Thinking out loud about getting under 20M unfactored exponents

[LaurV]

Originally Posted by chalsall

Would probably be more computationally efficient than additional P-1'ing.

I was talking P+1, not P-1. For diversity, and experiment. And for overtaking masser
Joking apart, there should be few "handsome" P+1 factors lurking there.

Just out of interest... Why did you ask for the sub-sub ranges you did?

Because in this table, the respective sub-sub-ranges are the ones with 200 and over candidates left. Killing few more candidates here is a small step ahead in "thinking out loud about getting under 200 unfactored exponents in each sub-sub-range". Currently, in the process of "thinking out loud about getting under 20M unfactored exponents (in gimps range)", we stop at "thinking out loud about getting under 2000 unfactored exponents in each xx.xM sub-range".

That is because every time you go one step "deeper", the difficulty increases exponentially.

The "200 expos in each xx.xxM sub-sub-range" would be the next step, which right now is utopic, that's why we don't do it, but you never know what Mrs. Future brings. For details, see the discussion above, in this thread (around this post). Finding more factors in the other 6 sub-sub-ranges is, in my opinion/mind/OCD, of a less importance, because they already have under 200 candidates left, so they did their job, and brought their contribution already, they can rest now. We will scratch those ranges only if we are not able to get 5 more factors in the 4 mentioned sub-sub-ranges.

[LaurV]

On Queen's drums and Freddie's voice:
Tu-du-DUM DUM DUM, tu-dum-dum-dum-du-rum,
Tu-du-DUM DUM DUM, tu-dum-dum-dum-du-rum...

M35191367 has a factor: 18989885268718641372929
found 1 factor for M35191367 from 2^74 to 2^75 (partially tested) [mfaktc 0.21 barrett76_mul32_gs]

Four more to go.
.

[LaurV]

Hey Chris, see what happens to you if you give me assignments!


Two more factors found, total four, two more to go. Working in 35.17 now, almost finished. We have 3 factors in 35M19 and 1 in 35M17. I will finish all 4 sub-sub-ranges regardless of the fact I find the remaining two factors or not, but if I do find them, I may slow down.

I have seen you pulled the 34M already, but honestly, this will need a lot of P-1 before going to a higher bitlevel. I am not ready yet to attack 34M.

Going down from 104M turret by turret, or lot by lot (in Forex trading, 1 lot, or 1 turret, is 100k USD), I have seen that the first "incomplete" range is 49.6M, and everything above it is completed. Good work guys! It seems a lot of effort went into this pet-project.

Is anybody working in 49.6M? I am thinking of some P+1 there, as I have few CPU cores free. Should I? (the P-1 there is quite high, I have seen 2 or 3 expos in 49.69M moving 2 bitlevels (74 to 76) in the last few days, but raising all the range to 76 bits would be a HUGE effort.

OTOH, Eth is comming back... (hopefully...)

[masser]

Originally Posted by LaurV

Is anybody working in 49.6M? I am thinking of some P+1 there, as I have few CPU cores free. Should I? (the P-1 there is quite high, I have seen 2 or 3 expos in 49.69M moving 2 bitlevels (74 to 76) in the last few days, but raising all the range to 76 bits would be a HUGE effort.

I see some of the remaining unfactored in 49.6M have B1<1M and B2<25M. Doing P-1 on those candidates to something with B1>2M B2>75M would make sense before any P+1.

P+1 in that range would be interesting, but it may not be cost-effective. Wayne might be right that the best way to knock out that range is to just throw A LOT of GPU firepower at TF.

[petrw1]

Originally Posted by masser

I see some of the remaining unfactored in 49.6M have B1<1M and B2<25M. Doing P-1 on those candidates to something with B1>2M B2>75M would make sense before any P+1.

P+1 in that range would be interesting, but it may not be cost-effective. Wayne might be right that the best way to knock out that range is to just throw A LOT of GPU firepower at TF.

Based on my personal 20 core CPU farm I rightly or wrongly determined to do further P1 where the current B1/B2 are less than about 500,000/1,500,000 and up them to 15,000,000/45,000,000.

Of course I (or anyone else) could always do more but even this has a calculated rate of 700 GhzDays/Factor and based on the current TF help I'm currently getting it seemed like a good point to hand it off to TF.

[chalsall]

Originally Posted by LaurV

Hey Chris, see what happens to you if you give me assignments!

Excellent... [/Burns]

Originally Posted by LaurV

I will finish all 4 sub-sub-ranges regardless of the fact I find the remaining two factors or not, but if I do find them, I may slow down.

Copy.

Originally Posted by LaurV

I have seen you pulled the 34M already, but honestly, this will need a lot of P-1 before going to a higher bitlevel. I am not ready yet to attack 34M.

I'm just bringing them in for whoever wants to take them (might be nobody). I next need to update the DCTF form so manual procurement is possible.

Anon seems to be doing some serious work to finish off 42.6M. I don't know if he's planning on then taking on 49.6M.

BTW, if anyone is interested, this is the report that Wayne and I have been working with the last few years (in addition, of course, to James' reports). It's quite computationally expensive, so this is generated once an hour at the top of the hour.

[masser]

We've now been factoring the 14.0M range for 15 months: 200 factors have been found, with a mix of TF, P-1, ECM and now a little bit of P+1. 43 factors to go for the under 2000 goal, so we are nearing the finish line.

I recently found the largest factor found as part of the 14.0M effort:
M14040307 has a 115.591-bit (35-digit) factor: 62556435002481120085623767582418577 (P-1,B1=3200000,B2=259200000)

I expect the following approximate returns to finish this range:

10 more P-1 factors
20 P+1 factors
15 mix of P+1, TF and whatever else seems productive (ECM, even more P-1?).

We have gotten some help lately from others (mostly VBCurtis) on the forum. Many thanks!

[VBCurtis]

Originally Posted by masser

We have gotten some help lately from others (mostly VBCurtis) on the forum. Many thanks!

Glad to help! My P-1 work continues on an old 2-core laptop, at about 36 hr/test. The machine that was running ECM is powered down during the heat, it will run occasionally during cool spells this summer.

[LaurV]

35.1M done. 77 factors found totally.

I will finish the remaining ~65 candidates to 75 before taking my toys and going home.
Chris, thanks a lot for serving me the right candidates.

[axn]

Yay

[petrw1]

Originally Posted by LaurV

35.1M done. 77 factors found totally.

I will finish the remaining ~65 candidates to 75 before taking my toys and going home.
Chris, thanks a lot for serving me the right candidates.

Thanks a bunch.
Come back any time