thanks for reply, so, if I use -y switch to do all factorization with yafu, it means that yafu will not use the settings from yafu.ini? (which is supposed to launch siqs, and not nfs). Also, related to the path of ggnfs, that is right in yafu.ini, and yafu(nfs) works well, but this problem appears only from aliqueit. Is it a different syntax to specify the path in aliqueit.ini? like, does it matter if I use / or \ or else? Can I give spaces in the path name? Do I need to use quotes? I think I am missing something here...

Confused...

Is there any posibility to use aliqueit just for combining the factors to get the next term of the sequence, but to do ALL the factorization work with an external program that [U][B]I[/B][/U] specify, launched from aliqueit with the parameters that [U][B]I[/B][/U] want? (please remark the stressed I's).

Thanks a lot in advance.

Is it possible that you have a second copy of yafu.exe in your aliqueit folder? And that this one has a different yafu.ini (with a wrongly specified gnfs path)?

P.S.: Depending on your hardware, there is a crossover line at approx. 95-98 digits: Around this crossover (say when you are not much more than ~2-3 digits off) siqs and gnfs are almost equally fast; above this crossover gnfs is faster. For a c108, for siqs you see, how long it will take, but gnfs (which was chosen automatically) is quite a bit faster.

Thanks for your fast reply, I fully understand what you say. I have to dig more on it. For my computer, yafu's "tune()" sets the "crossover" around 100-105 digits, depending of what I am doing with the computer during "tune" is running. Because at 108 is not a big difference, I prefer to use siqs, because of the fact that siqs is telling me in the beginning how many relations it needs, as I said, so I can have an idea of how much time it needs more till finish, and I can see some progress ("found xxx relations from yyy needed", and xxx is raising). For nfs I have no idea when (and if) it will finish. That is why I manually edited yafu.ini and played with the values, set the crossover higher. When I run yafu independently, it works. When it runs from aliquot program, it is quite arbitrarily, and this is frustrating, for example the C111 was factored with [B]siqs[/B] launched from yafu launched by aliqueit (who automatically selected siqs!), but sometimes, no idea what criteria he uses, he launches nfs. I believe that the selection is done by yafu, and the aliqueit just launches "yafu factor(xxxx)". I would prefer to be able to customize aliqueit in such a way to specify when I want "yafu siqs()" and when I want "yafu nfs()", eventually with other options (-threads xx, -noecm, etc). Slowly I started to figure them out by playing with the ini files and reading the yafu sources (I can not compile them, using an old VS2005, but I can look into them). Are there any aliqueit12 sources available?

By the way, there is no yafu in aliquot folder. I have a folder called "work", and inside there are other 4 folders, yafu (with 1.28.5 inside), msieve (with 1.49 inside), aliquot (with aliqueit12 indide), ggnfs (with all the lasieveXXX.exe, etc). I don't use python, I have perl in a different folder (used for pari). How the aliqueit.ini should look? (specially the lines where paths are specified). Does it need the python stuff? I am starting aliqueit with -y switch. Theoretically, it should pass everything to yafu and go to sleep until yafu reports some factors. The documentation on aliqueit is scarce, I don't know why it needs all the stuff it needs, in the ini file.

[QUOTE=LaurV;273351]thanks for reply, so, if I use -y switch to do all factorization with yafu, it means that yafu will not use the settings from yafu.ini? (which is supposed to launch siqs, and not nfs). Also, related to the path of ggnfs, that is right in yafu.ini, and yafu(nfs) works well, but this problem appears only from aliqueit. Is it a different syntax to specify the path in aliqueit.ini? like, does it matter if I use / or \ or else? Can I give spaces in the path name? Do I need to use quotes? I think I am missing something here...

Confused...

Is there any posibility to use aliqueit just for combining the factors to get the next term of the sequence, but to do ALL the factorization work with an external program that [U][B]I[/B][/U] specify, launched from aliqueit with the parameters that [U][B]I[/B][/U] want? (please remark the stressed I's).

Thanks a lot in advance.[/QUOTE]
I have a GUI for running Aliqueit at [URL="http://www.starreloaders.com/edhall/AliWin/AliWin.html"]AliWin2 - A GUI by Edwin C. Hall for Aliqueit by Mikael Klasson[/URL] that you might be interested in looking over. The page goes into a full explanation of a working install of all the programs. Unfortunately, I've not had an opportunity to test this on a Windows 64-bit OS machine, yet. But, you might be able to catch a couple ideas from the page. The source is also provided. There is a [URL="http://mersenneforum.org/showthread.php?t=13365"]thread about AliWin on this forum[/URL].

(I also have a page of [URL="http://www.starreloaders.com/edhall/AliWin/AliqueitLinstall.html"]Steps to install and set up Aliqueit on an Ubuntu computer[/URL] for linux users, but no GUI, yet.)

The AliWin2 page gives full instructions to achieve a working set of programs for running Aliqueit via AliWin2, but everything should still run manually, if desired.

To answer a few of your direct questions:

- For a Windows system, I have had to use \ instead of / in aliqueit.ini. (factmsieve.py still needs /)
- I always use full paths starting from C:\
- Quotes are a must if there are spaces within the path and optional otherwise.
- I have had trouble with Python distinguishing between c and C. (My ggnfs line has to have an upper case C.) I haven't run Perl. All my Windows paths now begin with upper case C:\...
- [URL="http://mklasson.com/aliqueit112.zip"]Aliqueit112.zip[/URL] from Mikael Klasson's page includes source files in the src folder.
- For all my setups I include YAFU and its ini in the aliqueit folder.

I hope at least some of this is helpful.

[QUOTE=LaurV;273365] Thanks for your fast reply, I fully understand what you say. I have to dig more on it. For my computer, yafu's "tune()" sets the "crossover" around 100-105 digits, depending of what I am doing with the computer during "tune" is running. Because at 108 is not a big difference, I prefer to use siqs, because of the fact that siqs is telling me in the beginning how many relations it needs, as I said, so I can have an idea of how much time it needs more till finish, and I can see some progress ("found xxx relations from yyy needed", and xxx is raising).
For nfs I have no idea when (and if) it will finish. That is why I manually edited yafu.ini and played with the values, set the crossover higher. When I run yafu independently, it works.
[/QUOTE]
As is frequently the case, I've probably mis-communicated the function of the crossover field in the tune_info lines of the yafu.ini file. The crossover is reported as a FYI value only - it is not used by the program to determine the point at which to switch to ggnfs. The crossover point is determined using the other values recorded in the tune_info lines, namely, the coefficients that determine the expected running time of each method (siqs,nfs) as a function of the digit size of the input. So playing with the crossover value would involve changing these coefficients (not recommended).

[QUOTE=LaurV;273365]
When it runs from aliquot program, it is quite arbitrarily, and this is frustrating, for example the C111 was factored with [B]siqs[/B] launched from yafu launched by aliqueit (who automatically selected siqs!), but sometimes, no idea what criteria he uses, he launches nfs. I believe that the selection is done by yafu, and the aliqueit just launches "yafu factor(xxxx)".
[/QUOTE]

Yeah, this should be what happens - when run with -y, aliqueit turns over control of to yafu's factor(), which will automatically determine the best way to proceed. The criteria is to minimize the expected time of a complete factorization given inputs predicting QS and NFS time as well as measurements of time spent doing other things like ECM. I don't know why a C111 would be chosen as optimal with siqs... I'd need to see the .ini file and the logfile to begin to understand that.

[QUOTE=LaurV;273365]
I would prefer to be able to customize aliqueit in such a way to specify when I want "yafu siqs()" and when I want "yafu nfs()", eventually with other options (-threads xx, -noecm, etc). Slowly I started to figure them out by playing with the ini files and reading the yafu sources (I can not compile them, using an old VS2005, but I can look into them). Are there any aliqueit12 sources available?
[/QUOTE]

Currently there is no way to specify a strict cutoff between siqs and nfs. It either happens automatically using tune_info entries or it happens at 95 digits with no tune_info entries. The aliqueit source is h[URL="https://sites.google.com/site/bbuhrow/home/aliquot-sequences/aliqueit112.zip?attredirects=0&d=1"]ere[/URL].

Thanks both for detailed help. I am looking right now into AliWin to see how I can make use of it.

@bb: I knew about the fact there is no way to specify a "hard" cutoff between nfs and siqs. I was talking about the possibility to specify [B]in aliqueit's[/B] ini file (or else) to [B]explicitly [/B]run "yafu siqs()" or "yafu nfs()" instead of "yafu factor()", and more generally, to be able to tell to aliqueit (or to other "wrapper" application like it) [B]what program to launch and [/B]with which parameters. Aliqueit is quite scarce on such feature, or not enough documented. There was no arguing about yafu, only about aliqueit. As you know, yafu is my preferred factoring toy. :smile:. Fast and neat. I fully understand how yafu works and I fully understood siqs algorithm, even did my own implementations for small numbers, to help myself understand what's going on. However, nfs algorithm is still slippery terrain for me. This is an additional reason why I prefer siqs, beside of what I explained before. We are generally afraid of things we don't fully understand :smile:

At long last the workers have finished all the c107s (4188 of them) on open sequences under 1M.

Now on to the c108s -- currently about 2600 to go, but of course there will be many more generated along the way.

From 660:[code] 890 . c181 = [COLOR="Red"]2^3 * 3^2 * 5[/COLOR] * [COLOR="Blue"]759492743757752077394167264747516318068472138837[/COLOR] * [COLOR="blue"]8577699773645811181373240691092327851472636312778873586096412527952983518022415774631214696619438304178458945171149397776984407487[/COLOR]
891 . c181 = [COLOR="red"]2^3 * 3^2 * 5 [/COLOR]* 149 * 92136793617291827118528192645303326285053567 * 1067720609074675907784135312004454076764365970682619148779619910073905157810431774214090863907675720930834522584769245460601147841503
892 . c182 = [COLOR="red"]2^3 * 3^2 * 5[/COLOR] * 31 * 59 * 21334548679760461 * 43469986346582693 * 751783956649954897933022237086010321 * 26113775754970230178461154432639855352846265143121917161867270820209283645793379460313545764882866347258143
[/code]Well, that didin't work out. Looks like someone cracked the c178, maybe ECM? The driver didin't break, though. It was at 890 on the last download.....

[URL="http://factordb.com/sequences.php?se=1&aq=11040&action=last20"]11040[/URL] got the downdriver at height 152.

[QUOTE=unconnected;280963][URL="http://factordb.com/sequences.php?se=1&aq=11040&action=last20"]11040[/URL] got the downdriver at height 152.[/QUOTE]
That's the second highest ever acquisition and the highest above 10^4. Congratulations!

Someone had quite a run with 81192, but I don't see it mentioned anywhere in the forum. From the worker logs:

updating 81192.1235 (was c112, d114, driver: 2^3 * 3): got 81192.3108, c51, d58, driver: 2^4 * 31

Now back up at 100 digits and 3800 terms, with 2^5 * 3. Anyone want to take credit?