Could the client save the work on a pause or shutdown?

[TiO2]

I frequently have to pause F@H to do other GPU-intensive tasks; unfortunately if I don't time this well I can lose a percent or two or work (since GPU units seem to revert to the last finished integer percent when I pause). Would it be possible to have the data automatically saved when the client is paused or is shut down?

[Joe_H]

The simple answer is no.

There are complications in stopping the processing of the WU data and having it in a consistent state usable for restarting after a pause or stop. One aspect is what data is on the GPU itself, and whether it has been replaced by other video data instead. The checkpoints are one point in time where the WU data is consistent, and verified by a sanity check run on the data in progress.

[MeeLee]

Joe,

Is it possible to implement a feat, to pause at the next feasible state for pausing the client?
When my system is doing upgrades, or I need to reboot the system, I'm ok to press pause and have the client run a few more seconds to a minute to finish the WU to the first available savestate.
It's not imperative to have the client pause the WU right away, but neither do I want to wait until the WU finishes, or each WU reaches a savestate.
Since save states on WUs (I presume) are different on each graphics card, it would mean I could reboot my system when one card just passed it's save state, and lose almost no data, I might on my other cards.

It would be beneficial to have such a feat.
Especially for the windows client, that sometimes force closes FAH for updates.

[foldy]

It would be something like the Finish button but "finish at next checkpoint".

[Mitzel]

I would love something like that too. Even the "finish at next checkpoint" would work much better than what we have currently.

[Joe_H]

Something like this has been proposed several times, there may even be an open ticket for it. But a brief look did not turn up one for me.

In the case of the CPU folding core A7 this has been done. On pausing the A7 core will attempt to write out a checkpoint before exiting, from what I have seen this works about 2/3 of the time. This is possible because the A7 core is based on a different processing method than GPU folding.

Now what you are looking for is this:

When my system is doing upgrades, or I need to reboot the system, I'm ok to press pause and have the client run a few more seconds to a minute to finish the WU to the first available savestate.

However due to the way the GPU core is coded, it often will not just be a short period of a few seconds to a minute, but 15 minutes or more before the next time for a checkpoint can be made.

[bruce]

I, too, have wished we had such an option.

Pause at the next checkpoint could be created in either CPU or GPU versions but the complexity of the code changes would be different.