resources? I'm sure we could find something interesting within your reach....

I believe bchaffin has a script that runs against all open, unreserved sequences below 1Mil. The last status update I saw, and I can't find the post ATM, they were around 110 digits (C110). So most sequences of interest are above 110 at this time.

Yes, he is running a project that works on the composites of unreserved sequences. Last update on the progress (as of July) his workers were working on the c106 range, with ~30% done.

I was looking into a process to *ping* sequences in the range of 1M-4M since they are full of C71-C100 just to keep the queue busy. If there is no work then interest will soon die.

I'd say you're probably right about that....need to keep the workers supplied with WUs.

But I held off implementing it for a few days because of another project.

Everybody, please let me know if I am stepping on anyone's toes.

RichD.

No problem here....

Originally Posted by bsquared

I didn't realize sequences were all up that high. I guess there is plenty of room for yoyo's project to operate in.

Yes, all open sequences have been calculated to at least 110 digits. And it is really nice to have the workers slogging away on everything that isn't reserved. There's nothing more boring than watching a sequence keep on going up with a driver....and it's nice to poll the DB and see a downdriver that developed "in the wild" every once in a while.

[Andi47]

I had to kill a GNFS-128 which was halfway through linalg for a PC restart, and now can't resume it:

I was using aliqueit.exe with the -y flag (for using yafu).

When I killed the job, it said something like that it had caught a ctrl-c signal, but then linalg kept running. When I hit ctrl-c a second time, the job was interrupted immediately.

When I tried to resume (starting aliqueit with -p -y -e flags), a hundred of unusual lines where scrolling over the screen (which error messages "must specify -R flag when savefile already exists"), and then I got an error message "can't open logfile for append".

Seems, that a GNFS-128, which was 98% done, is lost.

Is there a possibility to run aliqueit and pass a -R flag to yafu (or otherwise resume an interrupted yafu factorization while running aliqueit)?

[bsquared]

Originally Posted by Andi47

Is there a possibility to run aliqueit and pass a -R flag to yafu (or otherwise resume an interrupted yafu factorization while running aliqueit)?

Yes: you can use the yafu.ini file. Just add a line R=1 to it and it will run in ggnfs restart mode.