Aliqueit.exe discussion

[mklasson]

Originally Posted by 10metreh

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

Syd might have changed the syntax of the result page, causing headaches for aliqueit's parsing.

[Mini-Geek]

This request might be out of the scope of aliqueit.exe, but I think it'd be great if it had support for expressions and for reading in a list of numbers to factor. e.g. you could have a file named 3045.txt with this in it:

11018340991488369220774176542232958731662721978339859921879635728252983742300159
3045*2^20-1
3045*2^21-1
3045*2^22-1
...

or maybe even something to automate groups of things, like the ABC format used in many prime-searching apps:

11018340991488369220774176542232958731662721978339859921879635728252983742300159
ABC 3045*2^n-1
20
21
22

and then you'd use a command like: "aliqueit -f 3045.txt -o fac_3045.elf" and it factors each line in 3045.txt and writes it to the fac_3045.elf file.

Due to the lack of these two features (and the fact that aliqueit is the best program I know of to factor a number from start to finish with various methods), here's what I've been doing: (if you decide not to put this feature in, this may be useful to someone else doing bunches of factorizations )

1. Write a PARI/gp command to make a batch file with the expression evaluated already, e.g.

   for (x=16,240, write("3045.bat",concat("aliquot -q ",3045*2^x-1)))

   produces

   aliquot -q 199557119
   aliquot -q 399114239
   aliquot -q 798228479
   ...

2. Add commands like this to the end of the batch file:

   copy alq_*.elf alq_3045.elf
   aliquot -s 0 3045

   and of course move the batch file to the aliqueit directory
3. Make sure there are no other .elf files in the same directory (you can either move them to another folder or copy aliqueit.exe and all necessary files to a blank folder)
4. Run the batch file, wait, and enjoy the factors.
5. (optionally: delete all those .elf files to tidy up the folder)

To anyone: Let me know if you know of a better way to get the job done (wouldn't necessarily have to be with aliqueit, but I don't know of anything better).
The biggest problems with this, besides the inconvenience to do it all, is that aliqueit.exe has to start up again over and over, read its configuration over and over, print out its configuration over and over, and create and write to new .elf files over and over. (who knows how much time all that wastes!)

[jasonp]

Msieve can both parse expressions and perform a list of factorizations in batch mode. Just put multiple numbers in the input file. It should all work fine until you hit an input that requires NFS...

[bsquared]

YAFU can as well, although again, not for NFS inputs.

[Mini-Geek]

Originally Posted by jasonp

Msieve can both parse expressions and perform a list of factorizations in batch mode. Just put multiple numbers in the input file. It should all work fine until you hit an input that requires NFS...

Hm...this works, and is a LOT quicker for small inputs (I think about <50 digits, very roughly), so I'll definitely use it for that. But is the ECM in Msieve as fast as GMP-ECM? Does Msieve calculate the digit level it should factor to as precisely as aliqueit.exe does, or is it rough to factoring to the 5-digit level as the logs imply? aliqueit.exe calculates a different factor level for each incoming digit size, and it uses GMP-ECM, and your choice of YAFU or Msieve for the QS (for me YAFU's slightly faster).
YAFU has similar issues, I know that its ECM is far slower than GMP-ECM's implementation.
(this doesn't surprise me since GMP-ECM is built specifically for ECM, but it should still be considered when looking at overall speed)

[10metreh]

Originally Posted by Mini-Geek

Hm...this works, and is a LOT quicker for small inputs (I think about <50 digits, very roughly), so I'll definitely use it for that. But is the ECM in Msieve as fast as GMP-ECM? Does Msieve calculate the digit level it should factor to as precisely as aliqueit.exe does, or is it rough to factoring to the 5-digit level as the logs imply?

Msieve uses GMP-ECM. However, it's rough to the 5 digit level, so it isn't too good above 90 digits.

[mklasson]

It's a nice idea, and I agree it's a bit unnecessarily restricting to only let aliqueit factor aliquot sequences. I'll add it to the list now, but I expect it'll be a while before it gets removed.

[Mini-Geek]

Originally Posted by mklasson

It's a nice idea, and I agree it's a bit unnecessarily restricting to only let aliqueit factor aliquot sequences. I'll add it to the list now, but I expect it'll be a while before it gets removed.

Ok, thanks.

[mdettweiler]

Originally Posted by 10metreh

Msieve uses GMP-ECM. However, it's rough to the 5 digit level, so it isn't too good above 90 digits.

What about with the -e option? That makes msieve do full ECM.

[henryzz]

Originally Posted by mdettweiler

What about with the -e option? That makes msieve do full ECM.

thats what he is talking about

[Mini-Geek]

Originally Posted by mdettweiler

What about with the -e option? That makes msieve do full ECM.

Without -e, it only searches for 15-digit factors with ECM. With -e it searches for factors in 5-digit factor size chunks starting at 15 (15, 20, ...) which is almost always very inefficient. Compare with aliqueit which sets the factoring depth based on the 1-digit incoming number chunks, and sets the factor size within the thousandth based on that.