[QUOTE=Bdot;351021]
I just had a quick look over your script, and I'm quite interested in using it as I'm running a number of Linux machines too.

I'd like to refer you to [URL="http://www.mersenneforum.org/showpost.php?p=319728&postcount=53"]this[/URL] post where I briefly explained the file locking that mfakto uses for worktodo and results files. (details in [URL="https://github.com/Bdot42/mfakto/blob/master/src/filelocking.c"]filelocking.c[/URL] :smile:) This should help for directly accessing the files while mfakto is running.
[/QUOTE]

Thanks! I'll try to implement this, shouldn't be too hard :smile:

[QUOTE=TeknoHog;351027]I'll try to implement this, shouldn't be too hard :smile:[/QUOTE]

[URL]https://github.com/teknohog/primetools/blob/master/mfloop.py[/URL]

[QUOTE=TeknoHog;351038][URL]https://github.com/teknohog/primetools/blob/master/mfloop.py[/URL][/QUOTE]

Hmm, I think this approach has a timing issue as the lockfile testing and creation are not atomically.

I'm not really fluent in python, but from what I've just read, fd = os.open(lockfile, O_CREAT | O_EXCL) should do the trick. Iff it returns a positive number, then you got the lock (and can os.close(fd) and remove(lockfile) when done writing).

[QUOTE=Bdot;351206]Hmm, I think this approach has a timing issue as the lockfile testing and creation are not atomically.

I'm not really fluent in python, but from what I've just read, fd = os.open(lockfile, O_CREAT | O_EXCL) should do the trick. Iff it returns a positive number, then you got the lock (and can os.close(fd) and remove(lockfile) when done writing).[/QUOTE]

Thanks. It wasn't quite as simple in Python, but the basic idea works, now updated in Github.

Is there such as thing as mfakto for exponents <2M? Or do we have to run those on CPU?

There is a mfakt[COLOR=Red][B][U]c[/U][/B][/COLOR] for exponents between 2k and 1M, few of us use it but it is not "distributable", you have to ask Oliver for it. The reasons are multiple, first of all the work is not really needed, considering how much ECM that range had, there would be no factor to 2^100 or 2^180 bits or so (depending on the expo), except eventual ECM misses. Second, is slower (because not much optimization was put into it, and it doesn't really fit for expo/bitlevlel/classes on the mentioned range) - you better use your resources to do faster work on the LLTF front - that work is more needed. Again, due to the qty of ECM done, your chances to find factors at 60-70 bits (20-25digits) is minuscule, almost null. Third, only "trusted" people should report "no factors" in that range. Most probably the bitlevels will not be visited ever, and a missing factor would be missed forever (this is more of an "emotional/subjective" consequence, it has no "objective" negative impact, but you know how people are :razz:).

Okay. So is it a problem if my CPU runs them if I'm not "trusted" then?

[QUOTE=blahpy;351613]Okay. So is it a problem if my CPU runs them if I'm not "trusted" then?[/QUOTE]
Hehe, for sure there is no problem![B] You do what kind of work you like[/B]. OTOH, P95 is a bit "safer" in reporting the results, compared with mfaktc, where some teenager eager for credits can just edit the text file and send false results. With P95 one must have at least some knowledge of computing those checksums, and people with knowledge are generally more "mature" in their thinking. My opinion. One who knows, what and how, would not risk his reputation for few GHDays of credit.

[QUOTE=blahpy;351589]Is there such as thing as mfakto for exponents <2M? Or do we have to run those on CPU?[/QUOTE]

Perfect agreement to LaurV, just a little addition: Last time I checked (a few minutes ago), mfakto was accepting exponents >1M (and prime). So for your question about exponents <2M, you have at least half an answer :smile:.

Still, LaurV's comments about usefulness apply.
[code]
./mfakto -tf 1036769 70 71
...
got assignment: exp=1036769 bit_min=70 bit_max=71 ([B]230.65 GHz-days[/B])
Starting trial factoring M1036769 from 2^70 to 2^71 (230.65GHz-days)
Using GPU kernel "cl_barrett32_77"
No checkpoint file "M1036769.ckp" found.
^C
mfakto will exit once the current class is finished.
press ^C again to exit immediately
^Cmfakto will exit NOW!
[/code]Towards 2M, there are exponents TF'd to "only" 63, with less ECM. There you'd start with ~1GHz-days for the first bitlevel.

Uh-oh, I've developed a problem with my combination of MFAKTO x64 0.13 and the AMD Radeon 7770.

Ever since yesterday, it's been crashing periodically. At first I was getting only messages indicating that the AMD driver had crashed, but today I'm also getting notifications of "application crashes." Here is what the latest such notification looks like:

[B][SIZE=1][COLOR=#0066cc][SIZE=1][COLOR=#0066cc]Problem signature:[/COLOR][/SIZE][/COLOR][/SIZE][/B]
[SIZE=1]Problem Event Name: APPCRASH[/SIZE]
[SIZE=1]Application Name: mfakto_x64.exe[/SIZE]
[SIZE=1]Application Version: 0.0.0.0[/SIZE]
[SIZE=1]Application Timestamp: 5199fc1c[/SIZE]
[SIZE=1]Fault Module Name: amdocl64.dll[/SIZE]
[SIZE=1]Fault Module Version: 10.0.1016.4[/SIZE]
[SIZE=1]Fault Module Timestamp: 5065fc35[/SIZE]
[SIZE=1]Exception Code: c0000005[/SIZE]
[SIZE=1]Exception Offset: 0000000000f5102d[/SIZE]
[SIZE=1]OS Version: 6.1.7601.2.1.0.768.3[/SIZE]
[SIZE=1]Locale ID: 1033[/SIZE]

[SIZE=1][SIZE=2](There are 4 "Additional Information" lines containing long strings. I have removed them in case they pose a privacy issue -- don't know enough about this sort of thing to tell -- but will be happy to put them back in if assured that there's no risk.)[/SIZE]

[SIZE=2]The only possible causes I can think of are:[/SIZE]

[SIZE=2]1) It's been hot around here lately, so the GPU is overheating. (The PC is in the basement.)[/SIZE]

[SIZE=2]2) Something in the GPU is going bad after 8 months of 24/7 operation.[/SIZE]

[SIZE=2]Monitoring by GPU-Z indicates that GPU temps reach 79C, although GPU load is only 2% and the fan speed reaches only 46% of capacity at ~2545 RPM. The last time it crashed, it had been running for just 15 minutes.[/SIZE]

[SIZE=2]Any thoughts? Should I try MFAKTO x32 0.13 or 0.12 (either one) to see what happens? Another thing that comes to mind is to open up the case and see if the inside needs dusting.[/SIZE]

[SIZE=2]Thanks in advance for any clues or advice.[/SIZE]

[SIZE=2]Rodrigo[/SIZE]

[/SIZE]

Have you tried shutting it down for a bit and rebooting? I see 24/7 so not sure if that is what you do sometimes...