mfaktc: a CUDA program for Mersenne prefactoring

[tServo]

Originally Posted by storm5510

In this case, is it using Linux drivers or Windows drivers? I believe it to be the latter. I don't see a reason why it would not. I have no need to run a GPU program with it.

Perhaps this documentation might answer some questions.
It is from the latest Cuda toolkit ( 11.x ) and shows how to run the Cuda toolkit from WSL2.

https://docs.nvidia.com/cuda/wsl-user-guide/index.html

[kriesel]

Originally Posted by aheeffer

What are the odds? Mfaktc found two factors in the same 79-bit range:

387687056493187475315017 and 413121450787122123422543 divide M333114347

Slim, but cumulative. In a single attempt,

in the same bit level,
(bitlevel-1)/bitlevel * (ln(bitlevel/(bitlevel-1)))² / 2 =
78/79 * (ln(79/78))² / 2 = 8.03E-5.

In the same class c of the same bit level,
(bitlevel-1)/bitlevel * (ln(bitlevel/(bitlevel-1)))² / 2 / c =
78/79 * (ln(79/78))²/2/c. Assuming more-classes, c=960; 8.35E-8

Make enough TF runs (tens of thousands or more), and finding two factors in the same bit level occurring at least once becomes almost inevitable.

https://www.mersenneforum.org/showpo...82&postcount=5

I hit these recently:

Manual testing    104615629    F    2020-10-16 14:48    0.0    Factor: 75292655729231867474113 / TF: 75-76    72.7737
Manual testing    104615629    F    2020-10-16 14:48    0.0    Factor: 52035700039723086976271 / TF: 75-76    33.7870

That's once, in the past 6 months reviewed.

[James Heinrich]

It's harder to see from the data since many people (most?) will stop factoring immediately after finding a factor, so unless the factor pair is in the same class the second factor wouldn't be discovered.

[kriesel]

Originally Posted by James Heinrich

It's harder to see from the data since many people (most?) will stop factoring immediately after finding a factor, so unless the factor pair is in the same class the second factor wouldn't be discovered.

Right. Depends on settings in this part of the mfaktc.ini (my default settings shown)

# Allow to split an assignment into multiple bit ranges.
# 0 = disabled
# 1 = enabled
# Enabled Stages make only sense when StopAfterFactor is 1 or 2.
# Do not change this in the middle of a run which spans over multiple
# bitlevels, in this case mfaktc will ignore the checkpoint file and
# restarts from the beginning.
#
# Default: Stages=1

Stages=1


# possible values for StopAfterFactor:
# 0: Do not stop the current assignment after a factor was found.
# 1: When a factor was found for the current assignment stop after the
#    current bitlevel. This makes only sense when Stages is enabled.
# 2: When a factor was found for the current assignment stop after the
#    current class.
#
# Default: StopAfterFactor=1

StopAfterFactor=1

Computing credit is given as if the setting was StopAfterFactor=2. (Which is less than the computing performed when StopAfterFactor=1, in the case of factor(s) found.)

[James Heinrich]

Originally Posted by kriesel

Computing credit is given as if the setting was StopAfterFactor=2. (Which is less than the computing performed when StopAfterFactor=1, in the case of factor(s) found.)

I believe credit should be given appropriately depending whether the bit level was finished or not. If this is not the case please email/PM me a note/reminder with a screenshot of manual form output and I'll look into it.

[kriesel]

Originally Posted by James Heinrich

I believe credit should be given appropriately depending whether the bit level was finished or not. If this is not the case please email/PM me a note/reminder with a screenshot of manual form output and I'll look into it.

I've seen it happen repeatedly that TF factor found credit given is less than the full corresponding bit level. (Unless I've been reading it wrong somehow.)
Will email you next time I (a) hit a factor, and (b) remember to do the screen capture(s) and email. Forum PM does not support attachments IIRC.

[James Heinrich]

If you complete the bit level after finding factor it should give full credit, if you stop after factor found then scaled credit is appropriate. I'm out of town for 2 days but can double check the code when you email me.

[Viliam Furik]

Originally Posted by James Heinrich

I believe credit should be given appropriately depending whether the bit level was finished or not. If this is not the case please email/PM me a note/reminder with a screenshot of manual form output and I'll look into it.

I can confirm, mersenne.org gives credit only for work done before the factor is found if it's found. But neither I have a fresh factor, so I can't provide evidence immediately.

[LaurV]

Yep, it works as expected. See discussion around, where somebody complained the given credit for factor was only 70% or so. I have many factors which got lower credit than completing full bitlevel. Which is normal, as the work done to find the factor was less than the full bitlevel, sometimes very low if the factor was found in first classes.

[kruoli]

Originally Posted by LaurV

Yep, it works as expected.

Only if you stop after finding a factor! Yes, that's the default option, but continuing should also be credited, is what James says.

[LaurV]

Oh, you mean the other way around... Never tried that, because I always stop after class, but it should be easy to try, just take a factor line already reported and edit the asterisk. I will try when I reach home in about one hour (still at job).