(archive)MISFIT

[LaurV]

Originally Posted by swl551

auto fetch from GIMPS

[thinking] It may be the time for me to start testing this toy... [/thinking]

Can I fetch a TF assignment from GIMPS, modify the bitlevels in the worktodo file (to make mfaktc to TF few bits higher, this is what the interface snip shows in your picture), do all the work, and report the result only once, at the end? (I mean, for each exponent, don't report the intermediary work until the 73rd bit is done, otherwise I lose the assignment after primenet's bitlevel is done). Or stop auto reporting and do the report manually, at the end, I won't mind to write a batch to do that. If we can do this, then the gpu72 won't be necessary anymore. Otherwise, I may wait until your program can cope with chalsall's toys. I mean auto fetch from gpu72.

[swl551]

Originally Posted by LaurV

[thinking] It may be the time for me to start testing this toy... [/thinking]

Can I fetch a TF assignment from GIMPS, modify the bitlevels in the worktodo file (to make mfaktc to TF few bits higher, this is what the interface snip shows in your picture), do all the work, and report the result only once, at the end? (I mean, for each exponent, don't report the intermediary work until the 73rd bit is done, otherwise I lose the assignment after primenet's bitlevel is done). Or stop auto reporting and do the report manually, at the end, I won't mind to write a batch to do that. If we can do this, then the gpu72 won't be necessary anymore. Otherwise, I may wait until your program can cope with chalsall's toys. I mean auto fetch from gpu72.

It can do all the things you described and more. It has been able to manually fetch work from GIMPS for a while now with the click of a button. Fully automated fetching is in the works and almost done.

A note on the reporting of completed assignments. If you have Stages=1 in your mfaktX.ini MISFIT may report intermediate results depending on the timing of the results upload. You can turn off auto-results uploading and do it manually if you prefer. Also I could code to not report intermediate results, but it requires some overhead in determining what to export. Not a big deal except that it increase chance of IO collision with results, workToDo and mfaktC. MISFIT and MFAKTO both implemented the .LCK file-locking system to prevent IO collisions, MFAKTC has not. When MFAKTC does implement file-locking the risk of collision will be gone.

Reporting of intermediate results has never caused me a problem since the next upload will complete the range and the chance of someone getting the same exponent from GIMPS is very low in the few hours between uploads. If that not tolerable or ideal consider it a limitations for now. FLASHJH may have additional insight on this and he is the heavy hitter in the MISFIT community.

See additional screenshot covering results uploading and stalled process detection.

[flashjh]

Originally Posted by swl551

If the process does not start, meaning it cannot execute the external program, MISFIT will throw an exception and the Alert Panel will be displayed.

Ok. I was clicking update stat and it didn't run, but I never tried setting update to 5 min. I posted at 1852(L) and at 1946(L) it ran fine, I just didn't get to see it. As long as it's running, I'm happy. Actually, it tried to run a few more times but had nothing to do, so it returned 1 and MISFIT reported that it returned 255. Is 255 expected for a 1 (or do you just report 255 for any non-zero return)?

Seems to be perfect, thanks again.

[swl551]

Originally Posted by flashjh

Ok. I was clicking update stat and it didn't run, but I never tried setting update to 5 min. I posted at 1852(L) and at 1946(L) it ran fine, I just didn't get to see it. As long as it's running, I'm happy. Actually, it tried to run a few more times but had nothing to do, so it returned 1 and MISFIT reported that it returned 255. Is 255 expected for a 1 (or do you just report 255 for any non-zero return)?

Seems to be perfect, thanks again.

I return what the OS returns to me as the exit code. I do not override it with my own value.

if (P.ExitCode != 0)
                                PopSafeErrorForm(P.StartInfo.FileName + " returned error " + P.ExitCode.ToString());

[flashjh]

Cool. No big deal as long as we're not trying to use the exit code, but I'll look at the exit codes later.

[swl551]

2.3.0 beta 3 with Full Work Fetch is ready for closed beta. Please contact me directly for a copy.

[Dubslow]

Originally Posted by LaurV

If we can do this, then the gpu72 won't be necessary anymore.

Well, the point of GPU272 was to distribute lower-than-accessible-from-PrimeNet expos for TF by quick workers. Auto-assignment was never the goal. I'd thus argue that even with autofetch-from-GIMPS, it would still be a good thing to mix in some GPU272 exponents in there.

Of course, is MISFIT could auto fetch from GPU272, that would be supremely awesome...

[kracker]

Originally Posted by Dubslow

Of course, is MISFIT could auto fetch from GPU272, that would be supremely awesome...

++1

[swl551]

Originally Posted by Dubslow

....
Of course, is MISFIT could auto fetch from GPU272, that would be supremely awesome...

In red circle of the screenshot is where you put your GPU72 fetch utility and as long as it appends data to MISFITworkToDo.txt it will be fully integrated.

Beyond that there is nothing more I can do.

[Dubslow]

Well... as a suggestion for chalsall, if you merely implemented the same API as PrimeNet (perhaps with a few less restrictions w.r.t. things like HWID, GUID, etc.), then it should be almost trivial for swl551 to implement GPU272 autofetch (perhaps almost as simple as `sed -e s/http:\/\/v5.mersenne.org\//http:\/\/gpu72.com\//g` ). And it should be fairly easy (relatively) for you to implement such an interface, since you've already done the interface part (not the backend) for the Prime95 proxy.

[swl551]

Originally Posted by Dubslow

Well... as a suggestion for chalsall, if you merely implemented the same API as PrimeNet (perhaps with a few less restrictions w.r.t. things like HWID, GUID, etc.), then it should be almost trivial for swl551 to implement GPU272 autofetch (perhaps almost as simple as `sed -e s/http:\/\/v5.mersenne.org\//http:\/\/gpu72.com\//g` ). And it should be fairly easy (relatively) for you to implement such an interface, since you've already done the interface part (not the backend) for the Prime95 proxy.

Currently I execute HTTP gets to mersenne.org and parse out the response strings. I'd do the same for GPU72. Really they should have web services for this stuff.

Can you point me to documentation on

sed -e s/http:\/\/v5.mersenne.org\//http:\/\/gpu72.com\//g

as I have no idea what it is.