Polynomial selection stage question

Hello.
I'm trying to factor 154 digits number using python driver for GGNFS and MSIEVE. I'm also using GPU version of msieve on GeForce 470 GTX. Unfortunately I can run it only 14 hours in a day. But msieve sets time limit of polynomial selection to 287.50 hours.
To move to next stage I should get ".fb" file, that msieve generates at end of poly-selection. However I have ran msieve in a while some hours and got "project.dat.p" file with intermediate results.
So can I make up own ".fb"-file by hands based on best e-value from "project.dat.p" or this privilege given only to msieve?
My project.dat.p file: http://paste2.org/p/935980

[Batalov]

I had a similar problem.
Here's a possible solution - ProcessExplorer
It lets you put your process to sleep for the day and then wake it up when you are leaving for the night.

You can also suspend that pesky M$oft dialog "You have not rebooted your computer since the last update: reboot now / reboot later" that jumps out every few minutes and sooner a later gets under your mouse click.

[Batalov]

You can also find the best poly so far by doing

grep -n alp dat.p | sort +6rg | head
261:# norm 5.449549e-015 alpha -6.140948 e 3.020e-012
231:# norm 5.088774e-015 alpha -5.937988 e 2.892e-012

#-> that shows that your best polynomial start at line 261

Then do
tail +261 dat.p | head
and add it to your .poly file

[jasonp]

Originally Posted by Batalov

You can also suspend that pesky M$oft dialog "You have not rebooted your computer since the last update: reboot now / reboot later" that jumps out every few minutes and sooner a later gets under your mouse click.

Slightly off-topic, but this is also controlled by a registry entry that you can edit if you have administrative privileges on the machine. If not, you can always drag the window so that only its title bar is visible on your desktop; the warning is only repeated if you do something to the dialog box, so this will stop it annoying you temporarily.

[Random Poster]

Originally Posted by jasonp

Slightly off-topic, but this is also controlled by a registry entry that you can edit if you have administrative privileges on the machine. If not, you can always drag the window so that only its title bar is visible on your desktop; the warning is only repeated if you do something to the dialog box, so this will stop it annoying you temporarily.

Or you could just turn off automatic updates. There's not much point of getting updates if you aren't willing to reboot (or shut down) the computer afterwards, and updates are only released once per month so downloading them manually shouldn't be too difficult.

[frmky]

Should I worry about "error: size score computation failed" messages? I got two in the last run.

On a 178.6-digit number (i.e., log₁₀(N) = 178.6) I found the following polynomial scores after about 5 days x 4 GPUs = 20 GPU-days using Telsa C1060's:

# norm 1.307908e-17 alpha -8.505804 e 7.968e-14 rroots 3
# norm 1.182218e-17 alpha -8.638512 e 7.515e-14 rroots 3
# norm 1.164100e-17 alpha -8.585526 e 7.442e-14 rroots 3
# norm 1.252224e-17 alpha -8.335315 e 7.438e-14 rroots 3
# norm 1.076992e-17 alpha -8.016579 e 7.060e-14 rroots 3
# norm 1.060567e-17 alpha -7.846450 e 6.966e-14 rroots 5
# norm 1.065377e-17 alpha -7.716075 e 6.936e-14 rroots 5
# norm 1.061400e-17 alpha -8.249336 e 6.926e-14 rroots 5

How do these compare to your expectations?

[jasonp]

I've gone to a lot of trouble to get rid of those errors, but I guess they still crop up very rarely. It just means the code aborted two polynomials because the numerical integrator couldn't converge

[Batalov]

Ah, the famous "crap" got a new name?

[frmky]

Originally Posted by jasonp

I've gone to a lot of trouble to get rid of those errors, but I guess they still crop up very rarely. It just means the code aborted two polynomials because the numerical integrator couldn't converge

No worries. If you want to look into it, the number was 2,1195+ and I believe this is the line from the msieve.dat.m file:
17400 16121870647111176931 74778552341263563683890185954076125

The full errors were
integrator failed 2.031708e-21 6.429909e-25
error: size score computation failed
integrator failed 8.527355e-23 6.756194e-25
error: size score computation failed

[jasonp]

Originally Posted by Batalov

Ah, the famous "crap" got a new name?

This annoyance predates the GPU code; actually I hope one of jrk's patches removed the crap for good.

Greg: thanks, I'll try to find out what exploded here. Actually this is the first report of an integration error in a long time.

[jrk]

Originally Posted by frmky

On a 178.6-digit number (i.e., log₁₀(N) = 178.6) I found the following polynomial scores after about 5 days x 4 GPUs = 20 GPU-days using Telsa C1060's:

# norm 1.307908e-17 alpha -8.505804 e 7.968e-14 rroots 3
# norm 1.182218e-17 alpha -8.638512 e 7.515e-14 rroots 3
# norm 1.164100e-17 alpha -8.585526 e 7.442e-14 rroots 3
# norm 1.252224e-17 alpha -8.335315 e 7.438e-14 rroots 3
# norm 1.076992e-17 alpha -8.016579 e 7.060e-14 rroots 3
# norm 1.060567e-17 alpha -7.846450 e 6.966e-14 rroots 5
# norm 1.065377e-17 alpha -7.716075 e 6.936e-14 rroots 5
# norm 1.061400e-17 alpha -8.249336 e 6.926e-14 rroots 5

I'd say anything above about 7.7e-14 for a number that size is pretty nice.