Windows restarted while CUDALucas was running. Should I be worried?

[ICWiener]

Yesterday while I was at at work my home PC running Windows 10 decided to apply an update and restart on its own (despite me turning off automatic updates/restarts, but that's another story). CUDALucas was therefore automatically closed by Windows while it was computing.

When I restarted it, I saw that it detected a checkpoint file with a timestamp that corresponds to about the time of the automatic restart, and was able to continue from there. However, I am a bit worried that this checkpoint file might somehow be "corrupted" by the restart and therefore all subsequent computations will give a false result.

Is it worth re-loading my last checkpoint that I saved about a week ago and lose a week of calculations rather than letting CUDALucas continue and risk having a false result once it completes in several months' time?

[retina]

Yes, you should be worried that MS has so much control over "your" computer. Don't simply accept that MS can change "your"computer whenever they please, fix it.

[kriesel]

Originally Posted by ICWiener

Is it worth re-loading my last checkpoint that I saved about a week ago and lose a week of calculations rather than letting CUDALucas continue and risk having a false result once it completes in several months' time?

Probably no problem. I've stopped CUDALucas, mfaktx, and CUDAPm1 by CTRL-C hundreds if not thousands of times. As long as the save operation finished before the reboot, it should be fine. Windows by default waits for the app to close after signaling it to, equivalent to the user signaling it to stop.
I'd set CUDALucas to periodically save checkpoint files to disk though, and closer to hourly, than your current weekly or less often. And institute a regular automatic backup regime. What protection against power disturbances do you have in place?

[ICWiener]

Thank you @kriesel. I know that CTRL-C saves a new checkpoint file, I'm just not sure whether Windows truly waited for that operation to finish before killing the process.

Perhaps I'll revert to my last checkpoint file and lose a week of work, rather than compromise months of future work.

Thank you!

[henryzz]

Originally Posted by ICWiener

Thank you @kriesel. I know that CTRL-C saves a new checkpoint file, I'm just not sure whether Windows truly waited for that operation to finish before killing the process.

Perhaps I'll revert to my last checkpoint file and lose a week of work, rather than compromise months of future work.

Thank you!

If it wasn't fully written that should be obvious when loading the file.

Originally Posted by henryzz

If it wasn't fully written that should be obvious when loading the file.

That is a common problem in Windows...

Linux has the sync command which writes everything to disk (flush buffers).