Aliqueit.exe discussion

[schickel]

Originally Posted by Dougal

im having trouble with this at the moment.i keep getting this error when it goes to use yafu.

whats my problem??

I think your problem is right here. Aliqueit can't find the "aliqueit.ini" file that tells it where everything is. Since you're not starting the program from a shortcut, that means that the .ini file got unpacked to some other directory....

Look around and see if you can find it.

It's probably still under the tree....

[10metreh]

I think that, because Dougal has not started aliqueit from the aliqueit108 directory, aliqueit is searching in C:\Documents and Settings\X for aliqueit.ini. I agree with Mini-Geek: start it from the aliqueit108 directory.

BTW: there are a couple of adjustments to make in aliqueit.ini that I mentioned in the "Getting Started" thread - have you done them? (Except the figure of 93 for gnfs_cutoff is an old one - change it to around 89 for better results. Edited the Getting started post.)

[Mini-Geek]

Does anyone have some more updated factoring depths for aliqueit? For reference, here are the defaults (and what they mean), from aliqueit.ini:

//Formulae used to determine the maximum factor size we will do ecm to.
//For QS: <qs_k> * input_digits + <qs_m>
qs_k = 0.448
qs_m = -11.26

//For GNFS: <gnfs_k> * input_digits + <gnfs_m>
gnfs_k = 0.235
gnfs_m = 9.4

These are quite outdated, putting the meeting point of the two (where QS or GNFS will take about as long, and so are ECMd to the same depth) at c97. Since GNFS is now faster than QS down to c90 or lower, this must result in lots of inefficient ECM for small GNFS jobs. I could easily mess with the numbers until they match at, say, c90, (e.g. just change gnfs_m to 7.91) but that'd probably make for unintended side effects (too little ECM on larger jobs? still bad parameters for QS jobs?).

[paleseptember]

I am attempting to set up Aliqueit.exe (and associated mess) after a break from the project. I've downloaded ecm, ggnfs, yafu, msieve, sorted out the command line paths (nargghh!), and am crunching away at a random sequence to check that things appear to be working.

At what point should yafu change over to ggnfs? I'm getting screens involving:
====
Welcome to Yafu
====
*** Working *** (abbreviated)

I don't remember seeing these before. Have I screwed up something in my config, or has the program changed in the last six months or so?

[Mini-Geek]

Originally Posted by paleseptember

At what point should yafu change over to ggnfs?

Around 89 digits.

Originally Posted by paleseptember

I'm getting screens involving:
====
Welcome to Yafu
====
*** Working *** (abbreviated)

I don't remember seeing these before. Have I screwed up something in my config, or has the program changed in the last six months or so?

YAFU has (somewhat recently) changed to showing a rather terse screen output by default. You can adjust this with one or more -v commands, which you can add via the yafu.ini file. It also has support for multi-threading now. Something like this as yafu.ini will have multithreading and the verbose screen output of the older YAFUs:

threads=2
v=true
v=true

[bsquared]

Originally Posted by paleseptember

I don't remember seeing these before. Have I screwed up something in my config, or has the program changed in the last six months or so?

Yes, I'm not in front of my code right now but I seem to remember changing things so that the banner is always displayed. Above a certain digit limit I think aliqueit dumps all screen output to dev/null, so they will eventually disappear. Or, you can set the following in yafu.ini:

silent=true

[mdettweiler]

Originally Posted by Mini-Geek

Around 89 digits.

YAFU has (somewhat recently) changed to showing a rather terse screen output by default. You can adjust this with one or more -v commands, which you can add via the yafu.ini file. It also has support for multi-threading now. Something like this as yafu.ini will have multithreading and the verbose screen output of the older YAFUs:

threads=2
v=true
v=true

What exactly is the difference between one -v and two -v's? I've tried both before but there doesn't seem to be any difference upon cursory examination. (I personally prefer to run yafu with at least one -v most of the time so I can see the progress. Quite frankly, I would think it slightly more logical to, at the very least, make something that shows progress and ETA the default verbosity level, with the terse "Working" message triggered by a flag, but myabe that's just me. )

[bsquared]

Originally Posted by mdettweiler

What exactly is the difference between one -v and two -v's? I've tried both before but there doesn't seem to be any difference upon cursory examination. (I personally prefer to run yafu with at least one -v most of the time so I can see the progress. Quite frankly, I would think it slightly more logical to, at the very least, make something that shows progress and ETA the default verbosity level, with the terse "Working" message triggered by a flag, but myabe that's just me. )

The only difference lies in if you run factor(). Two v's prints out more time measurements/estimates for the adaptive ECM runs. One v only prints the progress.

Thanks for the feedback... I don't know until people tell me what they find annoying (or nice) :)

Progress for the SIQS is easy. What would you prefer to see as far as the P+1, P-1, and ECM runs, when using factor, as the default verbosity? Was it better before (i.e., what you get with one -v now)?

[mdettweiler]

Originally Posted by bsquared

The only difference lies in if you run factor(). Two v's prints out more time measurements/estimates for the adaptive ECM runs. One v only prints the progress.

Thanks for the feedback... I don't know until people tell me what they find annoying (or nice) :)

Progress for the SIQS is easy. What would you prefer to see as far as the P+1, P-1, and ECM runs, when using factor, as the default verbosity? Was it better before (i.e., what you get with one -v now)?

I kinda like what you get with one -v now, though I can see why that might not be ideal as a default setting; some might prefer to actually be able to see the last few lines of a sequence in aliqueit. How about something like this:

Starting SIQS on cx: <number>
===Sieving in progress (1 thread)===
x relations found, x needed (this line would be replaced with "QS complete, x relations found, elapsed time: x" at the end of this phase)
===Post processing in progress (msieve-1.38)===
Filtering: removing singletons (for example--replace with the current phase whenever a new one is begun, then "Filtering complete, elapsed time: x" at the end)
Linear algebra at x%, ETA x (replace with "Linear algebra complete, elapsed time: x" at the end)
Square root: dependency x in progress (replace with "Square root complete, xth dependency, elapsed time: x" at the end)
SIQS complete, elapsed time x
*** factors found ***
PRPx = x
PRPx = x
ans = 1

Not too verbose, yet you're still not flying blind. ECM/P-1/P+1 could be done in a similar level of verbosity--actually, those are already just about perfect for this kind of verbosity right now with -v.

[bsquared]

Originally Posted by mdettweiler

Not too verbose, yet you're still not flying blind. ECM/P-1/P+1 could be done in a similar level of verbosity--actually, those are already just about perfect for this kind of verbosity right now with -v.

I like that as a happy medium, and it shouldn't be too hard to do. People that like it nice and terse can always use "silent". It probably will be a few days before I can release new code, so if anyone else wants to weigh in, go ahead.

The last few posts are pretty much YAFU centric... maybe move them to the YAFU thread?

[EdH]

Sorry if this is somewhere I should have found it. I did look in a lot of places...

(Even sorrier if this is something I should know.)

Where can I find a list of the codes Aliqueit places in the sequence line it outputs to the console, i.e. un, ru?