Aliqueit.exe discussion - Page 29

mklasson · 2009-09-28, 15:45

Quote:

Originally Posted by zarabatana

Sorry for this novice question, but, is possible make aliquot use only yafu with this argument: yafu factor (num)

Not without manually altering the code and recompiling it, no.

Quote:

Originally Posted by zarabatana

yafu is faster than aliqueit using ecm.

Do you mean that calling "yafu factor(num)" and letting yafu do everything (trial,rho,squfof,p+-1,ecm,qs) factors a number faster than the current setup where aliqueit does trial&rho, gmp-ecm does p+-1&ecm, and yafu does siqs? This would surprise me as I thought gmp-ecm was much faster than yafu's p+-1&ecm. Do you have any particular input number for which this is true, or have I misunderstood you completely?

Bear in mind that you'd have to run several tests even for a single input number in order to get a reasonable estimate of the average time taken as p+-1 and ecm are both probabilistic.

Quote:

Originally Posted by zarabatana

If siqs can't be used with yafu use, use msieve.

I'm not sure what you mean here.

Quote:

Originally Posted by zarabatana

Here isn't the right place, but, why yafu use msieve 1.38 and not 1.43?

I'll happily leave that question to Ben, and I also think it's been answered earlier in this thread, but I don't think msieve 1.38 or 1.43 matters much in the context yafu uses it (only postprocessing QS).

(Oh, and hooray for yafu.ini btw!

)

bsquared · 2009-09-28, 16:18

Quote:

Originally Posted by mklasson

Do you mean that calling "yafu factor(num)" and letting yafu do everything (trial,rho,squfof,p+-1,ecm,qs) factors a number faster than the current setup where aliqueit does trial&rho, gmp-ecm does p+-1&ecm, and yafu does siqs? This would surprise me as I thought gmp-ecm was much faster than yafu's p+-1&ecm.

That would surprise me as well, as I know my ecm and p+-1 are not as fast. Maybe it would be true for certain cases where the speed difference is not too great (small inputs, or lucky ecm curves, or something) and the extra overhead aliqueit has in calling and parsing external programs makes up the difference... but these cases are no doubt rare.

Quote:

Originally Posted by mklasson

I'll happily leave that question to Ben, and I also think it's been answered earlier in this thread, but I don't think msieve 1.38 or 1.43 matters much in the context yafu uses it (only postprocessing QS).

(Oh, and hooray for yafu.ini btw!

)

Correct! Yafu doesn't call an external library, it contains a port of the msieve code (thanks to jason's very friendly licensing agreement, which I happily mimic). This porting was last done when msieve was at version 1.38, but it doesn't really matter for QS.

zarabatana · 2009-09-28, 16:22

Thank you for your response mklasson.
I make one single test: ran aliqueit and get one co factor with 85 digts.
With this co factor opened 2 prompts and ran:
1) aliqueit <seq>
2) yafu factor(cf)
Yafu finish first.

Ecm on aliqueit, on my system, isn't so fast as msieve and yafu.
Again, thank you for your attention and time.

Notes:
1) I use one Athlon 64 x2 4800; both (aliqueit and yafu) was using only one core.
2) the CF was

Quote:

7523910645664573083025092009318373358583041753880039331761557971634348725636632423965347736802467847647

zarabatana · 2009-09-28, 16:25

bsquared thank you for your attention and response too.
I didn't know that

Quote:

it doesn't really matter for QS.

I'm new here, so I'm sorry for any inconvenient.

10metreh · 2009-09-28, 16:26

Quote:

Originally Posted by zarabatana

Thank you for your response mklasson.
I make one single test: ran aliqueit and get one co factor with 85 digts.
With this co factor opened 2 prompts and ran:
1) aliqueit <seq>
2) yafu factor(cf)
Yafu finish first.

Ecm on aliqueit, on my system, isn't so fast as msieve and yafu.
Again, thank you for your attention and time.

Notes:
1) I use one Athlon 64 x2 4800; both (aliqueit and yafu) was using only one core.
2) the CF was

Was the <seq> 108416, or was it that large number?

zarabatana · 2009-09-28, 16:29

10metreh
Yes, the <seq> is 108416.
Thank you for your attention

10metreh · 2009-09-28, 16:34

Quote:

Originally Posted by zarabatana

10metreh
Yes, the <seq> is 108416.
Thank you for your attention

Aliqueit doesn't actually ever "finish" unless you stop it, it crashes, or the sequence terminates. Did aliqueit move onto the next line, and if not, did it call yafu?

zarabatana · 2009-09-28, 16:44

Sorry, I probably don't explain right.
I stop aliqueit when it found the cf that I told.
Then I open one new prompt and run, ate the same time yafu and aliqueit.
I hope at this moment any doubt was solved, but, if not, fell free to point my mistakes or ask for more explanations.
Thank you for your time.

mklasson · 2009-09-28, 17:38

Quote:

Originally Posted by zarabatana

Sorry, I probably don't explain right.
I stop aliqueit when it found the cf that I told.
Then I open one new prompt and run, ate the same time yafu and aliqueit.
I hope at this moment any doubt was solved, but, if not, fell free to point my mistakes or ask for more explanations.
Thank you for your time.

Like 10metreh pointed out, aliqueit will just continue working on the next number in the aliquot sequence instead of quitting. A more appropriate comparison would be to run

Code:

aliqueit -q 7523910645664573083025092009318373358583041753880039331761557971634348725636632423965347736802467847647

and

yafu factor(7523910645664573083025092009318373358583041753880039331761557971634348725636632423965347736802467847647)

as both programs will then exit after factoring that specific number. But again, you'll have to repeat the test several times (remembering to delete the .elf file in-between) and calculate the averages for each program to get a meaningful comparison. Should you happen to do that and find that yafu is indeed faster than I guess yafu's ecm to qs ratios are significantly better to compensate for gmp-ecm's superior raw speed.

bsquared · 2009-09-28, 17:42

Quote:

Originally Posted by zarabatana

Thank you for your response mklasson.
I make one single test: ran aliqueit and get one co factor with 85 digts.
With this co factor opened 2 prompts and ran:
1) aliqueit <seq>
2) yafu factor(cf)
Yafu finish first.

The CF you quoted has 103 digits while you said you ran the experiment on an 85 digit number... so something isn't consistent.

I suppose it's possible now that yafu is running considerable less ECM curves than aliqueit, and that is what is making it go faster. It would maybe be instructive to compare the respective log files (factor.log from the yafu only run and the aliqueit.log file with this input) for this comparison.

Quote:

Originally Posted by zarabatana

bsquared thank you for your attention and response too.
I didn't know that
I'm new here, so I'm sorry for any inconvenient.

No inconvience at all, thanks for your questions.

mklasson · 2009-09-28, 17:49

Oh, and be sure you're using a gmp-ecm executable that's optimised for your system. Most significantly, make sure you're using a 64-bit version if you're running a 64-bit OS. http://gilchrist.ca/jeff/factoring/ is nice.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Resuming aliqueit	johnadam74	Aliquot Sequences	4	2016-03-28 12:32
Apparent aliqueit issue with specifying factors	pakaran	Aliquot Sequences	2	2015-09-12 23:10
Using Several Instances of Aliqueit for a large gnfs job	EdH	Aliquot Sequences	6	2011-12-13 18:58
Setting up aliqueit	science_man_88	Aliquot Sequences	185	2011-11-08 12:18
Tried out aliqueit.exe: ggnfs failing	Greebley	Aliquot Sequences	35	2010-02-13 15:23

2009-09-28, 16:29	#314
zarabatana Sep 2009 24₁₀ Posts	10metreh Yes, the <seq> is 108416. Thank you for your attention

2009-09-28, 16:44	#316
zarabatana Sep 2009 2³·3 Posts	Sorry, I probably don't explain right. I stop aliqueit when it found the cf that I told. Then I open one new prompt and run, ate the same time yafu and aliqueit. I hope at this moment any doubt was solved, but, if not, fell free to point my mistakes or ask for more explanations. Thank you for your time.

2009-09-28, 17:49	#319
mklasson Feb 2004 2×3×43 Posts	Oh, and be sure you're using a gmp-ecm executable that's optimised for your system. Most significantly, make sure you're using a 64-bit version if you're running a 64-bit OS. http://gilchrist.ca/jeff/factoring/ is nice.