fond of a factor? Urn yourself to become remains

[Bdot]

Originally Posted by chalsall

If you give me the all of the ranges in question, I could quite easily spider them to find which candidates were TFed at the depths you've specified with mfakto using the two kernels in question.

Thanks. The complete list of ranges is:
1163264 - 1179648
2326528 - 2359296
4653056 - 4718592
9306112 - 9437184
18612224 - 18874368
37224448 - 37748736
74448896 - 75497472
148897792 - 150994944
297795584 - 301989888
595591168 - 603979776
1191182336 - 1207959552
2382364672 - 2415919104
Searching for barrett15_75 and mfakto_cl_71 with "to 2^71" or below is sufficient (these kernel names are unique to mfakto).

I've analyzed the bug thoroughly, also with respect to the other kernels and their limits. I did not find any further way to make the precomputing overflow the modulus, so to the best I can tell: this is it.

The problem is fixed and the fix being tested (I'll send it around to the testers later).

[chalsall]

Originally Posted by Bdot

Searching for barrett15_75 and mfakto_cl_71 with "to 2^71" or below is sufficient (these kernel names are unique to mfakto).

OK, the spidering is still underway, but the preliminary results are encouraging.

All of your specified ranges below 74,448,896 have been completed, and only the following match your specs:

,History,no factor for M2328107 from 2^61 to 2^62 [mfakto 0.08-Win mfakto_cl_71] by "lycorn" on 2011-11-30
,History,no factor for M2330837 from 2^61 to 2^62 [mfakto 0.08-Win mfakto_cl_71] by "lycorn" on 2011-11-30
,History,no factor for M2346413 from 2^61 to 2^62 [mfakto 0.08-Win mfakto_cl_71] by "lycorn" on 2012-01-23
,History,no factor for M2346413 from 2^62 to 2^63 [mfakto 0.08-Win mfakto_cl_71] by "lycorn" on 2012-01-23
,History,no factor for M2346413 from 2^63 to 2^64 [mfakto 0.08-Win mfakto_cl_71] by "lycorn" on 2012-01-23
,History,no factor for M2355799 from 2^61 to 2^62 [mfakto 0.08-Win mfakto_cl_71] by "lycorn" on 2012-01-23
,History,no factor for M2355799 from 2^62 to 2^63 [mfakto 0.08-Win mfakto_cl_71] by "lycorn" on 2012-01-23
,History,no factor for M2355799 from 2^63 to 2^64 [mfakto 0.08-Win mfakto_cl_71] by "lycorn" on 2012-01-23

And just for confirmation, you're not concerned about anything using kernel "mfakto_cl_barrett79" nor "mfakto_cl_barrett79_4". Correct?

There are also a few matches for "mfakto_cl_71_4" up in the 601M range. I'll report on that (and any other Spidy finds) once the other (much larger) ranges complete.

[Bdot]

Originally Posted by chalsall

OK, the spidering is still underway, but the preliminary results are encouraging.

All of your specified ranges below 74,448,896 have been completed, and only the following match your specs:

,History,no factor for M2328107 from 2^61 to 2^62 [mfakto 0.08-Win mfakto_cl_71] by "lycorn" on 2011-11-30
,History,no factor for M2330837 from 2^61 to 2^62 [mfakto 0.08-Win mfakto_cl_71] by "lycorn" on 2011-11-30
,History,no factor for M2346413 from 2^61 to 2^62 [mfakto 0.08-Win mfakto_cl_71] by "lycorn" on 2012-01-23
,History,no factor for M2346413 from 2^62 to 2^63 [mfakto 0.08-Win mfakto_cl_71] by "lycorn" on 2012-01-23
,History,no factor for M2346413 from 2^63 to 2^64 [mfakto 0.08-Win mfakto_cl_71] by "lycorn" on 2012-01-23
,History,no factor for M2355799 from 2^61 to 2^62 [mfakto 0.08-Win mfakto_cl_71] by "lycorn" on 2012-01-23
,History,no factor for M2355799 from 2^62 to 2^63 [mfakto 0.08-Win mfakto_cl_71] by "lycorn" on 2012-01-23
,History,no factor for M2355799 from 2^63 to 2^64 [mfakto 0.08-Win mfakto_cl_71] by "lycorn" on 2012-01-23

I'll redo them quickly (with the fixed mfakto AND mfaktc ;-) )

Originally Posted by chalsall

And just for confirmation, you're not concerned about anything using kernel "mfakto_cl_barrett79" nor "mfakto_cl_barrett79_4". Correct?

There are also a few matches for "mfakto_cl_71_4" up in the 601M range. I'll report on that (and any other Spidy finds) once the other (much larger) ranges complete.

That is correct, the barrett79 kernel is not affected. The trailing _<n> is just the vector size this kernel was used with, so all _<n> need to be included.

Thanks a lot for your help, and also big thank you to dabaichi and Axelsson for finding and reporting the bug!

[Axelsson]

The probability to trigger this bug is one in 128 and only with some distinct factoring depths... so what was the probability that I would run into it?

People that knows me know that with me in the mix the possibility to trigger the bug is close to unity! I like to experiment.

I break every program I get close to.



Thanks for fixing the code so fast!

/Göran

[chalsall]

Originally Posted by Bdot

I'll redo them quickly (with the fixed mfakto AND mfaktc ;-) )

OK. Spidy has found 691 up in the 601M range (all by Axelsson). Nothing else (yet) in the other ranges.

How did you want to handle this Bdot? Would it help if I sed'ed them into "Factor=..." assignment lines, and PM them to you? Or did you want the raw results lines?

[Axelsson]

It seems proper that I should rerun the results I reported. I already have an assignment list of the factors ready to feed to mfakto.

But maybe Bdot wants to double check it against mfaktc.

/Göran

[Bdot]

Originally Posted by chalsall

OK. Spidy has found 691 up in the 601M range (all by Axelsson). Nothing else (yet) in the other ranges.

How did you want to handle this Bdot? Would it help if I sed'ed them into "Factor=..." assignment lines, and PM them to you? Or did you want the raw results lines?

Originally Posted by Axelsson

It seems proper that I should rerun the results I reported. I already have an assignment list of the factors ready to feed to mfakto.

But maybe Bdot wants to double check it against mfaktc.

/Göran

If you could PM a "Factor=..." list to me and Göran, that would be great! I could put a Quadro2000 on it right now.

Göran, thanks a lot for helping me with this ... I'll send you the fixed binary soon - I'd like to run these test with both mfakto and mfaktc, just to be sure.

[Axelsson]

Hi Chalsall

Originally Posted by chalsall

OK. Spidy has found 691 up in the 601M range (all by Axelsson). Nothing else (yet) in the other ranges.

I found 692 in my list.
I've sent a list to Bdot already. If you can send me a raw list with numbers only then I can compare it with mine and see if there is any difference.

/Göran

[chalsall]

Originally Posted by Axelsson

I found 692 in my list.

Yes. 692 total, but one was done with "mfakto_cl_barrett72_4".

I'm about to send you and Bdot links to the raw results lines, and a SED'ed worktodo.txt file. I'll leave it up to you two to work out what ranges you each want to do.

[Axelsson]

Done!

I've re-run all the candidates I tested with mfakto 0.11 earlier and with the new 0.12 code I found 19 factors, two already known.

I even think the new program is a bit faster than the old one.



/Göran

[TObject]

M67389239 has a factor: 1389239703570981962393
k = 2^2 * 7 * 11 * 233 * 143631601 = 10307578214164