OFFICIAL "SERVER PROBLEMS" THREAD

[kladner]

Here's one for someone to run with and see what happens:
M23562031

I started to run it from 60-67, but even with two cores' affinity it was painfully slow. I restarted from 64, and it outran MISFIT's automatic stats update. No factors where I ran.

[James Heinrich]

Originally Posted by Madpoo

Here's one for someone to run with and see what happens: M23562031

[Tue Apr 28 10:20:23 2015]
UID: JamesHeinrich/mfaktc_GTX580, no factor for M23562031 from 2^1 to 2^70 [mfaktc 0.21 75bit_mul32_gs]

[petrw1]

Originally Posted by James Heinrich

[Tue Apr 28 10:20:23 2015]
UID: JamesHeinrich/mfaktc_GTX580, no factor for M23562031 from 2^1 to 2^70 [mfaktc 0.21 75bit_mul32_gs]

I was of the understanding that mfaktc does not work under 64 bits?
Or there is a special port but it works slowly?

Am I confused?

[James Heinrich]

Originally Posted by petrw1

I was of the understanding that mfaktc does not work under 64 bits? Or there is a special port but it works slowly? Am I confused?

You are confused
v0.20 did not support:
* exponents below M1,000,000
* GPU sieving below 2⁶⁴

v0.21 dropped the exponent limit from M1,000,000 to M100,000
v0.21 also extended GPU sieving below 2⁶⁴. It's faster than CPU sieving, but (significantly) slower than GPU sieving above 2⁶⁴.

[kladner]

Originally Posted by James Heinrich

You are confused
v0.20 did not support:
* exponents below M1,000,000
* GPU sieving below 2⁶⁴

v0.21 dropped the exponent limit from M1,000,000 to M100,000
v0.21 also extended GPU sieving below 2⁶⁴. It's faster than CPU sieving, but (significantly) slower than GPU sieving above 2⁶⁴.

Eek! I just realized that in my last round of experimentation I left v0.20 in residence. Time to fix that!

[Madpoo]

Originally Posted by kladner

I started to run it from 60-67, but even with two cores' affinity it was painfully slow. I restarted from 64, and it outran MISFIT's automatic stats update. No factors where I ran.

I picked one of my dev boxes with 12 cores and I'm having it do TF work on 583 exponents where only sannerud has done anything (at least in v5).

I don't have a fancy schmancy GPU so I'm only taking these up to 2^62.

Once that's done, is there anyone or ones interested in looking at this list for rechecking from 2^62 up to whatever? In almost all of them, sannerud reported them up to 2^67, 2^68 or 2^69 which is more than I'd want to bite off without a GPU or seven available. But at least you'll know they're at least done to 2^62 reliably.

[Mark Rose]

Originally Posted by Madpoo

I picked one of my dev boxes with 12 cores and I'm having it do TF work on 583 exponents where only sannerud has done anything (at least in v5).

I don't have a fancy schmancy GPU so I'm only taking these up to 2^62.

Once that's done, is there anyone or ones interested in looking at this list for rechecking from 2^62 up to whatever? In almost all of them, sannerud reported them up to 2^67, 2^68 or 2^69 which is more than I'd want to bite off without a GPU or seven available. But at least you'll know they're at least done to 2^62 reliably.

Is this a new list separate from the list in the other thread?

[Madpoo]

Originally Posted by Mark Rose

Is this a new list separate from the list in the other thread?

I don't know... I'd have to cross check. Maybe I can get to that tonight.

[Mark Rose]

Originally Posted by Madpoo

I don't know... I'd have to cross check. Maybe I can get to that tonight.

It could save you significant work. Another reason for saving duplicate TF work in the database.

[Madpoo]

Originally Posted by Mark Rose

It could save you significant work. Another reason for saving duplicate TF work in the database.

I just checked against that list. No repeats.

These new ones I'm looking at go from 23.2M to 49.9M, which is about where that other list George created starts out.

By the way, be sure to give George and James a wave and an "attaboy" for working on getting these v4 logs together. I don't know what format they were in originally but James wrangled them into a format that matched the v5 results and we got them added in to the database.

We're not totally done checking it all out for data issues but so far it all seems like it's in good shape.

[Madpoo]

Originally Posted by Madpoo

I just checked against that list. No repeats.

These new ones I'm looking at go from 23.2M to 49.9M, which is about where that other list George created starts out.

Attached is a list I generated of those exponents. I created worktodo.txt compatible entries for them with a starting point of 62-bits since that's what I got them all retested up to. I just put in 69 bits as the "to bits" since that's the furthest sannerud took any of them (some were 67, some 68).

Since these are exponents where I didn't see anyone do any other work at all, it doesn't really matter how far sannerud took them to... nobody took 'em any further.

There will still be some exponents to look at after these. Some of them will be where sannerud did checking of part of the range but someone else took it further. Those are a little more complex to sort out so I'll get to those later. This list ought to be a good start though for anyone interested.