Aliqueit.exe discussion

[Mini-Geek]

Originally Posted by Greebley

Also is there a maximum size for ecm curves done by aliqueit.exe? Is there ever an number that I would want to manually run more ecm at higher levels than aliqueit does? Not that I think I will be running much higher than the 124. That took a good while to complete. I am guessing a c130 would be measured in weeks?

From aliqueit.cc:

    int ecm_settings[][3] = {
        { 20, 11000, 74 },
        { 25, 50000, 214 },
        { 30, 250000, 430 },
        { 35, 1000000, 904 },
        { 40, 3000000, 2350 },
        { 45, 11000000, 4480 },
        { 50, 43000000, 7553 },
        { 55, 110000000, 17769 },
        { 60, 260000000, 42017 },
        { 65, 850000000, 69408 },    //quick, someone find a way for this to be useful!
    };

From default 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

At input_digits=237 for GNFS, that formula puts it above 65 digits. That means that the highest tier of ECM would only be exhausted on a c237 after an enormous amount of time (I'd guess years on a decent CPU) of ECM. No, you don't have to worry about it. (Now, might you want to reject aliqueit's assumptions somewhere before this and run more or less ECM? Certainly.)
(By the way, the QS and GNFS calculations meet at 97 digits.)

[Greebley]

I determined that I could run 35e7 (350 million) but not 40e7 due to the limit of 2 GB of memory, but if the 85e7 kicks in at 237 digits it doesn't sound like I need to worry about that. It also sounds like aliqueit covers any number I am likely to try to factor.

--

I was curious and tried aliqueit out on the 157 the forum is working on and just the polynomial finding would last 300 hours. I can see why it takes the whole group.

--

I think I saw that the beta msieve has degree 6 (and 4) as well as degree 5. How big does the number have to be for degree 6 become worthwhile - or is that not yet known?

[10metreh]

Originally Posted by Greebley

I determined that I could run 35e7 (350 million) but not 40e7 due to the limit of 2 GB of memory, but if the 85e7 kicks in at 237 digits it doesn't sound like I need to worry about that. It also sounds like aliqueit covers any number I am likely to try to factor.

The 85e7 will actually kick in earlier, at 216 digits. It will finish at 237.

I think I saw that the beta msieve has degree 6 (and 4) as well as degree 5. How big does the number have to be for degree 6 become worthwhile - or is that not yet known?

I think it's around 185 or 190 digits.

[Greebley]

As I understand it aliqueit uses a "probably prime" primality test and that some seq have been checked to prove the primality. Is this already done on all sequences?

If not, since I would be stepping over all the sequences checking for data errors, I could prove primality too. Is there a good way to do this?

Not that I would expect to find any errors, but it would be nice to be able to state that we know the sequences were all correct.

[axn]

Originally Posted by Greebley

As I understand it aliqueit uses a "probably prime" primality test and that some seq have been checked to prove the primality. Is this already done on all sequences?

The database does prove primality. So if the sequence manages to pass aliqueit's checks and is in the db (without any errors), it is fine.

[jasonp]

Originally Posted by Greebley

I think I saw that the beta msieve has degree 6 (and 4) as well as degree 5. How big does the number have to be for degree 6 become worthwhile - or is that not yet known?

It's not known (but will definitely be needed above 200 digits). Right now it also doesn't actually matter, if you're using msieve, because the degree 6 code isn't done yet. Nonethless the plan is to switch at around 180 digits, since the degree 5 code has a very difficult time for problems that large.

[Greebley]

Originally Posted by axn

The database does prove primality. So if the sequence manages to pass aliqueit's checks and is in the db (without any errors), it is fine.

Ah right, that's what worker 4 does all the time - 'primality proof'. Got it.

[Greebley]

The -s option no longer works. Anyone else having that problem?

[10metreh]

Originally Posted by Greebley

The -s option no longer works. Anyone else having that problem?

Fine over here.

[Greebley]

Actually does look like it is working but it no longer tells you the number of factors added. Did it get changed?

[10metreh]

Originally Posted by Greebley

Actually does look like it is working but it no longer tells you the number of factors added. Did it get changed?

Yep, that's gone here. I guess that's the database not aliqueit.